Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Language and the structure of Berkeley's world
(USC Thesis Other)
Language and the structure of Berkeley's world
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
LANGUAGE AND THE STRUCTURE OF BERKELEY’S WORLD
by
Kenneth L. Pearce
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHILOSOPHY)
Defended March 21, 2014
Copyright 2014 Kenneth L. Pearce
Contents
Contents ii
Abbreviations v
Acknowledgments vii
Introduction 1
Chapter 1: The Theory of Meanings 13
1.1 Overview of the Theory . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 The Port-Royalists . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3 Locke . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.4 Sergeant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Chapter 2: Berkeley’s Attack on Meanings 48
2.1 The Dialectical Structure of Berkeley’s Attack . . . . . . . . . . . . 50
2.2 The Case Against Abstraction . . . . . . . . . . . . . . . . . . . . . 54
2.2.1 The Phenomenological Appeal . . . . . . . . . . . . . . . . 55
2.2.2 The Impossibility of Abstract Ideas . . . . . . . . . . . . . 61
2.2.3 The Uselessness of Abstract Ideas . . . . . . . . . . . . . . 77
2.3 Causal Representation . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.4 Primitive Representation . . . . . . . . . . . . . . . . . . . . . . . 89
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Chapter 3: The Rudiments of Berkeley’s Positive Theory 94
3.1 General Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.2 Operative Language . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.3 Mathematical and Scientific Language . . . . . . . . . . . . . . . . 117
3.3.1 Arithmetic and Algebra . . . . . . . . . . . . . . . . . . . . 118
3.3.2 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.3.3 Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
ii
Chapter 4: A Commentary onAlciphron VII 131
4.1 Alciphron’s Critique of Religious Mysteries (xx1-4) . . . . . . . . . 133
4.2 Euphranor’s Refutation of the Theory of Meanings (xx5-7) . . . . . 139
4.3 Euphranor’s Theory of Language (xx8-11) . . . . . . . . . . . . . . 140
4.4 The Nature of Faith (xx11-13) . . . . . . . . . . . . . . . . . . . . 146
4.5 The Nature of Science (xx14-16) . . . . . . . . . . . . . . . . . . . 155
4.6 Euphranor’s Summary and Conclusions (xx16-18) . . . . . . . . . . 158
Chapter 5: Rule-Following 163
5.1 Implicit and Explicit Rule-Following . . . . . . . . . . . . . . . . . 164
5.2 Berkeley’s Internalist Requirement . . . . . . . . . . . . . . . . . . 179
5.3 The Conventional Rules of Language . . . . . . . . . . . . . . . . . 185
5.4 Inference Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
Chapter 6: Reference and Ontology 200
6.1 Sensible Qualities . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.2 Occult Qualities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.3 Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.3.1 Bodies as Linguistic Constructions . . . . . . . . . . . . . . 218
6.3.2 Alternative Interpretations . . . . . . . . . . . . . . . . . . 220
6.3.3 The Richness of Berkeleian Bodies . . . . . . . . . . . . . . 224
6.3.4 Immediate Perception . . . . . . . . . . . . . . . . . . . . . 226
6.3.5 Predication . . . . . . . . . . . . . . . . . . . . . . . . . . 228
6.4 Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
6.5 Spirits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
6.6 Some Ontological V ocabulary . . . . . . . . . . . . . . . . . . . . . 247
6.6.1 ‘Exists’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
6.6.2 ‘Thing’ and ‘Being’ . . . . . . . . . . . . . . . . . . . . . . 254
6.6.3 ‘Real’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.6.4 ‘Same’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6.7 Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6.8 Conclusion: Avoiding Immaterialism? . . . . . . . . . . . . . . . . 263
Chapter 7: Assent and Truth 268
7.1 The Nature of Assent . . . . . . . . . . . . . . . . . . . . . . . . . 271
7.2 The Nature of Truth . . . . . . . . . . . . . . . . . . . . . . . . . . 290
7.2.1 Berkeley and Deflationism . . . . . . . . . . . . . . . . . . 293
7.2.2 Berkeley and Pragmatism . . . . . . . . . . . . . . . . . . . 296
7.2.3 Berkeley and Carnap . . . . . . . . . . . . . . . . . . . . . 301
7.2.4 Realism about Truth, Holism about Meaning . . . . . . . . 306
iii
Chapter 8: The Linguistic Structure of Berkeley’s World 325
8.1 A Literal Language of Nature . . . . . . . . . . . . . . . . . . . . . 328
8.2 Phonology: Sensible Qualities . . . . . . . . . . . . . . . . . . . . 340
8.3 Lexicography: Co-Instantiation . . . . . . . . . . . . . . . . . . . . 346
8.4 Syntax: Causation and Laws . . . . . . . . . . . . . . . . . . . . . 357
8.5 Excursus on Commonsense and Natural Science . . . . . . . . . . . 363
8.6 Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
8.6.1 Informing and Instructing about Ideas . . . . . . . . . . . . 370
8.6.2 Informing about Other Finite Minds . . . . . . . . . . . . . 372
8.6.3 Informing about God . . . . . . . . . . . . . . . . . . . . . 380
8.6.4 The Interpretation of the Discourse of Nature . . . . . . . . 382
8.7 Conclusion: From Fleeting Ideas to Robust Structure . . . . . . . . 384
Chapter 9: Skepticism and Unperceived Objects 386
9.1 ‘In Opposition to Sceptics’ . . . . . . . . . . . . . . . . . . . . . . 387
9.1.1 Berkeley’s Skeptical Opponents . . . . . . . . . . . . . . . 388
9.1.2 Adequate Solutions . . . . . . . . . . . . . . . . . . . . . . 402
9.1.3 The Demon Game and the Independence Assumption . . . . 405
9.2 The Problem of Unperceived Objects . . . . . . . . . . . . . . . . . 411
9.3 Divine Language and Unperceived Objects . . . . . . . . . . . . . . 415
9.3.1 Deflating the Subjunctives . . . . . . . . . . . . . . . . . . 415
9.3.2 Deflating the Divine Ideas . . . . . . . . . . . . . . . . . . 423
9.4 Defeating the Demon . . . . . . . . . . . . . . . . . . . . . . . . . 428
Bibliography 431
Index 452
iv
Abbreviations
Alc Berkeley, George. (1732) 2010. Alciphron: or, the Minute Philosopher. In Jaffro,
Brykman, and Schwartz 2010, 17–274.
An Berkeley, George. (1734) 1992. The Analyst: or, A Discourse Addressed to an
Infidel Mathematician. In Jesseph 1992, 159–221.
BW Luce, A. A., and T. E. Jessop, eds. 1948–1957. The Works of George Berkeley,
Bishop of Cloyne. By George Berkeley. 9 vols. London: Thomas Nelson and
Sons.
CGB Hight, Marc A., ed. 2013b. The Correspondence of George Berkeley. Cam-
bridge: Cambridge University Press.
CSM Cottingham, John, Robert Stoothoff, Dugald Murdoch, and Anthony Kenny,
trans. 1984–1991. The Philosophical Writings of Descartes. By Ren´ e Descartes.
3 vols. Cambridge: Cambridge University Press.
DFM Berkeley, George. (1735) 1948–1957. A Defence of Free-thinking in Mathemat-
ics. In BW, 4:109–141.
DHP Berkeley, George. (1713) 2008. Three Dialogues Between Hylas and Philonous.
In Clarke 2008, 151–242.
DM Berkeley, George. (1721) 2008. An Essay on Motion. In Clarke 2008, 243–267.
EHU Locke, John. (1690) 1975. An Essay Concerning Human Understanding. Edited
by Peter H. Nidditch. Oxford: Oxford University Press.
Enq Hume, David. (1748) 1999. An Enquiry concerning Human Understanding.
Edited by Tom L. Beaunchamp. Oxford Philosophical Texts. Oxford: Oxford
University Press.
LW Locke, John. 1823. The Works of John Locke. 10 vols. London: Thomas Tegg.
MI Berkeley, George. 1987. George Berkeley’s Manuscript Introduction: An Editio
Diplomatica. Edited by Bertil Belfrage. Oxford: Doxa.
N Berkeley, George. 1989. Philosophical Commentaries. Edited by George Has-
son Thomas. The Philosophy of George Berkeley. Ohio: Garland.
v
NTV Berkeley, George. (1709) 2008. An Essay Toward a New Theory of Vision. In
Clarke 2008, 1–66.
PHK Berkeley, George. (1710) 2008. A Treatise Concerning the Principles of Human
Knowledge. In Clarke 2008, 67–149.
PO Berkeley, George. (1712) 1948–1957. Passive Obedience. In BW, 6:15–46.
Siris Berkeley, George. (1744) 1948–1957. Siris: A Chain of Philosophical Reflexions
and Inquiries Concerning the Virtues of Tar-water, and divers other Subjects
connected together and arising One from Another. In BW, 5:31–164.
THN Hume, David. (1739–1740) 2000. A Treatise of Human Nature. Edited by David
Fate Norton and Mary J. Norton. Oxford Philosophical Texts. Oxford: Oxford
University Press.
TVV Berkeley, George. (1733) 1998. The Theory of Vision, or Visual Language, Shew-
ing The Immediate Presence and Providence of a Deity, Vindicated and Ex-
plained. In Philosophical Works Including the Works on Vision, edited by Michael
Ayers, 277–304. London: J. M. Dent.
vi
Acknowledgments
I would like to thank, first and foremost, the members of my committee: Samuel Rick-
less for providing extremely detailed and helpful written comments, and for exhibiting a
healthy skepticism of my more contentious interpretive claims; Gideon Yaffe for impor-
tant philosophical insights, but especially for extremely helpful practical advice about
the process of composing a large work of this nature; Edwin McCann for fascinating
and wide-ranging conversations on the history of philosophy, science, and religion, as
well as more specific discussions of the material presented in this dissertation; and, of
course, James Van Cleve for supervising this dissertation, and constantly bringing to my
attention further philosophical questions raised by my work so far. I would also like
to thank the whole committee, and especially the outside member, Anthony Kemp, for
their patience and stamina in wading through the very large quantity of material here
presented.
In addition to the members of the committee, Janet Levin offered helpful feedback
on chapter 2, and Matthew Babb provided detailed written comments on chapter 5.
Portions of chapter 9 were presented at the International Berkeley Society group session
of the 2011 American Philosophical Association Eastern Division Meeting, the 2013
International Berkeley Conference in Krakow, Poland, and a departmental colloquium at
the University at Albany. I benefited greatly from discussions with audience members at
these events, especially Martha Brandt Bolton, Richard Brook, Georges Dicker, George
Pappas, and Luc Peterschmitt. I would also like to thank David Fussner, the author of the
L
A
T
E
X package biblatex-chicago, which I used for bibliography management. Fussner
made a number of improvements to his already excellent package in response to needs
that arose in the course of this project. Finally, I would of course like to thank my
wife, Lauren, for her patience and support throughout the long process of writing this
dissertation.
vii
Introduction
According to George Berkeley, “the doctrine of signs [is] a point of great importance,
and general extent, which, if duly considered, would cast no small light upon things, and
afford a just and genuine solution to many difficulties” (Alc,x7.16). As has long been
recognized (White 1955; Winkler 2005a; Brykman 2010, 407), this is certainly true
of Berkeley’s own philosophy. However, although this fact has been recognized, the
lesson has not been applied in a thorough and detailed way to the solution of problems
in Berkeley’s philosophy. The aim of this dissertation is to argue that Berkeley believed
that a proper understanding of signs – and, specifically, of language – could solve one
of the most central difficulties of his philosophy, namely, the problem of showing how
a world of fleeting ideas could exhibit the sort of robust structure attributed to physical
reality by commonsense and Newtonian physics.
The Problem: Structure
Berkeley’s ontology is extraordinarily sparse. It consists only of ideas and the minds
which perceive them. These ideas, I will argue in chapter 2, are conceived of as sense
data, or pure phenomenal ‘feels,’ with no intrinsic representational capacity. Further-
more, as Berkeley emphasizes (DHP, 205, 245-246), ideas are ‘fleeting’ and ‘variable’
1
– one is always succeeded by another. Little or nothing remains constant in our expe-
rience. Yet these ideas are meant to be the building blocks of human thought, and of
the world of bodies. How can these ever-changing sense experiences give rise to the
sort of structure we routinely attribute to the world, in both commonsense and natural
science? Further, given that these sense experiences are likewise the building blocks of
our cognition of the world, how is it even possible for us to make such an attribution?
Two sorts of structure are of particular importance: the co-instantiation of quali-
ties in single, enduring object perceived by more than one subject, and the relation of
physical causation.
1
Under both the former and the latter heading, we must be able
to address the relations between actually sensed qualities as well as those between the
theoretical qualities (and quantities) introduced by natural science. Given Berkeley’s
sparse resources, we must explain the very possibility of our belief in such structure
and, if Berkeley is to be a defender of commonsense, we must likewise give an account
of how, on his theory, such beliefs can be true.
The Solution: Language
Berkeley’s solution to this problem, I argue, lies in his theory of language. Berkeley re-
peatedly emphasizes the importance of the theory of signs, and specifically of language,
to his philosophy. In the Introduction to the Principles, he describes his critique of ab-
straction as a discussion of “the nature and abuse of language” (PHK, Introx6). Early
1. Spatio-temporal relations are likewise quite important, but will not be addressed in any detail. These
relations are somewhat less problematic, insofar as visual sensations are, according to Berkeley, ordered
in a visual space, and tangible sensations in a tangible space. Further, Berkeley holds that we derive
our notion of time from the succession of ideas in our own minds (PHK,x98). However, problems
remain. Insofar as the visual and tangible spaces we experience are simply relations within the momentary
experience of a single perceiver, under a single sense modality, this is far more impoverished than our
ordinary notion of space. Likewise, a single perceiver’s subjective time is certainly not the ordinary
notion of time. The latter issue is briefly addressed on pp. 252-253.
2
in the body of the Principles, he asserts that “an intuitive knowledge may be obtained of
[immaterialism] by any one that shall attend to what is meant by the term ‘exist’ when
applied to sensible things” (PHK,x3). More to the point, in the Three Dialogues, Hylas’s
assertion that the variability of ideas renders an immaterialist account of co-instantiation
impossible is alleged by Philonous “to have taken its rise from not rightly understand-
ing the common language of men speaking of several distinct ideas, as united into one
thing by the mind” (DHP, 245). Finally, in De Motu Berkeley says that realism about
forces and the other theoretical entities by means of which physicists attribute structure
to the world is a result of human thought “being obstructed by words which are poorly
understood” (DM,x1).
Berkeley’s linguistic solution, I argue, works at two levels. At the first level, it is
by the adoption of conventional rules that ideas, including words, become signs and ac-
quire the sophisticated representational content which makes the attribution of structure
possible. This is sufficient to explain how it is possible for us to represent the world
as structured. However, as I argue in chapter 7, Berkeley holds onto a realist concep-
tion of truth on which some sort of matching with objective reality is required. Thus if
our representation of the world as structured is to be a true (faithful) representation, the
world must somehow be structured. It is here that the second level of Berkeley’s solu-
tion enters the picture. According to Berkeley, the perceived world is itself a language
– or, rather, a discourse in a language. Berkeley intends this claim quite literally. It
is the linguistic structure of the perceived world that our thought and speech about co-
instantiation, physical causation, and other structural concepts aims to capture. In this
way, I argue, Berkeley succeeds in preserving the commonsense and scientific structure
of the perceived world.
3
Aims and Methodology
This dissertation is an exercise in the history of philosophy. History of philosophy,
as a discipline, might be described as intellectual history in the service of philosophy.
The historian of philosophy is part of the broader philosophical enterprise, struggling
to come to grips with some of the deepest and most difficult questions which have
been asked by human beings, and she hopes, by engagement with the history of hu-
man thought, to shed some light on these questions.
There are many different ways of approaching this task. In Anglophone philosophy,
it is presently customary to divide these approaches into two camps, known as analytic
and contextual (see Watson 2012).
2
One may say, rather crudely, that the analytic histo-
rian of philosophy approaches his task by reading the classic writings of the Great Dead
Philosophers and applying to them the same sort of logical analysis he would apply to
the work of a living philosopher. On the other side of this crude contrast, the contextual
historian approaches her task by immersing herself in the intellectual context of the time
and place in which some Great Dead Philosopher lived and wrote, in order to understand
how the Great Dead Philosopher’s ideas came about. Having drawn this crude contrast,
one may proceed to criticize the analytic historian as unhistorical, and the contextual
historian as unphilosophical.
Somewhat less crudely, one may position historians of philosophy on a spectrum,
with ‘pure’ intellectual history at one extreme and the casual use of quotations from the
Great Dead Philosophers in the course of philosophical argument at the other. It is then
incumbent on the historian to situate her project in some region of this spectrum and
explain why such a project is valuable.
2. Watson, however, prefers the term ‘historicist history of philosophy’ over the more common
moniker ‘contextual history of philosophy.’
4
To me it seems that every region of the spectrum has its value for the project of gain-
ing philosophical insight by means of historical insight. However, it seems to me that the
contextual must precede the analytic. I take this position because it is a presupposition
of my historical work that the philosophers being studied are indeed great philosophers,
and precisely insofar as they are great philosophers it is to be assumed that the positions
they actually held, and the arguments they actually made, as the philosophers themselves
understood them, are likely to be superior to the positions and arguments which emerge
from casual readings of a handful of well-known texts. Thus if we wish to gain max-
imum philosophical insight from the study of a particular Great Dead Philosopher, we
must begin by seeing things from that philosopher’s perspective, by understanding his
philosophical aims and concerns, and the particular arguments and positions to which
he was responding.
On the other hand, if we are to do history of philosophy, and not pure intellectual
history, we cannot stop here, for this contextual approach will usually leave the relevance
of the Great Dead Philosopher’s arguments and positions to our own concerns in doubt.
We must proceed to take a more analytic approach and show how these arguments and
positions can be brought to bear on contemporary issues.
I am in no way opposed to division of labor in these tasks. From the fact that the
enterprise of history of philosophy requires for its success both contextual and analytic
approaches, it does not follow that every individual practitioner must do both. Never-
theless, I have attempted both sorts of tasks in this dissertation. In particular, this dis-
sertation exhibits a constant movement from the contextual to the analytic. This can be
seen in the ‘big picture’ structure of the dissertation, in which the first four chapters are
focused primarily on the historical development of Berkeley’s theory of language, and
the last five focus on analysis and application. It can also, I hope, be seen on a smaller
5
scale throughout. My aim is to begin by understanding Berkeley in his own context,
before transposing his theories and arguments into ours. To achieve the latter task, I
have been liberal in offering comparisons between Berkeley and various philosophers
influential in the analytic tradition, especially Wittgenstein and Quine.
At this point, one must surely ask, why Berkeley? Does Berkeley really have so
much to say to contemporary philosophical concerns?
The answer, I believe, is a resounding ‘yes,’ and the comparisons to Wittgenstein and
Quine are among the reasons why. If my interpretation is correct, Berkeley anticipates
the (in)famous ‘linguistic turn’ in philosophy, and also the idea of ‘meaning as use.’ Fur-
thermore, Berkeley holds that language is essential to all general thought. These devel-
opments drive much of his metaphysics and epistemology. At the same time, the position
Berkeley ends up in has major differences from Wittgenstein and Quine, particularly in
Berkeley’s relatively traditional positions on God and the human mind. Thus one may
say that, from the perspective of analytic philosophy, Berkeley’s premises look surpris-
ingly contemporary, while some of his conclusions look surprisingly old-fashioned, and
others look just plain surprising. As a result it becomes extremely important to under-
stand just what is driving Berkeley’s arguments, whether the arguments work, what are
the costs of accepting the conclusions, and, should the conclusions prove too costly,
what can be done to escape them.
The principal project of this dissertation is sympathetic exposition, but it is exposi-
tion of arguments as well as positions, and this exposition aims to be a step toward an
evaluation of these arguments and positions, with the ultimate hope of shedding light on
the philosophical issues at stake. Furthermore, one has not really gained understanding
of a philosophical argument unless one has understood the assumptions on which the
6
argument is built and the possible ways of escape. As a result, although a full philo-
sophical evaluation of Berkeley’s arguments and theories is left for future work, I do,
at a number of points, indicate what I take to be the most pressing problems for Berke-
ley, and the best ways of escape for those who find Berkeley’s final theory untenable.
Additionally, the interpretation of Berkeley I provide is organized around his response
to what I take to be the single most pressing problem for his philosophy as a whole,
namely, the problem of structure.
Summary of the Chapters
I begin, in chapter 1, by identifying and explaining the view Berkeley describes as a
‘received opinion’ regarding language, namely, “that language has no other end but the
communicating our ideas, and that every significant name stands for an idea” (PHK,
Introx19). I argue that Berkeley is not, as is often assumed, targeting Locke exclusively.
Rather, Berkeley is correct in calling this a ‘received opinion.’ What Berkeley opposes
is the reification of meanings, and this tendency in philosophy can be traced all the way
back to Aristotle. I call the view Berkeley opposes ‘the Theory of Meanings.’ I examine
three versions of the Theory of Meanings from Berkeley’s immediate context – versions
developed by the Port-Royalists (Antoine Arnauld, Claude Lancelot, and Pierre Nicole),
John Locke, and John Sergeant – with the aim of showing that, despite their important
differences, Berkeley’s objections apply to all of them.
In chapter 2, I interpret Berkeley’s critique of abstraction in light of the results of
chapter 1. I argue that Berkeley is correct to characterize the critique of abstraction
as an argument about “the nature and abuse of language” (Introx6). Berkeley’s anti-
abstractionism is, in fact, a lemma in an argument against the Theory of Meanings. I
argue that Berkeley succeeds in showing that the ‘meanings’ posited by the Theory of
7
Meanings would have to be entities of a quite unusual, and metaphysically suspicious,
sort, and hence that Berkeley’s sparser theory is to be preferred if it is able to make
sense of the phenomena.
In chapter 3, I begin the task of reconstructing Berkeley’s own theory of language by
drawing together Berkeley’s views on a variety of specific uses of language in his writ-
ings prior to Alciphron (1732). In particular, I examine Berkeley’s theories of general
terms, operative language (i.e., the use of language to influence emotions, actions, etc.,
in the absence of relevant ideas), and mathematical and scientific language.
Chapter 4 provides a commentary on the seventh dialogue of Alciphron. I argue that
this dialogue provides a general theory of language, and not only a special-purpose ac-
count of certain sorts of religious discourse. Further, I argue that, contrary to Jonathan
Bennett, David Berman, and others, Berkeley, in Alciphron, recognizes no sharp dis-
tinction between ‘cognitive’ and ‘emotive’ language. Although he recognizes what may
be called ‘cognitive’ and ‘emotive’ aspects of meaning, he holds that all of our actual
language contains a mixture of both. This chapter also lays the foundation for my in-
terpretation of Berkeley’s mature theory of language by showing that Berkeley radically
rejects the Theory of Meanings and seeks to put in its place an account on which words
get to be meaningful when they are used according to rules as part of a public social
practice.
Chapter 5 is devoted to understanding the rules governing this social practice. I
argue that Berkeley recognizes that one can follow the rules of language in the absence
of explicit, articulable knowledge of those rules, and that one generally learns rules,
not by being taught explicitly, but by environmental conditioning. Nevertheless, rule-
following does require a certain sort of knowledge or understanding, namely, the ability
to ‘see’ what the rule requires in actual or hypothetical circumstances, and to do the
8
required thing because one ‘sees’ it to be required. A consequence of this is what I call
‘Berkeley’s Internalist Requirement’ – that is, his view that, in order to follow a rule, I
must first have some independent means of recognizing and distinguishing among the
circumstances in which the rule requires me to perform or refrain from actions. This
chapter further examines the sense in which the rules of language must be matters of
public convention and, finally, provides an exposition of Berkeley’s formalist theory of
inference.
In chapter 6, I apply the results of this study of Berkeley’s philosophy of language to
questions of ontology. In particular, I show how Berkeley’s immaterialist metaphysical
system follows from his theory of language. Along the way, I develop and defend a
novel interpretation of Berkeley’s theory of bodies on which bodies, like forces, owe
their existence and nature to the conventional rules of human thought and language.
If, however, the physical world is merely a creation of human thought and language,
one would be right to wonder in what sense our thought and speech about it can be
said to be true. This is the central question of chapter 7, in which I examine Berkeley’s
conceptions of assent (belief) and truth. I argue, on the basis of Berkeley’s treatment
of the subject in Alciphron, that Berkeley holds that belief, in general, is a matter of the
will rather than the understanding. More precisely, to have a belief is to have certain
dispositions to thought, feeling, speech, and action. This, however, renders the question
of truth all the more pressing, since it is unclear how these sorts of dispositions can be
understood to represent the world as being a certain way. I therefore survey a number of
candidates for Berkeleian theories of truth, including deflationist, pragmatist, and Car-
napian approaches, and argue that, while each gets something right, none is ultimately
satisfactory. I argue that Berkeley’s ultimate view combines a basically realist concep-
tion of truth with a radical holism about meaning in a way that is, in certain respects,
9
anticipatory of the Wittgensteinian tradition in the philosophy of religion. Neverthe-
less, Berkeley retains a sufficiently realistic conception of truth to require some sort of
corresponding objective reality.
The question of chapter 8 is, therefore, to what sort of underlying reality is our
talk of bodies answerable? I argue that this question can be answered by attending to
Berkeley’s theory of the physical world as a great discourse ‘spoken’ by God. This claim
of Berkeley’s is, I argue, intended quite literally. He takes the world to have a structure
quite similar to the structure of human languages. It is this linguistic structure – the
grammar of the language of nature – that body talk, in plain language and in natural
science, aims to capture.
Finally, in chapter 9, I apply the results of the foregoing chapters to two of the most
difficult questions in Berkeley scholarship: Berkeley’s claim to be an anti-skeptic and
defender of commonsense, and his views about the status of objects not presently per-
ceived by humans. I argue that Berkeley sees immaterialism as securing the gardener’s
knowledge of the cherry tree without the gardener having to be an immaterialist. This
is because immaterialism amounts to explicit, articulable knowledge of the rules gov-
erning the gardener’s linguistic practice. The gardener can, however, engage in that
practice correctly, and thereby gain knowledge that his cherry tree exists, without hav-
ing any meta-linguistic knowledge or belief of this sort. I then turn to Berkeley’s defense
of the correctness of the gardener’s belief that his cherry tree exists when no human is
in the garden. I argue that Berkeley holds that the truth of this claim is secured by the
fact that the conventional rules of the language of nature would require the tree to be
perceived under specified circumstances, and show how such an interpretation can make
sense of Berkeley’s notorious talk of divine ideas.
10
A Note on Terminology
A number of technical terms which are very important in the exposition of Berke-
ley’s philosophy are unfortunately used in very different ways by different scholars.
I have here attempted to use these terms in what I take to be the most historically well-
motivated senses.
Following G. E. Moore (and the tradition he aimed to refute), I use the term
‘idealism’ for the view that all of fundamental reality is mental or, if one prefers, spir-
itual (Moore 1903). Exactly what being ‘mental’ or ‘spiritual’ amounts to is perhaps
rather contentious, but for purposes of Berkeley interpretation one may get a sufficient
grasp of what is meant by ‘idealism’ by beginning with a Cartesian dualism of the men-
tal and the physical, and then asserting that only those things on the mental side of that
demarcation are metaphysically fundamental, so that the physical things, if they are to
exist at all, must be reduced to the mental.
Following John Sergeant, I use the term ‘ideism’ for the idea theory of perception
and mental representation endorsed, in some form, by all, or nearly all, of the ‘moderns’
prior to Reid. Confusingly, this view is sometimes called ‘idealism’ or ‘idea-ism.’ Be-
cause ‘ideism’ is not in common use, I have often favored phrases like ‘idea theory’ or
‘Way of Ideas.’
The terminology is further confused by the fact that it is far from clear that there is
anything of substance held in common by the various ideists. Reading the discussion of
the Lockean and Port-Royalist versions of ideism in chapter 1 of this dissertation, for
instance, one may get the impression that superficial, terminological agreement (they
all use the word ‘idea’ a lot) is masking fundamental disagreement. I suspect this is
the case, but insofar as the ideists were seen in their own time, by writers like Sergeant
11
and Stillingfleet, as an identifiable school of philosophers, it is useful to have a label for
them.
The term ‘phenomenalism’ is used to refer to the view that bodies and/or the qual-
ities of bodies owe their existence and nature to their being perceived or conceived by
finite minds such as ourselves. Note that on this usage analytic or subjunctive phenom-
enalism, the view that statements about bodies are equivalent in meaning to certain long
conjunctions of subjunctive conditionals about the perceptions of minds, is only one
possible version of phenomenalism.
Finally, I use the term ‘immaterialism,’ as Berkeley does, not for the denial of
matter, but rather as a name for Berkeley’s total metaphysical system. Immaterialism is
thus a particular form of phenomenalistic idealism.
12
Chapter 1
The Theory of Meanings
In order, Berkeley says,
to give a farther account how words came to produce the doctrine of abstract
ideas, it must be observed that it is a received opinion that language has no
other end but the communicating our ideas, and that every significant name
stands for an idea (PHK, Introx19).
This is the view Berkeley seeks to refute.
Here as elsewhere, commentators have typically assumed that Berkeley’s primary
or only target is Locke.
1
I will argue, however, that Berkeley intends his criticism to
be of much wider application. Indeed, Berkeley’s criticisms strike at the heart of an
entire tradition of theorizing about language which stretches back to Plato and Aristotle.
I will refer to the common core of this tradition as ‘the Theory of Meanings’ (capital-
ized), because it involves the postulation of a special class of entities which I will call
‘meanings.’
In this chapter, I will begin by explaining, in a general way, what is involved in the
Theory of Meanings. After this, I will explore in some detail three different theories
of meanings found in works which would have been familiar to Berkeley. The works
1. George Pappas affirms that Berkeley had targets other than Locke in mind – he identifies the
Scholastics and Malebranche, among others – but his discussion is nonetheless firmly focused on Locke
(Pappas 2000, 23, 31-32, 71-79). Other commentators do not mention the possibility of other targets at
all (see, e.g., Atherton 1987; Winkler 1989, chs. 1-2; Stoneham 2002, ch. 7). However, Martha Brandt
Bolton does emphasize the wider range of targets Berkeley has in mind (Bolton 1987, 63-66), and Julius
Weinberg places Berkeley’s attack in the context of the Scholastic background (Weinberg 1965, 5-13).
13
to be discussed are the Port-Royal Logic, John Sergeant’s Solid Philosophy, and, of
course, Locke’s Essay. These works span the full range of 17th century philosophy,
representing Cartesianism, Aristotelianism, and modern empiricism, respectively. In
the next chapter, I will develop an interpretation of the arguments in the Introduction to
the Principles which shows that Berkeley’s aim is not to refute some particular theory of
the nature of meanings, but rather to refute the Theory of Meanings as a whole. Thus I
will show that Berkeley’s critique applies to all of the theories discussed in this chapter,
and also to earlier theories, such as Aristotle’s, and later theories, such as Frege’s.
1.1 Overview of the Theory
The Theory of Meanings is an answer to the question, what does it take for a word to
be meaningful? The answer that it gives is a rather straightforward and intuitive one:
for a word to be meaningful is for that word to have a meaning. A ‘meaning’ is here
understood as a special sort of entity with which the word in question has a merely
conventional association.
In typical cases, a meaningful word is about some (actual or merely possible) things
in the world. To understand the word ‘apple’ is to grasp the meaning of that word, and
to grasp the meaning of that word is to think about apples. If, however, the Theory of
Meanings is not just to push the problem of meaningfulness one step back, by raising
the question of how our thoughts get to be about things, then the connections between
the meanings and the objects in the world must not be a matter of convention (see Figure
1.1). Thus, for instance, Aristotle holds that “spoken sounds are symbols of affections
of the soul” and, that although the sounds, being merely conventional, vary from one
culture to another, the affections of the soul are universally the same (Aristotle De Int. 1
16a4-7). The affections of the soul are intrinsically, by their very nature, about things in
14
Word
conventional
Meaning non-conventional
Object
Figure 1.1: The Theory of Meanings
the world, but words get to be about things in the world only in virtue of their conven-
tional connection to affections of the soul. Similarly, it is Frege’s view that it is merely
a matter of convention that a particular word is associated with a particular ‘sense,’ but
it is the very nature of that sense to pick out the objects it does (Frege [1892] 1960).
On this kind of model, successful communication, which is taken to be the central
purpose of language, occurs when a speaker uses her familiarity with the linguistic con-
ventions to ‘translate’ the meanings she has in mind into a sequence of spoken or written
signs which the hearer, being familiar with the same conventions, can then translate back
into the meanings, so that the hearer comes to have the same meanings as the speaker.
Thus, for instance, if the meanings are taken to be ideas (as on the theories of the Port-
Royalists and Locke), we may suppose that I am thinking of apples, and my thinking of
apples is constituted by my (occurrently) having an apple idea. In virtue of my famil-
iarity with English, I know that the word associated with that idea is ‘apple.’ I can now
utter that word and you, being likewise familiar with the conventions, will thereby come
to have the same idea I have.
It should be clear that the establishment of such conventions requires that speakers
have a prior grasp of the meanings in question: language learning, on this kind of pic-
ture, involves first having a separate grasp of the words (considered as mere sounds) and
15
the meanings, and then learning to associate them with one another. Thus a corollary
of the Theory of Meanings is that the introduction of language cannot expand the repre-
sentative power of thought, since we cannot talk about things in the absence of a prior
ability to think about them.
2
The central thesis of the Theory of Meanings is, then, that a word gets to be mean-
ingful by its conventional association with a special entity, its meaning, which in turn
is non-conventionally related to some (possible) objects in the world. Associated with
this view is a picture of language learning and communication which takes mental rep-
resentation to be prior to and independent of linguistic representation, while linguistic
representation is taken to be parasitic on mental representation. Furthermore, this sort
of representation and cognitive communication is taken to be the central purpose of
language.
1.2 The Port-Royalists
The Port-Royal Logic provided one of the most influential treatments of language in the
early modern period. The account of language advocated there is a paradigmatic version
of the Theory of Meanings.
The Logic, along with its companion piece, the Port-Royal Grammar, was composed
at Port-Royal Abbey in France in the middle of the 17th century. The Grammar was first
published in 1660 and the Logic in 1662. The primary originator of the theory put forth
in these books seems to have been Antoine Arnauld, though his collaborators, Claude
Lancelot for the Grammar and Pierre Nicole for the Logic, were intellectual giants in
2. Cf. Ayers 1991, 1:301: “For all but a few, in Locke’s time as before, the structure of thought is the
source of the structure of language.”
16
their own right I shall refer to Arnauld, Lancelot, and Nicole collectively as ‘the Port-
Royalists.’
According to the Logic, “we can generally say that words are distinct and articulated
sounds that people have made into signs to indicate what takes place in the mind” (Ar-
nauld and Nicole [1662] 1996, 74). The Port-Royalists hold that there are four basic
mental operations: conceiving, judging, reasoning, and ordering.
3
They call the things
operated on ‘ideas’ (Arnauld and Lancelot [1660] 1975, 65-68; Arnauld and Nicole
[1662] 1996, 23). These operations give structure to our thought, and this structure is
prior to language. Following Aristotle and Aquinas, the Port-Royalists hold that lin-
guistic conventions consist only in the linking of ‘articulated sounds’ with universal and
non-conventional ideas and mental operations (Arnauld and Nicole [1662] 1996, 28; see
Buroker 1993, 463-464).
4
The signification (meaning) of a word, according to the Port-Royalists, must be
something that “takes place in the mind.” They hold that “for an uttered or written
sound to signify is nothing other than to prompt an idea connected to this sound in the
mind by striking our ears or eyes” (Arnauld and Nicole [1662] 1996, 66). The theory,
then, is that, in successful verbal communication, the speaker uses conventional rules to
translate what is taking place in her mind into words, and, as a result of these words, the
same thing subsequently takes place in the mind of the hearer.
Although the Port-Royalists sometimes assert that the meanings of words must be
ideas (26, 66), the more general formulation, that words signify “what takes place in the
3. This list was inherited from the Medieval logicians. For a general treatment of the reception of these
Medieval doctrines in the 17th and 18th centuries, see Nuchelmans 1983. Nuchelmans devotes his fourth
chapter to the Port-Royal Logic. On the Port-Royalists’ treatment of the four operations, also see Buroker
1993, 456-458.
4. This view is explicitly associated with Aristotle and Aquinas in Arnauld (1684) 1775–1783, 585-
586. These themes were further developed in Scholastic sources with which the figures under discussion
here would have been familiar. For a brief summary, see Ashworth (1984) 1998, 185-187.
17
mind,” expresses their considered view, since they endorse “a most general distinction
among words into those that signify the objects of thoughts and those that signify the
form and the manner or mode of our thoughts” (Arnauld and Lancelot [1660] 1975, 67-
68; also see p. 122). The word ‘idea’ strictly speaking applies to the object or content of
thought, whereas the manner of thought is better called a ‘perception.’ However, as we
shall see below, it is Arnauld’s view that this distinction is purely semantic; metaphysi-
cally speaking the idea and the perception are identical. Nevertheless, this distinction is
crucial to the Port-Royal semantic theory, since it is the basis for the distinction between
verbs, on the one hand, and nouns and adjectives on the other (Arnauld and Lancelot
[1660] 1975, 122; Arnauld and Nicole [1662] 1996, 74, 78; for discussion, see Nuchel-
mans 1983,x4.3).
The most important operation (‘manner or mode’) of thought is judging, which the
Port-Royalists define as “the action in which the mind, bringing together different ideas,
affirms of one that it is the other, or denies of one that it is the other” (Arnauld and
Nicole [1662] 1996, 23). ‘Judging’ is thus used as a general term, encompassing both
affirmation and denial (82). Every indicative verb signifies affirmation; the copula sig-
nifies simple affirmation whereas other indicative verbs signify the affirmation of some
particular predicate.
5
Thus, for instance, the Latin verb ‘vivit’ signifies the affirmation
of living (79). That is, in the sentence ‘Petrus vivit,’ the noun signifies the idea of Peter
and the verb signifies the mental action of affirming the idea of living, so that the speaker
intends for the hearer to affirm the idea of living of the idea of Peter. This is brought out
more clearly in the paraphrase ‘Petrus vivens est’:
6
in this example, ‘Petrus’ signifies
5. The signification of verbs in other moods is discussed in the Grammar (Arnauld and Lancelot [1660]
1975, 136-138), which also contains a discussion of the mental acts signified by conjunctions and inter-
jections (168-169).
6. On this sort of paraphrase, see Aristotle Met., D7 1017a27-30; Arnauld and Nicole (1662) 1996,
79.
18
the idea of Peter, ‘vivens’ signifies the idea of living, and ‘est’ signifies the mental action
of affirmation which is to be performed on the preceding two ideas.
It should be clear from the preceding discussion that the expressive power of speech
is, according to this theory, in principle coextensive with the expressive power of thought
prior to the introduction of speech (see Buroker 1993, 457). We represent the world to
ourselves in thought by performing certain mental operations, primarily affirmation and
denial, on our ideas. Words can, by convention, be made to stand for any of these ideas
and operations, and this is the only way in which words can be meaningful (Arnauld
[1684] 1775–1783, 585-586; see Nadler 1989, 173-174). The expressive power of both
speech and thought will therefore depend on what ideas we have and what operations
we can perform.
With these considerations in mind, the Port-Royalists adamantly reject imagism.
Ideas, for the Port-Royalists, are conceptions in the intellect and not merely “images
formed in the brain” (Arnauld and Nicole [1662] 1996, 25-26; cf. CSM, 2:113), that is,
they correspond to what in Scholastic jargon were called ‘notions,’ in contrast to ‘phan-
tasms.’ These intellectual conceptions extend far beyond the possible objects of sensa-
tion or imagination, and do not originate from the senses, though they may be prompted
by sensory input (Arnauld and Nicole [1662] 1996, 28-30; cf. CSM, 1:303-305). A
particularly important way in which the intellect is able to extend its ideas beyond what
can be represented by mere images is abstraction. Abstraction is characterized by the
Port-Royalists as a process whereby we separate in thought what cannot be separated in
nature (Arnauld and Nicole [1662] 1996, 37-38). So, for instance, there cannot really be
“lines without width or surfaces without depth” in the world – not even in the corporeal
imagination – yet geometers succeed in forming ideas of just such things. This is done
19
by means of selective attention: they “consider the length without paying attention to
the width” (Arnauld and Nicole [1662] 1996, 38).
The foregoing account is well-supported by the Logic and the Grammar. However,
a complication is introduced by the theory of ideas Arnauld endorses in his famous
controversy with Malebranche. According to many scholars, in that context Arnauld
defends a version of direct realism (Cook 1974; Yolton 1975; Nadler 1989). Direct
realism, however, just is the denial of the claim that some entity such as an idea, phan-
tasm, or notion stands between the perceiver and the external object perceived (Nadler
1989, 11-13). The Port-Royal theory relies crucially on a distinction between mental
acts or operations, i.e. manners or modes of perceiving, and the mental objects (ideas)
operated on or perceived. In the debate with Malebranche, however, Arnauld appears to
treat ideas as identical to acts of perception, and so collapses this critical distinction. It
will be worthwhile, therefore, briefly to examine the consistency of the theory of mental
representation found in the Logic and the Grammar with the theory found in Arnauld’s
On True and False Ideas (Arnauld [1683] 1990).
On True and False Ideas aims to refute Malebranche’s theory of ideas as ‘representa-
tive beings,’ i.e. as entities “actually distinct from our mind as well as from the [external]
object” (63) by perception of which we (indirectly) perceive external objects. The book
opens with a list of seven methodological rules which Arnauld says our inquiry ought to
follow. The first and third of these are of special interest here:
The first [rule] is to begin with those things that are clearest and simplest,
and which cannot be doubted, provided one pays attention to them. . . . The
third is not to seek reasons ad infinitum, but to stop when we get to what
we know to be the nature of a thing, or what we know with certainty to be
a quality of it. One must not ask why extension is divisible, for example, or
20
why the mind is capable of thought, for it is the nature of extension to be
divisible, and that of the mind to think (Arnauld [1683] 1990, 50).
Arnauld then applies these rules in a brief exposition of his own theory of ideas before
launching into his critique of Malebranche. He begins by reciting Descartes’s cogito
argument to show that I exist and that it is my nature to think. He then asserts that “just
as it is clear that I think, it is also clear that I think of something, i.e. that I know and
perceive something. For that is what thought is essentially” (53). Arnauld’s view is
that the essence or nature of thought is to represent (Nadler 1989, 126-127, 177-178).
According to Arnauld’s third rule, when we have identified the nature of a thing, we
have reached the point where explanation comes to an end: the mind thinks because its
nature is to think, and thoughts represent because it is their nature to represent, and this
is as far as explanation can, or should, go. Thus, for Arnauld, the representative capacity
of thought must be taken as primitive (cf. 174).
Arnauld next produces an argument for the conclusion that ideas are modes of the
mind (and not Platonic objects, as Malebranche thought). He writes,
true modifications cannot be conceived without conceiving of the substance
of which they are the modifications; so if it is my nature to think, and I can
think of different things without changing my nature then these different
thoughts can only be different modifications of the thinking which consti-
tutes my nature (Arnauld [1683] 1990, 53).
Arnauld connects his view with Descartes’ definition of an idea as “that form of any
given thought, immediate perception of which makes me aware of the thought” (CSM,
2:113). Arnauld glosses this as the claim that the idea “is not really distinct from our
thought or perception, but is rather our thought itself insofar as it contains objectively
what is formally in the object” (Arnauld [1683] 1990, 72-73). In other words, what idea
21
I am having is just a matter of what sort of perceptual act I am performing. Arnauld (and,
on Arnauld’s interpretation, Descartes) can thus be seen as what we should nowadays
call an adverbialist: he holds that to have a particular idea is to perceive or think in a
certain manner. For instance, to have the idea of red is to perceive redly.
7
In Arnauld’s view, when I perceive (or imagine, or think of) something red, and thus
may be truly said to have (occurrently) an idea of red, what is happening is that my soul
undergoes a certain modification, the modification I have just called ‘perceiving redly’.
This modification just is a single mental state, and which sort of state it is is entirely a
matter of its intrinsic nature; it does not depend on a relation to another mental state,
a Platonic entity, or a concrete object to get its nature. This is what I mean by the
attribution of adverbialism to Arnauld. However, if this is to be an account of mental
representation, it must be the case that agents who exemplify such a mental state are (at
least sometimes) thereby related to objects in the world. This relation Arnauld takes as
primitive. It is entirely in virtue of the intrinsic nature of my mental act that I count as
perceiving red; no object distinct from that act plays any role in constituting its content
(cf. Nuchelmans 1983, 72).
Both of these things must happen in all thought: there must be a soul which is
the subject of the modification, and the modification must be representative in nature
(must have some content). To speak of perceiving and to speak of having ideas are two
different ways of describing this situation. ‘Perception’ and ‘idea,’ Arnauld says, refer
to “a single modification of our soul which necessarily contains [two] relations:” it is
related to the soul which perceives, and to the object which is perceived. The use of
‘perception’ emphasizes the former relation, and the use of ‘idea’ emphasizes the latter,
7. Arnauld’s interpretation of Descartes is defend by Nuchelmans 1983,x2.1 and Nadler 1989,x15.
22
but both words refer the same modification (Arnauld [1683] 1990, 66; cf. Cook 1974,
55-56).
Arnauld describes the difference in meaning between ‘perception’ and ‘idea’ as a
difference in the ‘most direct’ meaning or reference of the words (Arnauld [1683] 1990,
66). In the context of the Port-Royal semantic theory, this should probably be under-
stood in terms of the theory of ‘incidental ideas.’ Incidental ideas are ideas associated
with a word or phrase in the mind of the speaker or the hearer which are not part of the
“principal meaning of [the] expression” (Arnauld and Nicole [1662] 1996, 66). These
incidental ideas are of three types. The first type consists of ideas which, although not
part of the ‘principal meaning’ are nevertheless associated with the word by ‘common
use.’ Thus the Port-Royalists say that the ‘principal meaning’ of the phrase ‘you lied
about it’ is “‘you know that the contrary of what you say is true.’ But in common use
these words carry an additional idea of contempt and outrage” (66-67). The second type
are those which are idiosyncratic to a particular usage of the word on a particular oc-
casion. These ideas, we are told, are “prompted by the speaker’s tone of voice, facial
expression, gestures, and other natural signs” (67). The third type of incidental idea is
rather more complicated. These incidental ideas, the Port-Royalists say, are those “the
mind adds to the precise meaning of terms, for a specific reason. This often happens
when, after it conceives the precise meaning connected to the word, it does not stop
there when this meaning is too confused and general” (70). The star examples here
are demonstrative pronouns, and highly general nouns like ‘thing.’ In these cases, the
context usually determines some particular thing as the referent of the word or phrase,
and so the mind proceeds to think about the features of the thing referred to, in order to
conceive it more clearly, although none of these attributes is, strictly speaking, signified
by the word used.
23
Arnauld says that “the perception of a square has as its most direct meaning my soul
perceiving the square, whereas the idea of a square has as its most direct meaning the
square in so far as it is objectively in my mind” (Arnauld [1683] 1990, 66). This can
be taken in each case as the ‘principal meaning’ mentioned in the Logic (Arnauld and
Nicole [1662] 1996, 66). The idea of perception and the idea of idea are thus both,
according to the definitions in the Logic, ideas of modified things (30-31): when we
speak of perceptions, we consider a mind as perceiving, and when we speak of ideas,
we consider an object as perceived. Now both of these complex ideas contain the idea
of a particular mode of the mind, and, as Arnauld tells us in this very passage, that mode
“necessarily contains both of these relations” (Arnauld [1683] 1990, 66) – that is, an act
of perceiving must be some mind’s act, and it must be of something (though the thing
that it is of need not be present or even exist). Since the act of perception always involves
a subject and an object, it is natural to pass from the subject-as-perceiving to the object
of perception, or from the object-as-perceived to the subject of perception, so that the
idea of the object can be taken as one of the incidental ideas attached to ‘perception,’
and the idea of the subject can be taken as one of the incidental ideas attached to ‘idea.’
These can be regarded as incidental ideas of the third type, those which are frequently
added simply because they are readily available and allow a more distinct cognition.
Strictly speaking, we should thus say that the common thing signified by ‘percep-
tion’ and ‘idea’ is a perceptual event, which consists of a mind perceiving an object. A
perception is always someone perceiving, and an idea is always something being per-
ceived. Neither word, strictly speaking, refers to the act of perceiving alone
8
– rather,
8. Contrary to Cook 1974, 55-56.
24
‘perception’ refers primarily to the mind-as-perceiving, and ‘idea’ to the external object-
as-perceived (although this object need not actually exist). The primary or most immedi-
ate significations of the two terms overlap (both include the perceptual event), and what
the Logic calls their ‘entire meanings,’ “the entire impressions [they make] in the mind”
(Arnauld and Nicole [1662] 1996, 66), coincide. The ‘entire meaning’ is the complete
perceptual event, including subject, act, and object.
This theory of Arnauld’s is supposed, as we have already seen, to be an elaboration
on Descartes’s statement that an idea is a form of thought. Earlier, I argued that, given
how Arnauld understands this statement, he should be classified as having an adverbial
theory of mental contents, that is, as taking the different contents perceptions can have
as being characteristics of the perceptual act. However, Arnauld’s emphasis on the con-
nection between the idea and the external object – defining ‘the idea of a square’ as “the
square in so far as it is objectively in my mind” (Arnauld [1683] 1990, 66) – muddies the
waters somewhat. Fortunately, not long after his discussion of the meanings of ‘idea’
and ‘perception,’ Arnauld offers a clarification of this issue:
We must not confuse the idea of an object with that object conceived, at
least as long as one does not add ‘insofar as it is objectively in the mind’. For
being conceived, in regard to the sun that is in the sky, is only an extrinsic
denomination, i.e. only a relation to the perception which I have of it (67).
The Logic defines an ‘extrinsic denomination’ as a mode “taken from something that is
not in the substance, such as ‘loved,’ ‘seen,’ ‘desired,’ names derived from the actions
of something else” (Arnauld and Nicole [1662] 1996, 32). To say that being perceived
is a purely extrinsic denomination of the sun that is in the sky is to say that the change
from being perceived to not being perceived does not require any change in the intrinsic
features of the sun. Whether I perceive the sun or not is entirely a matter of the intrinsic
25
features of my mind (cf. Nadler 1989, 126-127, 146, 167-168, 177-178). This is also
the reason why it is possible to perceive an object which does not exist: being perceived
does not require having any intrinsic features.
This distinction between ideas and perception lacks metaphysical depth – the words
‘idea’ and ‘perception,’ in the end, provide two different ways of talking about the same
worldly states of affairs. Nevertheless, the distinction is sufficient for the uses to which
it is put in the Port-Royal semantic theory. That theory, essentially, connects nouns with
ideas and verbs with perceptions. Although, metaphysically speaking, all that is going
on in either case is minds thinking, nouns instruct us to consider external objects-as-
perceived,
9
whereas verbs tell us how to consider them. Thus in our example ‘Petrus
vivit,’ we are instructed to consider Peter, an external object, and living, a way external
objects can be, and to put them together in thought in a particular way, considering Peter
as living. To say that I affirm living of Peter is to say that I think of Peter and living
together in a particular manner, and this manner of thinking is, in Arnauld’s usage, one
of the varieties of perception. The instruction to consider an object in a particular way,
or to put multiple objects together in thought in a particular way, is the distinguishing
feature of verbs.
We can summarize Arnauld’s theory of mental representation as follows. In every
perceptual event, there is involved a particular act of perception, which is a modification
of some mind. This modification must be of a particular sort, or, as Descartes might put
it, it must have a particular form. It is regarding this form that Arnauld is a primitivist:
he holds that it is the nature of a given perceptual act to have a certain content and no
more can be said about the matter. Certainly no reductive analysis of what it is for a
perceptual act to have this or that content can be given; it is simply a brute fact that such
9. Here the ‘-as-perceived’ is essentially redundant, for we can only consider objects insofar as we
perceive them. This does not mean that I must attend to the fact that the object is perceived by me.
26
an act of perception has such a content. A perception is of an external object when that
object matches the content.
Is this a form of direct realism? On the one hand, Arnauld’s theory does not rec-
ognize the existence of a mental entity, distinct from the act of perception, which is
the immediate or direct object of perception. On the other hand, the theory does hold
that perceiving an object involves having a mental ‘copy’ of that object. The object is
‘copied’ into the very nature of the act of perceiving. Whether this should be called
‘direct realism’ is largely a terminological matter.
10
What is important for present purposes is that Arnauld’s theory of mental represen-
tation does not require a reevaluation of our earlier account of the Port-Royal theory
of language. Recall that, at the most general level, the Port-Royal theory of language
says that, prior to the introduction of language, we have certain ideas, and we can per-
form certain operations on these ideas, by which complicated structures of thought can
be constructed. The role of linguistic conventions is to attach arbitrary signs to our
ideas and the more complex structures formed out of them. Arnauld’s theory of mental
representation adds the further claim that ideas are not mental objects distinct from our
mental acts, but simply mental acts considered in terms of their representational content.
In (mentally) affirming that Peter lives, I perform three mental acts: I conceive Peter-ly,
I conceive living-ly, and I then use these two acts of conception as components in an act
10. For an illuminating discussion of the definition of ‘direct realism’ as applied to the interpretation of
Arnauld, see Hoffman 2002.
27
of affirmation.
11
The Latin sentence ‘Petrus vivit’ is a recipe for the construction of this
complex mental state. The mental state is the meaning of the sentence.
1.3 Locke
Although Noam Chomsky dubbed the Port-Royal program ‘Cartesian Linguistics’
(Chomsky 2009), there is in fact nothing distinctively Cartesian or rationalist about the
Port-Royalists’ theory of language. Their approach to the philosophy of language is
the Theory of Meanings, which they inherited from Aristotle: words are linked by con-
vention to pre-existing mental contents, and the aim of uttering words is that the hearer
might come to have the same mental contents as the speaker. The distinctively Carte-
sian elements of the theory are to be found only in the nature and origin of the mental
contents in question. Chomsky seems to think that what is ‘Cartesian’ about the theory
is its assumption of a universal structure of thought which is reflected in the structure
of language (78). There is, however, nothing distinctively Cartesian or rationalist about
this view. It was, as we have seen, endorsed by Aristotle. Furthermore, as we shall now
see, the central elements of the Port-Royal theory were also endorsed by John Locke.
Locke’s endorsement of the Theory of Meanings is evident in the very structure of
the Essay: after dispensing, in Book I, with the doctrine of innate ideas, Locke moves
on, in Book II, to give an account of the ideas we do have, where they come from,
and how we represent the world to ourselves by means of them. Only after this do we
11. The complex mental action can be represented using function notation as AF-
FIRM(CONCEIVE(Peter), CONCEIVE(Living)). Note that although the act of affirmation takes
as input two acts of conceiving, it is not about those acts. Rather, it is about their objects. If I want
to affirm, for instance, that my conception of a triangle is clear, then the act I must perform is AF-
FIRM(CONCEIVE(CONCEIVE(triangle)), CONCEIVE(clarity)). This notation is slightly misleading
insofar as it suggests that in conceiving I somehow operate on the external object, which in Arnauld’s
view is not the case. However, this misleading way of speaking seems unavoidable, since we have no
way of picking out any particular idea other than by saying what it is of.
28
arrive at Book III, ‘Of Words.’ This organization presupposes, with the Port-Royalists,
that the full range of expressive power is available in a sort of non-conventional ‘mental
language’ prior to the introduction of words.
12
Berkeley’s rejection of this picture is
very probably the reason for his remark that Book III of Locke’s Essay ought, by rights,
to have come first (N,x717; see Atherton 2007, 284-285; Brykman 2010, 407).
According to Locke, the use of language requires the ability “to frame articulate
Sounds,” and “to make [these sounds] stand as marks for the Ideas within [one’s] own
Mind, whereby they might be made known to others, and the Thoughts of Men’s Minds
be conveyed from one to another” (EHU,xx3.1.1-2; cf. LW, 4:413, 9:249). This is again
an endorsement of the Theory of Meanings.
Like the Port-Royalists, Locke usually says that all meaningful words stand for
ideas, but when he is speaking more carefully he carves out an exception for those
words he calls ‘particles’ (EHU,x3.7; cf. Ayers 1991, 1:22-23). These particles sig-
nify ways in which the mind puts its ideas together. The star examples are ‘is,’ which
signifies affirmation, and ‘is not,’ which signifies negation (EHU,x3.7.1). Lacking the
Port-Royalists’ interest in linguistics, Locke does not discuss what goes on in sentences
12. On ‘mental language’ in Locke and the preceding tradition, see Ashworth (1984) 1998, 185-188.
29
lacking an explicit copula, or give any kind of account of complex sentences. His re-
marks on language are, indeed, more a gesture in the direction of a theory than an actual
theory.
13
The theory at which he is gesturing could easily be that of Port-Royal.
14
Locke’s theory of mental representation is likewise structurally similar to the Port-
Royal theory: “the Object of the Understanding when a Man thinks” is termed an ‘idea’
(EHU,x1.1.8; cf. LW, 4:130-131) and thinking is said to consist in the exercise of
various faculties, assumed to be common to all human beings, which operate in one way
or another on ideas (EHU,xx2.9-11).
Locke does, however, differ importantly from the Port-Royalists on the question of
the nature and origin of our ideas. Locke, of course, rejects innate ideas and holds that
all ideas originate from either sensation or reflection. The question of the nature of
these ideas, for Locke, is vexed. Locke explicitly runs together items the Scholastics
and Port-Royalists had so carefully distinguished, saying that he “used [the word ‘idea’]
to express whatever is meant by Phantasm, Notion, Species, or whatever it is, which the
Mind can be employ’d about in thinking” (x1.1.8). The suggestion here that notions (i.e.
intellectual conceptions) cannot, or at least need not, be distinguished from phantasms,
could be taken to imply an imagistic conception of ideas: that is, since we clearly do
have mental images, the suggestion that all of the objects “the Mind can be employ’d
13. Thus I disagree with Ian Tipton’s remark that “whereas Locke has a very genuine interest in lan-
guage, Berkeley’s concern with language, though in its way no less genuine, is always subservient to his
interest in problems which are not essentially problems about words” (Tipton 1974, 141). There is not, I
think, any contrast between Locke and Berkeley here. Both Locke and Berkeley have a great deal to say
about language, but in both cases the ultimate aim is to solve certain problems in metaphysics and episte-
mology. Furthermore, in both cases the attention to metaphysics and epistemology is in turn motivated by
religious concerns. (Berkeley’s religious concerns are indicated on the title pages of nearly every work he
wrote; on the importance of religion, and especially the politics of religion, to the project of Locke’s Es-
say, see Rogers 1994; Jolley 1999, 10-12, 16-17.) This attitude is shared not only by Locke and Berkeley,
but by most philosophers in the period; indeed, insofar as the Port-Royalists differ, it is probably only due
to the involvement of Claude Lancelot who was really a grammarian and not a philosopher or theologian.
14. On the influence of the Port-Royal Logic on Locke, see Yolton 1975, 146, 153; Schaar 2008, 328-
329, 332.
30
about in thinking” are of the same kind could be taken to imply that they are all images
(cf. Ayers 1991, 1:45-48).
15
Locke’s use of a visual perception “in a thick mist” as an
example of an ‘obscure and confused’ idea in his second letter to Stillingfleet reinforces
this impression (LW, 4:221-222).
16
I will argue that this initial impression is ultimately
correct: Locke is indeed an imagist.
In order to evaluate the attribution of imagism to Locke, we must distinguish be-
tween two different views which might be called ‘imagism.’ The first view is the re-
semblance thesis, according to which at least some of our ideas literally resemble their
objects, in much the same way a photograph resembles its subject. This was how ‘phan-
tasms’ were understood in the 17th century, and Descartes seems still to think of the
images in the corporeal imagination in roughly this way, though he emphasizes that
they do not resemble their objects in all respects (CSM, 1:165-166; Ayers 1991, 1:27).
It is thus likely that those 17th philosophers, such as Hobbes and Gassendi, who be-
lieved only in phantasms endorsed the resemblance thesis (Hobbes [1651] 1985, chs.
1-2; CSM, 2:186-191; cf. Lennon 1988, 234-236; Ayers 1991, 1:45-49; Jacovides 1999,
471-473).
The second view which might be called ‘imagism,’ is sensationsalism. This is the
view, explicitly endorsed by both Hobbes and Hume, that “there is no conception in a
mans mind, which hath not at first, totally, or by parts, been begotten upon the organs
15. Marc Hight draws a different moral: according to Hight, this passage supports the claim that “Locke
uses the term ‘idea’ as a genus that covers a wealth of species” (Hight 2008, 100). Now, it was not
without reason that Peter Browne complained that Locke’s philosophy (and that of Locke’s fellow ideists)
amounted to no more than “the talking of Idea’s, and running endless divisions upon them” (Browne 1697,
3); a large part of Locke’s Essay is engaged in constructing a complex taxonomy of ideas. However, Locke
nowhere indicates that phantasms, notions, or species are different varieties of ideas. Locke’s ‘endless
divisions,’ in other words, do not line up with the established taxonomies in either Scholastic or Cartesian
theory.
16. Later, however, Locke says that talk of ideas as obscure and confused is a metaphor drawn from
vision (LW, 4:242).
31
of Sense” (Hobbes [1651] 1985, 85), or, as Hume puts it, that “all our ideas or more
feeble perceptions are copies of our impressions or more lively ones” (Enq,x2.5). What
Hobbes and Hume affirm, specifically, is that all of our ideas are copied from the senses.
Of course, like Locke, they hold that the mind can operate on these ideas in various
ways, to recombine them and so forth. What I am interested in, and what I will mean
by ‘sensationalism,’ is the somewhat weaker thesis that all of our ideas are of the same
sort as (inner and outer) sensations. This is clearly an important feature of the views of
Hobbes and Hume, for they both hold that non-sensory ideas are exactly like sensory
ideas, only ‘more feeble.’
It should be noted that these two doctrines are logically independent. On the one
hand, if sensations do not resemble objects, then ideas copied from sensations will not
resemble them either. On the other hand, one might think that we come to have pho-
tographic likenesses by some means other than sense. Finally, however, the two views
clearly can be held together.
Locke’s claim that primary quality ideas resemble their objects appears to commit
him to the resemblance thesis, and his empiricism appears to commit him to sensational-
ism. Thus, on the most straightforward reading of Locke, he is an imagist in both senses
(cf. Jacovides 1999). Nevertheless, many scholars have resisted the attribution of one
or both sorts of imagism to Locke.
Interpreters who reject the attribution of the resemblance thesis to Locke generally
do so on grounds that the inherent philosophical absurdity of the view is too great for it
to be attributed with any plausibility to a thinker of Locke’s stature (Bennett 1971, 106;
Curley 1972, 450-452). Those who reject the attribution of sensationalism to Locke
generally do so on grounds that it cannot account for all of the ideas which we have, and
which Locke acknowledges that we have.
32
One of the reasons the resemblance thesis seemed plausible to Hobbes and Gassendi
was likely their materialism (or, in Gassendi’s case, at least materialist leanings) in the
philosophy of mind (Ayers 1991, 1:45-49). If the mind is a physical thing, then there is
no reason (prior to empirical neuroscience) why the mental representation of a square
should not somehow involve a square-shaped region of the mind. Furthermore, on a sim-
ple type-identity theory like Hobbes’, there is no reason why the retinal image – which
is, indeed, very much like a photograph – cannot be identified with visual sensation.
17
Locke, however, was not a reductive materialist of this sort. Even in his notorious
discussion of thinking matter (EHU,x4.3.6), Locke gets only as far as considering prop-
erty dualism; he does not seriously consider full-blown reductive materialism of the sort
propounded by Hobbes (Jolley 1999, ch. 5).
18
If, however, mental properties are wholly
distinct from physical properties, then it is quite unclear how there could be anything
like a ‘photographic resemblance’ between them.
19
As a result of this obvious difficulty, scholars have exercised considerable ingenuity
in trying to reinterpret Locke’s claims about resemblance to avoid commitment to the
thesis that any ideas literally resemble their objects (Bennett 1971, 106-111; Curley
1972, 452-454; Campbell [1980] 1998; McCann 2011, 176-179). My own suspicion
is that these attempts are entirely too ingenious, and that Locke in fact meant what he
17. Of course, secondary quality sensations cannot involve resemblances of this sort if nothing but
matter in motion exists in the world. These sensations are therefore identified by Hobbes with ‘divers
motions’ in the brain (Hobbes [1651] 1985, 86).
18. Michael Jacovides argues, on the basis of remarks in Locke’s dispute with John Norris, that Locke
was in fact agnostic about whether ideas were corporeal (Jacovides 1999, 473-474). Even if Jacovides is
correct, it does not follow from this that properties like perceiving a red circle or being in pain are fully
materialistically reducible; a materialistic reduction is contemplated only for the (internal, immediate)
object of thought. This sort of position would indeed make the resemblance thesis rather more palatable.
19. Gassendi appealed to these sorts of concerns against Descartes (CSM, 2:234). Interestingly, Locke
seems to be raising a similar worry against Malebranche at LW, 9:219. For discussion, see Jacovides
1999, 478-479.
33
said: that the idea of squareness, a mental thing, is similar to squareness, a mode of an
external material object. In fact, as Michael Jacovides points out, Locke twice asserts
explicitly that the very same geometrical property is possessed by an idea and a physical
object (EHU,xx2.8.18, 4.4.6; Jacovides 1999, 475).
Sensationalism is, prima facie, even more strongly supported by an even more cen-
tral aspect of Locke’s philosophy, namely, his empiricism. Locke is quite insistent that
all our ideas must originate from either sensation or reflection (EHU,xx2.1.1-5), and
it is a short step from there to the conclusion that all ideas are copies of inner or outer
sensations. However, Locke has given those who wish to avoid attributing sensation-
alism to him considerably more resources to work with than those who wish to avoid
attributing the resemblance thesis. This is because, as has already been observed, Locke
follows the Port-Royalists in attributing to the mind the ability to perform a variety of
operations on our ideas. At least some of these operations produce new ideas. It is
therefore not implausible to hold that some of the ideas produced in this fashion are
unlike sensations. Indeed, it is difficult to see how highly abstract ideas, or relational
ideas, for instance, could possibly be like sensations. However, again, Locke’s appar-
ent assimilation of notions into phantasms, combined with the intellectual background
of Hobbes and Gassendi, provides good reason for regarding Locke as a sensationalist,
despite the philosophical problems this leads him into (Ayers 1991, 1:44-66).
In sum, it is quite likely that the simplistic interpretation of Locke, on which he is an
imagist in both senses, is correct: Locke holds both that some ideas resemble the external
objects they represent, and that all ideas are of the same sort as the impressions of inner
and outer sense. Neither version of imagism, however, answers the question of how
ideas represent. From the fact that (some) ideas resemble what they represent, it does
not follow that they represent by resembling. In fact, Locke seems to suggest just the
34
opposite. Prior to introducing the primary/secondary quality distinction, Locke defines
a ‘quality’ as “the Power to produce any Idea in our mind,” and one of the examples
of such a power he gives is roundness, a primary quality (EHU,x2.8.8).
20
The primary
qualities are, it seems, powers to produce ideas resembling the powers that produce
them, while secondary qualities are powers to produce ideas that do not resemble the
powers that produce them. This is further supported by Locke’s discussion of the reality
of simple ideas, where he says that the reality (i.e. veridicality) of simple ideas lies “in
that steady correspondence, they have with the distinct Constitutions of real Beings”
and this, he explicitly says, is the case “whether [the ideas] be only constant Effects,
or else exact Resemblances of something in the things themselves” (x2.30.2). In other
words, both primary and secondary quality ideas are ‘real’ because they are “constantly
produced” by the same real constitutions, which they represent.
21
In this text, Locke slides back and forth between speaking of ideas as ‘corresponding
to’ or ‘representing’ powers, qualities, and constitutions. As we have seen, the powers
and qualities are one and the same thing (though not all powers are qualities). The
constitution of an object is, however, something else: it is the unknown ‘real essence’
which grounds the powers of an object (x3.6.6).
22
Now it seems that Locke sometimes,
as inx2.30.2, fails to distinguish powers or qualities from the constitutions which give
20. Also see EHU,xx2.8.9 (eds. 1-3), 2.8.18. On the changes toxx2.8.9-10 in the fourth edition, see
McCann 2011, 167ff.
21. EHU,x2.30.2, and the causal theory of representation suggested by it, will examined in more detail
inx2.3, below.
22. Strictly speaking, the real essence is the constitution which is had in common by all of the objects
which we group, on the basis of their shared qualities, into a sort and which grounds those shared qual-
ities. Thus ‘real essence’ and ‘constitution’ are not strictly speaking synonyms, since it makes sense to
talk about the constitution of an individual, whereas it makes sense to talk about a real essence only in
connection with a sort. This complication can be ignored harmlessly in the present context.
35
rise to them. Recognizing this ambiguity can help to solve three problems which arise
when one takes Locke’s definition of qualities as powers seriously.
23
The first problem is that it is quite difficult to see how a categorical property, such as
phenomenal squareness, could resemble a power. If what is meant here is not the power
itself, but rather the ground of the power in the object’s constitution, then the matter is
less difficult, since the latter is presumably a categorical property.
The second problem is that, although Locke denies that we can have genuine knowl-
edge of the constitutions of objects, he regards the corpuscular hypothesis as preemi-
nently intelligible, and even likely to be true (EHU,x4.3.16). However, according to
the corpuscular hypothesis, the perceived qualities of bodies are grounded in the shape,
size, and motion – i.e., the primary qualities – of bodies too small to be detected by our
senses. Surely, however, a size or shape too small to be detected by our senses is not a
power to cause ideas in us (Jacovides 2007, 117)! If we consider length, for instance,
as an intrinsic feature belonging to the constitutions of bodies, and suppose that this
intrinsic feature comes in degrees, then it is perfectly intelligible to suppose that there
are degrees smaller (and also greater) than we can detect.
The third problem is that Locke sometimes contrasts primary and secondary qualities
by claiming that secondary qualities are, and primary qualities are not, mind-dependent
(EHU,xx2.8.17, 2.31.2).
24
This claim is apparently meant to follow from the fact that
secondary qualities are powers and powers are relations (Rickless 1997, 304-308). How-
ever, if all qualities are powers, then the argument would apply to primary qualities as
well as secondary qualities.
23. Jacovides sees a similar ambiguity in Locke’s use of ‘power’ and uses it to explain away Locke’s
definition of qualities as powers (Jacovides 2007, 111-112).
24. Rickless 1997 takes this to be the fundamental contrast between primary and secondary qualities.
36
My interpretation does have the consequence (given Locke’s understanding of pow-
ers) that, for both primary and secondary qualities, in the absence of perceivers, the
quality properly so-called (i.e., the power to cause the idea) would not exist. Like-
wise, in both cases, the mechanical constitution which (in the actual world where there
are human perceivers) grounds that power would still exist. However, because of the
resemblance between primary quality ideas and mechanical constitutions, mechanical
constitutions are described in primary quality terms (cf. McCann 2011, 176). Thus
when Locke asserts that “The particular Bulk, Number, Figure, and motion of the parts
of Fire, or Snow, are really in them, whether any ones Senses perceive them or no”
(EHU,x2.8.17), and that the “Configuration of Particles” which grounds the color of
porphyry persists in the dark, although the color does not (x2.8.19), he means only to
say that the mechanical constitutions persist. This is confirmed by the fact that Locke
refers, in these texts, to the configuration of the ‘parts’ and ‘particles’ of the substances
in question, and not to the primary qualities of the (macro) substances themselves.
The resemblance claim which Locke endorses is that primary quality ideas do, and
secondary quality ideas do not, resemble the real constitutions which ground the pow-
ers of objects to cause those ideas. This view, however, does not require us to deny
that primary qualities are ‘powers,’ provided that we take ‘power’ to refer here, not to
some dispositional property, but rather to a mechanical constitution which grounds a
dispositional property. As I have said, I do not think that Locke is entirely clear on
this distinction, so that he often sloppily runs ideas, powers, and constitutions together.
I suspect that one reason for this is that Locke takes quality terms in plain language,
words like ‘red,’ to be ambiguous between these three alternatives: that is, ‘red’ can
refer either to (i) the idea of red, (ii) the power to cause the idea of red, or (iii) the
(possibly disjunctive) mechanical constitution in virtue of which objects have the power
37
to cause the idea of red. This applies to both primary and secondary quality terms, the
only difference being that in the case of primary qualities (i) resembles (iii).
Admittedly, at some points in EHU,x2.30.2 Locke talks as if there are two ways of
corresponding to a real constitution, “as to Causes, or Patterns.” However, what Locke
actually says is that in either case “it suffices that [the ideas] are constantly produced by
[the constitutions].” Even for primary quality ideas, being constantly produced by some
particular constitution is sufficient for veridicality; resemblance is not required. Locke’s
view is that primary quality ideas represent and resemble, but no simple idea represents
by resembling. All simple ideas, for Locke, represent their causes.
25
As has already been mentioned, the mind is, according to Locke, capable of per-
forming a wide variety of operations on its ideas. Some of these, like judging, construct
complex mental states which are not themselves ideas. Others construct new ideas. The
most important of these, for our purposes, is, again, abstraction. It is often thought that
Locke is not entirely consistent on the subject of abstraction. It is claimed thatx4.7.9,
which Berkeley so frequently cites, is a sloppy formulation. It surely cannot be Locke’s
considered view that when we think of triangularity in general there is, in the mind,
some entity which, though triangular, is “neither Oblique, nor Rectangle, neither Equi-
lateral, Equicrural, nor Scalenon; but all and none of these at once.” More plausible
is the account ofx2.11.9, a selective attention account similar to the one given by the
Port-Royalists (see Winkler 1989, 39-42; Ayers 1991, 1:49, 1:242-263; Lennon 2007,
254). We shall return to the question of how exactly Locke’s account of abstraction is
to be understood in our evaluation of Berkeley’s attack on it in the next chapter. There,
I will argue that both Locke and the Port-Royalists hold that selective attention is part
25. Martha Brandt Bolton likewise emphasizes that “the representative character of all simple ideas,
images or not, rests on their causal connections;” however, Bolton is at least somewhat ambivalent as to
how seriously Locke’s resemblance thesis is to be taken (Bolton 2007, 83-84). The tenability of this sort
of causal theory of representation will be examined inx2.3, below.
38
of a process by which new ideas are constructed, and that this approach allows for a
consistent reading of Locke’s texts. For now, it suffices to observe that abstraction is
one of the most important operations the mind performs on its ideas, for it is by this
means that ideas become general, and it is only by having general ideas that we can
make meaningful use of general terms (EHU,xx3.1.3, 3.3.6-9).
This, then, is Locke’s theory. All simple ideas originate in either sensation or reflec-
tion, and represent the real constitution which causes them. Some of these ideas, the
primary quality ideas, resemble these real constitutions; others, the secondary quality
ideas, do not. All of these ideas are ‘images’ in the sense that they are all of the same
general sort as the impressions of inner and outer sense. A variety of operations can be
performed on these simple ideas, some of which generate new ideas, and some of which
generate judgments and other complex mental states. Language consists in the arbitrary,
conventional linking of public signs with these private mental states. Comparing Locke
to the Port-Royalists, we can see radical disagreement about the nature and origin of
ideas give way to substantial agreement about the role those ideas, once got, play in
language and thought. Locke, too, clearly endorses the Theory of Meanings.
1.4 Sergeant
John Sergeant combines the Theory of Meanings with an account of the nature of men-
tal representation which is radically at odds with ‘modern’ Cartesian and Lockean ac-
counts. Sergeant says that he “observ’d that Philosophy labour’d and languish’d under
many Complicated Distempers, (all springing from this way of Ideas) and that they were
grown Epidemical.” He therefore resolved “to Stub up by the Roots that Way it self ”
(Sergeant 1697, Epistle Dedicatory). The fundamental error involved in the Way of
Ideas is, according to Sergeant, its failure to distinguish adequately between notions and
39
phantasms (Sergeant 1697, Prefx19).
26
As was noted earlier, the Cartesians did indeed
recognize the ‘corporeal imagination’ and its representations as distinct from the pure
intellect and its representations. However, from Sergeant’s perspective, the Cartesians
make notions (intellectual representations) much too similar to phantasms (corporeal
images).
Sergeant says that the main point of dispute between himself and the ideists is
“Whether our knowledge is made by the Thing being in our Mind when we know it,
or an Idea or Similitude of it only” (Prefx25). In other words, Sergeant interprets the
ideists – both the Cartesians and Locke – as holding that thinking involves the presence
of a mental ‘copy’ of the external object in the mind. This entity is the ‘idea.’ As we
have seen, this description applies to Arnauld, who identifies the idea with the act of per-
ception, insofar as Arnauld’s adverbialism builds the features of the represented object
into the intrinsic nature of the act. The description applies much more straightforwardly
to Locke.
According to Sergeant, the ideist approach necessarily fails to distinguish appropri-
ately between notions and phantasms, and, as a result, fails to secure human knowledge.
Sergeant insists on four criteria for distinguishing notions from phantasms (Prefxx20-
24): (1) phantasms are sensible while notions are insensible; (2) notions, not phan-
tasms, are the meanings of words; (3) purely corporeal living beings (‘beasts’) can have
phantasms, but only spiritual beings can have notions; and, finally, (4) notions can, but
phantasms cannot, be general. All four criteria would be endorsed by the Cartesians as
part of the distinction between the representations of the pure intellect and those of the
corporeal imagination. However, as Sergeant recognizes, Locke’s attitude toward these
criteria is ambivalent at best. First, as we have seen, Locke explicitly announces that
26. Sergeant’s subtitle, ‘against the fancies of the ideists,’ is a pun on this criticism: the ‘fancy’ is both
the faculty which receives phantasms, and the faculty which invents fictions. See Sergeant 1697, 80-82.
40
he will use ‘idea’ to mean “Phantasm, Notion, Species, or whatever it is that the mind
can be employ’d about in thinking” (EHU,x1.1.8), or, as Sergeant sees the matter, that
he will use ‘idea’ “very Equivocally” (Sergeant 1697, 3). Locke, then, evidently does
not think the distinction on which Sergeant insists very important. Sergeant ultimately
thinks that Locke is an inconsistent imagist: according to Sergeant, Locke’s usual po-
sition is that we have only phantasms, but he sometimes illicitly imports notions (42,
160-162). This is not an especially charitable interpretation of Locke, but given both the
evidence for Locke’s imagism and the philosophical problems it causes him, Sergeant’s
reading is certainly defensible.
To be an imagist (in the ‘sensationalist’ sense) just is to reject criterion (1) and
hold that all mental representations are sensible. Locke does, however, come close to
endorsing criteria (2), (3), and (4). Locke holds that the key distinction between humans
and non-human animals is that humans have the power of abstraction (EHU,xx2.11.10-
11). Abstract ideas are necessary for general thought, and are the meanings of most
words (x3.3.1). Thus if for ‘notion’ we substitute ‘abstract idea,’ then Locke will regard
(2), (3), and (4) as, at least, very nearly correct.
27
It seems, then, that Sergeant’s criteria are endorsed consistently by the Cartesians
and occasionally and partially by Locke. However, Sergeant believes that the ‘ideists’
have failed to see a crucial consequence of criterion (2). Sergeant affirms that “the
Meanings of Words . . . are most evidently ... our Notions.” However, “we do not intend
or mean when we speak of any thing, to talk or discourse of what’s Like that Thing, but
of whats the same with it” (Sergeant 1697, Prefx21).
28
Sergeant expresses the argument
27. There is a further caveat to Locke’s endorsement of (3), as written, and that is his notorious uncer-
tainty about whether humans are really spiritual beings (EHU,x4.3.6). This is not important for present
purposes.
28. Daniel Flage radically misunderstands these claims of Sergeant’s when he takes Sergeant to be
claiming that “there is no resemblance between one’s notion of a thing and the thing of which one has
41
quite clearly in a later passage: “Notions are Meanings, or (to speak more properly)
what is meant by the words we use: But what’s meant by the words is the Thing it self ;
therefore the Thing it self is in the Meaning; and consequently in the Mind” (Sergeant
1697, 33). He goes on to illustrate his point vividly by means of the following example:
“when a Gentleman bids his Servant fetch him a Pint of Wine; he does not mean to bid
him to fetch the Idea of Wine in his own head, but the Wine it self which is in the Cellar”
(33).
29
Sergeant’s central claim is that the Way of Ideas necessarily introduces the Veil of
Ideas. If some mental entity stands between my thought and its object then, according to
Sergeant, I do not succeed in making cognitive contact with the external world. If I am
to think about external objects, then the objects themselves must be somehow available
to thought, or, as Sergeant prefers to put it, ‘in the understanding.’ This is precisely what
a notion is: “A notion is the very thing it self existing in my understanding” (27). Having
such notions – that is, having external objects in one’s understanding – is, according to
Sergeant, an absolutely indispensible condition of knowledge.
Sergeant is a direct realist in a much more straightforward, unambiguous sense than
Arnauld. Sergeant holds that the mind makes direct contact with the external world
without the mediation of any kind of mental ‘copies’ of the objects of thought. A crucial
question facing direct realists is how the mind manages to do this. What relation obtains
between the mind and an external object which results in the mind having a thought
a notion” (Flage 1987, 175). Sergeant is quite clearly arguing that one’s notion and the thing one has a
notion of are not merely resembling, but identical. As Sergeant explicitly recognizes, in a passage Flage
quotes, for the notion to be identical to its object is not for it to be unlike it, but for it to be “more than
Like it” (Sergeant 1697, Prefx21). Identity implies not only resemblance, but perfect resemblance. On
the role of resemblance in Aristotelian theories of perception, see Carriero 2009, 136-138.
29. DHP, 229 may be intended as a response to this passage from Sergeant.
42
about that object? How is it that, if thinking about and perceiving objects involves be-
ing somehow related to them, we can think about objects which are not present? Even
worse, how is it that we can think of objects which never have existed and never will ex-
ist? The ideists can provide relatively straightforward answers to these questions: some
mental particular, the idea, exists in the mind whenever we are thinking. A thought is
about some external object whenever the idea ‘matches’ that object, and this is some-
thing the idea can do even when the object is not present. Furthermore, there is no bar to
the existence of ideas which do not match any objects at all. The adverbialism endorsed
by Arnauld (and possibly Descartes) shares this structure, despite the lack of an immedi-
ate object standing between the act of perception and the external object. Perceiving an
object is still a matter of having a mental state that matches it, and perceptions that lack
external objects are still explained as perceptions that fail to match anything. However,
Sergeant, like Berkeley, is puzzled by the question of what this matching could amount
to. Furthermore, Sergeant anticipates Berkeley in arguing that the very structure of the
ideist theory makes it impossible to verify that this ‘matching’ occurs in even a single
case (Sergeant 1697, 31-32; DHP, 246).
Sergeant believes, then, that the notion (mental content) must be identical to the ex-
ternal object if we are to succeed in making cognitive contact with the world. Returning
to Sergeant’s own terminology, we may say that this claim faces two difficulties. First,
how can one and the same object exist both in the understanding and in the external
world? Second, how can we have notions where there are no corresponding ‘things
themselves’?
Before addressing Sergeant’s responses to these questions, it will be worthwhile to
pause a moment and compare Sergeant to Berkeley. Berkeley agrees with Sergeant that
if we are to have knowledge of bodies, they must somehow be ‘in the understanding.’
43
He also agrees with Sergeant that existing ‘ideist’ theories cannot meet this requirement,
and so cannot secure our knowledge of bodies. However, Berkeley’s distinctive theory
of bodies consists in his radical answer to the two questions just raised: the same object
cannot exist both in the understanding and in the external world, so, if we are to have
knowledge of bodies, bodies must exist exclusively in the understanding. Furthermore,
since our mental representations of bodies just are the bodies themselves, we cannot
have mental representations where there are no corresponding ‘things themselves.’
Sergeant, unlike Berkeley, attempts to give materialist answers to these two ques-
tions. In response to the first question, Sergeant first criticizes the Scholastic species
theory of perception (Sergeant 1697, 59-61),
30
then develops a theory which is not im-
portantly different from it. According to Sergeant’s theory, certain ‘effluvia’ flow from
bodies and eventually, by a complicated causal chain, reach
some Chief Corporeal Part in Man, which is immediately united with the
Soul, as the Matter with its Form, and, therefore, is Primarily Corporeo-
Spiritual, and includes both Natures. Whence, when that Part is affected,
after its peculiar Nature, Corporeally; the Soul is affected after its Nature,
that is, Spiritually, or Knowingly (66).
Now, “Those Effluviums sent out from Bodies, have the very Natures of those Bodies in
them, or rather are themselves Lesser Bodies of the Self-Same Nature” (69). As a result,
when the Corporeo-Spritual Part is affected in this way, the parallel effect in the soul is
30. See Aristotle De Anima,B1-G2; Aquinas Summa Theologica 1.84. Flage 1987, 174-180 repeatedly
characterizes Sergeant as a ‘Scholastic’; Sergeant’s discussion of the species theory shows that he would
object quite strenuously to that label. Sergeant remarks that ‘the Schools’ “undertook to explicate [Aris-
totle on sensory perception] and did it untowardly” (Sergeant 1697, 58). Later in the passage Sergeant is
more explicit: “The Schoolmen . . . when they are at a Plunge how to find out a Reason for any difficult
Point . . . create some Entity which God and Nature never made, and then . . . alledge ’twas that Entity
which did the business” (60). Sergeant disavows this practice and aims, by this means, to separate his
Aristotelianism from Scholasticism.
44
knowledge of that nature. The different senses are affected by objects in different ways,
with the result that the effluvia carried to a single sense convey the nature of the object
only partially (Sergeant 1697, 70). The thing which is both in the understanding and
in the external world is, on Sergeant’s view, an Aristotelian (immanent) universal. Ulti-
mately, then Sergeant’s direct realism extends only to universals, and not to particulars.
Furthermore, Sergeant’s direct realism is not combined with content externalism: what
is represented is entirely a matter of what universals are in the understanding, and the
instantiations of universals in the understanding are distinct from their instantiations in
external objects. Sergeant thinks that this nonetheless results in our making successful
cognitive contact with the external world, precisely because these are universals: the
whiteness in my understanding is numerically identical to the whiteness of the paper I
am looking at, and that is how I know the paper to be white.
31
To our second question, how we can have notions to which there is no corresponding
object, Sergeant responds that there is a sense in which we cannot. ‘Fancies,’ as he
calls them, occur when previously stored effluvia are “agitated disorderly” so that they
“imprint[] Incoherent Phantasms on the Seat of Knowledge” (80). Sergeant is not totally
clear on this point, but given his theory it seems that he must say that these are not
genuine notions, but counterfeits of some sort.
The notions which enter the soul by means of the effluvia are, according to Sergeant,
all of the raw materials for cognition. He takes Locke’s refutation of the doctrine of
innate ideas to have been decisive (62). He needs, therefore, to explain how, from these
materials, we can build all of our representations. One way he believes we can do this
is by negating or recombining our notions (77-78, 82-85). Unlike Locke and the Port-
Royalists, Sergeant does not need to introduce any special mechanism by which notions
31. On Sergeant’s theory of sensory perception, see Southgate 2000, 296-299.
45
become general; since the natures that enter the understanding are merely partial, they
are already general (Sergeant 1697, 79).
Despite this radically different account of mental representation which, as has been
noted, has important similarities to Berkeley’s own views, Sergeant adopts the Theory of
Meanings essentially unmodified. On Sergeant’s view, notions are entirely prior to lan-
guage, and “Words are good for nothing in the World but meerly and purely to Signifie”
(356).
1.5 Conclusion
The theory of mental representation, and especially sensory perception, was one of the
major subjects of debate in seventeenth century philosophy, and an enormous variety
of theories were proposed. Three widely differing views have been surveyed here.
However, these views share a common structure: each meaningful word is linked by
convention to some entity which is the ‘meaning’ of that word and which is, in turn,
non-conventionally linked to an external object in the world. It is in virtue of their con-
nection to these ‘meanings’ that words are meaningful. Radical differences about the
nature and origin of meanings are accompanied by genuine agreement on the role these
meanings need to play in a theory of the meaningful use of language.
Because of Berkeley’s focus on abstract ideas, it might be thought that his argu-
ments, in the Introduction to the Principles and elsewhere, are directed merely against
a particular theory of the nature of meanings – that is, against the view that meanings
are abstract ideas. In what follows, I will show that this reading is incorrect. Berke-
ley means instead to argue that no entity of any kind could possibly play the roles that
meanings are supposed to play in the Theory of Meanings. Berkeley argues for a radi-
cal rejection of the picture of thought and language which was common ground among
46
his predecessors. Furthermore, as I will show in chapter 3, the rejection of meanings
leads Berkeley to an even more radical thesis: the rejection of the priority of thought to
language. Our powers of mental and linguistic representation alike rely on conventional
rules for the use of arbitrary signs in trains of speech, reasoning, and action.
47
Chapter 2
Berkeley’s Attack on Meanings
The main philosophical argument of the Introduction to the Principles begins with the
announcement that it will be “proper to premise somewhat . . . concerning the nature and
abuse of language” (PHK, Introx6). Berkeley means this quite seriously: the famous
discussion of abstraction which follows is about the use and abuse of language (con-
trary to Atherton 1987, 45-47).
1
The question which particularly concerns Berkeley is
the question of how meaningful words relate to ideas.
2
Berkeley believes that serious
confusions are caused in reasoning with words both when we neglect to consider the
corresponding ideas where they exist (PHK, Introxx21-25) and when we look for cor-
responding ideas where they are not to be had (MI,x61; PHK,xx135-142; cf. Brykman
2010, 407).
Most of Berkeley’s explicit argumentation is directed against a view that might be
taken as a rather simplistic reading of either Locke or the Port-Royalists. This view is
stated in more detail in Alciphron, where Berkeley’s opponent says:
Words are signs: they do or should stand for ideas; which so far as they
suggest they are significant. But words that suggest no ideas are insignif-
icant. He who annexes a clear idea to every word he makes use of speaks
sense: but where such ideas are wanting, the speaker utters nonsense . . . He
1. Roberts 2007, 43-58 also emphasizes philosophy of language as the context for Berkeley’s attack,
and my interpretation is in substantial accord with his. However, I treat the matter in considerably more
detail than Roberts does.
2. Berkeley’s concern with this question grew out of the controversy about religious mysteries occa-
sioned by Toland 1696. See Berman 1994, 15-17; Pearce, forthcoming(a); and also below, p. 271.
48
who really thinks has a train of ideas succeeding each other and connected
in his mind: and when he expresses himself by discourse, each word sug-
gests a distinct idea to the hearer or reader; who by that means has the same
train of ideas in his, which was in the mind of the speaker or writer. As far
as this effect is produced, so far the discourse is intelligible, has sense and
meaning. (Alc,x7.2)
Although Alciphron is from a significantly later period in Berkeley’s career, there is ev-
ery reason to suppose that this is precisely what Berkeley had in mind when he referred
in 1710 to the view “that language has no other end but the communicating our ideas,
and that every significant name stands for an idea” (PHK, Introx19). Now, the view as
described presupposes ideism, and so does not directly apply to Sergeant (or Plato, or
Aristotle). It is also simpler than the theories given by paradigmatic ideists like Locke
and the Port-Royalists. I claim, however, that Berkeley does not exploit these features
of his exposition in his arguments. The view described is a harmless simplification, in-
sofar as Berkeley aims to attack a fundamental presupposition shared by nearly all of
his predecessors, and by the simplistic view which is the explicit target of criticism. In
other words, from Berkeley’s perspective, the mistake made by his predecessors is so
fundamental and pervasive that examining the details of their views is unnecessary.
I aim, in what follows, to vindicate this perspective both interpretively and philo-
sophically. I will argue that what Berkeley is attacking is the Theory of Meanings, the
view “that every name has, or ought to have, one only precise and settled signification”
(Introx18). The fact that, for Sergeant, the signification is not an idea, is not relevant,
nor is the fact that, for Locke and the Port-Royalists words sometimes signify mental
acts. Berkeley’s main point would apply just as well, for instance, to a theory which
took each name to signify some Platonic entity (cf. Roberts 2007, 51-53). There are
49
no ready-made ‘meanings’ existing prior to the introduction of language waiting to be
assigned symbols. Being meaningful need not involve having a meaning – or, in Berke-
ley’s language, words can be significant despite not signifying anything (cf. Browne
1733, 534). Meaning or significance comes about when agents use signs in certain
ways.
In this chapter, I will first outline in more detail the dialectical structure of Berke-
ley’s argument against the Theory of Meanings. After this, I will discuss the main
lemma in Berkeley’s argument, namely, the claim that there are no abstract ideas. Next,
I will discuss two tactics for escaping the argument: adopting a causal theory of rep-
resentation, and taking representation as primitive. I will argue that the most plausible
causal theories actually accept Berkeley’s main conclusion, that what a given mental
state represents depends on how it is used by the mind. Taking representation as prim-
itive is a way of escaping the argument, but primitive representation is mysterious and
unparsimonious. I conclude, therefore, that while Berkeley’s argument is not ultimately
decisive, it does put him in a very strong position, provided that he can give an alter-
native account of our cognitive and linguistic abilities. In the following chapters, I will
take up the question of whether Berkeley can do this.
2.1 The Dialectical Structure of Berkeley’s Attack
In order to understand Berkeley’s attack, it is necessary to distinguish the textual struc-
ture of the Introduction from the dialectical structure of the attack. By ‘textual structure’
I simply mean the way in which the presentation of ideas and arguments is organized
in the text. By ‘dialectical structure’ I mean the way in which arguments are deployed
against positions, objections against arguments, and so forth. Determining the textual
structure is fairly straightforward, but determining the dialectical structure is somewhat
50
more complicated, and the latter is what we must achieve if we are to understand what
each argument is meant to do, and whether it succeeds at its task.
The textual structure of the Introduction to the Principles is as follows. After some
prefatory remarks on aims and methodology (PHK, Introxx1-5), Berkeley introduces the
doctrine of abstraction, and describes different versions of that doctrine at some length
(Introxx6-9). He then raises his most often repeated consideration against abstraction,
which I will call the ‘phenomenological appeal’ (Introx10). He simply reports that,
upon introspection, he cannot find, in his own mind, any abstract ideas. He also claims
that “there are grounds to think most men will acknowledge themselves to be in [his]
case. The generality of men which are simple and illiterate never pretend to abstract
notions” (Introx10). Now, if Berkeley is right that the existence of abstract ideas is,
at least, not immediately and obviously confirmed by introspection, then one wonders
why anyone believes in such things in the first place. Berkeley says that he will therefore
“examine what can be alleged in defence of the doctrine of abstraction, and try if [he] can
discover what it is that inclines the men of speculation to embrace an opinion, so remote
from common sense as that seems to be” (Introx11). Berkeley now produces his main
arguments against abstraction (Introxx11-17). Finally, he reconstructs what he takes
to be his opponents’ argument for abstraction, and tells us that, since the conclusion is
false, we must reject one of the premises, and the premise to be rejected is the Theory
of Meanings, which, he argues, is independently implausible (Introxx18-20). Lastly,
Berkeley gives an account of the philosophical benefits which will follow from having
corrected the errors in question (Introxx21-24).
If, as I have suggested, we ought to take Berkeley at his word when he says that the
discussion is about the use and abuse of language, then we must takexx18-20 as the
climax to which Berkeley is building. If this claim about language is the main point,
51
then we can best understand the dialectical structure by reversing the textual order and
considering Berkeley’s opponents’ argument for abstraction first. That argument, as
Berkeley sees it, goes something like this:
(1) Every meaningful name signifies some one idea.
(2) Not every meaningful name signifies some one particular idea.
Therefore,
(3) Not all ideas are particular.
3
This is a fair reconstruction of an argument endorsed by Locke. Locke affirms explicitly
that “so far as Words are of Use and Signification, so far there is a constant connexion
between the Sound and the Idea; and a designation, that the one stand for the other:
without which Application of them, they are nothing but so much insignificant noise”
(EHU,x3.2.7). Furthermore, any shift in which idea is immediately signified by a word
is classified by Locke as an “abuse of Words” (x3.10.5). The proper use of words (or at
least names; note that in this discussion Locke neglects his qualification regarding ‘parti-
cles’) requires that some one idea be attached consistently as the immediate signification
of each word. This is an endorsement of (1).
According to Locke, although the proper use of words requires that each word be
consistently attached to one and only one idea, it is nevertheless crucial to “the perfec-
tion of Language” that “signs can be so made use of, as to comprehend several partic-
ular Things.” Locke says that this “advantageous use of Sounds was obtain’d by the
difference of the Ideas they were made signs of. Those names becoming general, which
are made to stand for general Ideas” (x3.1.3). Locke thus endorses (2), and draws the
conclusion that there must be general (i.e. non-particular) ideas.
3. For similar reconstructions of this argument, see Weinberg 1965, 26; Stoneham 2002, 224-225.
52
Unlike Locke, Sergeant and the Port-Royalists begin from general thought, and only
afterward introduce general terms. (Arnauld and Nicole [1662] 1996, 37-40; Sergeant
1697, Prefx24). However, they are committed to the two premises of the argument,
and it is not unreasonable to suppose that something like this was among their implicit
reasons for believing in abstract ideas.
4
A clarification is in order about the conclusion of the argument. Locke, like Berke-
ley, holds that “all things that exist are only particulars” (EHU,x3.3.6; cf. EHU,x3.3.1;
DHP, 192), and ideas, of course, exist. Locke would agree, then, that, for instance, the
abstract idea of triangle is a particular idea but, he holds, this idea nevertheless repre-
sents generally. As we shall see, it is precisely this divide between the idea’s own nature
and its representational content that Berkeley attacks: according to Berkeley, nothing
which is itself particular could possibly be intrinsically well-suited to represent gener-
ally.
The main part of Berkeley’s response to the argument for abstract ideas is a direct
rebuttal of the conclusion, which will be examined below. If the conclusion is false, then
at least one of the premises must be rejected. Berkeley argues that (1) is independently
implausible on a number of grounds, while (2) is an evident truth. Even in mundane
cases where a word can be substituted for an idea, it need not be substituted in order for
the claim to be understood. Berkeley motivates this with the example of a variable in
algebra, which does indeed stand for some particular quantity, although the user need
not know what that quantity is (PHK, Introx19). Indeed, although Berkeley does not
make this point explicitly, the project of solving an algebra problem is typically the
project of finding out what a particular letter stands for; one uses the notation in order
4. Sergeant actually frames these considerations as an argument against abstract ideas: since ideas
cannot be general, he claims, we must have notions instead. His reasons for holding that ideas (phantasms)
cannot be general are similar to Berkeley’s (see below). For Sergeant’s own account of general notions,
see Sergeant 1697, 79.
53
to find out what the notation signifies, hence one clearly does not need to know what
it signifies before one can use it. Furthermore, there are meaningful bits of language
which do not have this sort of connection to ideas at all, and there are meaningful uses
of language where the aim is to do something other than excite an idea, such as to evoke
an emotional or practical response (PHK, Introx20).
This is the central argument of the Introduction: no one idea could possibly be the
meaning of a general term. Nevertheless, general terms such as ‘triangle’ are paradig-
matic examples of meaningful bits of language. Furthermore, plausibly, there are many
other examples of meaningful uses of language which do not involve the speaker’s idea
being excited in the hearer. The correct response, then, is to reject the Theory of Mean-
ings.
The argument can be generalized as follows: no one entity could possibly be the
(one and only) meaning of a general term. Nevertheless general terms are meaningful.
Therefore, a term can be meaningful without a convention by which it is linked with
some one entity which is the meaning of that term. But of course terms do get their
meaning by means of certain linguistic conventions. The linguistic conventions which
make terms meaningful must therefore not consist merely in a one-to-one mapping of
terms to meanings. As we shall see beginning in the next chapter, Berkeley holds that
these conventions consist, instead, of rules for the use of the terms in question for a
variety of theoretical and practical purposes. An important consequence of this view is
that the adoption of such rules can actually increase the expressive power of thought.
2.2 The Case Against Abstraction
The case against abstraction occupiesxx10-17 of the Introduction. x10 contains what
I am calling ‘the phenomenological appeal,’ the claim that abstract ideas are not to be
54
found in introspection. xx11-17 contain Berkeley’s actual arguments. Commentators
have often either dismissed the phenomenological appeal or not mentioned it at all and
focused instead on the arguments.
5
This seems to be a sensible approach, since philoso-
phers are in the business of making and evaluating arguments. However, without paying
serious attention to the phenomenological appeal, we will not successfully grasp Berke-
ley’s thought on this matter. Berkeley believes that it is a datum of experience that we
do not have abstract ideas; for him, the arguments for the impossibility of abstract ideas
are merely a secondary support. Our task here will be to determine exactly what phe-
nomenological reflection Berkeley invites us to perform and what result he expects us
to find. After this, we will be in a position to examine Berkeley’s arguments against the
possibility of abstract ideas and use them to explain the phenomenological data. Finally,
we will examine Berkeley’s arguments to the effect that abstract ideas cannot play the
theoretical roles they are meant to play.
2.2.1 The Phenomenological Appeal
While the arguments against abstraction appear only rarely, the phenomenological ap-
peal appears again and again throughout Berkeley’s corpus. For instance, here are just a
few of the widely scattered occurrences:
5. The phenomenological appeal is mentioned dismissively by Flage 1986, 499; Atherton 1987, 49;
and Pappas 2000, 49. It is not discussed by Bolton 1987; Winkler 1989; or Stoneham 2002. Weinberg
denies that the phenomenological appeal is the basis of Berkeley’s anti-abstractionism (Weinberg 1965,
13), but later indicates that Berkeley’s philosophical investigation of abstract ideas is meant to explain the
results of the phenomenological appeal (24-25). I agree with this latter point, but if this is the purpose
of the investigation then Berkeley must have rejected abstract ideas on the basis of the phenomenological
appeal before undertaking it. Rickless 2012, 733-735 takes the phenomenological appeal much more
seriously, and in fact holds that it is the sole basis for Berkeley’s rejection of ‘generalizing abstraction’
(as distinct from ‘singling abstraction’). Although I do not deny that Berkeley recognizes a distinction
between different varieties of abstraction, the interpretation that I will be defending holds that he rejects
every one of these varieties of abstraction for the same basic reasons, so that the phenomenological appeal
and the impossibility arguments are meant to apply equally to generalizing and singling abstraction.
55
Now I do not find that I can perceive, imagine, or any wise frame in my
mind such an abstract idea [of extension] as is here spoken of. A line which
is neither black, nor white, nor blue, nor yellow, etc., nor long, nor short,
nor rough, nor smooth, nor square, nor round, etc., is perfectly incompre-
hensible. This I am sure of as to myself; how far the faculties of other men
may reach they best can tell . . . After reiterated endeavours to apprehend
the general idea of a triangle, I have found it altogether incomprehensible
(NTV,xx124, 126).
Hylas. But what say you to ‘pure intellect’? May not abstracted ideas
be framed by that faculty?
Philonous. Since I cannot frame abstract ideas at all, it is plain, I can-
not frame them by the help of ‘pure intellect’ . . . But for your farther sat-
isfaction, try if you can frame the idea of any figure, abstracted from all
particularities of size, or even from other sensible qualities.
Hyl. Let me think a little – I do not find that I can (DHP, 193-194).
But upon looking into my own mind, I do not find that I have or can have
these general abstract ideas of a man or triangle abovementioned, or of
colour prescinded from all particular colours.
6
Though I shut mine eyes,
and use mine utmost efforts, and reflect on all that passes in my own mind I
find it utterly impossible to form such ideas (Alc,x7.6).
As Berkeley indicates in the footnote to the Alciphron passage, his fullest published
treatment of the matter is in our main text, the Introduction to the Principles. There,
6. Berkeley’s footnote: See the introduction to a treatise concerning the Principles of Human Knowl-
edge, printed in the year MDCCX. Where the absurdity of abstract ideas is fully considered.
56
after a lengthy discussion of different sorts of abstraction, Berkeley describes exactly
what he believes he can, and cannot, conceive:
Whether others have this wonderful faculty of abstracting their ideas, they
best can tell; for my self I find indeed I have a faculty of imagining, or
representing to my self the ideas of those particular things I have perceived
and of variously compounding and dividing them. I can imagine a man with
two heads or the upper parts of a man joined to the body of a horse. I can
consider the hand, the eye, the nose, each by it self abstracted or separated
from the rest of the body. But then whatever hand or eye I imagine, it must
have some particular shape and colour. Likewise the idea of man that I
frame to my self must be either of a white, or a black, or a tawny, a straight,
or a crooked, a tall, or a low, or a middle-sized man. I cannot by any effort
of thought conceive the abstract idea above described. And it is equally
impossible for me to form the abstract idea of motion distinct from the body
moving, and which is neither swift nor slow, curvilinear nor rectilinear; and
the like may be said of all other abstract general ideas whatsoever. To be
plain, I own my self able to abstract in one sense, as when I consider some
particular parts or qualities separated from others with which, though they
are united in some object, yet it is possible they may really exist without
them. But I deny that I can abstract one from another or conceive separately,
those qualities which it is impossible should exist so separated, or that I
can frame a general notion by abstracting from particulars in the manner
aforesaid. Which two last are the proper acceptations of ‘abstraction’ (PHK,
Introx10).
57
As has been said above, Berkeley believes that the reason people think they have
abstract ideas is that they meaningfully use words which do not correspond to any one
non-abstract idea. Because this is the source of the error, Berkeley urges us, in order
to perform this phenomenological reflection correctly, to remove ‘the veil of words’
(see MI,x11). As Euphranor ‘entreats’ Alciphron, “be not amused by terms, lay aside
the word force and exclude every other thing from your thoughts, and then see what
precise idea you have of force” (Alc,x7.9; cf. An,xx36-37; DFM,x48). Alciphron is
further urged “to exclude the consideration of [force’s] subject and effects” (Alc,x7.9).
Returning from force (which will be discussed in more detail inxx3.3 and 6.2) to the
simpler case of the general triangle, we can see that what Berkeley challenges us to do
is to think of triangle without connecting it to any other ideas, whether ideas of words,
ideas of particular triangles, or processes of inference about triangles. The challenge
is to get the alleged abstract general idea of triangle by itself and hold it still. This,
Berkeley claims, cannot be done.
As Berkeley says, “Whether others have this wonderful faculty of abstracting their
ideas, they best can tell.” I can simply report that my own reflections agree with Berke-
ley’s. In the course of a train of thought, I evidently have the capacity to think generally,
and also to think of things not directly represented in sensory experience, such as (to use
some of Berkeley’s favorite examples) God and virtue. However, just as Hume famously
remarked that “when I enter most intimately into what I call myself , I always stumble
on some particular perception or other” (THN,x1.4.6.3), so I find that whenever I try
to stop my train of thought and hold one item still, by itself, for examination, the item
I find is always some mental image, a mental image which may be ‘hazy’ like Locke’s
vision through a thick mist (LW, 4:221-222), but is never indeterminate in the way an
abstract idea is supposed to be.
58
The phenomenological appeal has been criticized by Michael Jacovides. According
to Jacovides, Berkeley’s theory of representation ‘corrupted his capacity to conceive’
(Jacovides 2009). Jacovides is concerned primarily with the Master Argument (Berke-
ley’s challenge to the reader to conceive of unperceived sensible qualities), but his thesis,
if correct, ought to apply to the doctrine of abstract ideas in general since the Master Ar-
gument simply asks the reader to perform a particular feat of abstraction (PHK,x22;
DHP, 200). Jacovides’ thesis, thus generalized, is that it is precisely because Berke-
ley’s understanding of ideas rules out the possibility of abstract ideas by means of the
arguments to be discussed below that Berkeley found himself unable to frame abstract
ideas. If this is right, then, despite appearances, the arguments are really prior to the
phenomenological appeal after all.
Jacovides’ central claim is that, as a result of endorsing the theory of representa-
tion he endorsed, Berkeley truly was somehow intellectually defective. According to
this view, Locke, Jacovides, and others do in fact have the capacity to frame abstract
ideas, but Berkeley has lost this capacity. Admittedly, Jacovides’ explicit thesis is more
restricted than this: what Jacovides says is that “Because Berkeley believed that he
couldn’t conceive of unperceived bodies, he lost the ability to conceive of unperceived
bodies” (Jacovides 2009, 428). The more general thesis which would be needed in order
to apply Jacovides’ criticism to the rejection of abstract ideas in general is considerably
less plausible. Berkeley clearly can perform the cognitive tasks for which abstract ideas
are supposed to be required, so if the theory of abstract ideas is to be correct, it must
be claimed that Berkeley does indeed have them after all. We might therefore modify
Jacovides’ thesis and, instead of claiming that Berkeley lost the ability to frame abstract
ideas, claim that he lost the ability to discover his abstract ideas introspectively. Al-
though in Berkeley’s day it was usually taken for granted that anyone can easily and
59
infallibly know the contents of her own mind, this view has now been called into ques-
tion by philosophers and psychologists.
7
A denier of the infallibility of introspection
could simply claim that this is one of the cases in which introspection misleads. In the
vein of Jacovides’ original criticism, one could claim, in particular, that introspection is
theory-laden and that Berkeley therefore simply cannot see those mental contents which
do not accord with his theory of the mind.
Another plausible account of how the introspective appearances might mislead is
as follows. Consider the phenomenon of peripheral vision. Peripheral vision differs
from normal vision in a variety of ways. However, it is difficult to determine by direct
examination of one’s own case exactly what these ways are because when we focus
our attention on the periphery of our vision, we naturally tend to focus our eyes on the
formerly peripheral region. As a result, it can seem like our vision is the same all the
way out to the edges, when in fact it is not. Now perhaps something like this goes on in
thought as well. Perhaps there are certain ideas we have in the course of ordinary trains
of thought which resist the sort of focus here attempted. In other words, perhaps we
have mental contents other than images, but, when we make them the objects of intense
focus and scrutiny, they become (or are replaced by) images.
The arguments to be discussed below will provide further reason to reject abstraction
rather than trying to save it by one of these strategies. For now, it suffices to observe
that those of us whose phenomenological reflections agree with Berkeley’s have at least
some reason to join him in rejecting abstract ideas. If we believe that introspection is
fallible, this need only mean that the reason we have is defeasible.
The conclusion Berkeley draws from his phenomenological reflections is that the
fundamental building blocks of thought are, one and all, determinate sensory images,
7. For a particularly radical rejection of the infallibility of introspection, see Schwitzgebel 2008.
60
images of the sort that could actually occur in (inner or outer) sensation. Thus the
mind’s ability to recombine the materials received from the senses goes only as far as
the ability to construct possibly but not actually sensed images (cf. MI,x26). If this is
right, then there is no particular entity the mind can contain or ‘latch onto’ which could
serve as the meaning of terms like ‘triangle.’ We must somehow manage to use this
word meaningfully despite the fact that it does not have a meaning.
The challenge for Berkeley, then, is to explain how we manage to think about all of
the many things we are able to think about given only these sparse cognitive building
blocks, and this is closely intertwined with the overarching question of this dissertation,
namely, how the perceived world, which, according to Berkeley, is constructed from
these same sparse building blocks, gets to have the complexity and structure it exhibits.
The answer to both questions is found in Berkeley’s theory of language.
2.2.2 The Impossibility of Abstract Ideas
Berkeley tells Samuel Johnson
8
that the reason he rejects Locke’s “abstract idea of exis-
tence exclusive of perceiving and being perceived” is, quite simply, that he himself has
no such idea (CGB, 319). In other words, the phenomenological appeal is Berkeley’s
fundamental ground for rejecting abstraction. Berkeley’s arguments for the impossibil-
ity of abstract ideas, to be examined here, are meant to explain why we have no abstract
ideas (Weinberg 1965, 24-25), and also to convince others who may reject the phe-
nomenological appeal. However, as Berkeley recognizes, the main reason for positing
abstract ideas was always to explain certain cognitive and linguistic phenomena. The
real test of Berkeley’s view is, therefore, whether his own more parsimonious theory
8. The American philosopher (1696-1772), not the English lexicographer (1709-1784). For biograph-
ical information, see Hight 2013a, 607.
61
does a better job explaining the phenomena. In the Introduction, Berkeley argues both
that abstract ideas are impossible and that they cannot play the intended explanatory
roles. For reasons which will emerge, I think Berkeley’s argument for the logical im-
possibility of abstract ideas is better construed as an argument for the claim that abstract
ideas would be entities of an extremely mysterious sort, giving us all the more reason to
prefer Berkeley’s theory, if that theory can indeed explain the phenomena.
The primary argument against abstract ideas inxx12-17 is that they cannot do the
theoretical work that they are meant to do. However, coming to the end of his attack on
abstraction, Berkeley says that he has shown both “the impossibility of abstract ideas”
and that “they are of no use for those ends to which they are thought necessary” (PHK,
Introx21). The argument for the first conclusion, that abstract ideas are impossible, is
in fact rather difficult to find in the published Introduction.
9
Berkeley cannot mean to
refer to the phenomenological appeal, for this can, at best, convince me that I do not
have abstract ideas; it provides no reason for thinking they are impossible.
10
In response
to this problem, Willis Doney has argued that the phenomenological appeal is actually
meant to provide a premise in an argument for the impossibility of abstraction (Doney
1983). However, as Doney recognizes, the resulting argument would be quite weak
dialectically, since the abstractionists will not accept the results of the phenomenological
appeal (304-305). Furthermore, as we shall see, Berkeley does have arguments for the
impossibility of abstract ideas which are fully independent of the phenomenological
appeal.
9. Ian Hacking writes: “When you read Berkeley’s Introduction, you should be struck by a surprising
absence. Berkeley never argues that there are no abstract ideas!” (Hacking 1974, 39).
10. Perhaps, as Rickless says, introspection can even reveal that forming abstract ideas is impossible
for me (Rickless 2012, 734n10); this certainly seems to be what Berkeley says, though one worries that
Berkeley may be overstating his case. However that may be, Berkeley clearly recognizes that the phe-
nomenological appeal cannot show that forming abstract ideas is impossible for everyone.
62
The text of the published Introduction is quite compressed, with the result that it can
be quite difficult to tease out the specific arguments Berkeley has in mind. The fuller text
of the Manuscript Introduction is illuminating on this point. Here, Berkeley provides
two distinct arguments for the impossibility of abstraction.
11
The first is an argument
that abstract ideas would have impossible combinations of features. The second is an
argument that abstract ideas would represent their objects as having impossible com-
binations of features; since conceivability implies possibility, if the alleged objects of
abstract ideas are impossible, abstract ideas are likewise impossible.
Comparing the manuscript with the published text, it becomes apparent that, by the
time Berkeley wrote the published introduction, he believed that if he simply explained
clearly enough what abstract ideas were supposed to be, readers would have no difficulty
seeing that there couldn’t possibly be such things, and so he thought it unnecessary to
explain why there couldn’t be such things (cf. Jesseph 1993, 24-27). Evidently he was
right about this; until the twentieth century, Berkeley’s critique of abstraction was re-
garded as his most successful philosophical project, despite the fact that the manuscript
material was not widely available (Luce 1934, 126; cf. Bolton 1987, 61-62; Jacovides
2009, 417). Nevertheless, careful examination of the two arguments which are explicit
in the manuscript and implicit in the published version will help us to understand ex-
actly why Berkeley thought the doctrine of abstract ideas was not merely empirically
false but actually incoherent. The two arguments are closely connected, so I will begin
by explaining Berkeley’s thinking in each case, and postpone evaluation of the argu-
ments to the end of this section.
11. I omit discussion of a third argument, the argument about the general line (MI,x10), because I
believe Berkeley ultimately rejected it as circular: the argument relies on the heterogeneity thesis Berkeley
defended in NTV, but Berkeley’s case for the heterogeneity thesis involved the rejection of abstraction
(NTV,x122; Atherton 1990, 177-183).
63
The first argument is stated most explicitly in MI,x8, which corresponds to PHK,
Introx9. There, Berkeley says that in order to construct the abstract idea man, the mind
“leaves out that which is peculiar to each, retaining onely that which is common to all.”
However, he notes, it is common to all men to have color, but they do not have the same
color, so the abstract idea must be colored without having any particular color. Likewise,
it must have height, but not any particular height. Berkeley goes on to say, in a passage
which he later crossed out,
Suppose now I should ask whether you comprehended in this your Abstract
Idea of Man, the Ideas of Eyes, or Ears, or Nose, or Legs, or Arms, this
might perhaps put you to a Stand for an Answer, for it must needs make
an odd & frightful Figure, the Idea of a Man without all these. Yet it must
be so to make it consistent with the Doctrine of Abstract Ideas, there being
particular Men that want, some Arms, some Legs, some Noses &c.”
12
Shortly thereafter, Berkeley goes on to complain that the abstract idea of animal must
be “Without covering either of Hair, or Feathers, or Scales & yet ... not naked” (MI,x9;
cf. NTV,x125).
13
In the published version of this passage, Berkeley retains the statement that abstract
ideas have determinables without a corresponding determinate, and concludes, without
explanation that an idea of this type “is not easy to conceive” (PHK, Introx9). It is clear,
however, that the conclusion is meant to be drawn in the same way. The abstract idea
12. I have omitted a variety of markings and insertions Berkeley made to this passage in the manuscript.
13. Crossed out text omitted.
Berkeley’s argument here bears an interesting resemblance to an argument of Aristotle’s against Pla-
tonic forms (Aristotle Topics,Z6 143b11-31; Met.,Z14; cf. Weinberg 1965, 42-43). It is noteworthy that
in the Manuscript Introduction, Berkeley uses both of Aristotle’s examples, animal and line (MI,xx9-10),
though in the published Introduction the line has dropped out (PHK, Introx9). Aristotelian influence is
also suggested by Berkeley’s interesting use of the word ‘Homonymy’ in MI,x31. See below, pp. 96-98.
64
man must be colored without being any particular color, but there just can’t be a thing
like that.
The second argument for the impossibility of abstract ideas in the Manuscript Intro-
duction is the argument from impossibility to inconceivability. Berkeley writes,
It is, I think, a receiv’d Axiom that an Impossibility cannot be conceiv’d. For
what Intelligence will pretend to conceive, that which God cannot cause to
be? Now it is on all Hands agreed, that nothing Abstract or General can be
made really to exist. Whence it should seem to follow, that it cannot have
so much as an Ideal existence in the Understanding (MI,x14; cf. DFM,
xx45-46).
14
This text is the most explicit deductive argument Berkeley provides for the impossibility
of abstract ideas, and Kenneth Winkler apparently takes it to be Berkeley’s only argu-
ment (Winkler 1989, 28-35; cf. Weinberg 1965, 14-15). Although the argument is, I
think, distinct from the argument of MI,xx8-9 it is, as we shall see, closely connected.
Although this text has no direct parallel in the published introduction (it would have
come between sections 10 and 11), the argument reappears inx21 where, after quoting
Locke’s notorious ‘triangle passage’ (EHU,x4.7.9), Berkeley sarcastically remarks, “He
that can conceive such manifest Contradictions & Inconsistencys, ’tis fit he enjoy his
Privilege.” Although this sentence is removed, the section survives as PHK, Introx13. It
seems likely that, since, in the passage quoted, Locke had explicitly said that the abstract
triangle was “something imperfect that cannot exist” there was no need for Berkeley
to draw the obvious conclusion explicitly; instead, he repeats the phenomenological
appeal.
The argument is quite simple, and can be reconstructed as follows:
14. Some markings and insertions omitted.
65
(1) Only particular things can possibly exist.
(2) Only things that can possibly exist can be conceived.
Therefore,
(3) Only particular things can be conceived.
15
Premise (2) is further supported by an allusion to an argument from Descartes: God’s
powers of conception are a strict superset of mine, but whatever God can conceive God
can bring about, so whatever I can conceive God can bring about. It follows that nothing
I can conceive is impossible (CSM, 2:254).
16
Berkeley has, then, two main arguments. One is an argument that abstract ideas
would have impossible combinations of features, and the other is an argument that ab-
stract ideas would represent impossible combinations of features.
17
There are three
principles by which combinations of features can be ruled impossible:
The Principle of Non-Contradiction (PNC) For any feature F , no object is both F
and not-F .
The Principle of Excluded Middle (PEM) For any objectx and featureF , eitherx is
F , orx is not-F .
15. Similar reconstructions can be found in Flage 1986, 489; Winkler 1989, 33; and Hight 2008, 224.
Note that unlike Doney’s reconstruction, this argument only relies on the plausible and widely accepted
premise that impossibility implies inconceivability (i.e., by contraposition, that conceivability implies
possibility), and not its much less plausible converse (Doney 1983, 297). On the acceptance of this
principle by Berkeley’s opponents, see Flage 1986, 490-492. The converse, on which Doney’s argument
relies, is in fact explicitly disavowed by Berkeley (DHP, 232 [1734 ed.]). See Ott 2004, 411.
16. It is, of course, important to Descartes that the principle applies only to clear and distinct concep-
tion. Berkeley never makes this qualification.
17. Pappas 2000, 49-64 also recognizes two distinct, though related, arguments against abstraction in
Berkeley, which are similar to the two I discuss here.
66
The Principle of Determinacy (DET) If any object has a determinable property (e.g.
color), then that object has a corresponding determinate property (e.g. red).
Berkeley’s claim is that abstract ideas would both have and represent their objects as
having combinations of features which violate one or more of these principles. This, he
claims, is impossible.
Berkeley’s claim here is initially quite puzzling. It seems that what the proponent of
abstract ideas should say is that, if ideas should be regarded as genuine entities at all,
18
then a distinction is to be drawn between the features they have and the features they
represent. Locke, in fact, explicitly insists that although ideas are particular things, they
nevertheless represent generally (EHU,x3.3.11).
19
Furthermore, it is crucial to Locke’s
picture that, although every idea has the property being an idea, most ideas do not
represent their objects as being ideas. The collection of features ideas have must obey
all three of the principles, but (the abstractionist ought to say) the collection of features
they represent needs to obey only (PNC). From the fact that an idea does not represent
its object as having some property, it cannot be concluded that the idea represents its
object as lacking that property. So, for instance, the abstract idea of man represents the
property being colored without representing any particular color property, because all
18. Against ideas as entities, see Bennett 1971,x5; Yolton 1975; Mackie (1985) 1998. For criticisms of
this ‘de-ontologizing’ strategy, see Ayers 1991, vol. 1, chs. 5-8; Hight 2008,x4.1, et passim.
19. C. C. W. Taylor misinterprets this section of Locke’s Essay as an assertion of Berkeley’s thesis that
nothing can be intrinsically well-suited to represent generally (Taylor 1978, 106). Admittedly, Locke does
say that the “general Nature” of ideas is “nothing but the Capacity they are put into by the Understanding,
of signifying or representing many particulars” and that this is a “relation, that by the mind of Man is
added to them.” This, however, is not to deny that the mind, by the process of abstraction, makes for itself
ideas which are intrinsically well-suited to represent in this way. One might suppose, for instance, that
taking something as a sign of something else, rather than considering it for its own sake, is something the
mind does but, once the mind determines to consider an idea as a sign, what the idea will represent is fully
determined by its intrinsic nature. In fact, as Taylor notices (107-108), if Locke does not say that ideas can
and words cannot be intrinsically well-suited to represent generally, then he has no reason for thinking
that a general word must signify a general idea, rather than thinking, with Berkeley, that a general word
signifies the many particulars in its extension directly.
67
of its objects are colored, but they are not all the same color. Of course, on this view, the
idea does not represent the property being colored by exemplifying that property, for if
it did then, by (DET), it would have to exemplify some determinate color property, and
so would presumably represent all men as being the same color.
Although this seems to be the obviously correct approach, it is not available to any-
one who holds that representation is by resemblance. For two objects to resemble or be
similar to one another, they must have a property in common. This is so even in cases of
inexact resemblance: for instance, if two objects are similarly colored, then they must
both fit into a common color category (e.g. blue), even if they are not of precisely the
same shade. Likewise, if two objects are similar in length, then there must be some
range of lengths (e.g. between eight and ten inches) into which they both fall. Or, if
objects are similar in shape or proportion, there must be some (reasonably natural) class
of shapes, or range of proportions, into which they both fall.
20
If an idea is to repre-
sent an object by resembling it, then the idea and the object must have some features in
common, and surely the relevant features are just those the idea represents the object as
having (cf. Marusic 2009, 434).
If, however, ideas are real, determinate entities and they represent by resemblance,
they will not be able to represent generally. Since there is nothing in the intrinsic nature
of the idea to pick out some subset of its features as its representational content, it will
represent only those objects which resemble it exactly, that is, those that share all of its
features. For instance, if we consider again Locke’s idea of a man seen through a thick
20. Contrary to Jacovides 1999, 468-469. Jacovides’ gift shop Statue of Liberty shares with the real
Statue of Liberty such properties as being shaped like a woman holding a torch and being taller than it is
wide.
68
mist (LW, 4:221-222),
21
we should ask why such a mental image would be a represen-
tation of the man rather than, say, the mist. Similarly, it could be a representation of
gray, or of objects between four and seven feet tall. If an entity is fully determinate, as
all entities must be, then resemblance alone cannot make that entity a representation of
anything which does not resemble it exactly.
I argued in the previous chapter that Locke holds that ideas represent causally, and
the Port-Royalists take representation as primitive, so that these considerations do not,
strictly speaking, tell against them. However, Berkeley’s argument is not therefore a
strawman. As was mentioned above, Gassendi, for instance, held that ideas must resem-
ble in order to represent (CSM, 2:186-191). Furthermore, Sergeant is in an even worse
position than the resemblance theorist, for Sergeant insists not on resemblance but on
identity. As we have seen, the ‘identity’ in question is the instantiation of the very same
Aristotelian universal. So notions represent by having the feature they represent their
objects as having. If notions are fully determinate entities, obeying (PEM), and they
represent in this way, then every feature they do not represent their objects as having
is a feature they represent their objects as lacking. Again, there will not be representa-
tion without exact resemblance. Furthermore, although Locke does not hold that simple
ideas represent by resemblance, he still faces a similar issue, for he has what Martha
Brandt Bolton calls a ‘descriptive theory’ of complex ideas (Bolton 2007, 70). That is,
Locke holds that complex ideas represent what they do in virtue of the simple ideas they
are made up of. For each feature a complex idea represents its object as having, it must
contain the idea of that feature. However, Locke denies the existence of negative ideas,
holding that negative words signify the absence of positive ideas (EHU,x3.1.4). Thus
21. It should be noted that Locke says that this is an example of an obscure and confused idea, not of
an abstract idea, but it is supposed to be analogous to the case of the idea of substance, which is not only
obscure and confused but also highly abstract.
69
he would seem to be committed to the claim that a complex idea represents its object as
lacking some feature simply by failing to include the idea of that feature. But if this is
so, then we are back to the same problem: for any given feature and any complex idea,
the complex idea either includes or excludes the idea of that feature, so that it represents
its objects either as having or as lacking that feature. Thus the ‘description’ included in
the content of the idea must be complete.
Once this approach is accepted, allowing violations of the three principles would
in fact not help. If representation involves exact resemblance or complete description,
and the idea of man was colored without having any particular color (or included the
idea of color without including the idea of any particular color), then it would represent
its object as being colored without having any particular color. But of course no such
bizarre entity could possibly be a man (cf. Stoneham 2002, 232).
Kenneth Winkler has suggested a way of interpreting the ‘selective attention’ aspect
of Locke’s thought about abstraction which, he says, is likely Locke’s actual view, and
which escapes Berkeley’s arguments (Winkler 1989, 41-43, 46). According to Winkler,
Locke’s use of ‘idea’ is ambiguous between the object of thought (Locke’s official def-
inition; see EHU,x1.1.8) and the act of thinking.
22
For Winkler’s Locke, to have an
abstract idea is not really to have an idea of a particular kind, but to have an idea in a
particular way. A similar account is developed in more detail by Michael Ayers (Ayers
1991, 1:248-253, 259-263).
Interpretively speaking, Winkler’s proposal is problematic since even the Port-
Royalists, who, unlike Locke, explicitly and consistently adopt a selective attention
model of abstraction, take selective attention to be a process whereby a new idea is
formed. They write,
22. The same complaint about Locke’s usage is made by Sergeant 1697, 142.
70
Suppose, for example, I reflect that I am thinking, and, in consequence, that
I am the I who thinks. In my idea of the I who thinks, I can consider a
thinking thing without noticing that it is I, although in me the I and the one
who thinks are one and the same thing. The idea I thereby conceive of a
person who thinks can represent not only me but all other thinking persons.
By the same token, if I draw an equilateral triangle on a piece of paper, and
if I concentrate on examining it on this paper along with all the accidental
circumstances determining it, I shall have an idea of only a single triangle.
But if I ignore all the particular circumstances and focus on the thought
that the triangle is a figure bounded by three equal lines, the idea I form
will, on the one hand, represent more clearly the equality of lines and, on
the other, be able to represent all equilateral triangles (Arnauld and Nicole
[1662] 1996, 38, emphasis added).
In On True and False Ideas, Arnauld is even more explicit:
The philosopher Thales, having to pay twenty workers one drachma each,
counted twenty drachmas and paid each worker. He would not have been
able to do this unless there were at least two perceptions in his mind: one
of twenty men and one of twenty drachmas . . . Having some spare time
he began to reflect, and thinking about what the two perceptions or ideas
have in common, namely that there is 20 in both, he abstracts from what is
particular in them the abstract idea of the number 20 . . . This is a third idea
or perception (Arnauld [1683] 1990, 74).
Here, partial consideration, or ignoring particular circumstances, is clearly seen as part
of a procedure whereby a new type of idea is ‘conceived’ or ‘formed.’ Ayers explicitly
says that Locke’s selective attention account is “reminiscent of the Port Royal Logic”
71
(Ayers 1991, 1:251), but does not recognize that the Port-Royalists clearly regard ab-
straction as a procedure for constructing new ideas.
If we interpret those passages where Locke discusses selective attention (e.g. EHU,
xx2.7.9, 2.13.13) as endorsing a theory like the one adopted by the Port-Royalists, then
those passages need not be taken to be at odds with the passages where he clearly seems
to think of abstract ideas as a distinct kind of object of thought (e.g.xx3.3.6, 4.7.9). A
consistent interpretation of Locke’s thought on abstraction is possible if we take Locke
to hold that selective attention is part of a process whereby new ideas are formed.
Ayers argues against this type of view by pointing to passages likex2.13.13, where
Locke writes, “a partial consideration is not a separating. A Man may consider
. . . Mobility in Body without its Extension, without thinking of their separation” (Ay-
ers 1991, 1:251). Ayers evidently thinks that this passage shows that the abstract idea
of mobility is really an idea of a mobile, extended body, in which only the mobility is
considered by the mind. This, however, badly distorts Locke, as can be seen by consid-
ering the very next sentence: “One is only a partial Consideration, terminating in one
alone; and the other is a Consideration of both, as existing separately” (EHU,x2.13.13).
In other words, when Locke says that we cannot mentally separate the mobility of body
from its extension, he is not claiming that we cannot think of mobility without also think-
ing of extension; he is rather claiming that we cannot think of mobility and extension as
separate.
In fact, there is no way of interpreting Locke consistently on this point without at-
tributing to him more than one use of the word ‘separating.’ Locke says explicitly that
abstraction is a separating of ideas (x3.3.6), and also that it is not a separating of ideas
(x2.13.13). Since, inx2.13.13, Locke says that the separation he is talking about “is
a Consideration of both, as existing separately,” it makes sense to connect this with
72
another collection of texts in which Locke speaks of ‘separation’ of ideas. These are
the texts in which Locke uses the term ‘separation’ for the mental act of denial (EHU,
x4.5.6). If this is what Locke intends inx2.13.13, then he would be saying that to sep-
arate in the mind the mobility of body from its extension is mentally to deny extension
of a mobile body, or, in other words, to believe the proposition that a body is mobile
but not extended. This, according to Locke, is psychologically impossible, because the
mind immediately, intuitively perceives the agreement of the idea of body with both
mobility and extension, and this sort of knowledge is involuntary (x4.13.1). According
tox2.13.13, what we can do is think of a mobile body without thinking of it as extended.
We do this by means of an abstract idea which includes mobility but not extension. What
we cannot do is think of a mobile body as unextended. If this latter task were possible
it would involve conceiving the two ideas of mobile body and extension as separated.
Berkeley’s arguments are directed against the view that there is some special class
of entities which, by their very nature, represent generally. It is therefore true that, if
the Winkler-Ayers interpretation were correct, Berkeley’s arguments would not touch
Locke. Berkeley is, however, still in disagreement with the Winkler-Ayers Locke:
23
according to Winkler and Ayers, Locke holds that general thought involves having an
idea in a particular way. Berkeley, by contrast, holds that general thought involves using
an idea in a particular way. On this interpretation, Locke claims to have the ability to
get into a static condition of having an idea in a general way, whereas according to
Berkeley general thought can occur only in the context of a process of manipulating
ideas according to rules (seex3.1, below). Berkeley will thus reject the view Winkler
and Ayers offer on Locke’s behalf on the basis of the phenomenological appeal. The
23. Contrary to Ayers 1991, 1:150-151.
73
arguments for the uselessness of abstract ideas, to be discussed below, also gain traction
against the Winkler-Ayers Locke.
Central to the reading of Berkeley I have defended is what Martha Brandt Bolton
calls Berkeley’s theory of ‘idea-objects.’ According to this view, ideas are fully deter-
minate individual entities which are not intrinsically ‘of’ anything other than themselves
(Bolton 1987, 68-69; cf. N,xx660, 843). Critics of this view have claimed that it would
involve Berkeley in ‘circular reasoning’ (Muehlmann 1992, 54) or make “the argument
for idealism . . . just too easy” (Rickless 2013, 113). These criticisms stem from two
features of Bolton’s view: first, her claim that the theory of idea-objects is, for Berkeley,
‘fundamental’ (Bolton 1987, 68) and, second, her claim that Berkeley’s theory of idea-
objects is the basis for his argument for idealism (73-76). The objection against Bolton’s
view is that the theory of idea-objects is much too closely connected to idealism to play
the role of a fundamental assumption to be used as the basis of an argument for idealism.
This objection against Bolton’s ‘idea-object’ interpretation is mistaken in two ways.
First, when Bolton says that Berkeley’s theory of idea-objects is ‘fundamental’ she does
not mean that Berkeley is not in a position to give reasons in its favor, reasons to which
the materialist is obligated to give some weight. Second, idealism does not fall out of the
theory of idea-objects so trivially as Bolton’s detractors suppose. Both of these points
can, as I will now proceed to show, be seen in Bolton’s original essay.
As to the first point, Bolton says that Berkeley’s theory of idea-objects was “mo-
tivated, at least in part, by obscurities in the established view that ideas have intrinsic
representational contents” (75). These ‘obscurities’ are closely related to what Winkler
calls “Berkeley’s explanatory challenge” (Winkler 1989, 45-48). The Way of Ideas, as
Berkeley sees it, is meant to be an explanatory account of human thought. It is supposed
to begin from some fundamental building blocks and a small set of operations we can
74
perform on them and by this means explain how we manage to perform all of the cogni-
tive tasks we in fact perform. If the basic materials from which such a program begins
are even more in need of explanation than the phenomena they are supposed to explain,
then the program has failed. Berkeley believes that this is the case with the process of
abstraction and the abstract ideas which are meant to result from it. The process and its
results are, Berkeley thinks, not merely mysterious but impossible. The impossibility,
however, comes from Berkeley’s own theory of ideas (the theory of idea-objects), which
his opponents need not endorse. The fundamental point is this: Berkeley takes ideas of
the sort he believes in, namely, mental images, to be paradigmatically unmysterious en-
tities. They are easily discovered introspectively, and their natures are utterly transparent
to us (PHK,x25; cf. EHU,x2.29.5).
24
However, to suppose that ideas of this sort could
be abstract leads to absurdities, and no one has been able to explain adequately what an
idea of any other type would be. In other words, Berkeley’s theory of idea-objects is
‘fundamental’ in the sense that he does not attempt to offer any explicit, deductive argu-
ment for it. However, it is supported by reasons. The key reason is the mysteriousness
of the alternatives. Until Berkeley’s opponents can either de-mystify their theories or
show that there is no way to explain the phenomena without mysterious entities, it does
seem that there is good reason to prefer Berkeley’s assumptions to theirs.
As to the second point, Muehlmann characterizes Bolton’s interpretation as attribut-
ing to Berkeley the fundamental assumption that “it is impossible to distinguish between
ideas and their objects” (Muehlmann 1992, 54). Similarly, Rickless writes: “consider a
situation in which I am looking at, and so have an idea of, a table T. It is an immediate
consequence of the claim that my idea of T is identical to T that T is itself an idea”
24. This is not to say that there are no puzzles about ideas of this sort at all. For a book-length treatment
of the metaphysical problems faced by early modern idea theorists, with heavy emphasis on Berkeley, see
Hight 2008.
75
(Rickless 2013, 114). This, however, mischaracterizes Bolton’s view, by missing a dis-
tinction which is crucial, both to Berkeley and to Bolton: “Although the intrinsic object
of an idea is just the idea itself, Berkeley holds that an idea can come to represent some-
thing else” (Bolton 1987, 68). Berkeley’s view is that it is no part of the intrinsic nature
of an idea to represent anything which is not exactly like it, hence an idea intrinsically
represents (is ‘the idea of’) only itself. However, this does not trivially rule out materi-
alism, for the materialist may hold that the idea represents some non-idea extrinsically,
or, in Berkeley’s terms, that the idea signifies a non-idea. This, according to Berkeley,
actually happens when we think of spirits and their actions (see below,xx6.4-6.5). Fur-
thermore, it is not a trivial consequence of the theory of idea-objects that there are no
mind-independent triangles. According to Bolton, “when Berkeley writes that someone
who perceives something triangular has an idea of a triangle, he means the person per-
ceives an idea that is a triangle” (68). This is to say that all triangle ideas are triangles.
This, however, does not (by itself) entail that all triangles are triangle ideas.
25
Thus a materialist who accepts the theory of idea-objects would appear to have the
following path open to her: she could hold that although triangle ideas are intrinsically
‘of’ only themselves, they signify mind-independent material triangles. This significa-
tion relation might arise, for instance, from resemblance or causation. These possibili-
ties are not ruled out by the theory of idea-objects alone.
Since Berkeley has reasons for the theory of idea-objects which can be appreciated
by the materalist, and the theory of idea-objects does not trivially entail the falsity of
materialism, an interpretation which has Berkeley using that theory as the basis for an
25. In my view – though I cannot speak for Bolton – it is a mistake on Rickless’s part to use a table as
an example. This is because, in my view, no idea is a table, hence no idea is intrinsically ‘of’ a table. See
below,x6.3. This, however, does not affect any of the essential points made by either side, all of which
are just as easily made using Bolton’s triangle example.
76
argument for idealism does not convict Berkeley of begging the question against his ma-
terialist opponents. In fact, because of the complexity and sophistication of Berkeley’s
understanding of signification, and in particular the difficulties involved in securing the
meaningfulness of the discourses of physics, on the one hand, and theology, on the other,
while still ruling out talk of matter, Berkeley’s path from the theory of idea-objects to
idealism turns out to be quite complicated. It is, indeed, sufficiently complicated that
we will not arrive at idealism until the end of chapter 6.
My conclusion is that Berkeley’s theory of idea-objects should be regarded as ceteris
paribus preferable to the various non-Berkeleian theories of human thought on offer,
such as the ‘ideist’ theories of the Port-Royalists and Locke, or Sergeant’s Aristotelian
theory. But the ceteris paribus is important here: if Berkeley cannot explain human
thought at least as well as his competitors, then abstract ideas, however bizarre they
may be, will have to be accepted to save the phenomena. For this reason, Berkeley
goes on, in the succeeding sections of the Introduction, to present a two-pronged attack.
He argues both that abstract ideas cannot explain the phenomena they were meant to
explain, and that those phenomena can be explained with his own sparser resources (cf.
Flage 1986, 496-498). The latter task is, of course, of central importance to the present
study, since it is here that Berkeley begins to develop his novel theory of language.
However, before examining Berkeley’s positive theory, we shall consider Berkeley’s
arguments for the uselessness of abstract ideas, as well as two strategies for escaping
Berkeley’s anti-abstractionist arguments.
2.2.3 The Uselessness of Abstract Ideas
Abstract ideas, Berkeley says, were posited primarily to explain two phenomena: the
existence of meaningful general terms in language (PHK, Introxx11-14; MI,xx16-23),
77
and the possibility of general knowledge (PHK, Introxx15-17; MI,xx24-29). Berkeley
argues that abstract ideas, if they existed, would be “of no use for those ends to which
they are thought necessary” (PHK, Introx21), that is, that they could not explain the
phenomena in question.
Berkeley gives two reasons for supposing that abstract ideas cannot explain the
meaningfulness of general terms. The first is that the theory of abstract ideas does not
provide an adequate account of what it is for a proposition to be general.
The Newtonian claim that “the change of motion is proportional to the impressed
force” is general in the sense that “whatever motion I consider, whether it be swift or
slow, perpendicular, horizontal or oblique, or in whatever object, the axiom concerning
it holds equally true” (Introx11). According to the Port-Royalists and Locke, the propo-
sition that the change of motion is proportional to the impressed force is formed in the
mind by affirming the complex abstract idea of being proportional to the impressed force
of the abstract idea of change in motion. What Berkeley wants to know is how it comes
about that that proposition is about particular observed changes in motion. Locke’s an-
swer is that any particular observed change in motion ‘agrees with’ the abstract idea of
change in motion. We notice this sort of matching between the abstract idea and the
observed particular and therefore predicate of the particular whatever we had previously
predicated of the abstract idea. Berkeley, however, does not believe that the abstract
idea is actually doing any explanatory work in this story. What goes on in the case is
this: I’ve got a certain sentence of English, ‘the change in motion is proportional to the
impressed force.’ I see a particular phenomenon and recognize it as a change in motion,
which is to say, as an element of the extension of the English phrase ‘change in motion’
which occurred in my sentence. I therefore predicate de re of that phenomenon that it
78
is proportional to the impressed force. What is it that the abstract idea in Locke’s story
was doing that the English phrase ‘change in motion’ can’t?
Locke’s answer is that the abstract idea serves as a sort of template against which
to match phenomena to see if they are indeed changes in motion (EHU,xx3.3.14-19).
This, however, is precisely the feature of abstract ideas which is so puzzling. Berkeley
makes the point more simply, using a different example, in the Manuscript Introduction:
that the interior angles sum to 180
is universally true of particular triangles, not true of
some mysterious entity, The Universal Triangle (MI,x24).
The difficulty Berkeley poses here is essentially the same as one expressed by
Wittgenstein. Wittgenstein asks us to imagine a signpost telling us which way to go.
Of course, in order to follow the instructions, we have to know the convention about
signposts, namely, that they tell you to go in the direction the arrow points. Adding
more signposts, Wittgenstein insists, will never fully disambiguate the meaning of the
original signpost so that it can be used by someone who doesn’t know the convention
(Wittgenstein 1953,x1.85). Similarly, in order to use a lookup table, you must know
that it is a lookup table, and that lookup tables are read straight across (x1.86).
Suppose that the abstract idea of man were a list of features an entity had to have to
be a man. The list might contain the ideas rational and animal. What is one meant to
do with this list? Evidently, one is meant to match it up with all those things that have
the listed qualities, regardless of what other qualities they may or may not have. But
one needs this rule in order to use the abstract idea. So now, in order to use the word,
we need both the abstract idea and a rule for the use of the abstract idea in connection
with the word. It seems, however, that we could just as easily have connected a rule
directly to the word without making use of the intervening abstract idea. That is, we
could simply adopt the rule of applying the word ‘man’ to all and only those things that
79
are both rational and animals. This is, essentially, Berkeley’s own solution (seex3.1,
below).
26
If abstract ideas are not helpful in explaining general terms, perhaps they will be
thought to be more helpful in explaining general knowledge. In Berkeley’s time it was
still widely held, following Aristotle, that all genuine or ‘scientific’ knowledge (Lat.
scientia; Gr. âpist mh) is universal (PHK, Introx15). To know something in this strong
sense was to grasp its nature and be able to deduce a variety of conclusions from that na-
ture. A ‘nature’ is here understood as a universal. The paradigm case of such ‘scientific’
knowledge was Euclidean geometry. The Euclidean definitions and axioms were though
to explicate the nature of figure or extension, and Euclidean geometry is a collection of
deductions from those definitions and axioms.
For Sergeant, taking a more strictly Aristotelian line, the universal nature is a real
thing literally present in the mind and the deductions proceed from it. For the Cartesians,
we must instead apprehend the ‘true and immutable nature’ of the thing by pure reason
so that we have an idea in the mind which is a sort of copy of that nature (CSM, 2:44-
45). For Locke, we can only have this sort of knowledge in cases where our abstract
idea just is the nature of the thing, which is to say, in the case of mixed modes. This is
the case with the geometric natures (EHU,xx2.31.3, 4.4.5-6).
Berkeley’s objection here is essentially the same as his objection to the use of ab-
stract ideas to explain the meaningfulness of general terms. The question is “how we
can know any proposition to be true of all particular triangles” in the absence of “a par-
ticular demonstration for every particular triangle, which is impossible.” The answer is
supposed to be that the abstract idea (or universal nature) includes only those features
“in which all the particulars do indifferently partake” (PHK, Introx16). But, of course,
26. Jonathan Bennett raises this objection against Locke’s theory, but doesn’t seem to recognize its
connections to Berkeley (Bennett 1971, 24-25).
80
in order to carry out the proof and thereby come to know something about all trian-
gles, one needs to know that the features included in the abstract idea, the features one
was using, are only the features common to all triangles. Once one knows this, what
useful role is the abstract idea playing? One could just as easily use some particular
triangle and make sure that the proof appeals only to the correct features (PHK, Intro
x16). Furthermore, one need not use a ‘triangular idea’ in the course of performing the
proof at all: it suffices to know what features are common to all triangles and deduce the
conclusion from one or more of those features.
27
Note that these arguments are not escaped by the Winkler-Ayers interpretation which
says that Locke takes abstraction to involve having a particular idea in a certain way by
selectively attending to certain of its features (cf. Roberts 2007, 56-58). We can think
of this, rather crudely, as involving certain features of the idea (the features common to
all triangles) having a certain phenomenological ‘glow’ which singles them out from the
rest of the features. This sort of ‘glow’ will not solve the problems just discussed, for one
must still know what to do with the ‘glowing’ features. As a result, the having of such
an idea in such a way will not be a mental state which, in virtue of its intrinsic nature,
represents generally. General representation can occur only in the context of an ongoing
cognitive process which proceeds according to certain rules. Once this is admitted,
however, there is nothing left for abstraction to explain which cannot be explained just
as well or better without it.
27. The ideas of these features are, of course, further general ideas, so Berkeley needs to show that the
reduction can be carried out without circularity or regress. Berkeley’s attempt to do this will be discussed
inx3.1, below.
81
2.3 Causal Representation
Berkeley argues that no entity could be intrinsically well-suited to represent generally,
and he wants to infer from this that ideas represent only in virtue of how they are used
by the mind. In the next section, we will consider a last-ditch effort to save intrinsic
representation; here I want to consider an attempt to find a via media between the posi-
tion Berkeley attacks and the position he endorses. This position would claim, contrary
to the alleged result of the phenomenological appeal, that ideas have their representa-
tive content independent of their use in a train of thought, while denying that they have
this content intrinsically. The view I have in mind is a certain variety of causal the-
ory of mental representation. I will argue, however, that the apparent middle ground
here is unstable: causal theories which claim that the representative content of ideas
is independent of the use to which the mind puts those ideas face a variety of serious
objections which they are unable to answer. Any plausible causal theory will amount
to what we might describe as an externalist use theory of representation (as opposed to
the internalist use theory I will be attributing to Berkeley). These theories hold, with
Berkeley, that ideas get their representational content by being used by the mind accord-
ing to certain rules; however they hold, against Berkeley, that these rules may refer to
circumstances which are outside the agent and to which the agent has no independent
epistemic access. Berkeley does need to rule out these views if his argument for imma-
terialism is to succeed (seex5.2 and ch. 6, below); however, the existence of these views
does not challenge the success of the argument about language in the Introduction to the
Principles.
I argued inx1.3 that Locke endorses a causal theory of representation, so Locke’s
text is a reasonable place to begin our discussion. However, as we shall see, the key
text for Locke’s causal theory of representation shows that he has muddled together
82
several quite distinct ways in which ideas might represent in virtue of their causes. I
shall not attempt to sort out this muddle or identify some particular theory as Locke’s
considered view, nor will I attempt to harmonize what Locke says in this text with what
he says elsewhere. Instead, I will simply use Locke’s text as a jumping-off point for a
discussion of the philosophical issue of whether a causal theory can be used to escape
Berkeley’s arguments. In the ensuing philosophical discussion, I will show how attempts
to solve a variety of difficulties faced by any theory of the general sort Locke describes
will force that theory into a shape that looks more like the recent approach which Jerry
Fodor has dubbed ‘Wisconsin semantics’ (Fodor 1984). Finally, I will argue that this
sort of ‘Wisconsin’ causal theory is really a variety of use theory.
The principal support for attributing a causal theory of representation to Locke is his
argument that all simple ideas are ‘real’ (i.e. veridical). Locke writes:
Our simple Ideas are all real, all agree to the reality of things. Not that they
are all Images, or Representations of what does exist, the contrary whereof,
in all but the primary Qualities of Bodies, hath already been shewed. But
though Whiteness and Coldness are no more in Snow, than Pain is; yet those
Ideas of Whiteness, and Coldness, Pain, etc. being in us the Effects of Pow-
ers in Things without us, ordained by our Maker, to produce in us such
Sensations; they are real Ideas in us, whereby we distinguish the Qualities,
that are really in things themselves. For these several Appearances, being
designed to be the Marks, whereby we are to know, and distinguish Things,
which we have to do with; our Ideas do as well serve us to that purpose, and
are real distinguishing Characters, whether they be only constant Effects, or
else exact Resemblances of something in the things themselves: the reality
83
lying in that steady correspondence, they have with the distinct Constitu-
tions of real beings. But whether they answer to those Constitutions, as to
Causes, or Patterns, it matters not; it suffices that they are constantly pro-
duced by them. And thus our simple Ideas are all real and true, because
they answer and agree to those Powers of Things, which produce them in
our Minds, that being all that is requisite to make them real, and not fic-
tions at Pleasure. For in simple Ideas, (as has been shewn,) the Mind is
wholly confined to the Operation of things upon it.; and can make to it self
no simple Ideas, more than what it has received (EHU,x2.30.2).
A terminological note is in order as to the second sentence of this quotation. Locke
there uses the word ‘representation’ in a narrow sense, to refer only to what literally
re-presents something, i.e. presents it again, by being a copy or resemblance of it (cf.
LW, 4:75). Berkeley also sometimes uses the word in this narrow sense (see Winkler
1989, 14-21).
28
I will, nevertheless, continue using that word in the broad sense in which
philosophers typically use it today. In this sense, an idea represents whatever it is of or
about. Locke clearly holds that secondary quality ideas are ideas of certain features of
bodies.
In this passage, Locke argues for the claim that every simple idea represents a feature
that is instantiated somewhere in the world; that is, no simple idea is purely fictitious in
the way the complex idea of a unicorn is. Locke argues as follows. Since there are no
innate ideas, every idea must have come into the mind at some time; that is, there must
have been some time at which the mind changed from not having that idea to having it.
But no change can occur except through the exercise of some power capable of bringing
28. However, contrary to Winkler, Berkeley does not always use the word in this narrow sense. See,
e.g., MI,x20; NTV,x143; Alc,x7.11.
84
about that change (EHU,x2.21.4). Now, the mind “can make to it self no [new] simple
Ideas,” so the power which brings it about that the mind has a new simple idea must be
outside the mind. What the simple idea represents is this power, and the very fact that
the mind has the simple idea guarantees that this power exists outside the mind.
Although the latter part of the section suggests a more sophisticated theory, which
will be discussed in a moment, this argument suggests a very simple theory of repre-
sentation. According to this theory, each simple idea must come into the mind from the
outside at some time, and whatever caused that original idea token is what ideas of that
type represent for the rest of that mind’s existence.
This theory is indeed the sort of middle way we are seeking, since it holds that ideas
represent by having a certain causal history independent of how they are used by the
mind. Having this causal history is, of course, an extrinsic feature. The theory also
explains how the idea comes to represent a particular feature of the object rather than
the object as a whole: it is by the exercise of a certain power that the object causes the
idea, and the idea therefore represents just this power of the object, and not any of its
other features. The theory also deals with a pressing problem faced by certain other
causal theories, namely, how misrepresentation is possible (Fodor 1984, 234; Dretske
[1994] 2000, 213-216). On this view, the simple idea could misrepresent in either of two
ways: first, it could be caused by a different power than the one it represents. Second,
the mind could use this idea in thought or imagination to represent the power in question
as existing somewhere other than where it in fact exists.
Unfortunately, the very feature which generates this solution also gives rise to a very
serious problem for this simple theory: systematic misrepresentation can occur far too
easily. For instance, if a healthy infant with a perfectly normal visual system is born in a
room with white walls and a red light bulb, then the idea which for us represents red will,
85
for the rest of that infant’s life, represent white for it. Since its visual system is perfectly
normal, it has the red idea under all the same circumstances as we do. However, the
mental state that it gets in those circumstances will represent white rather than red. As
a result, it will end up thinking all red things are white.
29
Another famous problem for causal theories is also faced by the simple Lockean
view under discussion. This is the problem of intermediate causes (Grice 1961, 142-
144). In Locke’s terminology, we may say that a red object has the power to change the
behavior of light in certain ways, and the light has the power to effect the eye in a certain
way, and the eye has the power to effect the optic nerve in a certain way, and so forth.
Now, in fact, we do describe the light and the retinal image as ‘red,’ but the signal on the
optic nerve is certainly not red. Furthermore, the red object may have been made red by
something that is not itself red; for instance the red object may be iron and water may
have caused it to rust. Although the water is not red, it has a power to cause red ideas in
me; it can do so by making iron objects become red. There will be a causal chain of this
sort every time I perceive a red idea by sense, and therefore there will be such a causal
chain in my initial acquisition of the idea. If this is the case, however, then it is not clear
how the idea can become a representation of one of these powers rather than another.
The latter part of Locke’s discussion may be helpful with respect to these difficul-
ties. There, Locke speaks of a “steady correspondence . . . with the distinct Constitutions
of real beings.” Some more complex correlation might be thought to solve, or at least
mitigate, these problems. However, it is not clear how Locke’s argument can guarantee
a ‘steady correspondence.’ The argument I have described appears to guarantee only
29. As Samuel Rickless pointed out to me, since Locke holds that there are no colors in the dark (EHU,
xx2.8.19, 2.23.11; see Rickless 1997, 305-309), he presumably also thinks that things which are normally
white are actually red, rather than falsely appearing red, under red light. However, if Locke allows that
the idea of red can ever be caused by anything other than the quality redness (for instance, if he allows
that there can be phenomenally red hallucinations), then a similar, though perhaps more complicated, case
can be constructed.
86
that each token of a given idea is brought about by some power. Locke may be attempt-
ing to get around this difficulty by appealing to some sort of uniformity of nature thesis
when he says that our ideas are “the Effects of Powers in Things without us, ordained
by our Maker, to produce in us such sensations.” However, solving the above difficulty
in this way without departing from the simple theory would require ruling out aber-
rantly caused perceptions. Aberrantly caused perceptions do sometimes occur, so any
uniformity thesis which will do the needed work is false.
One response to this line of argument would be to admit that Locke’s claim that all
simple ideas are necessarily ‘real’ is too strong, for there is a way a simple idea could
fail to be real, namely, by not having a “steady correspondence . . . with the distinct
Constitutions of real beings.” On this kind of view, a simple idea would represent the
unique objective feature of the world of which it is a reliable indicator and, if there is no
such feature, would be (in Locke’s terminology) ‘fantastical’ rather than ‘real,’ which is
to say, essentially, that it would fail to refer.
One may, of course, worry about the skeptical consequences of this theory. However,
it faces an even more serious problem. It seems possible that some of our simple ideas
could represent disjunctive features of the world.
30
Consider the rather mundane case
of red-green color blindness. In this case, it seems that we should not say that the
individual’s visual system misrepresents red as being green, or misrepresents green as
being red, but that the visual system uses one idea to represent objects as being either
green or red. The representation of objects as green-or-red is precisely as reliable as a
normal perceiver’s representation of this disjunctive property; it is only in distinguishing
between green and red that the color blind individual suffers any disability. However,
once we admit the possibility of disjunctive representation, we have a serious problem:
30. In fact, John Campbell has argued that, for Locke, all secondary quality ideas represent disjunctive
features of the world (Campbell [1980] 1998).
87
given that nature is to some degree lawful, there will be some massively disjunctive
property which exactly describes all the ways the idea can be caused. The idea will then
represent this property, and misrepresentation will be impossible (Fodor 1984, 241-242;
Dretske [1994] 2000, 214, 220-222).
In the text we have been considering, there is a hint of another strategy, which might
mitigate this difficulty. Locke says that the “several Appearances [were] designed to
be the Marks, whereby we are to know, and distinguish Things, which we have to do
with.” In other words, he appeals to the purpose for which these powers were “ordained
by our Maker, to produce in us such sensations.” (This can also be seen as a hint in
the direction of Berkeley’s divine language theory. See chapter 8.) Although there is
certainly no appeal to God, the claim that the purpose or aim of a symbol, the role it
plays in the functioning of the organism, is critical to its content is the central element
in the Wisconsin semanticists’ response to this problem as well (Fodor 1984, 243-248;
Dretske [1994] 2000, 216-226). The view here is that some state of an organism comes
to be reliably (if imperfectly) correlated with something that organism “has to do with,”
i.e. something the organism has a practical need for information about. Through either
an evolutionary process or a learning process, the state then takes on the function of
conveying needed information to the organism.
Here, however, we no longer have the via media we were seeking; instead, Berke-
ley’s point has been conceded. It is in virtue of the complex rules governing the tokening
of that state within the organism that the state has the content it does. The Wisconsin
semanticists differ from Berkeley only in having an externalist view of the rules in ques-
tion. The organism aims to follow a rule of the form token state-typeS when and only
when external circumstanceC obtains. A person cannot, of course, consciously adopt
this as a goal, since the person knows (or believes) that C obtains only by being in
88
stateS. However, there are thought to be sub-personal purposive systems in the human
organism, and these systems have this aim. Thus the person is able to represent the cir-
cumstance because her perceptual apparatus follows a certain rule, and this despite the
fact that she may not know the rule, and even if she did it would be impossible for her
to follow it consciously or intentionally. Berkeley does reject this view in favor of an
internalist conception of the rules governing mental and verbal signs,
31
but his reasons
for rejecting it are not to be found in the arguments of the Introduction, and so further
discussion of this view will be postponed to later chapters. Here we may simply con-
clude that the causal theory does not provide any middle ground between the intrinsic
representation theories Berkeley attacks and use theories of representation, such as the
one he endorses.
2.4 Primitive Representation
If Berkeley’s conclusion is accepted, it seems that there is no alternative to the adop-
tion of some sort of use theory about mental and linguistic representation. However,
as we have seen, the argument depends on what Winkler calls “Berkeley’s explanatory
challenge,” that is, the assumption that we are looking for some sort of reductive expla-
nation of representation. As Winkler notes, some abstractionists may regard their claim
as purely descriptive rather than explanatory (Winkler 1989, 45-46). Such a philosopher
might take representation as primitive, denying that it is the sort of thing that admits of
explanation.
31. However, as we shall see, Berkeley does not hold that the individual must be able to articulate
explicitly the rule she is following. He is an ‘internalist’ about the rules only in the sense that the circum-
stances to which the rules refer – that is, the circumstance in which a rule instructs us to perform an action
– must be accessible to the agent. This corresponds roughly to what, in epistemology, William Alston has
called ‘internalist externalism’ (Alston 1988). On rule-following in Berkeley’s theory of language, see
ch. 5, below.
89
This approach is suggested by the assertion in the first chapter of the Port-Royal
Logic that “The word ‘idea’ is one of those that are so clear that they cannot be ex-
plained by others, because none is more clear and simple” (Arnauld and Nicole [1662]
1996, 25). The Port-Royalists go on to argue against imagism, which they take to be
an unsuccessful attempt at reductive explanation of mental representation. Furthermore,
as we have seen (x1.2), the primitivist approach is endorsed quite explicitly by Arnauld
in later writings. Arnauld’s view (and likely also the view of his collaborators) is that
idea-perceptions are sui generis mental acts, distinct from sensations and images, whose
nature is to represent.
This kind of view must take the property representing redness to be utterly distinct
from the property redness. In doing so, it is able to avoid the difficulties raised in Berke-
ley’s argument against the possibility of abstract ideas. Berkeley’s argument hinged on
the claim that the abstract idea of, e.g., apple can neither have nor lack the property red-
ness, for if it has that property it will represent its objects (i.e. all apples) as red, but if it
lacks that property, it will represent its objects as non-red. Furthermore, since all apples
are colored, the abstract idea of apple ought to be colored, yet it can’t be any particular
color, since not all apples are the same color. On the primitivist view, these problems
do not arise, for the negation of representing redness is not representing redness, which
is a completely different feature from representing non-redness. Thus the abstract idea
of apple can obey the three logical principles (PNC), (PEM), and (DET) (see above, p.
66). The idea will not itself be colored, and therefore (DET) does not require it to have
a particular color. It will not be red, nor will it represent either redness or non-redness.
There is thus no property which it must neither have nor lack, nor is there any property
it must both have and lack. Finally, it will have the property representing coloredness.
However, this property should not be regarded as a determinable.
90
There is, however, a very deep puzzle about this view, and that is the puzzle of just
how a representational property (e.g. representing redness) is related to the property it
represents (e.g. redness) (cf. Chisholm 1982, 40-42). This is analogous to the problem
of how the primitivist (‘magical’ or ‘unsound’
32
) theory of the representational content
of propositions can explain the ‘makes true’ relation which can obtain, and be known
to obtain, between the world and a particular proposition (Van Inwagen 1986, 202-207;
cf. D. Lewis 1986,x3.4). According to the view under consideration, the property
representing redness is a simple, primitive property utterly distinct from redness, but it
must of course be somehow related to redness, for redness is the thing that it represents.
There is not, of course, any sort of contradiction here; merely a mystery.
Sergeant and the Aristotelian tradition may in fact be able to do better in this respect,
by introducing, rather than indefinitely many primitive categorical properties, a single
primitive relation. I earlier asserted, on the basis of Sergeant’s claim that having a no-
tion involves the presence of a universal in the mind, that Sergeant was a resemblance
theorist, that is, that he held that in order to think of redness, one must have a red thing
in one’s mind. But another interpretation, which might be regarded as more charitable,
is possible. Sergeant says that “when that [corporeo-spiritual] Part is affected after its
peculiar Nature, Corporeally; the Soul is affected after its Nature, that is, Spiritually,
or Knowingly” (Sergeant 1697, 66). Perhaps what Sergeant intends is that a red object,
like an apple, bears one relation to the universal redness, and to have a notion of red-
ness is for the soul to bear a different relation to that same universal.
33
If this is what
32. David Lewis had dubbed the view which takes possible worlds to be abstract objects and takes
their representational content as primitive ‘magical ersatzism’ (D. Lewis 1986, 141). Regarding this
terminological choice, Peter van Inwagen declares, “I will not accept this dyslogistic name for the position
I propose to defend. I will call it Unsound Abstractionism, which is an acronym for Unscientific Naive
Superstitious Obscurantist Unenlightened Neanderthal Dogmatic Abstractionism” (Van Inwagen 1986,
201).
33. This is also a possible interpretation of Aquinas Summa Theologica, Iq84a1.
91
Sergeant intends, then he can at least explain how my representation is related to the
apple: the redness the apple instantiates is the very same universal as the redness my
soul perceives. These relations can both be taken as primitive.
In addition to requiring the introduction of a new primitive relation, this view has the
cost of committing one to realism about universals, a cost most early moderns would be
unwilling to pay. However, it does have an answer to the difficulty faced by the theory I
have attributed to the Port-Royalists as to the nature of the relation between the idea and
the feature it represents. Furthermore, this theory would only posit a single primitive
relation, rather than needing to posit indefinitely many primitive properties.
In any event, it must be admitted that either of these primitivist theories can escape
Berkeley’s criticisms and rescue abstract ideas (or Sergeant’s notions). The cost is the
introduction of a new primitive. Note, however, that these theories do not do away with
any of Berkeley’s primitives: the Port-Royalists accept “images formed in the brain”
(Arnauld and Nicole [1662] 1996, 25), and Sergeant accepts ‘phantasms.’ If, there-
fore, Berkeley can explain all the phenomena with his sparser resources, this will be a
significant advantage for his theory.
2.5 Conclusion
The Introduction to the Principles is an attack on meanings. That is, it is an attack on
the view that words get to be meaningful by being ‘linked up,’ by convention, with some
special class of entities, which can then be referred to as the meanings of those words
(see Figure 1.1, page 15). The core of Berkeley’s attack on this doctrine is his argument
against abstract ideas. That argument is meant to show that no object could possibly be
intrinsically well-suited to represent generally, and so that the kind of non-conventional
92
linkage which the Theory of Meanings imagines between meanings and objects is im-
possible. In order to show this, he considers how the proponents of abstraction describe
the ideas we form by this process. In addition to noting that he cannot discover any
such ideas in introspection, Berkeley argues that such ideas are in fact impossible, and
that, even if there were such things, they would still not be intrinsically well-suited to
represent generally.
I have given a qualified defense of this argument. Abstract ideas would be bizarre
and mysterious entities, and converting them to logical rather than psychological entities
would not make them any less so. They can, in the end, be saved only by the introduction
of new primitives; parsimony requires that we not introduce such primitives unless the
phenomena force us to do so. What remains to be seen is whether the phenomena do
force us to do so, that is, whether Berkeley can adequately explain our mental activities
using mental images alone.
93
Chapter 3
The Rudiments of Berkeley’s Positive
Theory
So far we have been focused on the negative project of the Introduction to the Principles,
where Berkeley criticized the Theory of Meanings. I have concluded that Berkeley’s
arguments show that we have strong reason to prefer a theory of the sort he wants to
offer, provided that such a theory can account for the phenomena (cf. Flage 1986, 496-
498). Beginning as early as the notebooks and Manuscript Introduction, Berkeley has a
strong interest in providing just such a theory. However, in this early period he proceeds
piecemeal, dealing with one sort of linguistic rule at a time. In this chapter, I examine
Berkeley’s treatment of three kinds of language: general terms, operative language,
1
and
mathematical and scientific language. As will become clear in the discussion to follow,
Berkeley does not recognize any sharp dividing lines between these domains; there is
considerable overlap. Nevertheless Berkeley discusses them separately, and it will be
convenient for us to follow him in this.
The theory of general terms is Berkeley’s replacement for the doctrine of abstract
general ideas. The theory of operative language is meant to show that Berkeley’s theory
can account for the meaningfulness of certain terms in moral and religious discourse
which do not correspond to ideas, and therefore cannot be accounted for by idea-based
versions of the Theory of Meanings. Berkeley believes that mathematical and scientific
1. I adopt this term from Williford 2003.
94
language provides another example of meaningful terms that fail to correspond to ideas,
though these are of a rather different sort than the terms employed in operative language.
In this chapter, I will provide an exposition Berkeley’s early remarks on each of these
types of discourse; in the next chapter I will argue that Alciphron 7 draws these threads
together into a general theory of language.
3.1 General Terms
Since Berkeley’s critique of the Theory of Meanings relied heavily on his claim that it
is unable to account for the meaningfulness of general terms, it is imperative that he
provide an alternative account of this phenomenon. Berkeley’s strategy is to let one
general term stand directly for any of several objects without the intervention of an idea.
In the Manuscript Introduction, this led Berkeley to deny that there were any general
ideas at all (MI,x20). However, in the published version, Berkeley instead states that
ideas can become general in the same way words do, namely, by the introduction of the
right sorts of conventional rules (PHK, Introx12; see Belfrage 1986b, 326-328).
Berkeley’s most detailed treatment of general terms is found in the Manuscript In-
troduction. Other than the reversal of Berkeley’s position on general ideas, the changes
from the manuscript to the published version appear mostly to have been made for the
sake of brevity, rather than changes of opinion. The focus here will, therefore, be on
the Manuscript Introduction, though the changes made to the published version will be
examined along the way.
95
Berkeley’s discussion begins with the claim that “General Words . . . become so
[only]
2
by being made to mark a Number of particular Existences.” (MI,x18). Sim-
ilarly, Berkeley writes in the published Introduction, that “the name ‘line’ . . . must be
thought to derive its generality from . . . the various particular lines which it indifferently
denotes” (PHK, Introx12).
3
More specifically, Berkeley’s view is that “one Word [is]
made the sign of a great number of particular Ideas, between which there is some like-
ness, & which are said to be of the same Sort” (MI,x19). Berkeley is thus endorsing
a version of resemblance nominalism about universals (see Armstrong 1989, ch. 3).
Berkeley is aware that, since resemblance comes in degrees, the extension of a word
may be vague, but he is untroubled by this consequence. He writes that sorts do not
“have any precise Bounds or Limits at all,” but this is unproblematic, since “Language
[is] made by & for the common Use of Men, who do not ordinarily take notice of the
Minuter & less Considerable Differences of Things” (MI,x19).
4
The basic theory, then,
is that we call something by a general name, e.g. ‘red,’ because of a resemblance (‘like-
ness’) between it and the other things to which we apply that name (cf. NTV,x128).
In a later section of the manuscript, Berkeley summarizes his view by writing that
“there is in Truth an Homonymy or Diversity of significations in every Name whatsoever
except only the proper Names” (MI,x31).
5
The use of the word ‘homonymy’ is clearly
2. The word ‘only’ is inserted above a caret mark.
3. Berkeley often says that a general term denotes all of the ideas in its extension. In this passage from
the published Introduction, Berkeley makes it clear that the general term denotes all of the ideas in its
extension not by standing for the class or collection of such ideas, but rather by ‘indifferently denoting’
each of the ideas in the class – that is, it stands for the ideas in its extension severally, not collectively. I
thank Samuel Rickless for drawing my attention to this important distinction.
4. Berkeley later crossed out this entire passage. However, the view that because natural language is
designed for practice rather than speculation it exhibits various harmless imprecisions recurs at PHK,x52
and DHP, 245-246. For more on these imprecisions, see below,x7.2.4.
5. Berkeley later crossed out the word ‘homonymy,’ but this was probably only because appearing to
know something about Aristotelianism was out of fashion among the ‘Moderns’; Berkeley also deleted
96
a reference to Aristotle, who writes, “When things have only a name in common and
the definition of being [or: ‘account of the essence’] which corresponds to the name is
different, they are called homonymous” (Aristotle Cat. 1 1a1-2).
6
However, Aristotelian
homonymy is not always mere equivocation. According to Aristotle, some objects are
systematically homonymous. (In Aristotle’s usage, objects are said to be ‘homonymous’
with respect to a particular word.) In these cases, although the ‘account of the essence’ is
not the same in each instance, the instances are systematically related to one core usage,
so that the use of the same word in each case is no mere coincidence. Aristotle illustrates
this point with the use of the word ‘healthy’ (Gr. Ígeinìc). He writes, “Everything which
is healthy is related to health, one thing in the sense that it preserves health, another in
the sense that it produces it, another in the sense that it is a symptom of health, another
because it is capable of it” (Aristotle Met., G2 1003a35-38). Christopher Shields calls
this phenomenon ‘core-dependent homonymy’ (Shields 1999, ch. 4): the various objects
which are called ‘healthy’ are so-called on account of their relation to a common ‘core,’
although each is related to the core in a different way.
The important point here for our purposes is that, in Berkeley’s theory of general
terms, just as in Aristotelian core-dependent homonymy, although there is no universal
which is shared by all of the members of the extension of the term, the term is not
merely equivocal, but is used in a systematic way across a wide variety of cases. As
Berkeley remarks, “It is one thing for to keep a name constantly to the same definition,
a passage at MI,x19a that showed awareness of the Medieval debates about universals, and one atx22
which quoted Aristotle – in Greek, no less! The only references to Aristotle and Scholasticism remaining
in the published version are mocking (PHK, Introxx17, 20). The other later changes to this passage are
not substantive.
6. Recent treatments of Aristotle’s theory of homonymy and its role in his philosophy include Shields
1999 and Ward 2008.
97
and another to make it stand every where for the same idea; the one is necessary, the
other useless and impracticable” (PHK, Introx18).
The definition provides a rule for the application of the word. The word, as Berkeley
likes to say, ‘denotes indifferently’ any object which satisfies the definition. However,
language learning would be impossible if the only way to know the meaning of a word
was to have an explicit definition of it in some language and understand that definition.
Berkeley should therefore hold that what is fundamental is the rule itself, of which the
definition is merely a statement. That Berkeley does in fact hold this is further confirmed
by the fact that, as we shall see in the next two chapters, the adoption of this view allows
his theory of general terms to form a coherent part of his total theory of language.
The rule for the use of the word ‘triangle,’ that it should be applied to any “Plain
Surface comprehended by three right lines” (MI,x32; cf. PHK, Introx18), tells us in
exactly what way the objects in the extension of ‘triangle’ must be alike: they must
resemble one another in, for instance, having the same number of sides. However, as
Berkeley points out, “in the Definition, it is not said, whether the Surface be great, or
small, black, or white, or Transparent [&c]
7
, whether the sides are long or short, equal
or unequal, or with what angles they are inclin’d to each other” (MI,x32). Again,
mature speakers are able to state all of these rules – the respects in which there must be
resemblance and the respects in which there may not be – but, if Berkeley’s theory is
to succeed, it must also be possible to follow the rule without being able to state it in
language. Berkeley must take our capacity to recognize objects as falling under concepts
as primitive (cf. Taylor 1978, 115).
It might be thought that the introduction of this primitive puts Berkeley’s theory in
the same boat as the Theory of Meanings. If we are to have primitive recognitional
7. Inserted above a caret mark.
98
capacities, why not have primitive representation? The answer to this is that the abstrac-
tionists already appeal to our ability to recognize similarities between particular objects
in their accounts of the formation of abstract ideas. Again, Berkeley’s claim is that if
we have the abilities which would be necessary to make and use abstract ideas (if such a
thing were possible), then we don’t need abstract ideas. If we have the ability to recog-
nize objects as resembling one another, and to distinguish between different respects in
which they resemble, an ability which the abstractionists say is needed to form abstract
ideas, then we can successfully use general terms by the direct application of this ability
without having abstract ideas.
This argument is, however, ineffective against Sergeant, for Sergeant does not be-
lieve in any process by which we form abstract ideas. Furthermore, as a realist about
universals, Sergeant can give an account of resemblance, rather than taking it as primi-
tive: objects resemble when they instantiate a common universal. As committed nomi-
nalists, most early modern philosophers would regard primitive recognitional capacities
of the mind as more parsimonious and less mysterious than Sergeant’s Aristotelian uni-
versals. Certainly the nominalist position involves fewer and less mysterious entities.
However, it does so at the cost of assigning to those entities (minds) mysterious pow-
ers or capacities. In terms of the simplicity of his ontology at this point, it seems that
Berkeley is better off than Locke and the Port-Royalists and at least no worse off than
Sergeant.
An important question about Berkeley’s theory which is discussed at length in the
Manuscript Introduction but passed over in silence in the published version is the ques-
tion of how we are to analyze sentences containing general terms, and this discussion
helps clarify Berkeley’s theory. Berkeley asks us to consider the sentence ‘Melampus
99
is an animal,’ where ‘Melampus’ is the proper name of some particular dog (MI,x34).
8
According to Berkeley’s theory, ‘Melampus,’ as a proper name, stands for this particular
dog, and ‘animal,’ as a general term, can stand for this particular dog or any relevantly
similar entity. Berkeley says that when I utter this sentence, “All that I intend to signify
thereby [is] only this. That the particular [creature]
9
thing I call Melampus has a right to
be called by the Name Animal” (x34). In other words, what this sentence conveys is that
the rules governing the use of the word ‘animal’ permit its application to Melampus.
Although the sentence signifies or conveys this information, it need not give rise
to an explicit meta-linguistic belief that Melampus is rightly called ‘animal. ’ Rather,
it should be understood as producing in the hearer who accepts it certain dispositions,
including the disposition to call Melampus ‘animal.’
10
Berkeley himself does not seem
to see this point clearly in the Manuscript Introduction; however, as will be discussed
in detail in chapter 5, below, Berkeley clearly recognizes the possibility of following
rules without having explicit knowledge of them, and it is required for the success of his
theory that most linguistic rules be of this sort.
If the hearer has acquired the rest of the relevant linguistic rules, a wide variety
of other dispositions will be created by her acceptance of ‘Melampus is an animal.’ For
instance, the hearer has acquired the rule that ‘animal’ is properly applied only to objects
to which the phrase ‘living thing’ also applies; the hearer will therefore be disposed,
under appropriate circumstances, to call Melampus ‘living thing.’ Furthermore, if the
8. Melampus was a famous seer in Greek mythology (Hornblower and Spawforth 2005, s.v. ‘Melam-
pus (1)’). So far as I know, there is no significance to Berkeley’s choice of this name in his example.
9. Brackets original.
10. On the linguistic practice of accepting or rejecting the assertions of others, see Brandom 1983. On
the relationship of this disposition to the meta-linguistic belief, see below,x7.1.
100
hearer also speaks Greek, she will be disposed to call Melampus ‘zÀ | on.’
11
This much
can be done using only the rules for general terms; as we shall see, this is only one of
the many sorts of linguistic rules discussed by Berkeley.
As has been noted, in the Manuscript Introduction Berkeley denies the possibility
of general ideas. By the time he wrote the published version, he had come to see that
this was a mistake. The reason it is a mistake is that Berkeley’s immaterialism has the
consequence that a word (or, rather, an utterance or inscription of a word) just is an idea,
and words can clearly be general (cf. Lennon 1988, 242-243). There is, then, no reason
why there should not be general ideas of other sorts. In the published Introduction,
Berkeley writes, “By observing how ideas become general, we may better judge how
words are made so” (PHK, Introx12). Berkeley in fact holds that words and ideas
become general in precisely the same way, that is, by the adoption of a rule of use which
allows the idea or word to be substituted indifferently for any of a broad class of ideas.
Berkeley describes the formation of general ideas as follows:
an idea, which considered in it self is particular, becomes general by being
made to represent or stand for all other particular ideas of the same sort.
To make this plain by an example, suppose a geometrician is demonstrating
the method of cutting a line in two equal parts. He draws, for instance,
a black line of an inch in length; this, which in itself is a particular line,
is nevertheless with regard to its signification general, since as it is there
used, it represents all particular lines whatsoever . . . And as that particular
11. As Samuel Rickless pointed out to me, the elimination of a language-independent mental contents
creates serious problems for translation. One way in which such problems can be mitigated (but by
no means solved) is by attention to the rules followed by bilingual speakers (Quine 1957, 7-8; 1960,
27-29, 47). If detailed accounts of reference and truth in Berkeley’s theory can be given (I undertake
these projects in chs. 6 and 7, respectively), then similarity or sameness of reference for words and truth
conditions for sentences may provide an adequate foundation on which to build a system of translation
rules, even in the absence of bilingual speakers. For more on translation, seex7.1, below.
101
line becomes general, by being made a sign, so the name ‘line’ which taken
absolutely is particular, by being a sign is made general (PHK, Introx12).
Berkeley here clearly states that the general idea of a line and the general term ‘line’
become general in the same way. However, there is a clear point of disanalogy which
some commentators have thought important: the general idea of a line is itself a line
and therefore resembles the objects it signifies, whereas the general term ‘line’ does not
(Bolton 1987, 69; Winkler 1989, 42-43; Jesseph 1993, 35; Jacovides 2009, 425). The
importance of this disanalogy is often defended by claiming that it is necessary for the
success of Berkeley’s arguments for immaterialism. For instance, Martha Brandt Bolton
writes, “If a mind sets an idea up to represent something else, the idea must share certain
features with its acquired object. Berkeley draws from this the important conclusion that
the acquired object must be another idea” (Bolton 1987, 69). This, however, is incorrect
for two reasons. First, the question of whether ideas must represent by resemblance is
quite irrelevant to immaterialism, since linguistic representation, for Berkeley, does not
depend on the having of ideas. If we can meaningfully talk about material substances
without having ideas of them, in the way we can meaningfully talk about spirits without
having ideas of them, then Berkeley’s arguments for immaterialism fail. This is the
thrust of what Phillip Cummins calls ‘Hylas’ Parity Argument’ (DHP, 233; Cummins
1982).
12
Constraining the representation of general ideas to what they resemble will not,
of itself, make the case for immaterialism, since words represent without resembling and
their representation is not parasitic on representation by ideas. Second, by overplaying
the importance of resemblance, these commentators underplay the parallel Berkeley
sees between general ideas and general terms. Berkeley’s remark that “as the [inch-line]
owes its generality, not to its being the sign of an abstract or general line, but of all
12. For more on this issue, see below, ch. 6.
102
particular right lines that may possibly exist, so the [word ‘line’] must be thought to
derive its generality from the same cause” (PHK, Introx12) strongly suggests that he
means to claim that the general idea represents in precisely the same way the general
term does.
Berkeley says that both the general term ‘line’ and the general idea of a line “derive
[their] generality from the same cause, namely, the various particular lines which [they]
indifferently denote[]” (Introx12). In other words, in both cases a conventional rule is
adopted which associates the sign with a broad class of ideas, each of which has (to use
the language of Berkeley’s ‘Melampus’ discussion) equal right to be associated with the
idea or term.
What, then, is the difference between a general idea and a general term? The differ-
ence is simply the nature of the associated rules. The rules for the use of a general idea
make reference to the idea ‘considered in itself’ and the rules for the use of a general
term do not. For instance, the rules for the use of the general idea of a triangle (Intro
xx15-16) tell us that this idea may be substituted for another idea, provided that idea
resembles it by (inter alia) having the same number of sides. Furthermore, the rules
tell me that I may perform demonstrations (Berkeley seems primarily to have Euclidean
constructions in mind) using the general idea of a triangle and apply the results of those
demonstrations to any of the ideas it stands for, provided that the demonstration makes
use only of certain features (e.g. the number of sides) and not of others (e.g. the mag-
nitudes of the angles). When using the general term ‘triangle,’ the rules never tell me to
operate on the word itself in any way.
The point here is that, contrary to the commentators mentioned, the resemblance
of the general idea to its objects is not doing important work as far as its ability to
103
represent them is concerned.
13
Representation (in the broad sense; see p. 84) does not
require resemblance for Berkeley; what it requires is that we be able to follow some
rules which associate the sign with its objects.
3.2 Operative Language
One of the tenets of the Theory of Meanings which Berkeley attacks is “that language
has no other end but the communicating our ideas” (PHK, Introx19). Against this view,
Berkeley cites three examples of other aims which speech may have: “the raising of
some passion, the exciting to, or deterring from an action, [and] the putting the mind
in some particular disposition” (Introx20). Much of the literature refers to Berkeley’s
discussion of these alternative ends of language as his theory of ‘emotive meaning’ (see,
e.g., Belfrage 1986a; Berman 1994, 143-148). However, the use of this term may sug-
gest that Berkeley is to be classified as an ‘emotivist,’ in the twentieth century sense of
that term, about ethics and/or revealed theology, a thesis which is quite controversial.
Furthermore, even some of those who use the term ‘emotive meaning’ acknowledge that
not all of the uses of language discussed under this heading have to do with emotions
(see, e.g., Belfrage 1986a, 644-645). Kenneth Williford therefore proposes, on the ba-
sis of Alc,x7.17, that we use the term ‘operative language’ instead (Williford 2003,
272n2).
14
Here, Berkeley says that instead of “the imparting or acquiring of ideas,”
speech may aim at “something of an active operative nature, tending to a conceived
good.” In the account of general terms, Berkeley argued that the rules governing the
use of words do not always associate each word with some one idea; instead, one word
13. Here I am in agreement with Brook 1973, 35.
14. Willford cites from the 1752 edition of Alciphron, where the relevant section is numbered 7.14.
104
can often ‘indifferently denote’ any of several ideas. In offering his theory of operative
language, Berkeley argues that there may be linguistic rules of an entirely different sort,
which do not involve connections between words and ideas at all.
Again, the fullest treatment is in the Manuscript Introduction. However, the pub-
lished version is much clearer as to Berkeley’s general picture, so the discussion here
will be framed around the published text and supplemented with discussion of the
manuscript.
Berkeley’s approach is to begin by trying to understand, in particular instances, what
language is being used to accomplish, and to proceed from here to conclusions about
meaningfulness or significance.
15
Proponents of ideist versions of the Theory of Mean-
ings assumed that language only ever had one end (or, at least, that there was only one
end of language worthy of serious philosophical discussion), and that end was the com-
munication of ideas. As a result, they badly misunderstood nearly the whole of human
language, for it is very rarely the case that this is the sole end of any utterance. the
The communication of ideas “is in many cases barely subservient, and sometimes en-
tirely omitted, when [the other ends of language] can be obtained without it as . . . not
infrequently happens in familiar use of language” (PHK, Introx20).
The proponents of the Theory of Meanings were probably led astray by the fact that
they had been focusing on the sort of language that occurs in philosophical treatises
whose ends are purely theoretical or speculative. This sort of language is the exception
rather than the rule: “words . . . were framed by the vulgar, merely for conveniency and
dispatch in the common actions of life, without any regard for speculation” (DHP, 246).
15. This strategy becomes explicit in Alc,x7.8; see below,x4.3. The role of the aims and intentions of
speakers in Berkeley’s theory is emphasized by Williford 2003. There are obvious proto-Wittgensteinian
elements to Berkeley’s approach here, but I will not pause to discuss them at present.
105
The use of language in speculation is an extension of language beyond its original do-
main in the common actions of life. As Wittgenstein famously put it, philosophy occurs
“when language goes on holiday” (Wittgenstein 1953,x1.38).
16
When we realize that language does not aim only, or even primarily, at the communi-
cation of ideas, we ought also to realize that a theory of meaning given entirely in terms
of ideas will necessarily be inadequate. Berkeley tells us that it “often happen[s] either
in hearing or reading a discourse, that the passions of fear, love, hatred, admiration, dis-
dain, and the like, arise immediately in [the] mind upon the perception of certain words,
without any ideas coming between” (PHK, Introx20). Berkeley is not here talking about
some particular restricted domain of discourse. “At first, indeed,” he continues,
the words might have occasioned ideas that were fit to produce those emo-
tions; but, if I mistake not, it will be found that when language is once grown
familiar, the hearing of the sounds or sights of the characters is oft immedi-
ately attended with those passions, which at first were wont to be produced
by the intervention of ideas that are now quite omitted (Introx20).
This is the sort of thing that can happen with just any word, including perfectly well-
behaved words like proper nouns (Introx20). By constant use and constant association
with a certain emotional reaction, the word can eventually produce the emotional reac-
tion without the intervention of ideas.
Berkeley is trying to make the case that one can understand a sentence without hav-
ing ideas corresponding to each of the words involved. However, it should be clear that
16. Thus Jonathan Bennett is doubly mistaken when he claims that Berkeley is interested in “divorcing
meaning from ideas” only for non-theoretical uses of language, and hence “only in respect of the periph-
ery of language” (Bennett 1971, 54). Bennett is mistaken in supposing that Berkeley’s critique applies
only to non-theoretical language, for nothing could be more theoretical than advanced mathematics and
Newtonian physics, but these are explicitly within the scope of Berkeley’s thesis (see below,x3.3). Fur-
thermore, Bennett assumes that non-theoretical (or, as he also calls them, ‘practical’) uses of language are
‘peripheral.’ This is a view Berkeley adamantly rejects.
106
one can have an emotional reaction to a word without understanding it. For instance,
some individual might hear a word in a language he does not speak repeated over and
over again while he is being tortured, and thereby come to have a negative emotional
response to it, without ever learning its meaning. If, then, Berkeley’s observations are to
support his point, there must be cases in which having the appropriate emotional reac-
tion is sufficient for understanding the utterance. This is why Berkeley talks here about
language accomplishing its ends, and, a little below, talks about the intentions of the
speaker. There are cases, that is, where the fact that the utterance elicits a particular
emotional response is sufficient for the accomplishment of the ends at which the utter-
ance aims. This, however, will still not be sufficient if the intentions of the speaker are
too idiosyncratic. For instance, if the torturer later repeats the hated word (which, let us
suppose, does not mean torture or anything like that), intending to elicit fear from the
victim, and the victim is indeed afraid, it does not follow that the victim has understood
the word. What is needed is an example where the rules of language authorize the use
of the word to elicit the emotional (or other) reaction in question.
In the published Introduction, Berkeley offers two examples:
May we not . . . be affected with the promise of a ‘good thing’, though we
have not an idea of what it is? Or is not the being threatened with danger
sufficient to excite a dread, though we think not of any particular evil likely
to befall us, nor yet frame to our selves an idea of danger in abstract? (PHK,
Introx20)
The first example is discussed at great length in the manuscript, where Berkeley is con-
cerned with the religious doctrine of an inconceivably great heavenly reward. He writes,
107
We are told [that]
17
the Good Things which God hath prepared for them that
love [him]
18
are such as Eye hath not seen nor Ear heard nor hath it enter’d
into the Heart of Man to conceive. What man will pretend to say these
Words of the Inspir’d Writer are empty and . . . insignificant? (MI,x36)
Berkeley’s example sentence, which is a reference to 1 Corinthians 2:9 (quoting Isaiah
64:4), could be given the simpler paraphrase, ‘for those who obey God, there is a reward
of which no human being now has an idea.’ Of course, however, we understand the
sentence now. Further, Berkeley denies that there can be an abstract idea of a reward:
every idea of a reward, is an idea of some particular reward. Since St. Paul himself says
that we have no idea of the ‘Good Things,’ and he gives us no description of them, it
cannot be his intention to excite in our minds an idea of the ‘Good Things’ in question.
Instead, Berkeley says, his aim was “to make [us] more chearfull and fervent in [our]
Duty” (x36).
Promising an unspecified reward is a perfectly ordinary use of language. The aim
of that use is to motivate someone to take the course of action to which the reward is
attached. It frequently accomplishes that aim. However, since the reward is left un-
specified, no idea of the reward is given to the hearer. Nevertheless, if the hearer is
motivated in the way the speaker intends, then the speech-act has been carried off suc-
cessfully: both the speaker and the hearer are using language competently, and language
is accomplishing its ends.
Consider an everyday use of language: a flier on a lamp post reading ‘Lost Puppy;
Reward for Return.’ The reward is unspecified. One might, nevertheless, have an idea of
some particular reward – say, a certain sum of money – on the basis of social conventions
17. Brackets original.
18. Inserted above a caret.
108
regarding appropriate rewards for returning lost puppies. On the other hand, one can
equally well understand the sign without having any such idea. The rule governing
reward assertions does not require that, upon hearing such an assertion, one should
form an idea of a reward. What it does require is that uponaccepting such an assertion
one be at least somewhat motivated to perform the action in question (find and return
the puppy). This motivation may be weak, even negligible, in force, and it may be
overridden by other considerations. It may therefore have little effect on feeling and no
effect on action. However, when a competent speaker accepts the assertion, the rules
require her to have this motivation, however small.
One can, of course, reject the assertion without calling one’s linguistic competence
into question. One might think the speaker is accidentally or intentionally misusing the
word ‘reward.’ Berkeley mentions two conditions for accepting the assertion: one must
think the speaker has it in her power to bestow a reward, and one must think that the
speaker is honest (MI,x37). A further condition, not mentioned by Berkeley is that
one must think that the thing the speaker is promising to bestow is in fact a reward, i.e.
something worth having. It may be that, unbeknownst to the speaker, the thing offered
is of no value to me. In this case, the speaker would be mistaken in her belief that the
item in question is a reward (or, at least, that it would be a reward for me). This is an
indication of the rules governing the speaker’s use of the word ‘reward:’ the speaker
must have it in her power to bestow some benefit on the hearer, and intend to do so if,
but only if, the hearer does what is requested. These rules governing the behavior of the
speaker and the hearer constitute the meaning of the word ‘reward’ in this context.
Two objections may be made to this account of the meaning of ‘reward,’ one inter-
pretive and one philosophical. Interpretively it might be objected that the word ‘reward’
109
ought to be covered by Berkeley’s theory of general terms. Isn’t ‘reward’ just a per-
fectly ordinary general term, applying to such diverse things as sums of money, ice
cream cones, and heavenly bliss?
The answer to this question is emphatically ‘yes,’ but this does not vitiate the in-
terpretation given. Shortly after the mention of rewards and dangers in the published
Introduction, Berkeley writes, “general names are often used in the propriety of lan-
guage without the speaker’s designing them for marks of ideas in his own, which he
would have them raise in the mind of the hearer” (PHK, Introx20). These different sorts
of rules may apply to one and the same word, and which rule can actually be applied
will depend on the context. So, for instance, one may say ‘the reward’ and thereby refer
to some particular object of which both speaker and hearer have an idea, or one may
use ‘the reward’ in reference to the heavenly reward of which no human presently has
any idea. (Note also that this use of language will still be meaningful even if there is no
afterlife.) This supports my earlier claim that Berkeley’s theory of operative language
is not restricted to some narrow domain of discourse (e.g. moral or religious language),
but is a theory about language in general.
19
The philosophical objection to the account of ‘reward’ is this: in laying out the rules
governing the use of the word ‘reward,’ I several times used ‘reward’ and its synonyms
(and in this I follow Berkeley). Does this not vitiate the account?
This, however, simply repeats a difficulty dealt with above, namely, that speakers
cannot formulate the rules of language until after they know a language. What this
means is that one must be able to follow a rule without being able to formulate it. It
19. The fact that operative terms are not part of a ‘bracketed discourse’ of any sort means that Berkeley’s
theory is not affected by Geach’s famous objection to expressivism (Geach 1965, 462-464): Berkeley is
able to agree with Geach that operative terms are governed by the very same inference rules as other terms,
and therefore can be ‘mixed and matched’ in compound sentences and patterns of inference. Berkeley’s
understanding of inference, and its application to operative language, will be discussed inx5.4, below.
110
suffices for the correct use of ‘reward’ that the speaker has the requisite powers and
intentions and the hearer is properly motivated. No meta-linguistic thoughts or beliefs
are needed.
One might worry, however, about how it actually comes about that people follow
these rules. This is not, after all, very much like learning the meanings of concrete nouns
by hearing people say the word and seeing them point to the object. Something more
complex must be going on. About how someone acquires the word ‘reward,’ Berkeley
writes,
When he was a Child he had frequently heard those Words used to him to
create in him an obedience to the Commands of those that spoke them. And
as he grew up he has found by experience that upon the mentioning of those
Words by an honest Man it has been his Interest to have doubled his Zeal
and Activity for the service of that Person. Thus there [has] grown up in
his Mind a Customary Connexion betwixt the hearing that Proposition and
being dispos’d to obey with chearfulness the Injunctions that accompany it
(MI,x37).
The process here is essentially the reverse of the torture victim case discussed above.
One is simply conditioned to respond to the promise of a reward in a certain way
(Berman 1994, 162).
20
The ‘reward’ case is the only case where Berkeley describes, in detail, how a lan-
guage learner acquires a word with operative meaning, and it is worth noting that, in this
case, the word has both operative meaning and what we might call ‘referential meaning.’
However, Berkeley’s account of the acquisition of the operative meaning here makes no
20. It is no coincidence that Alciphron raises the objection to religion from fatalism immediately after
Euphranor has finished expounding this theory (Alc,x7.19); the theory involves forms of conditioning
which seem, at least superficially, to be at odds with libertarian conceptions of free will.
111
mention of the referential meaning. This suggests that the same account can be applied
to words that have no referential meaning at all. Furthermore, the process of condition-
ing which is described does not appear to presuppose any other linguistic abilities. It
appears, then, that operative meaning can float entirely free of referential meaning.
‘Reward’ is an example of a general term which is associated with certain rules be-
sides the rule allowing it to be substituted for the idea of any particular reward, and in
which those rules often function properly in the absence of any particular reward-idea
for which it can be substituted. (Of course, when the reward is given and received, the
parties should recognize it as a reward, applying the general term rule.
21
) The point of
the utterance in the case where no such ideas are had is to motivate someone to do some-
thing, and to take a certain emotional attitude to the task (to do it with “Chearfulness
and Zeal and Perseverance” (MI,x37)).
‘Reward’ talk exhibits the first two ends of language Berkeley listed: the utterance
aims at “the raising of some passion [and] the exciting to . . . an action.” Berkeley’s next
example deals with the third and final end of language on Berkeley’s list, “putting the
mind in some particular disposition.” Words of this sort float free of referential meaning
in a stronger sense than ‘reward’ does. Berkeley writes, “when a Schoolman tells me
‘Aristotle hath said it’, all I conceive he means by it is to dispose me to embrace his
opinion with the deference and submission which custom hath annexed to that name”
(PHK, Introx20). This is, of course, a joke, the sort of jibe at Scholasticism which was
21. Thus Berkeley’s account of the heavenly reward can be regarded as a version of what James Ross
has called ‘eschatological pragmatism:’ the view that, for the eschatological claims of Christianity, “truth
is fulfillment of expectation,” or, in other words, that the eschatological claims of Christianity are true
if and only if “each [Christian], at the parousia, will ‘find his faith fulfilled,’ each finding that ‘This is
just what I expected’ ” (Ross 1988, 283-284). Berkeley’s account is by no means identical with Ross’s,
but they share this feature: that having correct faith regarding Christian eschatological doctrine consists
not in being able to imagine, or even describe, the eschaton, but, rather, in a disposition to recognize the
eschaton as the fulfillment of one’s expectations when one gets there. Berkeley, unlike Ross, emphasizes
that this belief must be regarded as meaningful now because of its motivational role.
On Berkeley’s relationship to pragmatism about truth more generally, see below,x7.2.2.
112
fashionable among the Moderns. However, the point is a serious one. It is possible,
Berkeley thinks, to associate truth so strongly with some combination of words that
one no longer thinks about the literal meaning of the words and simply accepts what
comes after them. One naturally thinks of religious and political ideologies as examples.
One might hear a phrase like ‘the Bible says . . . ’ or ‘we Liberals believe . . . ’ and
simply accept what comes after it without ever thinking about the Bible and what it
says or about Liberals and what they believe. Furthermore, such a habit might become
so widespread in a speaker community as actually to alter the meanings of the words.
One can imagine a linguistically isolated cult with an authoritarian leader who makes
pronouncements beginning ‘the Bible says . . . ’ (or, ‘God has revealed to me . . . ’) and
is unquestioningly and without further thought believed by his followers. In this case
‘the Bible says. . . ’ might come to mean ‘I authoritatively pronounce . . . ’ and the usage
might persist long after no one knew what the Bible was. Berkeley cannot, of course,
seriously suppose that ‘Aristotle hath said it’ functions like this among the Scholastics,
for the Scholastics engaged in all sorts of debates about what Aristotle said, but this is
the serious philosophical content of his anti-Scholastic jibe.
This example shows how a word or phrase might lose all connection with ideas and
still have a significant (if nefarious) use. Berkeley’s last example in the appears to be
an example of a case where words are dissociated from ideas altogether. This is the
analysis of moral discourse.
This portion of the manuscript must be handled with care as Berkeley made quite
significant revisions to it, and then deleted the entire discussion from the published text.
In the first stratum of the manuscript, Berkeley begins by asking what is the purpose of
telling a person “that Such an Action is Honourable and vertuous.” He then denies the
Lockean account of this assertion, according to which the aim is to excite in the mind
113
of the hearer the abstract ideas of the action and of honour and virtue together with a
perception of their agreement. In place of Locke’s rejected account, Berkeley says that
“this [is] the full Purpose namely that those Words should excite in the Mind of the
Hearer an esteem of that particular Action and stirr him up to the performance of it”
(MI,x41).
It is primarily on the basis of this text that Belfrage takes Berkeley (in the
Manuscript Introduction) to endorse the view that moral assertions are disguised impera-
tives.Manuscript Introduction—) Belfrage says that, according to Berkeley, the meaning
of ‘Such an Action is Honourable’ “is exactly the same as” the meaning of ‘Perform and
esteem such an action!’ (Belfrage 1986a, 645) Belfrage likewise asserts the equivalence
of ‘There are inconceivably pleasant joys in store for blessed souls in heaven’ with
‘Act in accordance with what Christian doctrine prescribes as being our duty!’ (646)
Belfrage’s strong claim of exact equivalence cannot be correct, for each phrase clearly
differs from the other in terms of its appropriate usage and the appropriate response to
it (cf. Stevenson 1937, 24-26). The promise of a reward, for instance, is supposed to
make us cheerful and zealous, but a command does not do this. If we take the command
as coming from someone with authority, or someone with the power to punish disobe-
dience, we will obey it (other things being equal), but it will not produce cheerfulness
and zeal as the promise of a reward will. What is true is that each pair of sentences,
according to Berkeley, has the same primary aim: one promises a reward for an action,
praises an action, or commands an action all in order to get someone to perform that
action. However, these are genuinely different methods of pursuing that end, and they
are different precisely because the sentences differ in meaning.
22
22. Furthermore, as Williford points out, it is hardly credible that Bishop Berkeley should hold that
claims about the heavenly reward have no truth-value (Williford 2003, 300-301). The problem of Berke-
ley’s being saddled with forms of anti-realism he would find objectionable will be discussed at length in
our treatment of reference and truth in Berkeley’s mature theory, chs. 6 and 7, below.
114
There are further problems. In subsequent revisions, Berkeley seems to have soft-
ened his position in this paragraph. He corrected the initial question to ask “whether
every time [a man] tells another that Such an Action is Honourable and vertuous” he is
trying to convey the ideas of honor and virtue into his hearer’s mind, and he added a
clause specifying that the man speaks “with an intention to excite [his hearer] to the per-
formance of” the action. All this suggests that Berkeley may have been tempted to make
the case of moral discourse more similar to the case of reward talk, by saying that we
sometimes had ideas of particular virtues (or perhaps particular virtuous acts) when we
talk about virtue, but the language also has a motivational use and an emotional content.
Nevertheless, Berkeley does seem, at least at some points in his career, to have held
some form of non-cognitivism about moral discourse. In his notebooks he writes, “We
have no Ideas of Vertues & Vices, no Ideas of Moral Actions . . . morality consisting in
the V olition chiefly” (N,x669). Since ‘volition’ (a synonym for ‘will’) is just another
name for spirit,
23
and there can be no idea of spirit (PHK,x27), the claim that “morality
consist[s] in the V olition chiefly” is probably meant to explain why we have no ideas of
the virtues and vices: each of them is a disposition of the will.
Berkeley’s discussion is, again, meant to explain how these particular words can be
meaningful without standing for ideas. The answer is that these words are uttered with
the intention of bringing about a certain disposition of will. This suggests the radical
claim that to accept or believe a moral proposition just is to have a certain disposition
of will. In fact, I will argue below (x7.1) that by the time he wrote Alciphron, Berkeley
held that to accept or believe any proposition is to have a certain disposition of will.
If Berkeley accepted this claim about moral discourse, then he would have a clear
picture of what goes on in the affirmation that a particular action is virtuous. When one
23. More precisely: the faculty of volition (will), which is here in view, just is the spirit. Individual
volitions just are actions. For details, see below,xx6.4-6.5.
115
sincerely asserts a proposition, one accepts that proposition and wants one’s hearer to
accept it. But, according to this view, accepting moral propositions has nothing to do
with ideas, and everything to do with the dispositions of the agent’s will. Now, Berkeley
does not say exactly what this disposition of will is, though his talk of ‘esteeming’ the
action and being ‘stirred up to the performance of it’ suggests that he does not hold the
implausible view that believing an action is virtuous is the same as being disposed to
perform it; rather, he holds that believing an action is virtuous involves taking a certain
practical and emotional attitude toward it.
Even in the earliest stratum, there is no support for Belfrage’s radical claim of exact
equivalence between ‘such an action is honourable’ and ‘perform and esteem such an ac-
tion!’ However, we may say instead that these two utterances are, for the early Berkeley,
related to one another in the same way in which the assertion ‘snow is white’ is related
to the imperative ‘believe that snow is white!’ The exact nature of this relationship is,
of course, a difficult question.
I have been emphasizing the importance of rule-following to Berkeley’s theory of
language. What the theory of operative language adds is that the rules we follow in
language will not always be simple associations of words with ideas. The rules may
require competent speakers, under certain circumstances, to feel an emotion, or perform
or refrain from some action, or adopt some general attitude or disposition of mind. In
order to determine what the relevant rules are, we must examine the aim or purpose of
an utterance. Words and phrases that accomplish their purpose should be regarded as
meaningful or significant, regardless of whether there is any such object as the ‘meaning’
or ‘signification’ of that word or phrase.
116
In this early period, Berkeley provides an analysis of how operative meaning comes
about which appears to make it possible that operative meaning could float free of refer-
ential meaning entirely. However, he remains ambivalent as to how far this has occurred
in our actual language, especially with respect to moral terms. All of the examples
which make it into the published introduction are examples in which ideas continue to
play a crucial role in the total meaning of the word, although there are important aspects
of meaning which would be missed by an entirely idea-based semantics.
3.3 Mathematical and Scientific Language
There are three domains of discourse discussed by Berkeley whose rules are quite differ-
ent from those involved in general terms or operative language. These are the domains of
arithmetic and algebra, geometry, and physics. Not all of these ‘discourses’ proceed, of
course, in natural language; much of the work is done in formal mathematics. However,
like natural language, formal mathematics is a system of conventional symbols governed
by complex rules, and Berkeley often uses mathematical analogies to illuminate his the-
ory of language (e.g. PHK, Introx19). At one point Berkeley even identifies ‘modern
algebra’ as “a more short, apposite, and artificial sort of language” (Alc,x7.17). As a
result, we may treat Berkeley’s account of mathematical symbols as part of his theory
of language.
I shall not attempt here to give a general interpretation of Berkeley’s philosophy
of math and science, but will instead focus on Berkeley’s analysis of how the symbols
function. Furthermore, we do not yet have a sufficient grasp of Berkeley’s philosophy
of language to tackle the difficult questions of truth, reference, and objectivity, so these
must be postponed. The aim here will be simply to explain what sorts of rules Berkeley
believes govern the use of signs in the three domains mentioned.
117
3.3.1 Arithmetic and Algebra
According to Berkeley the subject matter of arithmetic is a certain collection of signs
(PHK,x122). Equivalent sign systems may be constructed using English number words,
or tally marks, or Arabic numerals (x121). In his notebooks, Berkeley is quite explicit
in simply identifying numbers with their names or numerals (N,xx763, 767; Jesseph
1993, 109).
24
In the Principles matters are less clear, though Berkeley does assert that
“the number of any particular things is said to be known, when we know the name or
figures (with their due arrangement) that according to the standing analogy belong to
them” (PHK,x121). In other words, to know how many widgets there are just is to be
able to apply the number word or numeral which, according to the sign system we have
adopted, is appropriate to the collection of widgets. There is no pre-linguistic grasp of
the number two, whether as a Platonic entity or as an idea; there is only the word ‘two,’
and the numeral ‘2,’ and the rules which allow us to apply them to pairs of things. The
reason these rules are of interest to us is that by means of them the sign systems we
adopt “direct us how to act with relation to things and dispose rightly of them” (x122).
Number words and numerals, for Berkeley, begin as simple general terms: ‘two’ is
a name which indifferently signifies any pair of objects (Brook 1973, 148). Berkeley
imagines that numbers began from “single strokes, points or the like, each whereof was
made to signify an unit, that is, some one thing of whatever kind they had occasion
to reckon” (PHK,x121). The generality here is critical: the tally mark signifies just
any ‘unit’ – that is, any one thing. It is natural, then, that a pair of tally marks should
be used to signify any pair of things, and so forth. Other notations were developed
in order to make the representation of large numbers easier. The critical thing which
24. Although some earlier editions of Berkeley’s notebooks had the nonsensical reading “Numbers are
nothing but Names, never Words” for entry 763, more recent editions, corrected from the manuscript,
read “meer Words” instead. See Luce 1989, 307.
118
each notation must have in order to form the basis of a system of arithmetic is what we
should nowadays call a successor function: that is, upon seeing any numeral, we must be
able, by following some rule, to construct the numeral which should come next. This is
different from any linguistic rule we have encountered so far: it is a rule for constructing
new signs.
To know a numeral system is just to have the capacity to enumerate any arbitrary
‘collection of units’ (PHK,x120). Once such a system is in place we can go further and
construct arithmetic proper. That is, we can introduce rules for manipulating signs to
get from the number of one collection to the number of another, “and thus computing in
signs . . . we may be able rightly to sum up, divide, and proportion the things themselves
that we intend to number” (x121). The fact that we here ‘compute in signs’ is crucial:
in following the rules we manipulate the symbols directly without recourse to any other
ideas. Other ideas come in only at the beginning, when we set up the problem, and at the
end, when the result is applied – and of course there are cases where the application of
arithmetic is far less direct than this, and so the manipulation of symbols is even farther
removed from any ideas which can be said to be signified by the symbols.
Algebra adds two more kinds of signs which obey different rules. The first are the
ordinary variables. Berkeley observes that in algebra “though a particular quantity be
marked by each letter, yet to proceed right it is not requisite that in every step each
letter suggest to our thoughts that particular quantity it was appointed to stand for”
(Introx19). Berkeley says this in the course of making a point about how words work,
and he of course thinks that the thing he is pointing out is a quite common feature of
signs. However, this particular case makes Berkeley’s point especially easy to accept
because of an additional point regarding the use of variables. Within the context of
a single problem, a variable is much like a numeral, in that it indifferently signifies
119
any collection of units of a certain size.
25
However, the size is not known until after
the problem has been solved. That is, if one begins the problem by saying ‘let x be
the number of widgets in this box,’ ‘x’ stands not only for that collection of widgets,
but for any other collection equinumerous with the widgets. The aim of the problem
is, however, to discover how many widgets there are, which is to say, to discover that
numeral which is equivalent in signification to ‘x.’ This can be done by following the
rules of arithmetic. What this shows is that the actual, immediate ability to substitute a
word for what it signifies is not crucial to signification; it is possible for the rules to pick
out the signified ideas by some more complex means.
The second kind of sign is not mentioned in the Principles, though Berkeley men-
tions it in his notebooks (N,x764) and makes much of it in Alciphron. This is “the alge-
braic mark, which denotes the root of a negative square,” i.e. the imaginary constant ‘i’
(Alc,x7.17). One simply cannot have a collection ofi many objects, or an objecti units
long, and so the symbol ‘i’ does not signify anything. However, being significant does
not require signifying something. The symbol ‘i’ is significant because the definition
i
2
=1, together with the ordinary rules of arithmetic, allows us to make calculations
usingi which ultimately do lead to useful results that can be applied to the world (see
Pearce 2008, 162-163). In the Principles, Berkeley discusses only the arithmetic of nat-
ural numbers. His theory of geometry will allow him to give concrete application to the
rational numbers; however, since he takes space and time to be fundamentally discrete,
26
irrational numbers must occupy the same position asi: they ‘have their use in logistic
operations’ (see Alc,x7.17) and so are significant despite not standing for any objects.
25. In the notebooks, Berkeley says that “Algebraic Species or letters are denominations of Denomina-
tions” (N,x758). Presumably he means that the letter is a sign of the numeral. This is at odds with his
view in the Introduction to the Principles, where the letter stands not for a numeral but for ‘a particular
quantity.’
26. On space, see NTV, passim, along with the discussion of geometry, below; on time, see PHK,x98.
120
In the early (1707) Miscellanea Mathematica, Berkeley had compared algebra to
a game, like chess, and even constructed a game board for it, which he recommended
as a teaching tool (BW, 4:214-220; Jesseph 1993, 106-117). This provides a helpful
understanding of Berkeley’s thought on the workings of arithmetic and algebra: making
arithmetic and algebraic calculations is like moving pieces on a chess board according
to the rules of chess. The crucial limitation of this analogy, according to Berkeley, is
that the ‘games’ of arithmetic and algebra are meant to have a practical point, rather
than to be mere entertainment (PHK,x119). The rules which have been adopted are,
therefore, not purely arbitrary in the way the rules of chess are, but are subservient to
such practical purposes as determining the number of widgets in the box.
3.3.2 Geometry
Unlike arithmetic and algebra, geometry is, according to Berkeley, not merely a science
of signs or symbols. It is meant to give an account of space, and space is an empirical
concept. There is, however, a difficulty. The infinite divisibility of figures is a funda-
mental tenet of classical geometry, but Berkeley held, and argued at length in the New
Theory of Vision, that empirical space is discrete, so that any given perceived figure is
composed of a finite number of sensible minima (NTV,x54). Berkeley also provides a
very simple argument for this conclusion in the Principles:
Every particular finite extension . . . is an idea existing only in the mind, and
consequently each part thereof must be perceived. If therefore I cannot per-
ceive innumerable parts in any finite extension that I consider, it is certain
they are not contained in it (PHK,x124).
This view led Berkeley, early in his career, to a radical rejection of classical geometry
(Jesseph 1993, 45-69). At one point in the notebooks, Berkeley goes so far as baldly to
121
deny the Pythagorean Theorem (N,x500). Fortunately, Berkeley had rejected this absurd
view by the time he wrote the Principles, and he there attempted to use his theory of
signs to develop a theory which could accommodate classical geometry while denying
that there were any infinitely divisible idea-objects (Jesseph 1993, 69-78).
The difficulty here can be brought into focus by considering two particular facts
about Euclidean geometry: first, by a very simple Euclidean construction, any line seg-
ment can be bisected; second, the Pythagorean Theorem obtains. Taking the first case
first, we may say that if space is discrete and any line segment is composed of a finite
number of minima, then a line segment is successfully bisected if and only if there are
an equal number of the line segment’s minima on either side of the bisecting line. Now,
as Richard Brook points out (Brook 1973, 157), the bisecting line must be at least one
minimum in width. If it is exactly one minimum in width, then it must pass through
one of the minima in the bisected line segment. As a result, it can bisect the line seg-
ment only if the line segment contains an odd number of points. If the line segment
contained an even number of points, it could still be bisected (contrary to Brook) if the
bisecting line were two minima (or some other even number) wide. If the line segment
were only one or two points in length, then it could be bisected only by drawing a line
which passed through the entire segment, so that there were zero points on each side of
the bisecting line. Berkeley’s theory thus does not have the consequence that there are
line segments which cannot be bisected; rather, it has the consequence that bisecting a
line segment requires a certain sort of matching between the length of the line segment
and the width of one’s pencil. This is, needless to say, quite odd and in radical conflict
with classical geometry.
As for the second problem, as has been known from antiquity, the Pythagorean The-
orem entails the existence of incommensurable line segments. More precisely, it entails
122
that there are some triangles such that, no matter what units of length one chooses, one
cannot assign an integer length to every side. However, Berkeley holds that there is a
unit, namely, the sensory minimum, such that every sensed or imagined line segment is
an integer number of units long. This, again, is in radical conflict with classical geome-
try.
In the Principles, Berkeley aims to resolve these difficulties and reconcile his theory
of space with classical geometry. Berkeley’s solution is simple, but quite innovative.
According to Berkeley, the particular figures or diagrams considered in geometric cal-
culations are to be regarded as signs or symbols representing all particular figures of that
sort. So, Berkeley says, when a geometer is demonstrating the method for bisecting line
segments, “He draws, for instance, a black line of an inch in length” which “as it is there
used, . . . represents all particular lines whatsoever” (PHK, Introx12). In this way,
the geometer considers [lines] abstracting from their magnitude. Which
does not imply that he forms an abstract idea, but only that he cares not
what the particular magnitude is, whether great or small, but looks on that
as a thing indifferent to the demonstration (x126).
Although the particular line in question is finite in extension and, therefore, according to
Berkeley, has only finitely many parts, it is a sign of all line segments, of any magnitude
whatsoever. Berkeley says that,
Because there is no number of parts so great, but it is possible there may be
a line containing more, the inch-line is said to contain parts more than any
assignable number; which is true, not of the inch taken absolutely, but only
for the things signified by it (x127).
123
Berkeley’s idea seems to be this: just as in demonstrating that the sum of the interior
angles of a triangle is equal to two right angles we use some particular triangle, but with-
out making use, in our demonstration, of any of the features that distinguish it from other
triangles, so in geometry more generally, we consider line segments without considering
their magnitude. These two cases cannot, however, be precisely analogous, since if they
were our results would have to apply to all line segments, regardless of magnitude, when
in fact some of our results (for instance, those involving incommensurables) apply to no
line segments. Instead, Berkeley seems to think that any given geometric proof or con-
struction requires sufficiently long lines, but since the inch-line stands for all lines, and
lines can be arbitrarily long, there are some sufficiently long lines among those it stands
for. Thus our geometric results all contain an implicit proviso, ‘for sufficiently large
figures . . . ’ However, for some geometric results, such as those involving incommen-
surables, no figure is in fact sufficiently large. This is because incommensurable lines
have no finite ratio to one another, but Berkeley’s lines are all made up of finitely many
points. Still, the larger the (finite) figure the more nearly it is capable of approximating
the theorem (Jesseph 1993, 74).
Here geometry functions similarly to algebra. We perform certain operations (e.g.
the operations involved in Euclidean constructions) which we can only perform by con-
sidering the signs while ignoring the particular ideas they signify, and this allows us
to come to general conclusions which we can apply in practice to any particular fig-
ures we wish. As applied to particular figures, many of our results turn out to be mere
approximations. However, as Berkeley says, “whatever is useful in geometry and pro-
motes the benefit of human life, does still remain unshaken” (PHK,x131). Berkeley’s
contemporaries never thought that geometry could, in practice, be applied as more than
124
an approximation in the first place: most of them held, with Descartes, that no phys-
ical object is ever more than an approximation to the simple geometrical natures we
know (CSM, 2:262). Thus geometry, in its practical uses, was already acknowledged
to involve an idealization from the physical facts; what Berkeley adds is that there are
no objects of any sort – physical, psychological, or logical – of which the theorems
of geometry hold precisely. Instead of being a precise description of abstract objects,
Berkeley takes geometry to be an approximate description of the objects of sensation.
27
3.3.3 Physics
In the Principles, Berkeley gives a simple account of the aim of natural philosophy.
The natural philosophers, he says, “frame general rules from the phenomena, and after-
wards derive the phenomena from those rules” (PHK,x108). The ‘explanation’ which
takes place in natural philosophy does not involve assigning efficient causes, but simply
subsumes particular phenomena under general laws (PHK,xx104-106; Stoneham 2002,
108). The aim of this activity is to make practically useful predictions (PHK,x107). A
question, however, arises regarding the status and meaning of various assertions made
by natural philosophers. In the Principles, Berkeley criticizes Newton’s notions of abso-
lute space and time as meaningless because empirically empty, but uncritically accepts
the notion of force (PHK,xx110-117; Downing 2005, 236). In De Motu (1721), Berke-
ley sets out to rectify this oversight by carefully examining the use of theoretical terms
and ‘mathematical hypotheses’ in physics (Peterschmitt 2003, 184-185n2, et passim).
In doing so, he makes significant advances toward his mature theory of language.
27. Cf. Alc,x4.23: “Does any one find fault with the exactness of geometrical rules, because no one
in practice can attain to it? The perfection of a rule is useful, even though it is not reached. Many may
approach what all may fall short of.”
125
Berkeley’s concern in De Motu was to defend Newton’s mechanics from the charge
of unsavory metaphysical commitments which had been advanced by Leibniz and his
followers (see, e.g., Leibniz and Clarke [1717] 1969,xx9.118-123).
28
At the same
time, although Berkeley suppresses his immaterialism in De Motu (Downing 1995a,
198; 2005, 237), he wishes to defend his metaphysics and theology from any encroach-
ment on the part of the physicists (DM,xx42, 71-72; Peterschmitt 2003, 185-186, 197).
Berkeley shares the positivist aim of separating science from metaphysics for the pro-
tection of science, but, unlike the positivists, he regards this separation as likewise pro-
tecting metaphysics and theology (Popper [1953–1954] 1970, 140-141).
29
The linguistic element of De Motu is Berkeley’s argument for the claim that “‘Force’,
‘gravity’, ‘attraction’ and similar terms are useful for reasoning, and for calculations
about motion and moving bodies, but not for understanding the simple nature of motion
itself or for designating so many distinct qualities” (DM,x17). What concerns us here
is Berkeley’s account of the function of these terms, that is, the question of how they
manage to be useful without designating anything. In order to answer this question,
28. Earlier treatments of De Motu saw it primarily as an attack on Newton. See, e.g., Popper (1953–
1954) 1970, 130. However, it is now widely recognized that Berkeley’s main targets are Leibniz and his
followers. See, e.g., Jesseph 1992b,x2.3; Downing 1995a, 199; 2005, 238 and Clarke’s footnotes in DM,
xx2, 9, 15-20, 44, as well as Jesseph’s footnotes toxx2, 8-9, 14-20, 43-45 in Jesseph 1992. Berkeley may
also have theological criticisms of realism about physical forces in mind. See, e.g., Malebranche (1674–
1675) 1997,x6.2.3. Of course, Berkeley’s project will require a ‘rational reconstruction’ of Newtonian
mechanics, and Newton would likely have rejected many of Berkeley’s suggestions, so Berkeley should
not be seen as defending Newton or Newtonianism generally; what he aims to do is to show that Newto-
nian mechanics can be defended without defending Newtonian metaphysics (Peterschmitt 2003, 184-187,
196-197). Luc Peterschmitt aptly characterizes Berkeley as an ‘ultra-Newtonian’ who seeks “to defend
Newtonianism against its own metaphysical ghosts” (197). Cf. also Moked 1988, 62: “[Berkeley] is very
keen . . . to draw a distinction between the main scientific contributions of Newton . . . and those utterances
and tenets of the Newtonians, and of the great man himself, which may be regarded as metaphysical
assumptions.”
29. For a detailed account of Berkeley’s demarcation of metaphysics from natural philosophy and its
historical significance, see Schliesser 2005.
126
Berkeley provides an account of the practice of physics which fills in the details of the
simple account from the Principles. Berkeley writes that
in mechanical philosophy, the principles are said to be those things in which
are grounded and contained the whole discipline, those primary laws of mo-
tion which are confirmed by experience and are refined by reason and ren-
dered universal. Those laws of motion are appropriately called principles,
because both general mechanical theorems and particular explanations are
derived from them (DM,x36).
Here, as in the Principles, a ‘particular explanation’ is just the subsumption of some par-
ticular observation under a general rule (xx37, 69). Berkeley gives the following basic
account of scientific methodology. From experience, one notices certain generaliza-
tions. These are confirmed by further observation, then ‘refined by reason and rendered
universal,’ i.e. they are given some particular precise formulation which is proposed as
a fundamental law. It is possible either to apply the fundamental laws directly to the
explanation of particular phenomena, or else to derive ‘general mechanical theorems,’
i.e. useful generalizations of narrower scope than the fundamental law, from them, and
explain the phenomena using these. However, the laws of physics are not simple gen-
eralizations like ‘all ravens are black.’
30
Their structure is much more complex, and
often involves technical notions like ‘force.’ According to Berkeley, as in geometry so
“in mechanics, notions are initially established – that is, definitions, and first general
statements about motion – from which more remote and less general conclusions are
subsequently deduced by a mathematical method” (x38). The aim, then, is to begin with
a set of definitions and axioms to which a ‘mathematical method’ may be applied, in
30. Maudlin 2007, ch. 1, esp. pp. 10-14 makes a strong case for the claim that the assumption that laws
have this form is not a harmless idealization and has seriously misled philosophers in their thinking about
the laws of nature.
127
order to determine “the motions of any parts of the system of the world, and the phe-
nomena that depend on them” (DM,x38).
This theory is quite similar to the account of laws adopted by David Lewis, accord-
ing to which “a contingent generalization is a law of nature if and only if it appears as a
theorem (or axiom) in each of the true deductive systems that achieves a best combina-
tion of simplicity and strength” (D. Lewis 1973, 73). However, Berkeley does not say
that the ‘principles,’ as he calls them, of mechanics must be those of each system which
‘achieves a best combination of simplicity and strength;’ it is enough that they are part
of a system which we adopt and successfully use in predictions.
31
What is crucial for our purposes is that terms such as ‘force’ are introduced in formal
definitions for use, along with axioms, in a deductive system whose aim is to predict
phenomena. By such a deductive system, “the system of the world would be subjected
to human calculations” (DM,x66). It is precisely by their role in this system that terms
like ‘force’ come to have meaning (Peterschmitt 2003, 191). As Lisa Downing says,
these terms “function purely formally in the theory, like formal variables” (Downing
2005, 249). The terms in question are, then, quite similar to the constant ‘i’ which
we have previously discussed: they cannot be substituted for any particular ideas, but
instead are manipulated by mathematical rules and take their meaning from the role they
play in these calculations.
The word ‘force’ is, however, importantly different from the constant ‘i,’ and it is
here that De Motu constitutes a significant advance toward Berkeley’s mature theory of
language. ‘Force’ does not occur only in calculations; it can be used as an ordinary
31. There is, as Peterschmitt observes, a conventionalist strain to Berkeley’s philosophy of science
(Peterschmitt 2008, 31): it must be a system which we adopt. Peterschmitt suggests that Berkeley’s views
can, in this respect, be usefully compared with Poincar´ e’s. As James Van Cleve pointed out to me, there
are also similarities to Quine (Quine 1960, 19-25, 161). On the objectivity of natural laws, and of the
other elements of the perceived world, see chapter 8.
128
English noun, and is treated as if it were the name of a quality possessed by bodies. Yet
it is radically unlike ‘redness,’ in that one can point to particular instances of redness but
cannot, in the same way, point to particular instances of force.
32
The meaning of ‘force’
is fixed not by ostension, but by the definitions and axioms of Newtonian mechanics,
yet it is able to work its way into ordinary English and be applied to the world. Berkeley
cautions that we should not lose sight of the crucial difference between ‘force’ and ‘red-
ness.’ These two words are governed by different rules, and the difference is analogous
to the difference between ‘i’ and ‘2.’
3.4 Conclusion
In the Introduction to the Principles, Berkeley attacks the Theory of Meanings, an ap-
proach to mental representation, linguistic representation, and the relationship between
them which was endorsed by thinkers across the entire spectrum of seventeenth century
philosophy. According to the Theory of Meanings, there are certain entities, ‘meanings,’
which, by their intrinsic nature, represent, and the role of linguistic conventions is just to
link words to meanings. Berkeley’s primary argument against this view is his claim that
nothing could possibly be intrinsically well-suited to represent generally. As we have
seen, Berkeley’s argument can be escaped only by taking representation as primitive.
Considerations of parsimony therefore recommend searching for an alternative. In this
early period, Berkeley begins to develop such an alternative. His fundamental claim is
that understanding words is not a matter of linking them to meanings, but of follow-
ing the rules governing their use in language. To this extent, at least, Anthony Flew
is correct in regarding Berkeley as a precursor of the later Wittgenstein (Flew [1974]
32. The significance of this difference will be examined at length in chapter 6.
129
1993). In fact, anticipations of Wittgenstein can be found not only, as Flew observes, in
Alciphron VII, but throughout Berkeley’s career, beginning as early as the Manuscript
Introduction. It is, however, only in Alciphron 7 that Berkeley pulls the threads together
into something like a general theory of language. It is to this treatment we now turn.
130
Chapter 4
A Commentary onAlciphron VII
Having traced the development of Berkeley’s theory of language prior to 1732, we turn
now to Berkeley’s mature theory as contained, primarily, in Alciphron VII. I begin, in
this chapter, with a detailed commentary on that text before turning, in the following
three chapters, to an examination of the philosophical issues arising from Berkeley’s
theory. Beginning with this kind of textual focus will help to protect against anachro-
nistic readings of Berkeley’s concerns, arguments, and ideas. It will also, perhaps more
importantly, give me the opportunity to defend, from a textual and contextual perspec-
tive, my contention that there is such a thing as Berkeley’s mature theory of language
and that it is contained in Alciphron VII. This contention has come under attack from two
directions. First, Ian Hacking has argued at length that none of the British empiricists
had a ‘theory of meaning’ in the post-Fregean sense. According to Hacking, “theories
of meaning have to do with the essentially public features of language, whatever it is
that is common to you and me, in respect of the word ‘violet’, which makes it possible
for us to talk about the flowers in Knapwell wood” (Hacking 1974, 50). Although early
modern philosophers made casual remarks about the ‘common acceptation’ of words,
Hacking claims that they were always more concerned in their talk about meaning, with
the logical and epistemological facts about our private ideas than about these ‘common
acceptations.’
Second, some interpreters have seen Alcipron VII as engaged in a narrow apolo-
getic exercise, to show that the philosophy of language assumed in a certain objection
131
to Christianity is incorrect, without providing any comprehensive theory of language in
its place (Bennett 1971, 53-55; Berman 1994, ch. 6; Jakapi 2002b, 30-32). Interpreters
who have this narrow view of the significance of Alciphron VII differ greatly among
themselves as to what exactly Berkeley’s apologetic strategy is. The most influential
‘narrow’ interpretation is due to David Berman. Berman argues that Berkeley should be
seen as anticipating 20th century emotivism about moral and religious discourse. 20th
century emotivism was a theory about a particular narrow domain of discourse; it was
not a general philosophy of language. Similarly, according to Berman, one can find
in Alciphron a “linguistic distinction which [Berkeley] was the first to draw: between
cognitive statements which inform . . . and emotive utterances which produce emotions,
dispositions, and actions” (Berman 1994, 148). According to this interpretation, Berke-
ley gives only a fragment of a philosophy of language in Alciphron VII, dealing only
with the special case of (emotive) religious mysteries.
Against both of these claims, I will argue that, although Berkeley’s 18th century con-
text and his religious concerns make for certain differences as compared to later philoso-
phies of language, it is indeed one of Berkeley’s aims in this dialogue to “make sense of
our daily practice” of spoken and written language in general (Alc,x7.8). These, Berke-
ley recognizes, are public, social practices governed by conventional rules of speech,
thought, and action. Alciphron VII, I will argue, contains a ‘theory of meaning’ in what
Hacking identifies as the post-Fregean sense of that phrase. The theory of meaning in
question anticipates, in certain respects, the later Wittgenstein.
1
Since Berkeley’s aim
is to “make sense of our daily practice,” he focuses on how bits of language are used,
what the purposes of these uses are, and how they interact with other aspects of human
social behavior. Furthermore, although we can distinguish different types of linguistic
1. In regarding Alciphron VII as providing a general theory of language similar to Wittgenstein’s I
follow Flew (1974) 1993.
132
rules, Berkeley recognizes no sharp distinction between different domains of discourse
(cf. Olscamp 1970b, 148; also seex3.2, above).
I divide the relevant portion of Alciphron VII into six parts. In the first part (xx1-
4), Alciphron provides an argument against Christianity based on Locke’s theory of
language. In the second part (xx5-7), Euphranor summarizes Berkeley’s reasons for
rejecting Locke’s theory. Next, Euphranor presents his alternative theory of language
(xx8-11), and applies this theory of language to analyze two distinct, but similar, vari-
eties of assent to propositions, which he calls ‘faith’ (xx11-13) and ‘science’ (xx14-16).
In the sixth and final portion of the text (xx16-18), Euphranor summarizes the preceding
discussion and draws a number of general conclusions from it.
4.1 Alciphron’s Critique of Religious Mysteries (xx1-4)
Most of Alciphron is taken up by a debate between two Christian protagonists, Euphra-
nor and Crito, and their ‘freethinking’ opponents, Alciphron and Lysicles. The seventh
dialogue begins with the title character’s admission that the preceding dialogues have
provided “several probable reasons . . . for embracing the Christian faith.” However, Al-
ciphron insists, “those reasons, being only probable, can never prevail against absolute
certainty and demonstration.” More specifically, the arguments adduced for Christian-
ity rest on history and tradition and so, at bottom, on testimony, which is a ground of
probable belief, but never of certainty. As a result, if there were a demonstration of the
falsity of Christianity – if Christianity could be shown to contain not merely “things odd
and unaccountable” but “metaphysical absurdities and absolute impossibilities” – then
the probable arguments would be rendered irrelevant. “To be plain,” Alciphron says,
“no testimony can make nonsense sense: no moral evidence can make contradictions
133
consistent” (Alc,x7.1). Alciphron draws this account of the epistemological situation
from Locke (EHU,xx4.18.5-10), and Euphranor accepts it on Berkeley’s behalf.
2
After securing Euphranor’s promise that if Christianity can be shown to contain
nonsense or contradictions he will “have nothing more to say” in its defense, Alci-
phron indicates the particular charge he will make against Christianity. His accusation
is that among the essential teachings of Christianity are to be found certain doctrines,
the so-called ‘mysteries,’ which are really no more than “empty notions, or, to speak
more properly, . . . mere forms of speech, which mean nothing, and are of no use to
mankind” (Alc,x7.1). This accusation, and the argument which follows, is drawn from
John Toland’s notorious Christianity Not Mysterious (Toland 1696).
3
Toland’s book
begins with a lengthy account (without attribution) of Locke’s epistemology and philos-
ophy of language, then proceeds to argue that, given these theories, no meaning can be
assigned to the so-called ‘mysteries’ of the Christian faith. Toland defines a ‘mystery’
as “a thing of its own Nature inconceivable” (66). Because the thing is inconceivable,
the words with which we confess our belief in the mystery are not associated with any
ideas (135). But, according to the Theory of Meanings, if we have no ideas, the words
are meaningless and we might as well be confessing “that something call’d Blictri [has]
a Being in Nature” (138).
4
2. On Locke’s religious epistemology, and Berkeley’s endorsement of it, see Pearce, forthcoming(a).
3. On the importance of this work to Berkeley’s intellectual context, and its direct and indirect influ-
ence on Berkeley’s thought, see Belfrage 1985; Berman 1994, 11-17, 148-150; Pearce, forthcoming(a);
forthcoming(b),x8.
4. As Roomet Jakapi points out (Jakapi 2007, 189-192; cf. Jakapi 2002a, 404-406), discussions of
mysteries in this period, including those found in Toland and Berkeley, often slide back and forth between
several senses of ‘mystery,’ including: (a) certain (alleged) supernatural realities (e.g. the afterlife);
(b) certain propositions, beliefs, or teachings about those realities (e.g. the proposition, belief, or teaching
that there will be a bodily resurrection of the dead); and (c) certain traditional formulas in which those
beliefs and teachings are expressed (e.g. the statement, in the Apostles’ Creed, “I believe in . . . the res-
urrection of the body; and the life everlasting”). This can be observed in Toland’s switch from treating
134
Toland’s charge was that it was impossible genuinely to believe doctrines such as
Trinitarianism because there were no ideas behind the words which could be the ele-
ments of a ‘mental proposition’ (see EHU,x4.5.2). In Alciphron’s mouth, this is trans-
formed into the charge that Christianity prescribes certain “forms of speech, which mean
nothing, and are of no use to mankind” (x7.1). These are, in fact, two distinct charges,
although those who accept the Theory of Meanings will be likely to confound them.
First, it is charged that the words in question mean nothing. If my account of the The-
ory of Meanings and Berkeley’s opposition to it is accepted, then this phrase should
be taken quite strictly and literally as the accusation that there is no thing which those
words mean, or, equivalently, that there is no thing which is the meaning of those words.
Within the Lockean framework, this is essentially identical to Toland’s explicit charge,
that the words do not stand for any ideas. The second accusation is that the words in
question “are of no use to mankind.” For proponents of the Theory of Meanings, these
two charges can hardly be separated: the words in question have no meanings, hence
they are not meaningful, hence they are not useful. This is the charge which Alciphron
will press, and Euphranor will answer, in this dialogue. Euphranor’s strategy will be, es-
sentially, the reverse of Alciphron’s: he will argue that the words in question are indeed
useful, and if useful, then meaningful. However, Euphranor concedes that the words
have no meanings. Hence the conclusion which must be drawn is that being meaningful
need not involve having a meaning, that being significant need not involve signifying
something.
Alciphron begins pressing his charge by laying out the Theory of Meanings (Alc,
x7.2). Berkeley’s characterization of the Theory of Meanings should by now be familiar
and so need not detain us. Alciphron then claims that, despite the wide acceptance of
mysteries as inconceivable things to regarding them as meaningless words, a switch which is also reflected
in Alciphron’s presentation of the argument.
135
the Theory of Meanings as a theory, people often go on using words to which they can
connect no ideas, and this is a source of confusion and strife.
After this, Alciphron recites Locke’s account of knowledge as “the perception of the
connexion or disagreement between ideas,” and of opinion or faith as ‘weaker’ forms of
assent, which likewise involve joining or separating ideas (Alc,x7.3; see EHU,xx4.1.2,
4.14.4). Since all forms of assent involve the joining or separating of ideas, it is impos-
sible to assent to that of which we have no ideas. As a result, these meaningless words
cannot be expressions of genuine belief.
Having laid out Locke’s theory, Alciphron remarks, “It were needless to illustrate
what shines by its own light . . . My endeavour shall be only to apply it in the present
case” (Alc,x7.3). Locke’s theory, Alciphron supposes, is evidently true. If Christianity
is to be defended, Christian doctrine must demonstrate its Lockean credentials: the ideas
corresponding to each word must be exhibited. This, Alciphron is confident, cannot be
done.
At this point, Alciphron goes for the throat:
Grace is the main point of the Christian dispensation . . . Only one thing I
should desire to be informed of, to wit, What is the clear and distinct idea
marked by the word Grace? I presume a man may know the bare meaning of
a term, without going into depth of all those learned inquiries. This surely
is an easy matter, provided there is an idea annexed to such a term. And if
there is not, it can be neither the subject of a rational dispute, nor the object
of real faith . . . For there can be no assent where there are no ideas: and
where there is no assent there can be no faith: and what cannot be, that no
man is obliged to. This is as clear as any thing in Euclid (x7.4).
136
It is noteworthy that Berkeley’s presentation makes the radical nature of Toland’s argu-
ment clearer than Toland himself had:
5
if Toland is correct, it follows that no one has
ever been a Christian, for a Christian is, inter alia, one who believes in divine grace,
6
but no one has ever had any idea of divine grace, and no one can believe in what she has
no idea of.
Alciphron makes a point of going after a central element of the Christian faith, a
belief confessed by the ordinary church-goer, rather than “any nice disputed points of
school divinity” (Alc,x7.3); but why does Alciphron believe that ‘grace’ is meaningless?
Against the meaningfulness of ‘grace,’ Alciphron makes two points. First, he notes that
no point has created more controversy in the church than this doctrine of
Grace. What disputes about its nature, extent, and effects, about univer-
sal, efficacious, sufficient, preventing, irresistible grace, have employed the
pens of Protestant as well as Popish Divines, of Jansenists and Molinists,
5. The reason Toland does not put his point in this way is that he wants to argue, not that Christian
faith is false, unjustified, or impossible, but rather that ‘mysteries,’ such as the doctrine of the Trinity,
form no part of authentic Christianity. Berkeley has at least two reasons for framing Toland’s argument
differently. First, he refuses to countenance the possibility of a non-Trinitarian Christianity. Second, he
evidently thinks that, if Toland’s argument worked at all, it would work equally well against the concept
of grace, which is arguably even more central to Christianity (or at least orthodox Protestantism) than
Trinitarianism.
6. Samuel Rickless suggested to me that Locke, although he holds that Christians ought to believe in
divine grace, would object to making this part of the definition of ‘Christian.’ To this I reply, first, that
Locke describes Christianity as ‘the covenant of grace’ (LW, 7:103-104, 124, 155), as contrasted with
‘the Law’ or ‘covenant of works;’ hence he would not disagree with Alciphron’s claim that “Grace is the
main point of the Christian dispensation.” Nevertheless, it is true that Locke’s central contention in The
Reasonableness of Christianity is that there is only one article of faith necessary for salvation, namely,
that Jesus is the Messiah. It might therefore be thought that one may be a party to the ‘covenant of grace’
without actually believing in grace, and hence that Locke would deny that Alciphron’s argument leads to
the conclusion that no one has ever been a Christian. However, it is not clear that even this much is true,
for Locke glosses ‘Messiah’ as “the Saviour that was promised” (7:23; cf. 24-25), and this would seem
to build the concept of grace (i.e. God’s action to save the undeserving) into the very concept of Messiah,
which is crucially involved in Locke’s one article of Christian faith. Finally, it is to be noted that Locke’s
view here was widely seen as heterodox in the extreme (see, e.g., Edwards 1696; Milner 1700, ch. 14). I
think it is safe to say that Berkeley, as a member of the Anglican clergy, would find the inclusion of belief
in grace in the definition of ‘Christian’ unobjectionable.
137
of Lutherans, Calvinists, and Arminians, as I have not the least curiosity to
know, so I need not say (Alc,x7.4).
We may, for present purposes, safely follow Alciphron in refraining from inquiry into
the details of these disputes. It is, however, quite true that the proper understanding
of grace has been among the most contentious issues in Western Christianity from the
Reformation to the present. If the disputants did not share a common idea of grace,
and so were engaged in “mere verbal trifling,” that would explain the intractability of
the disputes (Alc,x7.4; cf. EHU,x3.10.12; PHK, Introxx21-22). On the basis of
introspection, Alciphron believes that this is indeed the case: “whenever I laid aside the
word Grace, and looked into my own mind, [I found] a perfect vacuity or privation of
all ideas” (Alc,x7.4).
Having dispensed with the alleged idea of grace, Alciphron goes on to question why
people thought there was such an idea in the first place.
7
For this he gives two reasons.
First, “a word often heard and pronounced is [commonly] believed intelligible, for no
other reason but because it is familiar” (x7.4). Second, Alciphron thinks that many of
our ways of speaking about grace are drawn from perfectly intelligible ways of speaking
about physical objects. Grace, in the sense under discussion, is meant to be “an active,
vital, ruling principle, influencing and operating on the mind of man, distinct from every
natural power or motive.” Alciphron observes that “Men speak of this holy principle as
of something that acts, moves, and determines, taking their ideas from corporeal things,
from motion and the force of momentum of bodies” (x7.4). Grace of course cannot
literally be a thing like that, but the similarity of the vocabulary creates, according to
Alciphron, an illusion of intelligibility.
7. Interestingly, this parallels Berkeley’s attack on abstract ideas in the Introduction to the Principles
(see above,x2.1).
138
Alciphron concludes that ‘grace’ is meaningless, and hence the so-called ‘doctrines
of grace,’ about which there is so much dispute, are not beliefs or doctrines at all, but
rather empty forms of words, which are both meaningless and useless. Furthermore, he
claims, “The same method of reasoning may be applied by any man of sense to confute
all other the most essential articles of the Christian faith” (Alc,x7.5). This completes
Alciphron’s case for the claim that, probable arguments notwithstanding, Christianity
must be rejected.
4.2 Euphranor’s Refutation of the Theory of Meanings
(xx5-7)
At this point Euphranor takes the reins and begins to question Alciphron about the The-
ory of Meanings and its consequences. Euphranor observes, “every time the word man
occurs in reading or conversation, I am not conscious that the particular distinct idea of
a man is excited in my mind” (x7.5).
8
This prompts Alciphron to summarize Locke’s
theory of abstract ideas, and claim that Euphranor does indeed have the abstract general
idea of a man in his mind each time he hears the word ‘man.’
In response to this, Euphranor summarizes the arguments against abstract ideas from
the Introduction to the Principles (Alc,x7.6). Alciphron is certain that there must be
some unnoticed error in Euphranor’s reasoning, since without abstract ideas general
names would be meaningless, with the result that there would be no general proposi-
tions and hence no general knowledge. This is, of course, precisely the concern that
Berkeley thinks led to the positing of abstract ideas in the first place (see above,x2.1).
8. Sections 5-7 were excised from the 1752 edition, and the subsequent sections re-numbered. All
references here are to the 1732 edition unless otherwise noted.
139
Euphranor responds by introducing Berkeley’s account of general names and ideas (see
above,x3.1). “[W]ords become general,” Euphranor says, “by being made to stand in-
discriminately for all particular ideas, which from a mutual resemblance belong to the
same kind . . . a particular idea may become general by being used to stand for or repre-
sent other ideas” (Alc,x7.7).
Alciphron, still wedded to the Theory of Meanings, takes this to mean that each
time a general name is pronounced, I have in mind some one particular idea. (For
instance, that every time I hear the word ‘man’ I must think of either Peter or Paul or . . . )
Euphranor, however, insists that this is phenomenologically implausible. Alciphron, on
reflection, agrees and admits that he was led to think this by the Theory of Meanings
and not by introspection.
This passage is noteworthy in that it asserts the independence of linguistic and men-
tal representation. Not only do general terms not stand for abstract general ideas; they
do not stand for Berkeleian general ideas either. Words and ideas can become general by
the very same process of adopting conventional rules, and there is no logical or temporal
priority between these two processes.
4.3 Euphranor’s Theory of Language (xx8-11)
At the end ofx7, Alciphron questions what other use words can have besides suggesting
ideas. At the beginning ofx8, Euphranor proposes to “inquire what [the use of words]
is? and see if we can make sense of our daily practice.” As I indicated at the beginning of
the chapter, I take this line seriously, and literally: this section and those which follow
aim to provide a theory which will make sense of the practice of spoken and written
language.
140
Euphranor begins by noting that words are a species of a broader genus, namely,
signs. Some light may therefore be shed on the nature of words by consideration of
some examples of simpler sign systems (cf. Wittgenstein [1958a] 2009, 105-106). The
first example Euphranor considers is the use of poker chips. These “are used,” he says,
“not for their own sake, but only as signs substituted for money, as words are for ideas.”
However, part of the point of having chips is that one need not “every time these counters
are used throughout the whole progress of a game . . . frame an idea of the distinct sum
or value that each represents.” Alciphron agrees that it is sufficient, for the counters to
serve their purpose, that “the players at first agree on their respective values, and at last
substitute those values in their stead” (Alc,x7.8).
The second example is an arithmetic calculation in which “the figures stand for
pounds, shillings, and pence.” Here again, Euphranor says, it is not necessary “through-
out the whole progress of the operation, in each step to form ideas of pounds, shillings,
and pence.” Alciphron agrees that “it will suffice if in the conclusion those figures direct
our actions with respect to things” (x7.8). This is, clearly, the sort of ‘computing in
signs’ Berkeley spoke about in his discussion of the use of mathematical symbols in the
Principles (PHK,x121; see above,x3.3.1).
From the poker chip case, Euphranor draws the conclusion that “words may not be
insignificant, although they should not, every time they are used, excite the ideas they
signify in our minds, it being sufficient, that we have it in our power to substitute things
or ideas for their signs when there is occasion” (Alc,x7.8). It should be noted, first,
that Euphranor here gives a sufficient condition for the meaningfulness of a word: if
there is something that can be ‘substituted’ for a word, then that word is meaningful. As
will emerge, Berkeley does not take this to be a necessary condition of meaningfulness.
Second, the thing substituted is described as a ‘thing or idea.’ Berkeley’s immaterialism
141
is not in play in Alciphron, so Euphranor should not be read as collapsing the distinction
between things and ideas as Philonous does (DHP, 244). Rather, Euphranor makes this
disjunctive claim in order to be neutral on whether the reference of words to things is
mediated by ideas as ideist versions of the Theory of Meanings suppose.
From the arithmetic case, Euphranor draws the conclusion “that there may be an-
other use of words, besides that of marking or suggesting distinct ideas, to wit, the
influencing our conduct and actions.” This does seem to be what Alciphron had con-
ceded in response to the case: he said that it would ‘suffice’ if the conclusion of the
calculation was action-guiding. Alciphron does not say what this would suffice for, but
Euphranor interprets Alciphron to mean that it would suffice for the usefulness of that
application of signs. Euphranor goes on to say that the influencing of our actions
may be done either by forming rules for us to act by, or by raising certain
passions, dispositions, and emotions in our minds. A discourse, therefore,
that directs how to act or excites to the doing or forbearance of an action
may, it seems, be useful and significant, although the words whereof it is
composed should not bring each a distinct idea into our minds (Alc,x7.8).
Since influencing actions is a legitimate use of signs in general, and words in partic-
ular, signs that accomplish this end are useful, regardless of the means by which they
accomplish it. There is a variety of means by which words do this. For instance, a
sentence could provide an explicit instruction, which would be adopted and followed
by the hearer, or it could put the hearer into some kind of affective state which would
tend to lead to the action. As long as this end is accomplished, Euphranor claims, that
application of words is useful and significant. It is evidently a background assumption
that a word or other sign is meaningful (or significant) whenever it accomplishes the
142
purpose for which it is intended (cf. DHP, 223).
9
Alciphron does not question any of
these claims. Thus Euphranor has at this point undermined Alciphron’s inferences: from
the fact that a word is not associated with any idea, it does not follow that the word is
meaningless or useless: meaningfulness need not involve having a meaning.
Next Euphranor argues that since ideas are “altogether inactive” there cannot possi-
bly be an idea of “an active mind or spirit.” The conclusion is “that those words, which
denote an active principle, soul, or spirit, do not, in a strict and proper sense, stand for
ideas” (Alc,x7.8) Of course, however, these words are meaningful.
10
Having presented three lines of argument against the Theory of Meanings, Euphra-
nor repeats the claim that the Theory of Meanings is the source of the doctrine of ab-
straction. Alciphron asks whether Euphranor rejects abstraction entirely, and Euphra-
nor makes the qualification familiar from the Introduction to the Principles (PHK, In-
trox10): he can separate in thought “those things that can really exist, or be really
perceived asunder,” but cannot “frame abstract general ideas” (Alc,x7.8). Alciphron
persists in endorsing the Theory of Meanings, at which point Euphranor urges him to
attempt to “frame an idea of number in abstract exclusive of all signs, words, and things
numbered.” Alciphron concedes that he cannot.
Euphranor now observes that “though, it seems, neither you nor I can form distinct
simple ideas of number, we can nevertheless make a very proper and significant use of
numeral names” (x7.8). The use of the word ‘proper’ here provides further evidence
against Hacking’s thesis that none of the British empiricists had a theory of meaning:
9. A question arises here as to who intended the word or sign for that purpose. For reasons discussed
inx3.2, above, I believe Berkeley means, or at least ought to mean, to refer to the collective intention of
the community by whose conventions the sign is governed. For further discussion, seex5.3, below.
10. In a footnote, Berkeley refers the reader to PHK, Introx20 and PHK,x135 for details.
143
‘propriety of speech’ or ‘propriety of language’ is connected by Berkeley and other writ-
ers of the period with the ‘common’ (or ‘vulgar’) ‘acceptation’ (or ‘use’) of language,
phrases which, as Hacking explicitly acknowledges, clearly have to do with the pub-
lic rules of linguistic usage (DHP, 216, 239-240, 247; EHU,xx3.9.8, 3.11.11, 3.11.24;
Hacking 1974, 47-49).
11
The question here is whether ‘propriety of language,’ i.e. the
public conventions governing the use of words, requires that one have a certain idea
each time one hears or says a particular word. The answer is ‘no.’
Alciphron had earlier claimed that ‘force,’ as contrasted with ‘grace,’ was a perfectly
meaningful word, corresponding to a clear and distinct idea (Alc,x7.4).
12
Euphranor
now suggests that we revisit this assertion and “examine what idea we can frame of
force abstracted from body, motion, and outward sensible effect” (x7.9). He doubts that
such an idea will be found. This corresponds to Alciphron’s second complaint against
‘grace,’ that no corresponding idea is to be found in introspection. Alciphron concedes
that the case of ‘force’ is the same as that of ‘grace.’
Next, symmetrically, we come to Alciphron’s first complaint, his claim that the ap-
parent intractability of the disputes about grace is evidence that ‘grace’ does not signify
any common idea. Euphranor lists a number of English and Latin terms closely con-
nected to force, and remarks that “no small controversies have arisen about the notions
or definitions of these terms” (x7.9). These are the same controversies which Berkeley,
11. Berkeley also uses the phrase ‘propriety of language’ at PHK, Introx20 and PHK,x115. Locke’s
uses of the term ‘propriety’ with this meaning are far too frequent for a comprehensive enumeration.
To list just a few examples, see EHU,xx1.1.5, 2.9.1, 2.21.16, 2.30.4, 2.32.1, 2.32.17, 4.4.9-10. In fact,
Locke’s Essay does not contain a single usage of the word ‘propriety’ with any other meaning than this.
12. Alciphron characterizes Locke’s theory as requiring a clear and distinct idea for a word to be mean-
ingful. This characterization of Locke’s view is drawn from Stillingfleet, and is, as Locke himself pointed
out (LW, 4:28-29), incorrect. (I have elsewhere marshaled evidence that the Locke-Stillingfleet corre-
spondence was a significant influence on Berkeley’s thought; see Pearce, forthcoming(a).) I am here
concerned primarily with Berkeley’s positive theory, and not his arguments against Locke, so I will not
pursue this issue farther.
144
in De Motu, had sought to deflate. These disputes do not, however, prevent it from being
that case that “there are very evident propositions or theorems relating to force, which
contain useful truths” (Alc,x7.10).
It is interesting that Euphranor here says that the theorems in question contain use-
ful truths. In Berkeley’s view, the job of the physicist is to “frame general rules from
the phenomena, and afterward derive the phenomena from those rules” (PHK,x108).
In the more sophisticated theory of De Motu, fundamental physics aims to discover
certain ‘principles’ from which “both general mechanical theorems and particular ex-
planations are derived” (DM,x36). One might therefore question whether Berkeley’s
theory requires the theorems themselves to be true, or whether false propositions with
true entailments might suffice for the sort of explanation at which fundamental physics
aims.
13
These issues will recur in chapter 8 where the role of these laws in the language
of nature will be considered. For now, however, the key question is whether any state-
ment containing ‘force’ (and, by analogy, ‘grace’) is on Berkeley’s view true, or whether
it is only the particular phenomena, stripped of ‘mathematical hypotheses,’ which count
as true. This question is answered at the end ofx10, where Euphranor says that there are
“divers true and useful propositions concerning [grace] as well as [force].” The question
of how these propositions manage to be true will be examined in more detail in chapter
7, below.
Euphranor’s main aim inx10 is to argue that our lack of an idea does not prevent the
word ‘force’ from serving its purpose in the discourse of physics. As a result, we may
conclude by analogy that the lack of an idea attached to ‘grace’ is no evidence that this
word fails in its intended purpose. Just as “by means of the doctrine of force, men arrive
13. A theory of the latter sort is defended by Cartwright 1983.
145
at the knowledge of many inventions in Mechanics, and are taught to frame engines, by
means of which things difficult and otherwise impossible may be performed,” so
grace may be the object of our faith, and influence our life and actions, as
a principle destructive of evil habits and productive of good ones, although
we cannot attain a distinct idea of it, separate or abstracted from God the
author, from man the subject, and from virtue and piety its effects (Alc,
x7.10).
Euphranor now summarizes the results of the discussion. It has been concluded that
“although terms are signs, yet . . . those signs may be significant, though they should not
suggest ideas represented by them, provided they serve to regulate and influence our
wills, passions, or conduct.” It follows from this, Euphranor says, “that the mind of
man may assent to propositions containing such terms” (x7.11). Here we are speaking
of assent to a ‘verbal proposition,’ i.e. a sentence. Such assent occurs when the mind
is “directed or affected” by the words in the intended manner. This confirms my earlier
account of ‘reward’ talk (see above,x3.2): to assent to the sentence, i.e. to accept the
speaker’s assertion, is to be motivated as the speaker intends. This kind of assent does
not require the having of any ideas. As a result, it is possible to assent to sentences
containing ‘force’ or ‘grace:’ to do so is not to have any ideas at all, but to be in a certain
affective or motivational state, or to have certain dispositions to feeling and action.
4.4 The Nature of Faith (xx11-13)
After summarizing his conclusions about the uses of language and the conditions for the
significance of language, Euphranor enters into a discussion of the nature of faith. This
discussion is important to Berkeley’s concerns insofar as it provides a direct response to
146
the Toland argument from the beginning of the dialogue by showing that it is possible
to have faith in grace and the Trinity after all. It is relevant to our concerns insofar as
faith is or involves assent to certain claims.
14
From the results of the preceding sections,
Euphranor will argue, against Locke and Toland, that it is possible genuinely to assent
to verbal propositions (sentences) whose terms do not correspond to ideas in our minds.
Although perhaps prepared to accept the conclusion about grace, Alciphron had
balked at the suggestion that it was possible, in the same way, to believe the doctrine of
the Trinity. Here Euphranor argues that it is. One assents to the doctrine of the Trinity
when the “doctrine of a Creator, Redeemer, and Sanctifier makes proper impressions on
[one’s] mind, producing therein, love, hope, gratitude, and obedience, and thereby be-
comes a lively operative principle influencing [one’s] life and actions” (Alc,x7.11). For
Berkeley, a primary aim of religious discourse is moral motivation (Pearce, forthcom-
ing[b],xx4-5, 8-9). It is sufficient for the meaningfulness of this discourse that it should
accomplish this end. It accomplishes this end by getting people into certain affective
and/or volitional states. The people who get into those states can be said to assent to the
proposition – that is, to take it to be true.
Euphranor now argues ad hominem, extracting from Alciphron the concession that
there is faith among freethinkers as well. One may find, among the freethinkers, “a
fine gentleman or lady of fashion, who are too much employed to think for themselves,
and are free-thinkers at second-hand” (Alc,x7.11). These people may assent either to
the doctrine of fate or the doctrine of chance (the two alternatives to the doctrine of
providence) on the basis of testimony without having ideas of either fate or chance.
14. Jakapi 2002b argues that the nature of assent is in fact Berkeley’s main concern in these discus-
sions. I certainly agree that this is one of the main issues, but I take Berkeley to be more serious about
understanding language in general than Jakapi does. Williford and Jakapi 2009 does read the dialogue
as attempting to provide a general theory of language, and Jakapi has confirmed that this does indeed
represent a revision of his previous view (personal correspondence, August 27, 2013).
147
Despite their lack of ideas, their assent to these propositions may “produce real effects,
and show itself in the conduct and tenor of their lives, freeing them from the fears
of superstition, and giving them a true relish of the world, with a noble indolence or
indifference about what comes after.” Analogously, Christians may “believe the divinity
of our Saviour, or that in him God and man make one Person . . . so far as for such faith
or belief to become a real principle of life and conduct.” The effect of this ‘principle’ is
that the believer will “submit to [Christ’s] government, believe his doctrine, and practise
his precepts” (Alc,x7.11).
Here Euphranor makes a further claim: not only is it possible to assent without hav-
ing ideas, it is both possible and rational to assent despite unsolved puzzles in the doc-
trine. The puzzles about the Trinity and the Incarnation are analogous to puzzles about
“the principle of individuation in man” and “human personal identity.” Again, Alciphron
balks: “Methinks there is no such mystery in personal identity” (x7.11). Locke’s con-
sciousness theory, Alciphron claims, is perfectly adequate. Euphranor responds with the
familiar argument, more usually associated with Reid (Reid 1786, ch. 3.6), that Locke’s
theory gives rise to violations of the transitivity of identity. Euphranor wishes to draw
from this the conclusion that unresolved puzzles about personal identity do not render
belief in the persistence of persons irrational and so, by analogy, the presence of un-
resolved puzzles about (e.g.) the Trinity do not render belief in the Trinity irrational.
Alciphron rejects this line of argument: “you will never assist my faith by puzzling my
reason.” This, however, should not lead the reader to suppose that Berkeley takes the
line of argument to be defective. Rather, Alciphron’s rejection plays an important role
in the literary arc of the seventh dialogue, and the work as a whole. The freethinkers
pride themselves on their adherence to reason, and especially on following reason rather
than authority. However, Alciphron’s ally Lysicles has just suggested that “it would be
148
more prudent to abide by the way of wit and humour, than thus to try religion by the dry
test of reason and logic” (Alc,x7.11). As we approach the end of the work, the so-called
‘freethinkers’ grow more and more stubborn and dogmatic in their assertions. Under
the weight of Euphranor’s rational defense of Christianity, Alciphron and Lysicles show
themselves more committed to their infidelity than to reason, so that the very accusation
they had made against the faithful is retorted back on them.
Ignoring Alciphron’s remark, Euphranor asserts, “There is . . . a practical faith, or
assent, which shows itself in the will and actions of a man, although his understanding
may not be furnished with those abstract, precise, distinct ideas” (x7.12). Indeed, even
the abstractionists must acknowledge this, for they admit that these abstract ideas are
“above the talents of common men,” yet practical faith is to be found among these
‘common men.’ Practical faith thus must not be a matter of having such ideas.
15
Here Crito intervenes to inveigh against those “who confound Scholasticism with
Christianity.” The Scripture, Crito observes, seems little concerned with these sorts
of philosophical abstractions. Upon even a cursory examination of the Bible, “Every
one, whose understanding is not perverted by science falsely so called, may see that the
saving faith of Christians is quite of another kind, a vital operative principle, productive
of charity and obedience” (x7.12).
Alciphron, however, has a ready response. The Bible, to be sure, rarely uses techni-
cal philosophical jargon and does not seem to be concerned with engaging believers in
abstract speculation. However, the Bible is not the only authoritative doctrinal standard
to which Crito and Euphranor (and Berkeley) are bound: “What are we to think then
of the disputes and decisions of the famous Council of Nice, and so many subsequent
Councils? What was the intention of those venerable Fathers the Homoousians and the
15. On the kind of practical faith which is possible for those with limited understanding, cf. LW, 7:128-
132.
149
Homoiousians?” (Alc,x7.12) The Council of Nicea (AD 325) was famously, and quite
literally, a dispute over an iota. The winning (Athanasian) side affirmed that the Son
wasåmooÔsion, or of the same substance or essence as the Father, while the losing (Ar-
ian) side wished to affirm only that the Son was åmoioÔsion, or of similar substance or
essence to the Father. This is, by all accounts, an abstract, technical distinction.
To this Crito responds,
Whatever their intention was, it could not be to beget nice abstracted ideas
of mysteries in the mind of common Christians, this being evidently impos-
sible: nor does it appear that the bulk of Christian men did in those days
think it any part of their duty, to lay aside the words, shut their eyes, and
frame those abstract ideas (x7.12).
16
This, Crito insists, is yet another point of analogy between theology and physics: the
aim of the disputants in the debate about force is likewise not to settle on a particular
abstract idea. Crito does not say what the point of the debate about force is, but he does
give an account of the theological dispute. First, he says that in the ‘management’ of the
controversies “human infirmity must be supposed to have had its share” (x7.12). That
is, even the orthodox may have been unduly uncharitable to their opponents, and some
portions of these disputes may truly have been ‘purely verbal,’ in the sense that there
was nothing at stake in selecting one formulation over another. Crito does not, however,
hold that all of these disputes were pointless. Instead, he says, “the main end was not,
16. There is at least some evidence that Crito (and hence Berkeley) is mistaken about the history here.
Hilary of Poitiers, who was not present at the Council, but lived through the period of intense controversy
which followed and was an important disputant on the Athanasian side, insisted that the word ‘åmooÔsion’
was not so important as the belief it was used to express, which belief, he claims, both he and the entire
Christian Church, held before learning to express it with this word. Indeed, he even holds that language is
necessarily inadequate to express the faith of the Church. See Hilary of Poitiers On the Councils,xx65, 69,
79. Although the matter is not totally clear, this could easily be taken to suggest a language-independent
intellectual grasp of some concept denoted by ‘åmooÔsion.’ Incidentally, one of the main aims of this
work of Hilary’s is to answer the charge that the dispute about the iota is purely verbal.
150
on either side, to convey precise positive ideas to the minds of men, by the use of those
contested terms, but rather a negative sense, tending to exclude Polytheism on the one
hand, and Sabellianism on the other” (Alc,x7.12).
17
In support of this contention, Berkeley cites the fifth century Church history of So-
zomen.
18
He could just as easily have cited Athanasius himself who argues at length that
the aim of the Council was “to do away with the irreligious phrases of the Arians” and
that, due to the Arians’ interpretive ingenuity, this could be done only by the use of the
un-Scriptural word ‘åmooÔsion’ (Athanasius Defence of the Nicene Definition,x5.19).
Crito’s claim, which he takes to be supported by Sozomen’s account, is that ‘åmooÔsion’
gets its meaning from what it excludes. For one proposition to exclude another is for
the first to entail the falsity of the second, and these entailments, according to Crito, are
the primary ‘sense’ which was ‘conveyed’ by the use of the word ‘åmooÔsion’ – that is,
it they form part of that word’s meaning. This makes the comparison with force much
clearer: ‘åmooÔsion’ is like ‘force’ and other ‘mathematical hypotheses’ in physics in
that it gets its meaning (at least in part) from inference rules which connect it to other,
antecedently meaningful, bits of language.
Crito’s remarks about the ‘åmooÔsion’ radically undermine Berman’s interpretation.
Berman reads Berkeley as holding that religious mysteries are Ayer-style emotive utter-
ances (Berman 1994, 155). Ayer holds quite explicitly that it is, strictly speaking, im-
possible to contradict a purely emotive utterance. This is because emotive words stand
for ‘mere pseudo-concepts’ which do not stand in any logical relations to anything (Ayer
17. Sabellianism, also known as ‘modalism,’ is the view that the Father, the Son, and the Holy Spirit
are three ‘modes’ in which the one God is experienced by the Church. It denies that Trinitarian ways of
speaking have any ontological significance.
18. Berkeley erroneously cites book 2, chapter 8. The passage Berkeley has in mind is certainly book 2,
chapter 18 (contrary to Clarke, who says that Berkeley “should refer in general to Book I” (Clarke 2008,
306 note d)).
151
1952, 107). According to Crito, the one who affirms the ‘åmooÔsion’ contradicts both
the polytheist and the Sabellian. Hence ‘åmooÔsion’ does not stand for a mere pseudo-
concept, and the claim that the Son isåmooÔsion with the Father is not a purely emotive
utterance in Ayer’s sense.
19
In the face of Crito’s defense of the ‘åmooÔsion,’ Alciphron is undeterred. “But what
shall we say,” he asks,
of so many learned and ingenious divines, who from time to time have
obliged the world with new explications of mysteries, who, having them-
selves professedly laboured to acquire accurate ideas, would recommend
their discoveries and speculations to others for articles of faith?
Crito responds by disavowing entirely this whole project of “the explication of mysteries
in divinity.” The sort of ‘explication’ Crito has in mind is, apparently, the attempt to gain,
and give to one’s readers, distinct ideas corresponding to the terms in which the mys-
teries are expressed. This, Crito says, is “as fruitless as the pursuit of the philosopher’s
stone in chemistry, or the perpetual motion in mechanics” (Alc,x7.12), but it does not
follow from this that theology as a whole, or even revealed theology, conversant about
mysteries, is meaningless. The words in which the mysteries are expressed are perfectly
meaningful, but meaningfulness does not require direct connection to ideas.
19. Daniele Bertini also attributes to Berkeley the view that “it would be nonsensical to reason about the
Trinity” (Bertini 2010, 132). Admittedly, Berkeley does come close to saying this at one point, but it is
necessary to attend to Berkeley’s exactly phrasing: “to pretend to demonstrate or reason any thing about
the Trinity is absurd” (N,x584). The rest of the note is about demonstrating religious doctrines. I take
it Berkeley’s point is simply that no part of Trinitarian dogma can be demonstrated, not that no part of
Trinitarian dogma could be used as a premise in a demonstration. Indeed, the main point of the whole note
is that there can be demonstrations in revealed religion, although the ‘principles’ of those demonstrations
will be matters of faith (cf. Aquinas Summa Theologica, Iq1a8).
Bertini in fact takes an even stronger line than Berman, insofar as he rejects Berman’s view that Berke-
ley recognized such a thing as a ‘scientific’ natural theology in addition to emotive revealed theology.
Bertini’s view, like Berman’s, is unable to account for Crito’s claim that the ‘åmooÔsion,’ i.e. the central
element of Trinitarian dogma, entails the falsity of Sabellianism and polytheism.
152
Crito now suggests that the general approach which has been developed can be ap-
plied to other religious mysteries as well. He discusses, as an example, the doctrine of
original sin. We are, he says, unable to frame any idea of original sin, “or of the man-
ner of its transmission” – that is, of the manner by which Adam’s sin can lead to the
guilt and corruption of his progeny. Nevertheless, “the belief thereof may produce in
[one’s] mind a salutary sense of [one’s] own unworthiness, and the goodness of [one’s]
Redeemer: from whence may follow good habits, and from them good actions, the gen-
uine effects of faith” (Alc,x7.13). Sentences expressing the doctrine of original sin must
therefore be understood as among those whose purpose is “the influencing our conduct
and actions . . . by raising certain passions, dispositions, and emotions in our minds”
(x7.8). More generally, Crito insists that faith is “placed in the will and affections rather
than in the understanding, and producing holy lives, rather than subtle theories” (x7.13).
This, Crito says, is quite clear as long as “the Christian religion is considered as an in-
stitution fitted to ordinary minds, rather than to the nicer talents . . . of speculative men;
and our notions about faith are accordingly taken from the commerce of the world, and
practice of mankind, rather than from the peculiar systems of refiners.” In short,
one that takes his notions of faith, opinion, and assent from commonsense,
and common use, and has maturely weighed the nature of signs and lan-
guage, will not be so apt to controvert the wording of a mystery, or break
the peace of the church, for the sake of retaining or rejecting a term (x7.13).
The emphasis here is again on ‘common use’ and the ‘practice of mankind.’ What
we must examine, Crito says, is how the word ‘faith’ operates in ordinary language.
When these empirical facts are combined with the correct theory of language, we will
be able to see that when ordinary believers speak of having faith in the mysteries taught
by the Church, the thing they are talking about is possible. (Crito, and Berkeley, would
153
add that, given the arguments of the earlier dialogues, we will also be able to see that
the thing ordinary believers call ‘faith’ is desirable and correct; on how Berkeley would
have us go about assessing such claims, seex7.2.4, below.) Now, it is certainly part of
ordinary linguistic practice to describe ordinary churchgoers as assenting to, or having
faith in, mysteries such as the Trinity or original sin, despite the fact that no one is under
the illusion that these people have distinct abstract ideas associated with these terms.
This fact is deserving of more careful examination, which promises to make clear to us
the nature of “faith, opinion, and assent.”
There is, of course, a distinction between sincere and insincere professions of re-
ligion, and this is very much a part of plain language and commonsense. However,
the sincerity of one’s assent to (for instance) the Nicene Creed is certainly not to be
judged by one’s possession of an abstract idea corresponding to the phrase (in the En-
glish version) “of one substance,” for then there would be no sincere professions at
all. Yet surely the plain language distinction between genuine and feigned assent is
tracking some difference.
20
Crito holds, quite plausibly, that in ‘common use’ and ‘the
practice of mankind’ the distinction between genuine and feigned religious faith is a dis-
tinction between instances of verbal assent which are accompanied by certain practical
attitudes and actions, and those that are not. The one who sincerely or genuinely assents
to the doctrine of the Incarnation is the one who “submit[s] to [Christ’s] government,
believe[s] his doctrine, and practise[s] his precepts” (Alc,x7.11), and the one who sin-
cerely or genuinely assents to the doctrine of original sin is the one who has “a salutary
sense of his own unworthiness, and the goodness of his Redeemer” (x7.13). These peo-
ple are genuinely different from those who merely mouth the words, and this difference
20. With this should be compared Philonous’s observation that, quite independently of any metaphys-
ical theory, there is an empirical or experiential distinction between dreaming and waking which is left
perfectly intact after the rejection of material substrata (DHP, 235).
154
has empirical upshots which are available as a practical basis for the ascription of faith
to them. It is, in short, a part of the public, conventional rules of language.
Crito now attempts to add to the case for “the efficacious necessary use of faith
without ideas” by arguing that belief in a future state is efficacious in preventing people
from carrying out ‘wicked projects,’ “although it be a mystery, although it be what eye
has not seen, nor ear heard, nor has it entered into the heart of man to conceive” (Alc,
x7.13). The empirical claim about motivation is, of course, open to question, but the fact
that it is taken to be relevant is significant: Crito assumes that if talk about an afterlife
produces this effect, then it must be admitted to be both meaningful and useful. This is
Berkeley’s conclusion about religious discourse in general.
4.5 The Nature of Science (xx14-16)
At the end of Crito’s speech, Alciphron raises a new objection: “It seems, Euphranor
and you would persuade me into an opinion . . . that a man need not renounce his reason
to maintain his religion. But, if this were true, how comes it to pass, that, in proportion
as men abound in knowledge, they dwindle in faith?” (x7.14). This objection leads to a
discussion of the relationship between faith and knowledge. Euphranor claims that “The
objections made to faith are by no means an effect of knowledge, but proceed rather from
an ignorance of what knowledge is.” Since both faith and knowledge “imply an assent
of the mind,” they must be understood together, and when they are so understood the
freethinkers’ objections collapse.
Paradigmatic instances of knowledge are those involved in sciences (in the early
modern sense of that word), such as arithmetic or geometry. Euphranor gives the fol-
lowing account of science:
155
To trace things from their original, it seems that the human mind, naturally
furnished with the ideas of things particular and concrete, and being de-
signed, not for the bare intuition of ideas, but for action or operation about
them, and pursuing her own happiness therein, stands in need of certain
general rules or theorems to direct her operations in this pursuit: the sup-
plying which want is the true, original, reasonable end of studying the arts
and sciences. Now these rules being general, it follows that they are not
to be obtained by the mere consideration of the original ideas, or particu-
lar things, but by the means of marks or signs, which, being so far forth
universal, become the immediate instruments and materials of science (Alc,
x7.14).
A science, according to Euphranor, is a linguistic object. Generality can be had only “by
the means of marks or signs,” and this generality is essential to science. The conventions
which give these marks or signs their meaning are general rules by which we operate on
ideas for the purpose of pursuing happiness. “It is not therefore by mere contemplation
of particular things, and much less of their abstract general ideas, that the mind makes
her progress, but by an apposite choice and skillful management of signs” (x7.14). In
what, then, does the knowledge of a science consist? Euphranor tells us,
one, who can neither write nor read, in common use understands the mean-
ing of numeral words, as well as the best philosopher or mathematician.
But here lies the difference: the one who understands the notation of
numbers, by means therefore is able to express briefly and distinctly all
the variety and degrees of number, and to perform with ease and dispatch
several mathematical operations, by the help of general rules (xx7.14-15).
156
Scientific knowledge, in Euphranor’s view, turns out to be a kind of know-how. It con-
sists in practical competence with a certain sign system and certain rules for the practical
application of that system. The mathematician knows more signs and is adept at more
rules for manipulating them than is the non-mathematician. Euphranor remarks that
it is evident both that these operations have a “use in human life” and also “that the
performing them depends on the aptness of the notation.”
At this point, Euphranor enters into a discussion of the development of arithmetic.
He begins by claiming, quite explicitly, that arithmetic cannot be developed until after
the introduction of language. This claim flies in the face of the Theory of Meanings by
denying that mathematics can be represented in thought without the use of language.
The introduction of words and symbols for numbers and operations on them is not
merely putting into words a previously existing mental mathematics; it is in fact the
first creation of mathematics.
Once language has been introduced, it is possible to introduce “the use of names, by
the repetition whereof in a certain order [one] might express endless degrees of num-
ber.” This, Euphranor says, “would be the first step towards” arithmetic. This, of course,
repeats the previously discussed theory of the Principles, according to which the distinc-
tive and essential element of a numeral system is the existence of a successor function
by which signs for arbitrarily large numbers can be generated (see above,x3.3.1).
“The next step,” Euphranor says, “would be, to devise proper marks of a permanent
nature, and visible to the eye” (Alc,x7.15), or, in other words, a system of writing.
This system “would, in proportion as it was apt and regular, facilitate the invention and
application of general rules, to assist the mind in reasoning and judging, in extending,
recording, and communicating its knowledge about numbers.” Berkeley no doubt has in
mind the advantages of computing in Arabic numerals rather than, for instance, Roman
157
numerals. The use of this notation allows for much simpler rules for performing the four
basic arithmetic operations. In performing these operations, Euphranor says, “the mind
is immediately occupied about the signs or notes, by mediation of which it is directed
to act about things . . . without ever considering the simple, abstract, intellectual, general
idea of number” (Alc,x7.15). Again, just as in the Principles, this is a process of
“computing in signs” (PHK,x121).
In the science of arithmetic, Euphranor says, “the names [of numbers] are referred
to things, the characters [i.e. numerals] to names, and both to operation” (Alc,x7.15). It
is, in other words, the role that they play in computation that gives number words and
numerals their meaning.
Arithmetic and algebra, Euphranor says, are sciences par excellence, and can there-
fore be used as a model to understand sciences, and the knowledge of them, in general.
In particular, “all sciences, so far as they are universal and demonstrable by human rea-
son, will be found conversant about signs as their immediate object, though these in the
application are referred to things” (x7.16). In other words, the science itself is a system
of signs manipulated according to its own internal rules, but what makes it a genuine
science, and not merely a notation game, is that some of its theorems have practical
applications.
4.6 Euphranor’s Summary and Conclusions (xx16-18)
At this point, Euphranor summarizes the theory which has been defended. He begins
with a discussion of the importance of the use of signs to human knowledge. He says,
158
it is certain we imagine before we reflect, and we perceive by sense before
we imagine: and of all our senses the sight is the most clear, distinct, vari-
ous, agreeable, and comprehensive. Hence it is natural to assist the intellect
by the imagination, the imagination by the sense, and the other sense by
sight (Alc,x7.16).
This is done by the creation of signs apprehensible by the imagination for use by the
intellect, of signs perceivable by the senses for use by the imagination, and so forth.
This is the reason, for instance, for “the use of models and diagrams” in math and
physics, as well as for the use of metaphors drawn from physical things to describe
minds. Here Euphranor speaks of the use of signs to assist our thinking; as we saw in
the detailed discussion of arithmetic, there are very important sorts of thinking which
cannot be conducted without this sort of assistance. The conclusion is that “the doctrine
of signs [is] a point of great importance, and general extent, which, if duly considered,
would cast no small light upon things, and afford a just and genuine solution to many
difficulties” (x7.16). This, of course, is the central thesis of this dissertation: Berkeley’s
theory of signs, and specifically of language, is a “solution to many difficulties” in his
philosophy.
Euphranor now provides a list of theses about signs which have been defended
(x7.17). It will be helpful to comment on each of these theses individually.
Signs “do not always suggest ideas signified to the mind.” As we have seen, Eu-
phranor has claimed both that some signs never suggest ideas, and that no sign becomes
meaningful by always suggesting an idea; even those signs that frequently suggest ideas
can be used meaningfully without speaker and hearer having the idea in question in
mind.
159
“When [signs] suggest ideas, they are not general abstract ideas.” Signs, of course,
cannot do this, for there are no general abstract ideas to be suggested.
Signs “have other uses besides barely standing for and exhibiting ideas, such as
raising proper emotions, producing certain dispositions or habits of mind, and di-
recting our actions in pursuit of that happiness, which is the ultimate end and
design, the primary spring and motive, that sets rational agents at work.” The
essential point here is that linguistic conventions do not refer only to ideas.
21
The con-
ventions governing the use of a sign may also require us to perform a certain action, be
motivated in a certain way, or experience a certain feeling.
“Signs may imply or suggest the relations of things; which relations, habitudes or
proportions, as they cannot be by us understood but by the help of signs, so being
thereby expressed and confuted, they direct and enable us to act with regard to
things.” This thesis, which Berkeley added to the list in the 1752 edition, had not been
given much discussion in the earlier dialogue, and so it is not possible for us to know
in any detail what Berkeley had in mind. However, it is quite clear that Euphranor is
claiming that relational thought in general is possible only by means of signs. Again we
can see, on this view, just how impoverished human thought would be without language.
“The true end of speech, reason, science, faith, assent, in all its different degrees, is
not merely, or principally, or always the imparting or acquiring of ideas, but rather
something of an active, operative nature, tending to a conceived good; which may
sometimes be obtained, not only although the ideas marked are not offered to the
mind, but even although there should be no possibility of offering or exhibiting
21. On the role of convention in Berkeley’s theory of language, see below,x5.3.
160
any such idea to the mind.” This possibility is illustrated by the example of “the
algebraic mark, which denotes the root of a negative square.” This mark, Euphranor
says, “has its use in logistic operations, although it be impossible to form an idea of any
such quantity.” Euphranor’s point is quite general: “what is true of algebraic signs, is
also true of words or language, modern algebra being in fact a more short, apposite, and
artificial sort of language” (Alc,x7.17). It is, in other words, a widespread feature of
language that it aims to shape actions and feelings, and it sometimes does so without the
mediation of ideas. Note that this is the case not only for faith, but for “speech, reason,
science, faith, assent, in all its different degrees.” Clearly, then, this is not a thesis
to be confined to religious mysteries; it is a theory of language. The theory says that
words (or, more generally, signs) get to be meaningful when they are associated with
conventional rules which are useful for some purpose, and that to use them competently
is to follow the rules. The case in which these rules allow direct substitution between
words and ideas is a special case confined to a small portion of human language. This is
hardly surprising, for the having of ideas is not an end in itself; having ideas is meant to
contribute to larger projects and, ultimately to general human flourishing. Language is
likewise a tool for these ends, and the linguistic rules require the having of ideas only in
the narrow range of cases where this is necessary for language to accomplish its ends.
Finally, Euphranor summarizes his conclusions about science and of faith. It is not
the aim of either of these “to obtain precise ideas,” and when people mistakenly adopt
this end, “they will be sure to embarrass themselves with difficulties and disputes,” ex-
amples of which can be found in both science and religion. Our failure to obtain ideas
of infinitesimals in no way renders geometry either useless or meaningless; rather, ge-
ometry “governs and directs the actions of men, so that by the aid or influence thereof
161
those operations become just and accurate, which would otherwise be faulty and uncer-
tain” (Alc,x7.18). Since the fact that signs are useful and action-guiding is sufficient for
their meaningfulness in geometry, it must likewise be sufficient elsewhere, including in
religious mysteries.
At this point, Alciphron raises one final objection: “According to this doctrine, all
points may be alike maintained. There will be nothing absurd in Popery, not even tran-
substantiation.” It is noteworthy that Crito strenuously objects to this allegation. The
theory that has been presented does not have the consequence that there are no stan-
dards of correctness for religious mysteries. “This doctrine,” Crito insists, “justifies no
article of faith which is not contained in Scripture, or which is repugnant to human
reason, which implies a contradiction, or which leads to idolatry or wickedness of any
kind.” This, unfortunately, does little to clarify the question of what the standards of cor-
rectness for articles of faith, or for assertions more generally, are in Berkeley’s theory.
Euphranor and Crito had earlier defended the Bible’s claim to be regarded as a divine
revelation; whatever the relevant standards are, God can be presumed to meet them. It is
not clear what it would be for a doctrine to be ‘repugnant to human reason’ or ‘contain
a contradiction’ besides for it to lead to puzzles of the sort that have earlier been dis-
missed as unimportant. The last criterion, at least, is rather clear: since part of the aim
of these assertions is to generate good behavior, those that generate bad behavior are to
be rejected. These questions about the standards for accepting or rejecting assertions
will be addressed in chapter 7, below.
Alciphron does not challenge Crito’s defense. Instead, at this point he drops the
argument against religious mysteries and proceeds along a different track. Since the
discussion of language ends here, I will leave off my commentary and proceed to the
philosophical explication and evaluation of Berkeley’s theory of language.
162
Chapter 5
Rule-Following
So far, I have defended two interpretive theses about the role of rules in Berkeley’s phi-
losophy of language. The first is the claim that a word’s being significant or meaningful
consists in the existence of certain conventional rules by means of which the word be-
comes practically useful within the speaker community. The second is that speakers
and hearers need not be able to articulate the rules they follow. The aim of this chap-
ter will be to gain a clearer understanding of the nature of these rules, what it means
to follow them, and what it means for them to obtain as conventions in a particular
community. I begin by discussing the question of how, according to Berkeley, it comes
about that a particular agent follows a particular rule. By understanding Berkeley’s an-
swer to this question, we will understand what, according to Berkeley, is involved in
following a rule, and how Berkeley might respond to Wittgensteinian skeptical wor-
ries about rule-following. From this discussion will emerge a substantive constraint on
the sorts of rules it is possible for agents to follow, a constraint I will call ‘Berkeley’s
Internalist Requirement,’ by analogy to a certain form of internalism in epistemology.
Berkeley’s Internalist Requirement requires, in brief, that prior to acquiring the rule the
agent be capable of recognizing the conditions for the rule’s applicability. This seems a
modest requirement, but, as I will argue in chapter 6, it is a key premise in a series of
arguments establishing Berkeley’s immaterialist ontology. After examining the process
of rule acquisition and Berkeley’s Internalist Requirement, we will have a sufficiently
strong grasp of Berkeley’s general understanding of rules and rule-following to explore
163
Berkeley’s understanding of the rules of language. We will be particularly concerned to
show that Berkeley holds the rules of language to be a matter of public convention and to
understand the precise meaning of this doctrine. Finally, we will examine an especially
important class of linguistic rules, the rules of inference.
5.1 Implicit and Explicit Rule-Following
“Two ways there are of learning a language,” Berkeley observes, “by rule or by practice”
(PHK,x108 [1710 ed.]).
1
Those who learn a language ‘by practice,’ Berkeley writes,
will often understand the language “without . . . being able to say by what rule a thing
is so or so.” The one who learns ‘by practice’ is, however, at an advantage since “it is
very possible to write improperly, through too strict an observance of general grammar-
rules” (PHK,x108; cf. Wittgenstein [1958a] 2009, 116-117). This is because the rules
actually followed by the native speakers, especially in informal contexts, are far more
complicated and flexible than the rules the grammarians have written down. The suc-
cessful language learner, even if she initially learned the language ‘by rule,’ eventually
develops a ‘feel’ for the language which allows her to make appropriate exceptions to
the rules she has learned. This does not mean that she is in a position to correct the
grammarian; grammarians have a special expertise which goes beyond mere linguis-
tic competence. Writing down the rules is, in other words, a more difficult task than
following them.
This is the case with the extremely complicated rules of natural language, and in-
deed, more generally, with most conventions that grow up ‘organically’ rather than being
formally instituted. There are, however, simpler cases, such as the formal notation of
1. This point is less explicit in the 1734 edition, but is still present. On a possible reason for the change
see Stoneham 2013, 221.
164
arithmetic, in which the use of the signs is governed by rules which can be stated with
full precision.
Let us distinguish between three different ways in which the actions of an agent may
be related to a rule. First, the agent’s action may conform to the rule merely accidentally,
as when a child moves a chess piece at random, but the movement she makes happens to
be permitted by the rules of chess. We will call this ‘rule-compliance.’ Second, the agent
may have conscious awareness of the rule, and the explicit intention to follow it, like a
novice chess player who must say to himself ‘this is the bishop; it moves diagonally’
before deciding what move to make. Call this ‘explicit rule-following.’ The final case is
that of the experienced chess player who, looking at the board configuration, ‘sees’ only
the legal moves and decides among them without ever (consciously) thinking about the
rules of the game. Call this ‘implicit rule-following.’
The fluent speaker of a language follows its rules implicitly. The point that Berkeley
is making in the passage under discussion is that there are two very different ways in
which it can come about that someone follows a rule implicitly: in some cases, the
individual starts out by following the rule explicitly, and in other cases not. My task
here will be to argue that what Berkeley calls ‘suggestion’ just is the implicit following
of rules of a certain sort and to explain the different ways, according to Berkeley, this
can come about.
Berkeley’s clearest treatment of suggestion is to be found in the late (1733) work,
The Theory of Vision . . . Vindicated and Explained. In this work, Berkeley is concerned
to distinguish suggestion from inference, something he had not done clearly in earlier
works (Winkler 2005a, 161n6). According to the theory put forth in the Theory of
Vision Vindicated, to say that ‘blue’ suggests blue and ‘red’ suggests red is to say that
“upon hearing the words blue or red, [the appropriate colors will] be apprehended by
165
the imagination” (TVV,x10). Suggestion is the province of the senses, as contrasted
with inference and judgment, which are the province of the understanding (x42). As a
result, what is suggested by sensation may be said to be a mediate object of sensation,
but what is inferred from sensation is not an object of sensation (TVV,xx11-12; cf.
DHP, 174-175).
Two further facts about suggestion are emphasized in Berkeley’s earlier works: it
is involuntary, and it happens so quickly as to go unnoticed. The transition from the
suggesting idea to the suggested idea is “swift and sudden and unperceived” (NTV,
x145), and “it is not in our power to keep out the [suggested idea], except we exclude
the [suggesting idea] also” (x51).
Berkeley lists four ways in which one idea can come to suggest another, “by likeness,
by necessary connexion, by geometrical inference, or by arbitrary institution” (TVV,
x14; cf. TVV,x39). In Alciphron, a similar list is found, but with only three items:
“necessary deduction to the judgment, . . . similitude to the fancy, . . . [or] experience,
custom, and habit” (Alc,x4.10; cf. Rickless 2013, 46). The Alciphron passage can be
taken as giving the same list as the Theory of Vision Vindicated if we take ‘necessary
connexion’ and ‘geometrical inference’ to be two varieties of ‘necessary deduction to
the judgment.’ I will discuss each of Berkeley’s three relations in turn, beginning with
‘necessary deduction to the judgment.’
TVV,x14 is embedded in a prolonged argument against the view that we mediately
perceive by sight physical objects which are the efficient causes of our immediate visual
perceptions. ‘Necessary connexion’ is standard early modern causal vocabulary. It is
true that it is not always a causal term for Berkeley; for instance atx39, a passage which
repeats the taxonomy with which we are concerned, Berkeley says that, between the im-
mediate and mediate objects of vision, “there is no . . . relation of similitude or causality,
166
nor any necessary connexion whatsoever.” However, if these taxonomies really are to
correspond to the relevant alternatives Berkeley attempts to rule out in TVV, and if the
two passages are to be seen as giving the same taxonomy, then ‘necessary connexion’
inx14 must correspond to ‘causality’ inx39, and the other possible kind of ‘necessary
connexion’ inx39 must be the sort found in ‘geometrical inference.’ Alciphron lumps
these together under the heading ‘necessary deduction to the judgment.’ To summarize:
there is a broad use of ‘necessary connexion’ in which it stands for a genus of which
causality and logical (‘geometric’) entailment are species, and this is the way it is used in
x39. However, there is also a narrow use in which ‘necessary connexion’ is a synonym
of ‘causality,’ and this is how it is used atx14.
In Berkeley’s view, it is strictly speaking impossible that there should be an efficient
causal relation in the case at issue, since neither an idea, nor anything like an idea,
can ever cause anything (TVV,x11; cf. PHK,x25). However, causality is ruled out
only as a hypothesis about “how one idea comes to suggest another belonging to a
different sense, how things visible suggest things tangible, [and] how present things
suggest things remote and future” (TVV,x14). This is because, in these cases, both of
the relata are ideas. One can nevertheless legitimately infer that one’s ideas have a cause
(PHK,x26; TVV,xx11-12). This cause Berkeley calls ‘God.’
2
The causal relation here
establishes a connection between perceived ideas and God which can form the basis
of a suggestion relation, whereby the ideas in question, in addition to suggesting other
ideas, suggest to us the thought of God. As a result, once we have been sufficiently
indoctrinated with Berkeley’s philosophy, we will indeed come “to see the sovereign
Lord of all things” (albeit in a rather attenuated sense) when we look at the creation
2. Following Bennett 1971,x35, this has come to be known as Berkeley’s ‘Passivity Argument’ for the
existence of God. For an analysis of the argument, with references to the relevant secondary literature,
see Pearce, forthcoming(b),x1.1.
167
(PHK,x148, emphasis added). Bringing about this vision of God is, indeed, central to
the religious aim of Berkeley’s philosophical writing.
3
Although this seems to clash with Berkeley’s insistence that “the Power or Cause
of ideas is not an object of sense, but of reason” (TVV,x12), the conflict is merely
apparent. Berkeley’s point here is that God cannot at first come to be known by sense,
in the way colors, shapes, and familiar objects can. Rather, it is only after God has been
inferred, by reason, that the suggestion relation which makes God (again, in a rather
attenuated sense) a mediate object of sensory perception can get off the ground. This
is in line with Philonous’s claim that Hylas can mediately perceive Julius Caesar by
perceiving a painting or statue only if he has some prior knowledge of Caesar (DHP,
203-204).
The competing theory of vision Berkeley has in mind when he speaks of ‘necessary
connexion’ (in the narrow sense) says that the mediate objects of vision are suggested
by the immediate objects due to a causal connection between them. In speaking of
‘geometrical inference,’ on the other hand, Berkeley refers to the Cartesian theory which
he had refuted at length in the New Theory of Vision.
4
On this theory, it is mathematical
computation rather than causal inference which takes us from the suggesting idea to the
suggested idea. In our discussion of inference, below (x5.4), the difference between
these two types of inference will become important. Here it suffices to say that in both
3. Berkeley is, of course, careful to distinguish this sort of vision of God from Malebranchean vision
in God (PHK,x148; DHP, 213-215). The contrast is explicit in Alciphron’s summary of the conclusion
to which Euphranor has driven him by means of the divine language argument:
I was aware, indeed, of a certain metaphysical hypothesis, of our seeing all things in God
by the union of the human soul with the intelligible substance of the Deity, which neither
I, nor any one else could make sense of. But I never imagined it could be pretended, that
we saw God with our fleshly eyes, as plain as we see any human person whatsoever, and
that he daily speaks to our senses in a manifest and clear dialect (Alc,x4.14).
4. On the Cartesian targets of NTV, see Atherton 1990, Part 1.
168
of these cases the relation would have to be discovered by reason before suggestion can
occur. In the remainder of this section, I will follow Alciphron in lumping these together
and calling them cases of ‘deduction.’
We do not need to say very much about the second relation, similitude. As we have
already discussed (x3.1), Berkeley holds that the mind has a primitive capacity to engage
in acts of comparison, and by these acts it finds some ideas to be similar and others not.
The third and most important relation is the one Alciphron calls “experience, custom,
and habit.” This relation is the most important since Berkeley classifies our knowledge
of both human language and the language of nature under this heading.
In the Theory of Vision Vindicated, Berkeley calls this category ‘arbitrary institu-
tion’ (TVV,x14), and this may suggest that the connections between ideas in this cat-
egory must be instituted by someone. This, however, cannot be Berkeley’s view, since
Berkeley recognizes that the conventions of language are not traceable to explicit acts
of stipulation or institution, and he takes the suggestion relations involved in language
to fall under this heading. Thus it is incorrect to say, as Samuel Rickless does, that the
association of ideas in this category traces to a ‘stipulation’ or ‘act of will’ (Rickless
2013, 46). This is sometimes the case: for instance, the connection between visual ideas
and the tactile ideas they suggest is arbitrarily instituted by God (PHK,xx30, 36; DHP,
231), and this is no doubt the reason why Berkeley uses the term ‘arbitrary institution’
to refer to this category in TVV. However, Berkeley’s ultimate view is that, in this kind
of case, “one idea is qualified to suggest another, merely by being often perceived with
it” (TVV,x68; cf. NTV,x25; DHP, 204). It may or may not be the case that some agent
intentionally brought it about that such ideas were often perceived together. In vision,
this is the case: God intentionally associates visual ideas with tactile ones. However,
this intention of God’s is not directly responsible for the suggestion relation. Rather,
169
God’s intentional activity brings about the constant conjunction of the ideas, and the
constant conjunction results in the suggestion relation obtaining.
There are, then, three phenomena which can lead to one idea suggesting another:
deduction, similitude, and experience. The next question is how it comes about that, of
two ideas related in one of these ways, one comes to suggest the other. Here Rickless
holds that “The imagination is responsible for the mental act of associating the ideas
that are [so] connected” and, once this has been done, the faculty of imagination sees
to it that “the thought of P simply triggers the thought of Q, without the employment of
actual ratiocination” (Rickless 2013, 47). In Rickless’s view, P cannot suggest Q to me
unless I have previously performed an act of association.
This interpretation is rendered quite implausible by Berkeley’s use of words like
‘habit’ and ‘custom’ in connection with suggestion. One does not form a habit by de-
ciding to, or by performing some special act, distinct from the act which is to become
habitual. Rather, one forms a habit by performing a certain action repeatedly. This re-
sults in that very same action becoming habitual. As Euphranor says, speaking of both
English and the language of vision, “there must be time and experience, by repeated
acts, to acquire a habit of knowing the connexion between sign and things signified; that
is to say, of understanding the language, whether of the eyes or of the ears” (Alc,x4.11,
emphasis added). If I repeatedly pass from having idea P to having idea Q, I will form
the habit of moving from P to Q. Why exactly I passed from P to Q in the initial cases is
unimportant; however I did it, if I do it enough times I will form a habit. The stronger
the habit, the more the process becomes (nearly) instantaneous and involuntary.
In the ‘arbitrary institution’ case, one develops the habit by environmental condition-
ing, and this contrasts with the likeness and deduction cases. Comparison and deduction
are general purpose psychological mechanisms by which we tend to pass from one idea
170
to another, in our own private thinking. They are, for the most part, under voluntary
control. But sensory perceptions are not subject to voluntary control (PHK,x29). Thus
in sensory perception we often pass involuntarily from one idea to another.
5
Since there
are all sorts of regularities in the ideas which succeed one another in our sensory per-
ception, we are conditioned to certain expectations – which is to say, we are conditioned
to imagine the thing which is going to come next.
This interpretation is confirmed by the independent evidence which we have already
seen in favor of the claim that just this type of conditioning plays a vital role in Berke-
ley’s account of language learning (x3.2, above). Linguistic competence, for Berkeley,
involves suggestion (NTV,x51; DHP, 174). It is due to our conditioning that words
suggest to us the things they do.
We are now prepared to return to Berkeley’s distinction between two ways of learn-
ing language. The person who learns language ‘by practice’ has the same sort of com-
plex, external conditioning which forges the suggestion relations for children learning
their first language. The person who learns the language ‘by rule’ has some explicit
rules in mind, which she follows consciously and intentionally to produce the correct
linguistic behavior. By doing this repeatedly, she will forge the same suggestion rela-
tions as the one who learns ‘by practice,’ and it is then that she has become fluent (cf.
Roberts 2007, 155n112).
6
What Berkeley calls ‘suggestion’ is habitual movement from thinking about one
thing to thinking about another. These suggestion relations are instances of implicit rule-
following: Berkeley’s theory of vision is a theory about the rules by means of which
5. This important contast between the ‘arbitrary institution’ case and the other two is also emphasized
by Rickless 2013, 46.
6. This is, of course, simplified, since it is unlikely that anyone would ever become truly fluent in a
natural language entirely ‘by rule’ (e.g. by reading dictionaries and grammar manuals) without any of the
other sorts of conditioning.
171
we see distance.
7
Berkeley says little about other forms of rule-following. His most
extended discussion is the treatment of the word ‘reward’ in the Manuscript Introduction
(again, seex3.2, above).
8
On the basis of that discussion, we are justified in supposing
that other sorts of rules can be treated analogously. Implicit rule-following is habitual
action.
In order to engage in explicit rule-following, one must have explicit understanding
of the rule. In order to have such understanding, one must have already acquired rules
for the proper use of the symbolism in which the rule is formulated. Hence one must
have learned the use of some signs (not necessarily amounting to all of what we would
ordinarily call a language) ‘by practice’ before one can begin to learn others ‘by rule.’
Some implicit rule-following must thus be prior to any explicit rule-following.
9
This completes my interpretation of Berkeley’s understanding of rule-following. The
remainder of this section is devoted to addressing a Wittgensteinian philosophical worry:
what distinguishes the rule which is being followed from the indefinitely many distinct
rules which are complied with (Wittgenstein [1958b] 2009, 224-228, 260-263; 1953,
xx1.143-155, 185-190)? That is, in Kripke’s terms, how do I know that I am adding and
not quadding (Kripke 1982)? Even more radically, what fact about me, or the world, can
possibly make it the case that I am adding and not quadding? If these questions cannot
be answered, radical semantic indeterminacy will result (Quine 1957; 1960,x16). We
7. Berkeley uses the word ‘rule’ only in talking about his opponents’ theories (NTV,x40; TVV,x64),
but it is clear that what he is trying to do is to replace these incorrect rules with the correct ones.
8. Berkeley’s discussion of moral rules in Passive Obedience and the early dialogues of Alciphron
could perhaps be brought to bear here, but moral rules are very different in kind from the rules with which
we are presently concerned, and in any event the nature of rule-following is not addressed in any detail in
these texts.
9. This fact was widely remarked upon in 20th century work in the philosophy of logic and language.
See, e.g., Quine (1935) 1976, 103-106; Wittgenstein 1953,xx1.84-87; Quine (1954) 1976, 115; Dummett
(1973) 1978, 217.
172
must therefore proceed to ask, what is the best that can be said on Berkeley’s behalf in
response to this question?
In light of the various other parallels between Berkeley and Wittgenstein, perhaps
the first point that bears emphasizing here is that Berkeley shows no evidence of having
had the slightest inkling of this sort of problem. This is due to a difference of attitudes in
metaphysics about which things may be taken as primitive and which must be ‘reduced,’
and a closely related difference in epistemology about which notions are intelligible in
themselves and which must be explained. Berkeley, along with many other early modern
philosophers, takes agents and their actions – including actions taken in obedience to
rules – to be basic, both in the order of knowledge and in the order of being. Thus
Berkeley would simply be untroubled by his inability to give an account of the sort
whose absence is lamented by Wittgenstein and Quine.
Understanding this difference of background assumptions will help to clarify a num-
ber of issues. First, it will help us get clearer on exactly how far Berkeley anticipates
20th century philosophers such as Wittgenstein and Quine and where the resemblance
ends. Second, it may suggest strategies whereby we ourselves can avoid the difficulties
raised by Wittgenstein and Quine. Finally, the results of this discussion will allow us to
understand in more detail exactly how the rules of language come to be instilled in the
members of a community.
Let us begin by setting the problem out in its Quinean form. Quine famously frames
the issue as one which arises in attempting to translate one language to another, or to
determine which word in one language is used in the same manner (means the same) as
a word in another. We are tempted to think that words which can be defined ostensively
pose an easier case than others, but even here we are in trouble, for no possible pattern
173
of ostension can distinguish between rabbits, the property of rabbithood, rabbit-stages,
and undetached rabbit parts (Quine 1957, 5-7; 1960, 51-52).
It gets worse. The reason we cannot distinguish these cases is that they share the
same ‘stimulus meaning’ (Quine 1960, 32-33) – that is (to a first approximation), they
are prompted by the same external stimuli. However, as Wittgenstein might point out,
we cannot be sure, on the basis of finitely many cases, that any two words are even
stimulus synonymous, for it is always possible that the rules being followed diverge in
certain as yet unobserved cases. Wittgenstein worries that this will apply even in my
own case: I may not know which rule I am following (Wittgenstein 1953,xx1.186-188;
[1958b] 2009, 261-262).
Berkeley is not in a position simply to dismiss these worries out of hand. On the
interpretation I have defended, Berkeley is (implicitly) committed to agreeing with
Wittgenstein that I may not be able to articulate the particular rule I am following. Fur-
thermore, he is committed to agreeing that, even if I am able to articulate the rule, I
must do so in language, and hence a question arises about what rules I am following in
interpreting the formula which states the rule (Wittgenstein 1953,xx1.189-190). As a
result, the ability to state the rule we are following is no solution to the problem at issue.
Without committing to any particular interpretation of Wittgenstein or Quine, we
can identify two distinct problems here, one epistemological, and one metaphysical.
The epistemological problem is, how can anyone ever know what rule any agent is
following? The metaphysical problem is, what makes it the case that a particular agent
is following one rule rather than another?
Berkeley, clearly agrees with both Wittgenstein and Quine that any attempt to de-
termine which rule I am following must involve more than simple introspection. As a
result, the epistemological problem certainly does arise for him. The New Theory of
174
Vision is, after all, an empirical (not merely introspective) investigation of the rules we
humans follow in interpreting visual stimuli. Berkeley’s aim is to discover the rules ac-
cording to which the proper and immediate objects of vision – visual shapes and colors
– suggest the mediate objects – tangible ideas, and in particular tangible “distance, mag-
nitude, and situation” (NTV,x1). He does so by identifying the various data available as
inputs to the process (the proper objects of vision) and correlating them with the outputs
(perceptions of tangible distance, magnitude, and situation), forming empirical hypothe-
ses about the rules connecting the inputs to the outputs. Because we cannot immediately
and infallibly grasp the rules in introspection, the hypotheses we form about the rules we
are following suffer from a general problem effecting all of empirical science, namely,
the underdetermination of theory by observation.
Berkeley’s contemporaries were aware of this sort of difficulty. Underdetermination
is, after all, precisely what made Newton’s famous ‘rules of philosophizing’ necessary
(Newton 1729, 2:202-205). Newton’s rules are meant to be of use in determining which
among the many empirically adequate hypotheses ought to be adopted.
In De Motu, dealing with mechanics, Berkeley’s solution to problems of underdeter-
mination appears to be pragmatic and conventional. In these cases, he is willing to allow
metaphysical underdetermination: there is no fact of the matter about whether Newton
or Torricelli is correct (DM,x67). However, when speaking of minds, we are engaged,
not in mechanics, but in metaphysics, and hence we are dealing with genuine efficient
causation, about which Berkeley certainly wants to be a realist. As a result, there ought
to be some objective metaphysical facts in play here; the question is, what are these
facts, and where do they come from?
According to Berkeley, the only intelligible notion of efficient causation is the inten-
tional action of an agent (see below,x6.4). Only agents genuinely make things happen.
175
As a result, if any regularity is to obtain in a manner which is something other than mere
coincidence, the explanation for the obtaining of this regularity must be located in some
agent’s will. In Siris, Berkeley responds to an objection to this view:
It must be owned, we are not conscious of the systole and diastole of the
heart, or the motion of the diaphragm. It may not nevertheless be thence
inferred that unknowing nature can act regularly, as well as ourselves. The
true inference is that the self-thinking individual, or human person, is not
the real author of those natural motions. And, in fact, no man blames him-
self if they are wrong, or values himself if they are right. The same may be
said of the fingers of a musician, which some object to be moved by habit
which understands not; it being evident that what is done by rule must pro-
ceed from something that understands the rule; therefore, if not from the
musician himself, from some other active intelligence, the same perhaps
which governs bees and spiders, and moves the limbs of those who walk in
their sleep (Siris,x257, emphasis added).
10
Berkeley here clearly claims that genuine agency is required for genuine (i.e. non-
accidental) regularity.
11
More specifically, he claims that such regularity can only be
the result of an agent following a rule. Thus if there are movements of our bodies which
we do not produce by our own agency, then they must be produced by some other agent.
The use of the word ‘understands’ here is troubling given Berkeley’s recognition of
the possibility of implicit rule-following, and the difficulty of discovering what rules we
10. The status of ‘bees and spiders,’ as well as other non-human animals, will be examined inx8.6.
11. This provides an answer to the objection, raised by Brook 1973, 27, that Berkeley’s account of
suggestion contradicts his claims about the passivity of ideas. The movement from one idea to another
in suggestion is ‘done by rule’ hence it must be done by some mind; the suggesting idea does not simply
(efficiently) cause the second idea to spring up.
176
are following. However, this is made significantly less troubling by the explicit defini-
tion of ‘understanding’ Berkeley provides a few pages earlier: “We know a thing when
we understand it; and we understand it when we can interpret or tell what it signifies”
(Siris,x253). This ‘interpretation’ is explicitly connected with prediction: “According
to Socrates, you and the cook may judge of a dish on the table equally well, but while
the dish is making, the cook can better foretell what will ensue from this or that manner
of composing it” (x253), and for this reason the cook is said to ‘understand’ the rules of
cooking: not because he can state those rules, but because he foresees what outcomes
will follow according to them. In this case, the rules we are talking about are not the
rules the cook himself follows, but the rules God follows whereby the cook’s actions
have predictable outcomes (cf.x160). However, this is the same kind of understanding
which, according to Berkeley, is required for non-accidental regularities. In order for a
regularity to obtain non-accidentally, it must obtain as a result of the action of an agent
who is able to see, in each circumstance, what the rule requires of her. It does not require
the ability to state a rule.
12
One of the central conclusions Berkeley draws from the claim that only rule-
following can underlie genuine regularity is that the laws of nature can be nothing other
than the rules followed by God (PHK,xx30, 36; DHP, 231; cf. Boyle [1686] 1991, 180-
185). What this shows is that, for Berkeley, rule-following plays the same metaphysical
role which in many other systems, both before and after this period, is played by the
dispositions of (inanimate) objects, or by ‘governing’ laws of nature. It is a fundamental
12. Incidentally, Wittgenstein quite explicitly takes the notion of ‘understanding a rule’ in this way
(Wittgenstein 1953,xx1.150-155).
177
metaphysical posit meant to explain the uniformity of nature, and hence the rationality
of induction.
13
This provides an answer, of sorts, to the metaphysical worry: in order to explain the
rationality of induction, we must simply take it as given that there is a metaphysical fact
of the matter, even if it is unknowable. If, however, we do not wish to join Berkeley
in endorsing irreducible agent causation, we will have a rather more complicated story
to tell. The laws of nature, whatever their origins, will presumably ground facts about
the dispositions of agents in counterfactual scenarios, which will give us significantly
more determinacy than we would otherwise have. However, Quine’s argument works
precisely by noting that no physical dispositions could possibly distinguish the differ-
ences of meaning in question, and so tell us which rule we are following. Some headway
can be made by allowing (as Berkeley does) appeals to private mental states, but given
Berkeley’s extraordinarily sparse theory of mental contents this will surely not allow us
to draw all of the fine distinctions Quine mentions. Here there is a genuine question
about whether the different formulas actually specify different rules, since there is no
possible circumstance in which they differ in their guidance. This, however, is a ques-
tion about the individuation of rules, and not a question about whether there is a unique
rule which the agent follows.
The epistemological worry is more difficult to defuse. Given finite data, how can we
be justified in believing, of ourselves or others, that we follow one particular rule rather
than another? As has already been indicated Berkeley sees this as a matter for empirical
13. Cf. Reid (1788) 2010, 38, 251. For discussion of Reid’s view, see Pearce 2012,x3. For an argument
of this sort in favor of ‘governing’ (anti-Humean) laws of nature, see Armstrong 1983,x4.5. It is important
to note, and Armstrong is quite clear on this point, that the contemplated strategy will not work as a way of
persuading an (imagined) inductive skeptic of the rationality of induction; rather, it assumes the rationality
of induction as a datum to be explained, and argues that one theory (Armstrong’s own ‘governing’ theory
of laws) is preferable to another (the so-called ‘Humean’ theory of laws), since the former can explain
this datum while the latter cannot.
178
inquiry; it cannot always be settled reliably by introspection. Empirical inquiry is a
messy business, and our best theories are always prone to revision. However, if we
are not to succumb to inductive skepticism, we must accept that we do sometimes have
reasonable beliefs about future events, and this must be based on some rules we thought
were followed in the past.
In any event, as will become clear in chapter 9, Berkeley’s anti-skeptical project
does not require that we have firm, explicit, precise knowledge of these sorts of rules. It
suffices that we know them ‘by practice’ – that is, that we know how to follow them. By
this means, we gain knowledge of the order of nature; that is, of the various ideas which
we will get if we take various courses of action. Having explicit beliefs about what
the rules are can be helpful here, but it can be helpful even if the beliefs are relatively
tentative and imprecise. What is important is that there are rules, and we successfully
follow them.
5.2 Berkeley’s Internalist Requirement
The question of what an agent must know in order to follow a rule has, in recent times,
been most carefully explored in connection with a special class of rules known as epis-
temic principles. As the matter is usually formulated, an epistemic principle is a univer-
sal generalization of the form:
for any subject S and proposition p, if condition C is satisfied, then p is
justified forS (see Van Cleve 1979, 75).
Thus, for instance, Descartes is said to endorse the epistemic principle, “For allP , if I
clearly and distinctly perceive thatP , then I am certain thatP ” (66).
179
The debate between internalists and externalists in epistemology is a debate about
the status of these principles. Internalists assert, and externalists deny, that the subject
must have some sort of access to something about the principle in order to gain knowl-
edge (justification, certainty, etc.) by means of it. There are many different internalist
theses, which differ from one another along two basic dimensions: the sort of access the
subject must have (for instance, actual knowledge, potential knowledge, introspective
awareness, etc.), and to what the subject must have access. For each of these internalist
theses, there is a corresponding thesis about rule-following. Rules, like epistemic prin-
ciples, can be formulated as universal generalizations. Rules are generalizations which
have, or purport to have, normative force. To follow a rule is to instantiate that gener-
alization because of its actual or perceived normative force. Thus, for instance, it is a
rule of chess that bishops move only diagonally. This generalization is normative, and
an agent who is following (rather than merely complying with) the rules of chess re-
frains from moving bishops in any other fashion because of the normative force of this
generalization.
For present purposes, we will be concerned with two specific internalist theses about
rules:
Rule Internalism (RI) In order to follow a rule, the subject must first be capable of
explicit knowledge of the rule.
14
Condition Internalism (CI) In order to follow a rule, the subject must first be capable
of recognizing the condition of the rule’s applicability.
15
14. This corresponds to internalism type 2bii in the taxonomy of Van Cleve 2003, 45-46.
15. This corresponds to Van Cleve’s internalism type 1bii.
180
By ‘the condition of the rule’s applicability,’ I mean the situation in which the rule
instructs the agent to do something. This is to be distinguished from the condition
in which the rule is in force. The condition of the rule’s applicability depends on its
content, while the condition in which it is in force depends on the sort of normative force
the rule possesses. It is a rule of chess that whenever one checks the opponent’s king
one must announce this fact. The rule is applicable whenever one checks the opponent’s
king; it is in force whenever one is playing chess.
16
Both (RI) and (CI) are theses about the preconditions for rule-following. They claim
that one cannot genuinely follow (as opposed to merely comply with) a rule unless a
certain condition is first satisfied. The question at issue is, what must be going on in
the agent in order for the normative force of the generalization to play the right kind of
explanatory role in the agent’s actions? Internalists of various stripes claim that, in order
for this to happen, some aspect of the rule must have some kind of epistemic status for
the agent.
This debate is important since, as I have been arguing, it is a consequence of Berke-
ley’s theory that the rules we follow can expand our possibilities for mental representa-
tion. In particular, all general thought requires rule-following (see above,x3.1). Since
all rules must be general (must specify some repeatable condition and some repeatable
action to perform when the condition occurs), one cannot think about rules unless one is
already following some rules (cf. Wittgenstein 1953,xx1.84-87; Sellars 1954, 204-206).
It is precisely for this reason that Berkeley must reject (RI): until one begins to follow
some rules, one is not capable of so much as (explicitly) thinking about any rules. Thus
it must be possible implicitly to regard a generalization as having normative force, in
16. In the terminology of Schroeder 2014, the condition of the rule’s applicability is the condition in
which one is in the rule’s scope, while the condition in which the rule is in force is the condition in which
one is in the rule’s jurisdiction.
181
such a way that its perceived normative force helps to explain one’s actions, without
even being capable of explicitly formulating that rule.
This is closely analogous to problems that plague certain varieties of epistemic in-
ternalism: in order to gain knowledge, one must follow some justification rule, but
(according to these forms of internalism) one cannot follow a justification rule unless
one already has some knowledge (Van Cleve 1979). Accordingly (RI) is to be rejected.
I have already argued at some length that Berkeley does indeed reject it.
I will now argue that Berkeley accepts (CI), and identify his reasons for accepting
it.
17
This redeems the promissory note fromx2.3, above, by showing that Berkeley
has grounds for rejecting the causal theories of representation there discussed. (CI)
generates substantive constraints on the kinds of rules we can follow, which in turn
generate constraints on the possible range of linguistic and mental representation which,
as I will show inx6.7, below, plays an important role in Berkeley’s argument for the
claim that Locke’s talk of ‘material substratum’ is meaningless.
The truth of (CI) is a basic presupposition of Berkeley’s argument in the New Theory
of Vision. In the early sections of that work, Berkeley argues from the fact that we are
unaware of the lines and angles involved in Cartesian geometric optics to the conclusion
that we do not perceive distance by means of these lines and angles (NTV,xx9-13; cf.
Alc,x4.8). The rules involved in Cartesian geometric optics tell one to form certain
judgments of distance about the objects one sees depending on the angles the objects
make to the eye. What Berkeley is arguing is that these cannot possibly be the rules we
follow in forming judgments of distance by vision because we are capable of forming
those judgments correctly regardless of whether we are capable of recognizing the con-
ditions of the applicability of these rules. Berkeley does not dispute that we comply with
17. The analogous position in epistemology to Berkeley’s rejection of (RI) and acceptance of (CI) is the
position William Alston dubbed ‘internalist externalism’ (Alston 1988).
182
the Cartesian rules; indeed, he explicitly concedes that we do (NTV,x78; TVV,xx31-
32, 37, 43). Berkeley’s claim is, instead, that since we have no access to the conditions
for the applicability of these rules, they cannot be the rules we follow.
Although (CI) is presupposed by Berkeley, it is not difficult to construct his (implicit)
argument on its behalf, based on the account of suggestion developed in the previous
section. Rule-following, for Berkeley, is a learned behavior. In order for it to be learned,
one must have the conscious experience of passing from the condition of the rule to the
commanded action on many occasions. But one cannot be conscious of this transition
unless one is conscious of the condition. Hence the acquisition of the rule presupposes
consciousness of the obtaining of the condition.
This line of argument does not tend to support (RI) for, as we have seen, there are
any number of mechanisms by which one can pass from one idea to another without
explicit awareness of a rule. Once an agent has done this enough times, she develops a
habit of acting in a certain way. This habitual action can be a form of rule-following.
Now, according to the Wisconsin semanticists, the causal theorists discussed inx2.3,
some of the rules we follow which make mental representation possible may be learned
in this way. However, they cannot all be of this sort, for many of the rules represent to us
conditions of which we have no independent grasp. That is, according to the Wisconsin
semanticists, my perceptual representation of a white object in front of me consists in
my being in a state of some typeS. I cannot learn to follow the rule token state-typeS
in the presence of white objects, for I have no way of recognizing the presence of white
objects which is independent of the correlation between the presence of white objects
and my being in a state of type S. However, according to the Wisconsin semanticists,
some sub-personal system within me follows that rule. Some rules may be learned by
the individual organism but, according to the Wisconsin semanticists, the most basic
183
rules must be evolved by the species. As a result of such evolution, causal laws ensure
the reliable correlation between the internal state of the organism and the state of its
external environment. The internal state thereby comes to represent the environmental
factor with which it is connected.
As we have seen, Berkeley agrees that an idea may suggest its (efficient) cause.
However, we must remember how, for Berkeley, this is possible. An idea comes to sug-
gest its cause only when we have repeatedly inferred the cause from the idea, something
that is always necessarily a conscious process (NTV,x19). The rule can therefore be ac-
quired only after one has the conceptual sophistication to perform causal inferences and
some independent means of having thoughts about the cause. As a result, the possibility
of ideas suggesting their causes, as it is understood by Berkeley, does not contradict
(CI).
Since Berkeley holds that only agents can possibly follow rules, the type of causal
representation involved in Wisconsin semantics is, on his view, impossible. However,
first, Berkeley’s views on this subject are now widely rejected, and we have not uncov-
ered any independent argument for those views. As we will see in chapter 6, Berkeley
does have an argument for his position, but it is not an independent argument: it relies
on certain constraints on mental and linguistic representation which are derived from
(CI). As a result, (CI) cannot be derived from Berkeley’s rejection of causal activity by
non-agents. Second, it seems that, even granting all of Berkeley’s assumptions, causal
representation of this sort could be saved from this objection by resorting to occasion-
alism and holding that God follows the rule of causing me to tokenS in the presence of
white objects.
In defense of (CI) and against the Wisconsin semanticists, I suspect Berkeley would
pose the following question: how is it that the causal correlations in question bring it
184
about that I represent, or have knowledge of, the causes of my ideas? That is, if it
is not I who follow the rules, then how am I the one who is in the representational
state?
18
This amounts, in the end, to no more than pumping our internalist intuitions,
but it must be acknowledged that, in the end, philosophy must often come down to the
pumping of intuitions. Furthermore, the intuition does seem to have some force, even in
our 21st century context, and whatever force it has supports only (CI) and not the self-
defeating (RI). That I should be the one who ensures that I perform actionA whenever
circumstanceC arises does seem to require that I recognizeC and do A as a result of
that recognition; it does not seem to require that I be able to state the rule do A in C.
If mental or linguistic representation requires rule-following, then it seems that the one
following the rules is the one doing the representing. If, then, I am to have mental or
linguistic representations, I must follow certain rules, and I can follow these rules only if
I have some independent grip on the conditions of their applicability. This is Berkeley’s
Internalist Requirement.
5.3 The Conventional Rules of Language
The rules of language are rules governing a particular, public, social practice in which
we are all immersed. Berkeley’s discussion in Alciphron VII aims to make sense of this
“daily practice” (Alc,x7.8). Because language is communal in this way it is governed
18. William Lycan makes a similar point in a rather different context. According to Lycan, conscious-
ness is constituted by the operation of an ‘internal monitor’ which has the function for the thinker of
monitoring its internal states. On his view, it is (at least conceptually) possible that there could be sub-
personal systems which are conscious. Thus whether I am conscious of a monitored state depends on
whether the monitor is performing its monitoring function for me. If it is monitoring my internal states
for one of my sub-personal systems (or, in a more fanciful example given by Lycan, monitoring my inter-
nal states for CNN), then it will not follow that I am conscious of those internal states (Lycan 1995, 5-6).
What I am suggesting, on Berkeley’s behalf, is that in the same way a state functions as a sign for me only
if I follow some rule for its use, and this requires that I am aware of the condition for its application. It is
not sufficient that one of my sub-personal systems (if I have such systems), or God, should do so.
185
by certain norms. That is, despite the fact that the rules of language are in one sense
radically arbitrary (DHP, 247), it is nevertheless possible to follow the wrong rule (PHK,
x108). “Common custom is the standard of propriety in language” (DHP, 216).
19
If
language is to achieve its ends, the members of the speaker community must follow (at
least roughly) the same rules. Furthermore, we must be aware that this state of affairs
obtains; that is, we must follow the relevant rules with the expectation that others will
likewise follow them.
David Lewis helpfully compares the kind of coordination needed here to the decision
of which side of the street to drive on (D. Lewis 1969, 6, 40-41, 44-45, 58-59, 65-66;
1975, 6). Neither side of the street is intrinsically more eligible to be driven on; the
decision is radically arbitrary. However, if I am to be able to travel quickly and safely
to my destination, I must follow the rule of driving on a particular side and confidently
expect others to follow the same rule, and my expectation must be satisfied.
That coordination of this sort is essential to language was already recognized before
Berkeley (Arnauld and Nicole [1662] 1996, 58-63; EHU,xx3.2.4, 3.9.4, 3.9.8). For the
proponents of the Theory of Meanings, the coordination involved was simply that the
speaking of the word should bring it about that the hearer has the same mental content
as the speaker. For Berkeley, the rules and aims are more complicated, but the basic
point still stands. If, for instance, I am to use ‘reward’-talk to motivate you to perform a
certain action, I must presuppose that you have internalized certain rules for the use of
the word ‘reward’. From the fact that language seems to work, we can infer that we do
indeed presuppose something of this sort, and that this presupposition generally leads
to correct behavioral expectations. We must here answer three questions: (1) How are
these behavioral patterns reliably brought about in speaker communities? (2) What kind
19. Hylas is the speaker here, but this is a concession he makes at Philonous’s prompting.
186
of knowledge, belief, or presupposition about the rules followed by others is required?
and, (3) How do we gain this knowledge, belief, or presupposition?
The first question has already been answered. Rules of this sort are acquired by
environmental conditioning. An individual who lives in a society where the conventions
obtain is likely to receive such conditioning, in precisely the way Berkeley describes in
his account of the acquisition of ‘reward’-talk (see above,x3.2).
The second and third questions pose greater difficulties. We have already seen that
the rules of language cannot presuppose the ability to articulate rules, and that, for
Berkeley, one who is incapable of articulating the rules is incapable of thinking about
them. Furthermore, we have seen that there is some cause for pessimism about our abil-
ity to know exactly which rules are followed. This is a matter for empirical investigation,
and hypotheses regarding it are always to be regarded as rather provisional. It simply is
not the case that every language user knows (explicitly) which conventions obtain in her
society.
20
I have been saying that some sort of knowledge, belief, or presupposition is required.
Although Berkeley never addresses this question explicitly, what he needs to say, if his
theory is to work, is that presupposition is the correct option. The particular kind of
presupposition which will fit most neatly into Berkeley’s theory is what we might call
practical reliance. When we plan our actions, we make a variety of, often implicit, and
often complicated, assumptions about the behavior of others. That is, I have beliefs
of the form ‘if I do this, she will do that,’ which need not be backed by an explicit
belief that the other is following a specific articulable rule (though, of course, it may be
backed by such a belief). This expectation can amount to another form of suggestion:
20. Lewis places quite strong knowledge requirements on parties to conventions. A variety of problems
stemming from these requirements are pointed out by Burge 1975. However, even Lewis recognizes that
the knowledge in question will not, in general, be explicit (D. Lewis 1975, 24-25).
187
the thought of the other in a given situation suggests to me the thought of her acting a
particular way. Since these are suggestion relations, they can arise from environmental
conditioning, just like the other suggestion relations involved in language acquisition.
Simple conventions like which side of the road to drive on are usually accompanied
by an explicit (dispositional) belief that the convention obtains. However, complex con-
ventions, like the syntactic rules of English, or the rules of polite dinner party conduct,
are often followed by people who are not capable of articulating them: one learns the
convention, as Berkeley says, by practice. Part of what one learns is to rely on others
(mostly) following (approximately) the same rules as oneself.
5.4 Inference Rules
In both De Motu and Alciphron, Berkeley repeatedly emphasizes that certain words (and
other signs) are meaningful because certain inferences can be drawn from them. The
paradigm case is the use of ‘force’ in Newtonian mechanics. A wide variety of other
cases, including the instances of religious language with which Berkeley is concerned
in Alciphron, are also explicated in this way. The rules of inference must therefore be
regarded as among the most important rules in Berkeley’s philosophy of language.
My aim in this section will be to examine Berkeley’s view of the rules of inference
and the relation of this view to his general theory of language. I will be especially
concerned with inferences involving operative language, since these have often been
found problematic.
Let us begin by considering the sorts of inferences which are, on Berkeley’s view,
permitted by the rules governing general ideas and general terms. When one reasons
with general ideas, one is using particular, maximally determinate ideas as signs to rep-
resent classes of resembling ideas. One does this by following conventional rules. By
188
following these rules in, for instance, constructing geometrical demonstrations about
triangles, one “may consider a figure merely as triangular, without attending to the par-
ticular qualities of the angles or relations of the sides” (PHK, Introx16 [1734 ed.]); this
is the sort of abstraction Berkeley allows. What this amounts to, however, is simply
that “neither the right angle [of the particular triangle I consider], nor the equality, nor
determinate length of the sides, are at all concerned in the demonstration” (Introx16).
I argued above (x3.1) that Berkeley must accept in humans a primitive capacity not
only to recognize resemblance and difference between ideas, but to distinguish between
different respects of resemblance and difference. This is precisely what is involved in
this sort of reasoning with the general idea of triangle. The rules for the use of the
sign tell us in what respects other objects must resemble this idea in order to count as
members of the signified class. It is in these respects that ‘agreement’ must be found if
something is to be predicated of triangles universally. Thus, for instance, the rules re-
quire that every triangle be a two-dimensional figure; this is one of the required respects
of resemblance. Suppose one uses some non-triangular particular idea (say, a particular
square) to be, by signification, the general idea of a two-dimensional figure. One can
recognize the agreement between these two general ideas by noticing that, in determin-
ing whether this particular triangle was a two-dimensional figure, one used only the
respects of resemblance shared by all triangles.
Now let us consider the word ‘triangle.’ The rule for the use of the word ‘triangle’
permits its application to any “plane surface comprehended by three right lines” (Intro
x18). In the simple case of drawing the immediate conclusion ‘this is a triangle,’ one
simply applies the rule, which is, of course, consistent with being unable to state it. One
also, however, knows how to use the phrase ‘two-dimensional figure,’ and can follow the
rule for that phrase to conclude (ostending the same object) ‘this is a two-dimensional
189
figure.’ From these two sentences, using only the rules for general terms, one may
conclude, ‘the triangle is a two-dimensional figure’ (where ‘the triangle’ refers to the
particular triangle one earlier ostended).
One may wish to universalize and come to the conclusion, ‘every triangle is a two-
dimensional figure.’ One could presumably do this by induction on particular instances,
but in fact our certainty that all triangles are two-dimensional figures is much greater
than the level of confidence such induction could (justifiably) produce. An actual (de-
ductive) demonstration of this proposition may work in either of two ways. First, it may
work in essentially the same fashion as the parallel inference involving general ideas,
discussed above. That is, one may attend to the process of identifying something as a
triangle, and notice that nothing could, by this process, be identified as a triangle unless
that thing were a two-dimensional figure. This, however, involves meta-linguistic aware-
ness of the rules of a sort that has been determined to be problematic, at least if what we
are after is certainty (and surely we are certain that every triangle is a two-dimensional
figure if we are certain of anything at all).
There is a better account available. According to this account, the rule for the appli-
cation of ‘triangle’ to perceived objects is only one of the conventional rules associated
with that word. Other rules connect it directly to other words, including the phrase ‘two-
dimensional figure.’ There are, then, at least three distinct rules:
21
a rule specifying that
all and only objects of a certain sort may be described as ‘triangles,’ a rule specifying
that all and only objects of a certain sort may be described as ‘two-dimensional figures,’
21. In fact, there will be a great many more, as many as the conceptual connections between distinct
words. I do not think this is a problem. First, as I will argue in chapter 7, Berkeley subscribes to a radical
holism about meaning, so he should not hold that we learn words one at a time by learning, all at once,
the full set of rules associated with them. Second, I have argued that explicit awareness of the rules is not
required. Third, this approach does not place any higher demand on our cognitive capacities than does
the alternative. Even if these connections were empirical beliefs, we would still be committed to the view
that we have an enormous number of (dispositional) beliefs about conceptual connections.
190
and a rule specifying that everything described as a ‘triangle’ may be described as a
‘two-dimensional figure’. These rules together make the word ‘triangle’ meaningful.
This account moves any skeptical worries we might have back up to the meta-
linguistic level where they belong: we have no doubt that all triangles are two-
dimensional figures, for we are following a rule authorizing precisely this assertion (a
rule, that is, providing that any triangle, or group of triangles, including the group of all
triangles, may be described as ‘two-dimensional figures’).
That Berkeley does in fact have rules of this sort in mind is confirmed by the analogy
he draws between natural language and arithmetic and algebra (Alc,x7.8). In the dis-
cussion of arithmetic in the Principles, Berkeley refers vaguely to “methods of finding
from the given figures or marks of the parts, what figures and how placed, are proper
to denote the whole or vice versa” which it is possible to ‘contrive’ once one has a nu-
meral system. He describes the process of following these rules as “computing in signs”
(PHK,x121). Similarly, in Alciphron, Euphranor says that in bookkeeping, money is
considered only at the beginning and end of a given ‘operation;’ in between, one works
directly with the symbols (Alc,x7.8). The same is true with natural language inference:
in the usual case, one simply manipulates the words according to rules which connect
one word or string of words to another.
The rules of algebra permit one to simplify expressions by (e.g.) substituting ‘12’
for ‘7 + 5’. Each of these expressions has rules connecting it to the world in certain
ways, like the separate rules for recognizing triangles and two-dimensional figures: one
can recognize groups of 12, 7, or 5 objects, and can combine a group of 7 objects with a
group of 5 objects to make one larger group. The rules tell us that this larger group will
be a group of 12 objects, for the rules say that ‘12’ is substitutable for ‘7+5’. There is,
191
thus, a strong analogy between the rules for simplifying expressions in algebra and the
rules of inference involving general terms in plain language.
Recognizing these sorts of special rules that directly authorize the substitution of
words without explicit consideration of the relationship between the different rules gov-
erning the two words thus preserves our certainty about the correctness of our applica-
tions in particular cases and also preserves the parallel Berkeley sees between natural
language and arithmetic and algebra. As indicated above, however, a skeptical worry
at the meta-linguistic level remains: we may worry that there could be a situation in
which our linguistic conventions give contradictory instructions. By way of illustration,
imagine that we had a rule specifying that whatever is described as a ‘plant’ may be
described as a ‘rock.’ Then, for any given plant, our rules would tell us that it was both
permissible to call it ‘rock’ (since it is called ‘plant’) and impermissible to call it ‘rock’
(since it fails the resemblance tests for rock-hood). In natural language, this does indeed
happen, and we either live with the situation (if the incoherence is not too glaring), or
revise our conventions (see below,x7.2). An issue of this general sort is behind Locke’s
famous query, “Whether a Bat be a Bird or no” (EHU,x3.11.7): our conventions may be
indeterminate, may not cover every case, or may in some cases even conflict with one
another.
Although such imperfect conventions serve us fine for practical purposes, in the sci-
ences we must eliminate such difficulties, as far as possible, by introducing stipulative
definitions (EHU,x3.11.12; cf. Arnauld and Nicole [1662] 1996, 60-63). Unlike ordi-
nary language, the use of stipulative definitions in scientific language does involve ex-
plicit, meta-linguistic beliefs: one learns the formal, scientific vocabulary ‘by rule,’ and
this makes one, justifiably, more confident that it has the formal properties one desires,
such as giving a consistent and unambiguous verdict on the applicability of the word
192
to every conceivable case. Of course, since the definitions are stipulated in ordinary
language, one can still imagine that there will be some hitherto unnoticed contradiction,
but this is not an unrealistic radical skeptical hypothesis; the search for unnoticed con-
tradictions or other problems in formal systems is part of mathematical and scientific
practice (cf. Bacon [1620] 2000,x1.59).
In addition to the general term rules, which parallel the rules for simplifying ex-
pressions in algebra, there is another class of inference rules, those having to do with
quantifiers and logical connectives. These sorts of structural or syntactic rules may be
compared with the inference rules related to ‘=’ in algebra. These rules permit us to
operate on the sentence as a whole. For instance, the rules governing ‘=’ tell us that
we may choose any arithmetic operation whatsoever and perform it on both of the ex-
pressions joined by ‘=’ to produce another correct (assertable) equation. They also tell
us that we may substitute the expression on one side for the expression on the other.
Analogous rules can be seen as governing the logical vocabulary, rules of the sort one
can find in any introductory logic text.
All of these inferences proceed by conventional rules: we adopt the convention of
using a certain sign system and manipulating it in a certain way. Furthermore, they
are arbitrary in the sense that there are indefinitely many equally eligible sign systems.
However, it does not follow that there are no reasons favoring one system over another
for, as Berkeley emphasizes, a sign system is a bit of technology which is useful to us for
getting around in the world. Some systems are superior to others, as Arabic numerals are
superior to tally marks (PHK,x121). Other systems are downright useless. In this way,
all sign systems – and hence for Berkeley all sciences, including those we customarily
regard as a priori, since all sciences are sign systems (Alc,x7.16) – must answer to the
‘tribunal of experience’ (cf. Quine 1951, 38). If we adopt bad conventions, which allow
193
us to derive such sentences as ‘0 = 1’ (while keeping enough of our actual rules the
same) our sign system will become useless, for it will ultimately tell us that anything
may be substituted for anything else, and will make empirical predictions which will not
be satisfied. Thus we may, in designing a sign system, adopt the convention of asserting
just anything, but a bad convention of this sort, one that does not help us navigate the
world of experience, may lead to the collapse of the system.
The collapse may not be radical. There is no reason why we should not be able to
modify the system as needed to render it more useful. Depending on the nature of the
system and the nature of its failing, this may involve a tacit adjustment by the native
speakers of a language through natural language change, or it may involve an explicit
stipulation. If an explicit stipulation is used, it may create a special scientific dialect of
the sign system, or it may, through wider adoption, alter the language itself. Natural
languages may well survive with ill-defined rules for inferences involving the logical
vocabulary just as they survive with ill-defined rules for general terms. In actual fact,
the English connectives ‘or’ and ‘if’ are arguably ambiguous, the former between inclu-
sive and exclusive interpretations, the latter between the various different conditionals
identified by philosophers.
22
We are now in a position to give at least the beginning of a Berkeleian response
to Geach’s objection to expressivism (introduced above,x3.2). Geach famously com-
plained that ‘anti-descriptive’ theories of the meaning of moral terms could not explain
the validity of the following inference:
If doing a thing is bad, getting your little brother to do it is bad.
Tormenting the cat is bad.
Ergo, getting your little brother to torment the cat is bad (Geach 1965, 463).
22. These issues will be discussed in greater depth inx7.2.4, below.
194
The theories to which Geach is objecting hold that such sentences as ‘tormenting the
cat is bad’ do not assert anything, or predicate some property, badness, of the action
tormenting the cat, but rather condemn that action, and hence such sentences do not
express propositions. Geach’s objection is that ‘bad’ can be used in complex sentences
such as the first premise of the argument. Uttering this sentence does not appear to be
an act of condemnation. Yet ‘bad’ must be used in the same sense here as in the rest of
the argument, or the argument would be invalid.
I have been arguing that, on Berkeley’s theory, the rules for inferences of this sort
should be understood as rules for ‘computing in signs,’ just like the rules of algebra.
The ‘operative’ rules which attach to ‘bad’ do not prevent it from functioning within the
sign system as a perfectly ordinary adjective. The difference between ‘bad’ and ‘red’
resides in how the words are ‘cashed out’ at the end of the ‘game.’ In other words,
one reasons with operative terms in exactly the way one reasons with terms that denote
sensory qualities; the difference between them is a difference in what it means to assent
to the conclusion. If one concludes that the desk is black, one thereby comes to expect to
have a black sensory experience when looking at the desk. If one concludes that getting
one’s little brother to torment the cat is bad, one adopts some sort of negative affective
or volitional attitude or stance to that action.
23
None of this changes the essential nature
of reasoning which is, for Berkeley, simply a matter of manipulating symbols according
to rules.
This also allows for a unified account of complex sentences: complex sentences are
simply ‘inference tickets,’ taking their meanings from the inference rules to which their
logical connectives attach them. These rules authorize one, under certain circumstances,
23. The details of Berkeley’s meta-ethics are far from clear; seex3.2, above.
195
to assert or deny the simple sentences from which the complex sentence in question is
constructed, as well as various other sentences.
Is it the case, on this theory, that sentences such as ‘all triangles are two-dimensional
figures’ and ‘if Jose is either a Republican or a Democrat, and he’s not a Republican,
then he’s a Democrat’ are true by convention? I have argued so far that our linguistic
conventions commit us to the assertability of such sentences, and that if it were found to
be practically infeasible to go on permitting the assertion of such sentences, this would
call for an alteration in our language. Furthermore, I have argued that it is central to
Berkeley’s view that words get to be meaningful in virtue of these sorts of rules of use.
Finally, the necessary materials for responding to Quine’s famous argument against the
view that the logical truths are true by convention have been provided above, for Quine
himself acknowledges that if we could explain “wherein an adoption of conventions,
antecedently to their formulation, consists” then the theory of truth by convention would
be tenable (Quine [1935] 1976, 105; cf. Quine [1954] 1976, 115), and this we have done
(though only by rejecting some of Quine’s fundamental assumptions).
Nevertheless, we are not, at this point, in a position to assert that these sentences
are, for Berkeley, true by convention. This is because all we have said so far is that
our conventions authorize us to assert them. We have not yet discovered a Berkeleian
account of the nature of truth or its relation to the permissibility of assertion. If we were
to identify truth with assertibility, we would bring Berkeley quite close to Carnap: since
assertibility is relative to a particular sign system, and sign systems are conventions
which we adopt or reject on pragmatic grounds, truth would, on this proposal, inherit
these features. A Berkeleian sign system would turn out to be a Carnapian linguistic
framework, and metaphysical questions would turn out to be purely verbal (see Carnap
1950, and alsox7.2.3, below).
196
It is, however, possible to resist this move, for it is not obvious, either as a matter
of Berkeley interpretation or as a matter of fact, that to adopt a convention whereby a
certain sentence is assertible is to adopt a convention whereby that sentence is true. It
is not obviously impossible to adopt a convention permitting the assertion of falsehoods
in sufficiently rare cases. This could be done either by accident or on purpose. For a
possible example of the intentional case, consider a conventional rule which permitted
us to assert falsehoods if, but only if, doing so would save one or more innocent hu-
man lives (and would not endanger any other innocent lives).
24
It seems that cases of
this sort are sufficiently rare that the ends of language would not be undermined if we
followed this rule and (tacitly) expected others to follow it. For a possible example of
the accidental case, consider terms purporting to denote the theoretical entities of a false
scientific theory. It may well be that the linguistic conventions which governed the use
of ‘phlogiston’ authorized the assertion of ‘combustion involves the release of phlogis-
ton;’ nevertheless, that sentence was false all along, even before the community rejected
‘phlogiston’ discourse.
The rules of inference we have so far discussed are part of our linguistic conventions,
and are partly constitutive of the meanings of the words involved. These rules authorize
certain assertions. However, the question of whether such authorization guarantees that
the assertions will be true must await a more general inquiry into the nature of truth,
which will be the topic ofx7.2.
In closing this chapter, I wish to consider two inference patterns which play a central
role in Berkeley’s metaphysics, but which resist the sort of treatment we have given in
the cases discussed so far. The first of these is the inference from a change to an efficient
cause of that change. The second is the inference from the existence of an idea to the
24. Rules of this sort are much discussed in the literature on Kantian ethics. See, e.g., Korsgaard 1986,
328-330.
197
existence of a mind that perceives that idea. The first type of inference is needed to
secure the existence of God (PHK,x26); the second type of inference is needed to secure
the mind-dependence of bodies (x4) and the existence of other finite minds (see Pearce
2008, andx8.6, below).
If these two patterns of inference are authorized by the actual rules of language, no
special problems are raised for Berkeley’s first-order project. It is certainly true that
in Berkeley’s intellectual context causal inference was one of the recognized means of
justifying assertions. Berkeley can therefore use this kind of inference in justifying his
assertions to his contemporaries. But consider: ‘spirit,’ ‘cause,’ and ‘God’ are not con-
nected to ideas in the way words like ‘red’ are. They appear to be theoretical terms
connected indirectly (if at all) with experience, much like ‘force’ and, of course, Berke-
ley does explicitly compare theological discourse to talk about ‘force’ in Alciphron.
However, it is Berkeley’s view that “‘Force’, ‘gravity’, ‘attraction’ and similar terms
are useful for reasoning, and for calculations about motion and moving bodies, but not
for understanding the simple nature of motion itself or for designating so many quali-
ties” (DM,x17). Berkeley would certainly not say the same thing about ‘spirit,’ ‘cause,’
or ‘God.’ The point of Berkeley’s demarcation between physics and metaphysics (see
above,x3.3.3) was supposed to be to have a realist understanding of metaphysical dis-
course and an anti-realist understanding of physical discourse. Physics, in Berkeley’s
view, involves the adoption of conventional rules for the manipulation of symbols in
order to make useful predictions. Berkeley’s metaphysical claims are meant to make
contact with reality in a deeper way than this, but it is far from clear how, within Berke-
ley’s philosophy of language, this is possible.
198
Despite Berkeley’s emphasis in Alciphron on the similarities between the discourses
of theology and physics, it is clear from De Motu that he intended to draw a sharp dis-
tinction between them. The nouns used in metaphysics and theology pick out objectively
existing entities, and true assertions in these areas attribute to those entities features they
really, objectively have. The nouns in physics, on the other hand, are merely placehold-
ers in a conventional notational scheme, waiting to be ‘redeemed,’ directly or indirectly,
for ideas. It is to the examination of this distinction that we now turn.
199
Chapter 6
Reference and Ontology
Central to the project of De Motu is a distinction between what we may call ‘genuine
referring expressions’ and ‘quasi-referring expressions.’ Genuine referring expression,
like ‘red,’ are used to label objects which exist and have their natures independently of
the sign system. Quasi-referring expressions are syntactically just like genuine referring
expressions, but differ semantically in that they do not label objects in the way genuine
referring expressions do.
1
Thus Berkeley says quite explicitly that “‘Force’ . . . is used
. . . as if it signified a quality” (DM,x5, emphasis added).
2
Quasi-referring expressions
can be meaningful and can be used to express truths despite the fact that they are not
used to label anything (see Alc,x7.10). Quasi-referring expressions are not nonsense,
and it is not the introduction or use of quasi-referring expressions which, according
to Berkeley, is the cause of philosophical error. Errors stem, instead, from confusion
between genuine referring expressions and quasi-referring expressions (DM,x6).
Berkeley need not (and, in my view, should not) object to such English sentences
as “Gravitational attraction is one of the things referred to by ‘force.’” Instead, he
can merely say that the plain English verb ‘to refer’ is ambiguous and, when doing
philosophy, it is important that we distinguish between its two senses, i.e. between what
1. A similar distinction has more recently been defended by Hofweber 2009.
2. In the original Latin, ‘as if’ is in fact not ‘quasi’ but ‘tamquam.’ I will refrain from introducing the
atrocious barbarism ‘tamquam-referring expression.’
200
I am calling ‘genuine reference’ and ‘quasi-reference.’
3
As a result, I will make no
attempt to avoid using such locutions as ‘talk about forces;’ locutions such as this one
make perfectly good sense, on Berkeley’s view as I understand it, but one must realize
that such talk is not about anything in the way talk about red is about something. Insofar
as forces can be said to exist at all, they exist as an artifact of our scientific theories. Red
things, on the other hand, exist quite independently of any sign system we adopt.
Berkeley thus divides terms which are syntactically like referring expressions into
three categories: some are genuine referring expressions, others are quasi-referring ex-
pressions, and others are nonsense. This distinction forms the heart of Berkeley’s meta-
ontology: one incurs an ‘ontological commitment’ when, and only when, one attempts
to use a word or phrase as a genuine referring expression. The aim of this chapter is
to understand Berkeley’s meta-ontology and the way in which his first-order ontology
follows from it. Along the way, we will have occasion to address a number of contro-
versial issues in Berkeley’s ontology, including the nature of bodies (x6.3) and spirits
(x6.5). Making use of resources from our study of Berkeley’s philosophy of language,
I will explicate Berkeley’s theory of spiritual substance, and defend a novel interpreta-
tion of Berkeley’s account of bodies, according to which bodies, like forces, are mere
quasi-entities which ‘exist’ as a result of the way we use language to structure our sense
experience. This raises a question about what Berkeley means in claiming to defend the
‘reality’ of bodies. This and some related questions about ontological vocabulary will
be discussed inx6.6. Finally, inx6.7, I will defend Berkeley’s claim that it follows from
his philosophy of language that ‘matter,’ ‘material substratum,’ and other related terms
are nonsense.
3. Berkeley explicitly endorses this kind of move with respect to words like ‘cause’ and ‘force’ (Siris,
xx154-155, 220).
201
6.1 Sensible Qualities
The simplest terms to understand in Berkeley’s philosophy of language are those, like
‘red,’ which denote sensible qualities. Here Berkeley’s metaphysics and epistemology
create no difficulties for the application of his philosophy of language: the theory of
general terms works precisely as envisioned (see above,x3.1).
The reason for such straightforward application is that sensible qualities are imme-
diately and unproblematically ‘present to the mind.’ In our earlier discussion of general
terms, we spoke rather liberally of words being applied to objects, of objects being
called by various names, and of recognizing objects as falling under concepts. What
happens when we are trained in the use of a word like ‘red’ is that we acquire the (im-
plicit) rule: ‘red’ may be applied to red things.
More properly, and more informatively, in order to comply with Berkeley’s Inter-
nalist Requirement (above,x5.2), the rule for ‘red’ should be given by reference to an
exemplar or exemplars, rather than by use of the word ‘red.’ Of course, in the ordinary
case, the rule is not explicitly given at all. What we are concerned with here is the
most perspicuous way of stating the rule which the subject follows implicitly. Since the
subject learned the rule by means of exemplars, the rule is, in essence, ostensive: it is
a rule that says that things are to be labeled ‘red’ because they resemble those things.
Since the rule is followed implicitly, there is no need to have an occurrent thought of
any exemplar every time one follows it. However, the exemplars are required for the
learning of the rule. Before one can learn the rule, one must be able to recognize things
that resemble the exemplars, and this requires having the exemplars ‘before the mind.’
202
On the other hand, once one has acquired the rule, one simply labels objects ‘red’ auto-
matically, without the need for any ‘mental machinery’ at all.
4
It is the ability to follow
this rule, engaging in acts of labeling things ‘red’ that secures the status of ‘red’ as a
genuine referring term.
In light of the broadly Wittgensteinian approach to language which I have been at-
tributing to Berkeley, this account of the use of ‘red’ faces three difficulties. The first
difficulty has to do with the sort of ostension involved in the rule. The second has to do
with what constitutes an act of labeling. The third and final difficulty has to do with how
words can be used to label private sense experiences if words are part of a phenomenon
(language) which is essentially public in nature.
I said that the rule the agent follows in using the word ‘red’ relies essentially on
certain exemplars which played a role in the agent’s learning of the rule. However, as
Wittgenstein points out, unless one has some prior sense of what sort of word is being
defined – for instance, whether a color word, a number word, or a proper name of an
object – the introduction of exemplars will not be sufficient for learning the proper use
of the word (Wittgenstein 1953,xx1.29-30). Wittgenstein makes this point in the course
of criticizing the idea “that learning language consists in giving names to objects. Viz,
to human beings, to shapes, to colours, to pains, to moods, to numbers, etc.” where this
“naming is something like attaching a label to a thing” (x1.26). The point Wittgenstein
is making is not that naming or labeling is not a genuine or legitimate use of language
(he explicitly recognizes the existence of ‘labeling’ language-games;x1.27). Rather,
Wittgenstein is making two points about the naming or labeling use of words: first,
that it is not so pervasive and important in language as a whole as we are prone to
4. Cf. A. D. Woozley: “Berkeley is making a general point about symbols (which he calls signs), that
not only does intelligent and intelligible handling of them not require a concomitant shadow sequence of
images in the stream of consciousness, but it does not require any accompaniment at all” (Woozley 1976,
431-432).
203
think and, second, that it is not so straightforward and unproblematic as we are prone
to think. Wittgenstein’s first point is clearly and explicitly recognized by Berkeley,
although Berkeley perhaps does not travel so far in the opposite direction as Wittgenstein
does. However, I know of no text in which Berkeley recognizes the second point.
I suggested above (x3.1) that Berkeley recognizes a primitive faculty of comparison,
which allows us both to recognize ideas as similar or dissimilar, and to distinguish re-
spects of similarity. The categorization of a word which Wittgenstein says we need to
undertake before we can learn the word’s meaning by an ostensive definition is, essen-
tially, a matter of knowing what respect of resemblance (resemblance in color, resem-
blance in number, etc.) to look for in deciding when to apply the word. How exactly this
works is a thorny issue which Berkeley, as I have said, does not face squarely because he
does not see it clearly. If, however, we are willing to accept Berkeley’s primitivism about
comparison then a relatively simple solution presents itself: given a collection of exem-
plars, we simply compare them along the various dimensions of similarity which come
naturally to us until we find a respect in which they all resemble. (These ‘respects’ are
not universals. They are different comparison operations the mind can perform. Thus
the operation of color-comparison is both metaphysically and epistemologically prior to
color.) We then apply the label ‘red’ to those things which resemble the exemplars in
that respect wherein the exemplars resemble one another.
Our second question is, what is this act of labeling or calling? One can answer this
question, in Wittgensteinian fashion, by calling attention to a particular language-game
– in fact, to one that is literally a game played by parents with young children. The
parent holds up a flash card of a particular color, and the child calls out the name of the
color. The child is, in this game, engaged in an act of labeling her idea, of calling it
‘red.’
204
Now, Wittgenstein would be quick to point out that this kind of thing can only take
place in a broader linguistic context. I do not think Berkeley needs to deny this. What is
involved in Berkeley’s distinction between genuine reference and quasi-reference is not
the claim that there are acts of labeling which operate outside any linguistic context. It
is, rather, that there is a use of words which involves attaching them to things of which
we have a pre-linguistic grasp, and that, among all of our meaningful categorematic
terms, only a small, privileged class have this kind of use.
This, however, brings us up against the third and final difficulty. I said that the
child responding to the flash cards is labeling her ideas. However, if we have admitted
that this labeling can take place only in a broader linguistic context, and that language
is essentially a public social practice, then how is it possible that the child should use
words to label her private ideas? Should we not say, instead, that she labels the card?
When I argued, inx5.3, that Berkeley recognizes that language is a social practice
governed by public conventions, I was not arguing that Berkeley accepts anything like
Wittgenstein’s ‘private language argument,’ or that Berkeley accepted Wittgenstein’s
more radical thesis that there can be no private rule-following of any kind (Wittgenstein
1953,x1.202). Indeed, my treatment of rule-following inx5.1 supposed what Wittgen-
stein would never grant: that there is a metaphysical fact of the matter about what rule is
being followed independent of social facts such as what actions the community corrects.
Given this view about rules, which stops short of some of Wittgenstein’s more radical
conclusions, we can say that the broader context which is required for an utterance to be
an act of labeling has to do with the system of rules being followed by the agent. Absent
such a context, it is impossible to say whether, for instance, an utterance of ‘red’ is an
act of labeling (describing) the card which is in front of one, or asking for the red card
(in the belief that the card in front of one is non-red), or something else. However, there
205
is no bar here to the following of a system of rules whereby an agent would use a sign
to label her private experience.
This, however, does not yet fully answer the publicity objection, for we must not
only explain how an individual agent could make use of a sign which would be a label
for her private experience. We must go farther and show how language can be used to
label private experience, and we must do so in a way that respects the status of the rules
of language as public conventions.
The view I defended inx5.3 was that the rules of language are public conventions in
the sense that we all follow them and rely on others to follow them. This does not require
that every instance of following every rule is directly publicly observable. My labeling
of my individual ideas can be done in obedience to a public convention provided that I
follow a rule which is such that the community follows it and relies on others to follow
it. It is possible for rules for the labeling of private sense experiences to gain this status
if, but only if, we all have beliefs about what ideas others are having (cf. Craig 1982,
552-558). If you say ‘that looks red’ then, given the conventions surrounding the word
‘looks,’ I will take you to be using ‘red’ to label a visual idea you are having. If I have
an independent belief about what your present visual ideas are like, then I can figure
out what sorts of ideas you label ‘red.’ Note, however, that this process depends, first,
on our having some prior source of information about the ideas of others and, second,
on there being more to language than just the labeling of ideas. I have already said a
great deal about the second requirement. As to the first requirement, it is necessary first
to explain how we can even so much as conceive of ideas as existing in the minds of
others. I will undertake this task in my account of our talk about spirits and their actions
inxx6.4 and 6.5, below. It is further necessary to explain how we go beyond merely
conceiving the possibility of others perceiving certain ideas to come to particular beliefs
206
about the goings on in other minds. Inx8.6 I will argue that this information is provided
by the language of nature. Since it is only by means of the language of nature that we
know the existence of other finite minds, it is hardly surprising that human language –
language by means of which we interact with these other finite minds – should depend
on the language of nature in this way.
All idea theorists would agree that, in showing the flash card to the child, the parent
is giving the child red ideas. The Theory of Meanings holds that the child’s use of ‘red’
can somehow ‘pass through’ the red idea to attach to the mind-independent red object.
5
This is possible because the idea naturally, independent of any learned conventions or
rules, serves as a label for red things. This, however, is precisely the aspect of the Theory
of Meanings Berkeley attacked (see above, ch. 2). For Berkeley ideas are not intrinsi-
cally representational, although they can be used representationally, by the following of
rules. Since we lack prior independent access to mind-independent, external red things,
Berkeley’s Internalist Requirement prevents us from learning a rule whereby ideas, or
words, would label such things. Nevertheless, if the parent has the belief that such and
such course of action will lead to the child’s having red ideas, then the game can go on,
quite independently of the assignment of any representational content to the red idea.
I have already argued that even genuine referring terms are associated with other
rules besides the rules for reference. In the case of ‘red,’ one of the other rules will
allow us to assert, for instance, ‘red is a color.’ This use of ‘red’ takes us beyond the
use of words to label items in our immediate, present experience. From ‘red is a color’
we can infer ‘everything red is colored,’ and then likewise apply the name ‘colored’ to
5. The Cartesians hold that in the case of ‘red,’ a secondary quality term, this ‘passing through’ is
an error (see below, pp. 393-394), but they nevertheless hold not only that it is possible, but that it is
common. Furthermore, in the case of primary quality terms like ‘square’ the Cartesians take the ‘passing
through’ to be perfectly legitimate. Matters are, of course, complicated by the direct realism of Arnauld
and Arnauld’s interpretation of Descartes. See above,x1.2.
207
red objects, and so forth. This is what Berkeley is getting at with Euphranor’s poker
chip analogy (Alc,x7.8): the initial reports of our immediate experience are analogous
to buying the chips, the reasoning which proceeds by ‘computing in signs’ is analogous
to playing the game, and the resulting sentences with immediate observational import
are analogous to ‘cashing in’ the chips at the end.
The conclusions, I said, have ‘immediate observational import;’ but of course if
they were things we could presently observe to be true, then we would not need to
arrive at them by reasoning. What the conclusions really do, at least in typical cases, is
create observational expectations. These expectations are meant to be action-guiding.
As a result, even the most straightforwardly descriptive uses of language (e.g.,‘this is
red’) turn out, on Berkeley’s view, to have a crucial ‘operative’ component: although
we buy our chips with observations, we often cash them in for actions. It is sufficient
for the meaningfulness of a word that it should have some connection to observation
or action and thereby serve some useful purpose. However, not all meaningful words
are genuine referring expressions. What secures the status of sensible quality terms
as genuine referring expressions is the rule whereby they are directly applied to those
things to which they refer. This does depend on a prior grasp of the individual objects
to which the word is applied, though it does not depend on a prior concept (or abstract
idea) of redness. Since all general thought depends on signs, one cannot think about
red things in general until one has some sign by which to designate them; however, our
ability to learn the use of such a sign depends on our prior ability to apprehend individual
red things.
208
6.2 Occult Qualities
There are many terms which are like ‘red’ in the positions they can occupy in sentences,
and in the rules by which they are used in ‘computing in signs,’ but which nevertheless
lack the kind of direct applicability characteristic of sensible quality terms. Berkeley
calls the entities to which these terms purport to refer ‘occult qualities’ and argues that,
strictly speaking, there are no such things (DM,xx4-6, 17). Nevertheless, he does not
call for the elimination of all such vocabulary; instead, he defends the view that some
vocabulary of this sort, such as the word ‘force,’ should be retained on account of its
usefulness:
if, by considering this doctrine of force, men arrive at the knowledge of
many inventions in Mechanics, and are taught to frame engines, by means
of which things difficult and otherwise impossible may be performed; and
if the same doctrine, which is so beneficial here below, serves also as a key
to discover the nature of the celestial motions; shall we deny it is of use,
either in practice or speculation, because we have no distinct idea of force?
(Alc,x7.10)
The examples of occult qualities Berkeley discusses at the greatest length are the
‘mathematical hypotheses’ of physics.
6
Berkeley says that, in physics as in many other
areas, we are “obstructed by words that are poorly understood” (DM,x1). These words,
such as ‘solicitude,’ ‘effort,’ ‘striving,’ ‘force,’ ‘gravity,’ and ‘attraction’ (xx3, 5-6, 17),
7
6. Following standard 18th century usage, Berkeley uses the term ‘mathematical hypothesis’ to refer
to certain entities introduced to aid calculation: force, impetus, etc. The hypotheses are the things, not the
propositions about them. See Peterschmitt 2003, 191-192, 197.
7. The Latin words are ‘solicitatio,’ ‘nisus,’ ‘conatus,’ ‘vis,’ ‘gravitas,’ and ‘attractio,’ respectively.
Note that inx3 Berkeley writes ‘nisus sive conatus,’ implying that he takes ‘nisus’ (‘effort’) and ‘conatus’
(‘striving’) as synonyms.
209
are introduced because they are “useful for reasoning, and for calculations about motion
and moving bodies” (DM,x17); “however, when they are used by philosophers to signify
certain natures . . . which are not the objects of the senses . . . then they eventually give
rise to errors and confusions” (DM,x6; cf. Siris,x250).
Strictly speaking, talk about forces is just another kind of talk about motion (DM,
xx6, 22; Siris,x240), and motion is a sensible quality (DM,x43). The introduction of
mathematical hypotheses serves, in part, to “make notions or, at least, propositions [i.e.,
sentences] universal” (x7). In reality, “bodies are moved to or from each other, and
this is performed according to different laws” (Siris,x234). However, with the help of
the “general notions and propositions” made possible by mathematical hypotheses, the
physicist is able to construct “universal theorems of mechanics” by means of which “the
motions of any parts of the system of the world, and the phenomena that depend on them,
become known and determined” (DM,x38). In order to do this, the physicist needs to
talk about more than just bodies and motions; the laws of physics can be formulated
only by the help of certain quasi-referring terms, such as ‘force.’
Occult quality terms can be used in reasoning in much the same way as sensible
quality terms. The mathematical hypotheses of physics are typically introduced with
explicit and precise definitions, axioms, or postulates, and this makes them conducive
to formal, rigorous computation. Physics is, however, supposed to involve “sensation,
experimentation, and geometrical reasoning” (x1, emphasis added);
8
‘geometrical rea-
soning’ alone is not enough. Furthermore, physics is meant to “direct us how to act and
8. Luce, Jesseph, and Clarke all translate ‘experientia’ as ‘experience’ in this passage. I take ‘exper-
imentation’ to be more accurate to Berkeley’s intentions. In English it is not clear, in this context, what
the difference between ‘sensation’ and ‘experience’ is. However, even in classical Latin, ‘experientia’
can mean “a trial, proof, experiment” (C. T. Lewis 1890, s.v. ‘experientia’). I believe Berkeley means
here to use ‘sensus’ to refer to casual, unstructured observation, and ‘experientia’ to refer to the kind of
organized, intentional experimentation undertaken by early modern natural philosophers such as Robert
Boyle.
210
teach us what to expect” (Siris,x234). If physics is to be a genuine science and not a
mere notation game, the ‘chips’ must be ‘purchased’ at the beginning and ‘cashed in’
at the end, originating from observation, and ending in actions or expectations, just like
sensible quality terms.
Berkeley is fairly explicit about how this works: “no force is itself felt immediately,
nor can it be known and measured otherwise than by its effect” (DM,x10; cf. Siris,
x250). In other word, there are certain ‘known and measured,’ i.e. sensible, qualities
which are the ‘effects’ of a force. These are effects of a particular force in the sense that
that force is their physical cause, which is just to say that, within our physical theory,
they are conceived of as following from that force. Within the physical theory, the laws
connect the force to observable phenomena, and it is on the basis of those phenomena
that we attribute the force to the body. This also works in the other direction: once we
attribute a certain force to a body, we can predict other effects besides the one on the
basis of which we attributed the force in the first place. In this way, these mathematical
hypotheses guide our actions and expectations.
9
There is no way of grasping even particular individual instances of force until one
has introduced the term ‘force’ within the formal structure of the theory. It is not merely
the case that the word ‘force’ has no meaning outside the theory; it is rather that force
has no essence or nature (DM,x67) and, indeed, does not even exist (x39) apart from its
role in the theory.
Berkeley has one simple reason for holding this view: because of our lack of a
language- or theory-independent grasp of force, we cannot make the word ‘force’ mean-
ingful by using it to label forces (cf. Downing 1995a, 205-208). That is, Berkeley’s
9. For further details, see below,x8.4.
211
Internalist Requirement (above,x5.2) prevents us from following the rule ‘force’ may
be applied to forces, or any rule like it.
Berkeley’s Internalist Requirement implies that, in order to use ‘force’ to refer to
forces, we would need to be able to think of forces (at least individually) prior to the
acquisition of the word ‘force’ and the conventions surrounding it. In Berkeley’s view
this cannot be done. Perhaps, however, we could secure reference by defining ‘force’
in terms we already understand. Three strategies for doing this present themselves: we
could define ‘force’ by the distribution of forces in the world, we could define it by
the causal role forces play, or we could give an analysis of ‘force’ in terms of sensible
qualities.
The first strategy is something like the ‘Ramsey sentence’ approach to theoretical
terms (see D. Lewis 1972, 2009). The Ramsey sentence of a theory is constructed by
conjoining all of the claims the theory makes, replacing all of the theoretical terms with
variables, and introducing existential quantifiers to bind the variables. We then claim
that, if the Ramsey sentence is true (or at least approximately true
10
) each theoretical
term of the theory refers to whatever assignment of the corresponding variable made
the Ramsey sentence true (or approximately true). We can demonstrate the application
to our present case more simply by switching from ‘force’ to ‘mass,’ since mass is a
scalar quantity possessed by a single body. The Ramsey sentence for ‘mass’ would
say something like ‘9M(the acceleration of a body is directly proportional to the force
exerted on it and inversely proportional to the degree to which it possesses M & the
gravitational force between any two bodies is directly proportional to the product of the
degree of possession of M by each of the bodies & . . .),’ where the ellipsis would be
filled in with the rest of the things our theory says about mass. We would also need
10. Berkeley might say: true to some degree. See below,x7.2.
212
to replace all of the other occult quality terms (e.g. ‘force’) with variables, and add the
relevant additional quantifiers. This will complicate matters significantly, but the general
idea can be grasped without working out the example in further detail. We would then
say: if the Ramsey sentence is true, or nearly true, then whatever assignment for M
makes it most nearly true is the referent of ‘mass.’ I described this above as defining
‘mass’ in terms of the distribution of mass in the world because what it essentially does
is to look at the circumstances in which our theory tells us to attribute mass to objects
and says to let ‘mass’ refer to whatever it needs to in order to make those assignments
correct.
If this is to result in ‘mass’ being a genuine referring expression, rather than merely a
quasi-referring expression correctly predicated of the objects in question, the word ‘pos-
sess’ must be understood in a metaphysically loaded sense: we are not talking merely
about ‘mass’ being correctly predicated of objects, but about objects possessing some
objective feature. Now, for Berkeley, there are no abstract ideas of quality, property,
attribute, inherence, etc. There are sensible qualities, on the one hand, and there are
actions and powers on the other (see below,x6.4). As will be discussed below, it is a
conceptual truth, according to Berkeley, that action and power require will, but it is also
a conceptual truth that neither body nor matter can have will, thus if mass is a power
or action its attribution to body is incoherent. But, ex hypothesi, mass is not one of the
sensible qualities we already know. Hence, it must be a sensible quality we don’t know.
However, positing such a quality would explain nothing, and is not supported in any
way by the evidence which supports the theory.
If instead we follow the second strategy and try to define ‘mass’ or ‘force’ based
on what they cause, we fare no better. If we mean simply physical causation, then this
213
proposal does not differ from the last one, since physical causation is mere regularity.
11
If, however, we mean genuine efficient causation then, again, the proposal is incoherent
since this would involve attributing volition to bodies.
Finally, we could try to define ‘mass’ or ‘force’ in terms of sensible qualities known
to us. This, Berkeley says, is simply to cease using ‘mass’ and ‘force’ in the way
physicists do, and so to lose out on the benefits of physical theory, since physicists
distinguish force and mass from any sensible qualities (DM,x5; Alc,x7.10).
None of the obvious methods for securing genuine reference for ‘mass’ works.
Berkeley believes, more generally, that there is no way of securing genuine reference
for ‘mass,’ or any other occult quality term. Nevertheless, some of these terms have a
well-defined use in the discourse of Newtonian physics and, furthermore, are extremely
practically useful. The conclusion to draw is that they are quasi-referring terms.
Berkeley’s elimination of the contemplated methods of securing genuine reference
depends on highly controversial metaphysical views which will be discussed and evalu-
ated in more detail below. However, Berkeley’s most important point about theoretical
language of this sort can be appreciated whether or not one accepts those assumptions. It
is this: the rejection of the extra-theoretical existence of these entities in no way threat-
ens the use or purpose of the language of physics (DM,x66). In fact, the illusion that
these are genuine referring expressions leads to all sorts of “well-known disputes that
have greatly exercised learned men” (x15), while the rejection of that illusion neatly
resolves these disputes, thereby rendering the discourse more useful to its intended pur-
pose of helpfully guiding our expectations and actions. Thus the assumption of some
kind of linkage between the application of the word ‘force’ to an object and some objec-
tive feature that object has plays, or at least should play, no role in the rules surrounding
11. Or something near enough. Seex8.4.
214
the use of that word. This is what it means for ‘force’ to be a quasi-referring expression:
although it is syntactically, and hence inferentially, like ‘red,’ the rules of its usage do
not tell us to use it to label any one particular feature.
It does not follow from this that ‘force’ is eliminable from physics. Berkeley does
say that “general and abstract terms . . . were in fact invented partly to abbreviate speech,
and partly by philosophers for teaching purposes” (DM,x7). However, we must remem-
ber that a science, for Berkeley, just is a sign system (Alc,x7.14). There is no such thing
as Newtonian physics without ‘force’ – the imagined construct would be an entirely
different sign system.
12
Because of the generality they provide, if “the system of the
world” is to be “subjected to human calculations” (DM,x66), one is going to need some
mathematical hypotheses, and the mathematical hypotheses of Newtonian physics are
the best we’ve got (cf.xx38-39).
13
Thus by Quine’s criterion, Berkeley is ‘ontologically
committed’ to forces (Quine 1948): a scientific theory which Berkeley accepts contains
ineliminable quantification over forces.
All that follows from this is that Berkeley is implicitly committed to the rejection
of Quine’s criterion. The question which remains is: what alternative is Berkeley in a
position to offer? How does Berkeley manage to get committed to red but not to force?
The answer is that, for Berkeley, a word purports to be a genuine referring expression
(and hence its use is ontologically committing) only when we adopt a rule which gives
that word a labeling use. However, it can be shown that every attempt to use ‘force’ to
label things either fails (through failure to make contact with the alleged thing), or else
12. Such alternative sign systems are, of course, possible, as is shown by the alternative formulations of
classical mechanics developed in the 19th century.
13. John Foster says that Berkeley “goes astray” by failing to notice that “the sensory order can only
be adequately specified by reference to the physical theory which it prompts us to accept” (Foster 1985,
87-88). This mischaracterization of Berkeley stems from a lack of attention to the later works, De Motu,
Alciphron, and Siris.
215
violates the other rules for using the word ‘force’ (as a technical term of physics). Thus
what we have here is not a mere question of parsimony or theory choice, but a question
of coherence: as long as one is committed to using ‘force’ the way the physicists do,
any attempt to label anything with it results in contradiction.
14
Thus, as much as one
might be committed to sentences such as ‘there are forces’ – or, to use an example
from Berkeley’s own text, “The pure æther or invisible fire contains parts of different
kinds, that are impressed with different forces” (Siris,x162) – and as much as one may
not be able to avoid quantifying over forces in this way, nevertheless, it is not possible
(without contradiction) to incur a genuine ontological commitment to the (quasi-)entities
physicists (quasi-)call ‘forces.’
15
6.3 Bodies
According to Berkeley, the physical realist’s mistake is to think that “‘force’, ‘gravity’,
and similar words . . . are used to signify certain natures” (DM,x6). This is a linguistic
confusion (x1) which leads to pointless disputes which interfere with the real purpose of
physics, which is to “direct us how to act and teach us what to expect” (Siris,x234). We
will be better able to go about the business of physics if we pay attention to the use of
14. This characterization of the conclusion of De Motu agrees with Downing 1995a. I have attempted
in this section to show why Berkeley, given the rest of his system, accepts the conclusion, and so have
appealed freely to Berkeley’s other works. Downing provides an excellent account of the dialectic of
De Motu and, in particular, Berkeley’s strategy for persuading Cartesians – who could not be assumed
to accept, or even be familiar with, his other doctrines – of the conclusion that ‘force’ and other similar
terms do not (genuinely) refer.
15. It is worth noting that Quine himself begins with a distinction between those expressions which
are genuine names and those which are not, and then proceeds to give his famous account in terms of
quantification. What I am suggesting is that Berkeley needs to reject the move from naming to quan-
tification: for him, the distinction between genuine names and quasi-names is the fundamental point.
Furthermore, Berkeley rejects the claim that whether a word is a genuine name depends on the availabil-
ity of paraphrases (Quine 1948, 25-28). Whether a word is a genuine name depends, instead, on whether
the conventions governing its use instruct us to use it to name something.
216
such words and thereby come to realize that these things “have no stable essence in the
nature of things” (DM,x67).
In the Three Dialogues, Berkeley’s character Philonous gives a strikingly similar
account of the error of the materialist.
16
Philonous had argued in the first dialogue that
perceptual relativity lands the materialist in either contradiction or skepticism. “Suppose
now one of your hands hot, and the other cold, and that they are both at once put into
the same vessel of water, in an intermediate state; will not the water seem cold to one
hand and warm to the other?” (DHP, 179). If the materialist trusts her senses, she must
conclude that the water is both cold and warm – a contradiction. Thus, if she is to avoid
contradiction, the materialist must not trust her senses. But distrust of the senses is,
for Berkeley, the hallmark of skepticism (PHK,xx40, 101; DHP, 167, 173, 211, 237,
244-245).
In the third dialogue, Hylas argues that Philonous’s (i.e., Berkeley’s) view faces the
same objection: we perceive the same body as having many different, contradictory
sensible qualities, often at the same time. If, then, the existence and nature of body
is given in sensory perception, the same thing has contradictory qualities. Philonous
responds:
What . . . if our ideas are variable; what if our senses are not in all circum-
stances affected with the same appearances? It will not thence follow, they
are not to be trusted, or that they are inconsistent either with themselves or
anything else, except it be with your preconceived notion of (I know not
what) one single, unchanged, unperceivable, real nature, marked by each
16. Following Berkeley, I use the term ‘materialist’ to refer to those who believe that sensible qualities
inhere in a mind-independent ‘material substratum.’ The ‘physical realist’ is one who takes the theoretical
terms of physics to have ontological import, i.e., to be genuine referring expressions.
217
name; which prejudice seems to have taken its rise from not rightly under-
standing the common language of men speaking of several distinct ideas,
as united into one thing by the mind. And indeed there is cause to sus-
pect several erroneous conceits of the philosophers are owing to the same
original, while they began to build their schemes, not so much on notions
as words, which were framed by the vulgar, merely for conveniency and
dispatch in the common actions of life, without any regard to speculation
(DHP, 245-246).
The materialist is confused about plain language in just the same way the physical
realist is confused about the formal language of physics. Notice specifically, three paral-
lels between Berkeley’s diagnoses of the errors of the physical realist and the materialist:
(1) both err in supposing that the words in question designate stable ‘real natures’ ex-
isting independently of the sign system; (2) both make this error because they are in the
grip of the Theory of Meanings; (3) in both cases, the grip of the Theory of Meanings
is to be broken (in good Wittgensteinian fashion) by attention to the practical purpose
of the discourse. Given the parallel diagnosis, it is reasonable to expect a parallel cure.
This, I will now argue, is exactly what we find in Berkeley’s statements regarding the
nature of bodies: bodies, like forces, are mere quasi-entities whose existence and na-
ture are the creations of our linguistic conventions. ‘Body’-talk differs from ‘force’-talk
only in that humans have been using it from time immemorial (cf. James 1907, lecture
5; Carnap 1950,x2).
6.3.1 Bodies as Linguistic Constructions
There are two crucial passages in which Berkeley lays out his view of the nature of
bodies. In the first of these, Berkeley says that when
218
several [sensible qualities] are observed to accompany each other, they
come to be marked by one name, and so to be reputed as one thing. Thus,
for example, a certain colour, taste, smell, figure and consistence having
been observed to go together, are accounted one distinct thing, signified by
the name ‘apple’. Other collections of ideas constitute a stone, a tree, a
brook, and the like things (PHK,x1).
This lines up with a passage from the Dialogues where Philonous says that:
men combine together several ideas, apprehended by divers senses, or by
the same sense at different times or in different circumstances, but ob-
served however to have some connection in nature, either with respect to
co-existence or succession; all which they refer to one name and consider
as one thing (DHP, 245).
A body, according to Berkeley, is ‘constituted by’ certain ideas which “are united
into one thing (or have one name given them) by the mind” (249). Note, however, that,
although at the beginning of the Principles passage Berkeley says that “they [the sensible
qualities] come to be marked by one name,” his explanation of how this occurs does not
involve the collection being called by that name, nor does it involve any of the ideas in
the collection being called by that name. Rather, Berkeley says that the “colour, taste,
smell, figure and consistence” are “accounted one distinct thing,” and that thing is called
‘apple.’ Furthermore, Berkeley’s use of the phrase ‘and so’ strongly suggests that the
ideas in question come to be ‘reputed as one thing’ by means of their association with a
name.
In the Dialogues, Berkeley says that these various ideas are “refer[red] to one name
and consider[ed] as one thing.” Berkeley does say that the use of the name is somehow
tied up with the ideas which are grouped together by their “connection in nature,” and
219
this implies that the use of the word ‘apple’ provides a way of talking about the “colour,
taste, smell, figure and consistence.” It does not, however, imply that the word ‘apple’
is a label for these ideas, either individually or collectively. Rather, ‘apple’ provides a
way of talking about these ideas in the same way that ‘force’ provides a way of talking
about motion.
17
Introducing the ‘thing language’ gives us a way to organize and predict
our ideas, a useful way of structuring the deliverances of the senses, helping us to get
around in the world.
18
It may be objected that Berkeley does sometimes say that a body (or thing) is a ‘com-
bination’ of ideas or sensible qualities (NTV,x109; PHK,xx12, 38). My interpretation
does not, however, deny that the existence of a body consists in certain ideas being com-
bined together. What I deny is that this combination is some pre-linguistic entity waiting
to receive a label, like a sensible quality. Instead, it is by the conventions for the use of
the name that the ideas are combined.
6.3.2 Alternative Interpretations
The standard scholarly interpretations of Berkeley’s account of bodies divide into two
categories, which may be called ‘subjunctive interpretations’ and ‘idea interpretations’
(cf. Winkler 1989, ch. 7; Dicker 2011, ch. 14). According to subjunctive interpreta-
tions, Berkeley (like later subjunctive phenomenalists) takes statements about bodies to
17. On ‘force’ as a tool for talking about motion, see DM,xx6, 22; Siris,x240.
18. Similar ideas are expressed by Quine 1948, 35-37 and Carnap 1950,x2. Tipton 1974, 210 appears
(though somewhat ambiguously) to be making a similar suggestion about Berkeley’s view of bodies.
Elsewhere, however, Tipton says that, on Berkeley’s view, ‘body’ talk, though convenient, “involve[s] a
radical distortion of the facts” (223). Here I must disagree. The facts are not distorted by the ordinary
use of ‘body’ talk, but by the philosophers’ assumption that the names of bodies are genuine referring
expressions.
220
be equivalent in meaning to statements about what humans would perceive under cer-
tain conditions. According to idea interpretations, Berkeley identifies each body with
some idea or collection of ideas. Standard interpretations, of either family, fail to take
seriously Berkeley’s remarks about the function of the names of bodies. Furthermore,
they fail to get the epistemological facts about Berkeleian bodies right: Berkeley wishes
to hold, on the one hand, that we (already) have certainty about the existence and nature
of bodies by means of our senses but, on the other hand, there is still much more to
be learned by empirical investigation. Standard interpretations cannot hold these two
epistemological theses together.
Subjunctive Interpretations
The failure of the subjunctive interpretation of Berkeley’s theory of bodies is closely
connected to one of the best-known philosophical objections to subjunctive phenome-
nalism. The subjunctive phenomenalist holds that the claim that there is a body in front
of me is equivalent in meaning to some set of subjunctive conditionals about what hu-
man perceivers would perceive under specified circumstances. Thus, to know the first
is to know the second. However, we never know more than a few of the conditionals
which would be involved in such a translation. If all there is to a body is the handful
of conditionals we know, then commonsense is radically mistaken about the nature of
bodies, but if there is more to a body than this, then, on this interpretation, it turns out
that I do not in fact know that there is a desk in front of me. The subjunctive interpreta-
tion thus fails to capture what Berkeley takes to be our epistemic situation with respect
to bodies.
In addition to the question of whether we really know as many conditional claims
as, according to the subjunctive phenomenalist, we ought to know, Quine famously
221
raised a second worry about this kind of view: those conditionals we do know can only
be stated by means of ‘body’ talk (Quine 1948, 36-37). This was supposed to show
that subjunctive phenomenalism, as a program for reducing bodies to sense data, was a
failure, since subjunctive phenomenalism will not allow us to do away with ‘body’ talk.
In this respect, Berkeley, on the interpretation I am defending, has a more sophis-
ticated and plausible theory than the version of phenomenalism criticized by Quine. I
have already argued (x6.2) that, on Berkeley’s view, one need not be able to paraphrase
away talk about a putative class of entities in order to deny that such talk is ontologically
committing. Rather, one incurs an ontological commitment only by adopting rules for
the use of a word as a label.
On my reading, Berkeley denies that it is the purpose of ‘body’ talk to label anything.
Instead, Berkeley holds, ‘body’ talk is simply a tool for capturing certain practically
important regularities in our sense experience. Berkeley can agree with Quine that we
could not get around the world without such a tool, but this indispensability argument,
for Berkeley, will bear no ontological weight: in Berkeley’s view, body talk is nothing
more than a way of capturing these regularities.
Part of the way ‘body’ talk accomplishes its purposes is indeed by instilling in us
certain expectations at varying levels of detail, regarding what we would perceive in
counterfactual circumstances. Nevertheless, knowledge of ‘body’ statements need not
involve detailed knowledge of the ideas we would perceive in other circumstances. (Ex-
actly what is involved in such knowledge will be examined in more detail in chapter 9.)
222
Idea Interpretations
Idea interpretations subdivide into two categories, which we may call ‘divine’ and ‘hu-
man.’ According to divine idea interpretations, each body is identified with some idea
or collection of ideas had by God. As is well-known, although this approach has some
support in Berkeley’s text (PHK,xx6, 48; DHP, 212, 230-231, 248, 254), it brings back
all of the skeptical problems of representative realism which Berkeley insists his view
avoids (Mabbott 1931; Foster 1982, 29-32; Dicker 2011, 268). Divine idea interpreta-
tions make the veridicality of our ideas depend on some kind of ‘matching’ between our
ideas and God’s, but how can we know that the ideas really do ‘match’?
19
Human idea interpretations, by contrast, identify each body with an idea or collec-
tion of ideas had by humans. An interpretation of this sort has recently been defended by
Samuel Rickless, who takes bodies to be complex ideas assembled by human perceivers
(Rickless 2013, 45-46, 123). This interpretation receives strong textual support in the
New Theory of Vision:
By the application of his hand to the several parts of a human body [the
blind man] had perceived different tangible ideas, which being collected
into sundry complex ones, have distinct names annexed to them. Thus one
combination of a certain tangible figure, bulk, and consistency of parts is
called the head, another the hand, a third the foot, and so of the rest. All
which complex ideas could, in his understanding, be made up only of ideas
perceivable by touch (NTV,x96, emphasis added).
This passage causes two serious problems for my interpretation. First, it explicitly as-
sociates bodies (specifically, parts of human bodies) with complex ideas and, second,
19. Winkler 1989, 216-224 offers a number of additional criticisms of this interpretation.
223
it explicitly associates these complex ideas with ‘collecting,’ ‘combining,’ and ‘nam-
ing.’
20
The most straightforward reading of this passage would take Berkeley to be
claiming that the blind man first collects various tangible ideas into one complex idea,
then labels that complex idea (e.g.) ‘head.’ This would make ‘head’ a genuine referring
term.
Although this is the most straightforward reading of this particular passage, it cannot
be Berkeley’s considered view. It cannot be the case that the blind man’s complex idea
of the head is the head, for the head ought to include visual ideas which, Berkeley
explicitly says, the blind man does not have. This is only a vivid example of a more
general problem facing the complex idea interpretation: no human has all of the ideas
which make up (e.g.) a cherry (cf. Hight 2007b, 86-87). As a result, no human can
combine all of these ideas into one complex idea.
6.3.3 The Richness of Berkeleian Bodies
Whereas divine idea interpretations undermine Berkeley’s response to the skeptic, sub-
junctive interpretations and human idea interpretations fail to recognize the richness
of Berkeleian bodies: the bodies we know ‘contain’ more ideas than we have ever per-
ceived, imagined, or expected. This issue is addressed explicitly in the Dialogues. Hylas
objects, “Why is not the same figure, and other sensible qualities, perceived in all man-
ner of ways? And why should we use a microscope, the better to discover the true nature
of a body, if it were discoverable to the naked eye?” (DHP, 245). The objection is that,
according to commonsense, bodies are much richer than momentary perception. That
is, they have many more qualities than can be perceived at any one time. Furthermore,
20. I thank Samuel Rickless for directing my attention to this passage and the difficulties it causes for
my interpretation.
224
there are contradictory ideas attributed to the same body, as when a body looks smooth
to the naked eye but rough under a microscope. Philonous responds that if “every vari-
ation [in our ideas] was thought sufficient to constitute a new kind or individual, the
endless number or confusion of names would render language impractical” (DHP, 245).
Thus the practice of language necessarily requires that “men combine together several
ideas” so as to “refer to one name and consider as one thing” the heterogeneous objects
of sight and touch, as well as the heterogeneous objects perceived by microscopes and
by the naked eye (245). This combining allows us to say, in plain language, that “we
use a microscope, the better to discover the true nature of a body,” despite the fact that,
strictly speaking, the object (idea) seen with the aid of the microscope is utterly distinct
from the object (idea) earlier seen with the naked eye. By discovering “what ideas are
connected together” we learn “the nature of things” (245, emphasis added), that is, of
bodies (see Atherton 1991,xx4-6).
The examination of a body under a microscope allows us to make a discovery about
what ideas are ‘combined’ in that body. But this implies that there are more ideas ‘com-
bined’ in the body than the ideas I have experience before using the microscope. My
claim, supported by the explicit appeal to language in this passage, is that the con-
ventions regarding the names of bodies can determine, in advance, whether a given
‘new’ idea is to be attributed to the body in question – e.g., whether this never-before-
experienced taste is to be called ‘the taste of the cherry’ – and that it is by these rules,
which command the classification of this idea as a cherry idea, that the various ideas are
combined into the cherry. In this way, by means of the conventional rules of language,
we have already combined the ideas before we have perceived them. The combination
is thus a mere quasi-entity: it owes its existence to our linguistic conventions.
225
If, however, this is Berkeley’s considered view, then what are we to make of NTV,
x96? We may begin by noting that Berkeley there says “different tangible ideas, . . . being
collected into sundry complex ones, have distinct names annexed to them.” Berkeley
does not say: “sundry complex ideas are called by distinct names.” Rather, the grammat-
ical subject is ‘different tangible ideas.’ These ideas are said to have names ‘annexed’ to
them. This leaves the exact relation between the ideas and the names extremely unclear.
This passage does clearly imply that which ideas go together into complex ideas figures
into the explanation of which names are annexed to which ideas. My interpretation can,
however, accommodate this fact. In the Dialogues, Philonous says that ideas are ‘com-
bined together’ because they are “observed . . . to have some connection in nature, either
with respect to co-existence or succession” (DHP, 245). When the blind man feels, e.g.,
a head, he feels at once (co-existing) “a certain tangible figure, bulk, and consistency”
(NTV,x96). This is one complex tangible idea. It is, in part, because of the co-existence
of those simpler
21
ideas in a single sensory experience that all of them are attributed to
the same body, and so ‘combined together’ by our linguistic practices.
6.3.4 Immediate Perception
I have argued that bodies, in Berkeley’s view, are mere quasi-entities, like forces. It
will certainly be objected to this interpretation that there is at least one powerful con-
trast, for Berkeley, between forces and bodies: bodies are immediately perceived (PHK,
xx38, 95; DHP, 230) and forces are not (DM,xx4-5, 10). Furthermore, since bodies are
immediately perceived, they would seem to be available to be labeled.
21. Kenneth Winkler has argued convincingly that Berkeley does not believe in absolutely simple ideas,
since he holds that such ideas would be abstract. See Winkler 1989, ch. 3.
226
Berkeley’s claim that bodies are immediately perceived has caused a great deal of
confusion, because it is difficult to understand how this can be reconciled with his insis-
tence in other texts that only ideas are immediately perceived (see, e.g., PHK,x1; DHP,
175; Alc,x4.10).
22
These difficulties are exacerbated by the fact that it is unclear exactly
what Berkeley means by ‘immediate perception’ (Winkler 1989, 149-154; Pappas 2000,
ch. 6). On what I take to be the most plausible reading,
23
mediate perception, for Berke-
ley, is perception by suggestion. That is (to a first approximation), mediate perception
of B occurs when, because of some prior connection between A and B, perceiving A
leads me to think ofB. Thus when Berkeley says that bodies are (sometimes) perceived
immediately, he means that when I perceive (e.g.) a table, it is not the case that I first
perceive some non-table idea and, as a result of this, think of the table.
My interpretation can accommodate Berkeley’s claim that perception of bodies is
(sometimes) immediate. On my reading, the table is a quasi-entity arising from the
rules for the use of the word ‘table,’ and, by those rules, the visual ideas we experience
are attributed to (predicated of) the table. When I have a certain brown sensory idea, I
am seeing the table, not in virtue of seeing an idea which suggests the table, but in virtue
of seeing an idea which, by the rules for the use of the word ‘table,’ is attributable to the
table.
I can thus agree that on Berkeley’s view we perceive tables, but not forces, imme-
diately. A force can only ever be inferred from perception of motion. Thus a transition
must always be made from thinking about motion to thinking about force. However, no
such transition is necessary in perceiving a body, and this is because bodies and forces
are related to ideas in different ways.
22. For discussion, see Pitcher 1986; Winkler 1989, 155-161; Pappas 2000, 172-208; Hight 2007b,
94-105; Atherton 2008b; Rickless 2013, ch. 2.
23. Recently defended by Rickless 2013, ch. 1.
227
That bodies are immediately perceived means only that they are perceived without
suggestion. Bodies can be perceived without suggestion because perceiving an idea
attributable to a body is constitutive of perception of the body, and this is the case re-
gardless of whether the particular perceiver actually can or does attribute the idea to the
body. The fact that my perception of the body is immediate in this sense does not imply
that I have the kind of pre-linguistic awareness which would be necessary for names of
bodies to be genuine referring expressions. There is, in other words, no reason why a
mere quasi-entity cannot be immediately perceived, as Berkeley holds that bodies are.
6.3.5 Predication
The word ‘apple,’ like the word ‘force,’ is a bit of technology for helping us navigate
the world of sense experience. Like ‘force,’ ‘apple’ becomes a genuine word, rather
than merely a sound, and becomes a useful piece of technology, by playing a role in
a sign system governed by conventional rules tied to perception and action. However,
‘apple,’ unlike ‘force,’ is first and foremost a subject of predication. Although ‘force’
is grammatically a noun, within Newtonian physics every force must be attributed to a
body. This is done, not by speaking of the body ‘being’ a force, but by speaking of it
‘exerting’ or ‘imparting’ a force, or ‘striking’ another object with a certain force. Sen-
sible qualities are also ordinarily attributed to bodies, though here we use the ordinary
idiom of predication. Thus we speak both of the apple being red, and of the (mythical)
apple striking Newton’s head with such and such a force.
With respect to the predication of sensible qualities, Berkeley says that “to say a
die is hard, extended, and square, is . . . only an explication of the meaning of the word
‘die’” (PHK,x49). The usual interpretation of this remark is that Berkeley is endorsing a
class membership analysis of predication (Muehlmann 1992, 181-182; Stoneham 2002,
228
246-249); that is, a die is a class of ideas, and the sentence ‘a die is hard, extended, and
square’ means that hardness, extension, and squareness are among the ideas included in
that class.
This cannot be, strictly speaking, correct, for hardness, extension, and squareness
would be abstract general ideas, and hence Berkeley rejects them. Presumably, what the
commentators in question actually mean is that the die includes hard ideas, extended
ideas, and square ideas, ‘hard,’ ‘extended,’ and ‘square’ here being general terms which
can be used to name the ideas in question. If the sentence is interpreted generally, it will
be saying that every collection of ideas which is correctly called ‘die’ includes ideas of
these sorts; if it is interpreted particularly, then it will be saying that some particular die
includes ideas of these sorts.
The difficulties here are well-known. Berkeley was a nominalist and would not have
liked to be committed to classes (Muehlmann 1992, ch. 2, et passim; Stoneham 2002,
x7.2). Furthermore, Berkeley would need to include in the classes ideas never actually
perceived, but he explicitly denies the existence of such ideas (DHP, 234; Bennett 1971,
x32; Pappas 2000, 195; Hight 2007b, 93; Rickless 2013, 96-97). These difficulties could
be avoided by appeal to divine ideas, but this would create other difficulties related to
Berkeley’s response to skepticism (see below, ch. 9).
The thesis that ‘die’ is a quasi-referring term, like ‘force,’ when combined with the
interpretation of Berkeley’s philosophy of language I have been defending, provides a
solution to these difficulties. For Berkeley, to ‘explicate the meaning’ of a term is to
provide instructions for how to use it. The sentence ‘a die is hard, extended, and square’
provides instructions for the use of ‘die’ in much that way the formula ‘F = ma’
provides instructions for the use of the word ‘force.’ ‘F = ma’ tells us that, when a
body accelerates (undergoes a change in its state of motion) a force equal to the product
229
of the mass and the acceleration has been applied to it. If the acceleration is due, for
instance, entirely to gravity, then we attribute the force to the body (or system of bodies)
whose center of mass the accelerated body is accelerated towards. This is, on Berkeley’s
theory, simply a rule about the correct use of the word ‘force’ (seex8.4, below).
The rules for the use of the word ‘die’ are, in a way, the reverse of the rules for
‘force,’ since a die is a thing (subject) rather than a quality. What the rule tells us is that
‘hard,’ ‘extended,’ and ‘square’ are among the sensible qualities which can be predicated
of a die. In the ‘thing language,’ all sensible qualities
24
must be attributed to some thing.
This is a tool by which we organize sensible qualities. Thus, for instance, I have a
tangible square idea and I ask ‘what am I feeling?’ – that is, to what thing should this
idea be attributed? I am looking for a noun to put in subject position. What the rule
tells me is that ‘die’ is a candidate noun. Thus upon having the square idea, I may assert
‘the die is square.’ In making this assertion, I would be claiming that the square idea is
nomically connected to certain other ideas, among them some extended ideas and hard
ideas. (Of course, no idea could be either hard or square without being extended, so
there is some redundancy in Berkeley’s example.) It is by these very rules of usage that
the mind groups the qualities together: for the qualities to be “united into one thing” just
is for them to “have one name given them” (DHP, 249; cf. Beardsley 2001, 259).
The fluent speaker of the language has habitual connections between the words and
various ideas. Thus, if one asserts ‘the die is square,’ the fluent speaker will come to
expect a square idea to be nomically connected with certain other ideas. Berkeley’s
apple (PHK,x1) and cherry (DHP, 249) examples can be handled similarly.
24. Or at least all those that are taken to be ‘real,’ as opposed, e.g., to after-images.
230
In this fashion, Berkeley can handle both particular and universal predications.
When I say ‘this desk is black,’ I am affirming that certain black ideas are nomically con-
nected with ideas that have the organization assumed by the rules for the word ‘desk.’ As
a result of this information, competent speakers will expect to receive certain other ideas
together with the black ideas in question. Similarly, if I assert ‘all ravens are black,’ I am
leading you to expect, whenever you have one of the other raven ideas, to have certain
black ideas.
Contrary to what Berkeley says, these statements appear to go beyond mere expli-
cation of the words ‘desk’ and ‘raven’ (Brook 1973, 80n7). Furthermore, it seems that
they must go beyond mere explication, for if they did not, we would be in danger of
rendering these statements analytic, and hence necessary (Stoneham 2002, 257-259).
Berkeley is in need of an account of contingent predications.
Two answers to this difficulty are possible. First and simplest, it could be pointed
out that Berkeley does not say, in PHK,x49, that he is providing a general account
of predication. On this interpretation we would say that Berkeley is simply giving an
illustrative example, and has chosen to illustrate his point with essential predication.
The second answer would say that Berkeley does not actually recognize this kind of
sharp distinction between analytic, necessary statements on the one hand and synthetic,
contingent statements on the other.
25
On this reading, in saying ‘all ravens are black’ I
would be suggesting that the word ‘raven’ should only be used to label black things. If
I assert sincerely, then I am already following the rule I am proposing, so it is already a
fact about my use of the word ‘raven,’ and hence a fact about what I take the meaning
of ‘raven’ to be, that it labels only black things. If we were to encounter something like
a raven in all respects except color, we would have to respond by revising our rules in
25. This would, of course, be an important addition to the list of respects in which Berkeley anticipates
Quine (see Quine 1951).
231
some way, but we would have significant freedom as to how to revise our rules. It may
be indeterminate whether such revision involves changing the meaning of ‘raven.’
The kind of indeterminacy envisioned here is not indeterminacy about what rules
are followed by particular individuals – the kind of indeterminacy I argued against in
x5.1. It is, rather, indeterminacy about which, among the rules individuals follow, are
linguistic conventions. I expect most others in my community to believe that all ravens
are black, and hence to follow a rule whereby they do not label non-black things as
‘ravens’ (on the relationship between belief and rule-following, see below,x7.1). Since
we all follow this rule and rely on others to follow it, it amounts to a convention. What
I am suggesting is that there may be no sharp line between conventional belief and the
conventional rules of language. The plausibility of this suggestion will be reinforced
when I argue, inx7.2.4, that Berkeley recognizes the possibility that the very structure
of our language may (and, to some degree, does) exhibit systematic mismatch with the
world.
The language of bodies is no more (or less) than a new way of talking about sensible
qualities. We group the sensible qualities together by attributing them to a common
subject, which is to say, predicating them of some noun phrase. This language, like
other advances in notation, provides us with new and different ways of thinking, which
allow us to represent complicated regularities in our sense perceptions which we would
never be able to represent without this technology. Furthermore, once we have the
language of bodies, we can develop the language of physics, and thus attribute ‘occult
qualities,’ such as force, to these bodies, thereby further expanding the range of natural
regularities we can represent. However, as has already been mentioned, forces, and
many of the other occult qualities of Newtonian physics, are not attributed to bodies by
means of a simple copula, as sensible qualities usually are; rather, they are attributed by
232
means of verbs such as ‘exerts,’ verbs borrowed metaphorically from the language of
action (DM,x3), to which we now turn.
6.4 Actions
According to Berkeley, “spirits and ideas are things so wholly different that, when we
say ‘they exist’, ‘they are known’, or the like, these words must not be taken to signify
any thing common to both natures” (PHK,x142). Indeed, at the beginning of the Prin-
ciples, Berkeley seems to imply that, although there are such things as spirits and their
actions, we have no knowledge of them: “besides all that endless variety of ideas or
objects of knowledge, there is likewise something which knows or perceives them, and
exercises divers operations . . . about them” (x2). Spirit is not a thing known, but rather
a thing which knows, and these here seem to be mutually exclusive. As Berkeley later
says, ideas “cannot represent unto us, by way of image or likeness, that which acts”
(PHK,x27; cf. N,x684). As a result, neither spirit nor its acts are, strictly speaking
objects of knowledge; the self is rather a subject than an object.
26
Nevertheless, “in a
26. That this subject/object contrast is fundamental to Berkeley’s thought is, I take it, the main thesis
of Bettcher 2007. For an interpretation along these general lines which is perhaps somewhat less radical
than Bettcher’s or my own, see Winkler 2011.
It may be objected to my interpretation that PHK,x1 mentions among the ‘objects of human knowledge’
things “perceived by attending to the passions and operations of the mind.” There has been a great deal
of scholarly controversy over this clause (see Tipton 1974, 74-79; Flage 1985, 421-422; Ayers 2005,
47-48; Winkler 2011, 233-234). I interpret Berkeley here to be drawing a contrast between the objects
of knowledge and the subject of knowledge. The former are all ideas. This commits me to the view
that, for Berkeley, there are, in addition to our notions of the mind and its actions, ideas “perceived by
attending to the passions and operations of the mind.” Note that (contrary to Winkler) this does not carry
the un-Berkeleian implication that the operations of the mind are ideas, or that there are any ideas of the
operations of the mind. On this interpretation, the ideas we perceive when we attend to the passions and
operations of the mind are every bit as passive as the ideas of (external) sense, hence they cannot represent
(by resemblance) the mind or its actions. They must, like all ideas, be objects of the mind’s acts; perhaps
some pleasures or pains or moods are like this. In Berkeley’s notebooks, Locke’s ‘uneasiness’ seems to
be an idea of reflection (N,x653). For my purposes, nothing much turns on this issue, except insofar as it
might be thought to cast doubt on my general interpretation of PHK,xx1-2, so I will not discuss it further.
233
large sense, indeed, we may be said to have an idea of spirit, that is, we understand
the meaning of the word” (PHK,x140 [1710 ed.]). In the 1734 editions of the Princi-
ples and Dialogues, Berkeley introduced a technical use of the word ‘notion’ to capture
this ‘large sense’ of ‘idea’ and distinguish it from the narrower sense (xx89, 142 [1734
ed.]).
27
Actions and spirits are known only in a sense of ‘known’ entirely different from
the sense in which ideas and bodies can be said to be ‘known’. This is what Berkeley
means in those passages where he contrasts spirits with objects of knowledge (x2), or
even explicitly denies that we know spirits (N,xx576, 576a, 829): when we assert ‘ideas
are known,’ the thing we are saying about ideas is something which could not possibly
be true of spirits or their actions.
28
Nevertheless, we are able to speak intelligibly of
actions and spirits – that is, we have notions of them. Having notions is, for Berkeley,
tightly connected with understanding words (PHK,x27).
29
In what follows, I shall have
little to say, explicitly, about notions and will focus instead on how we manage to talk
about spirits and their actions. Following the same pattern as our discussion of qualities
and bodies, I will first discuss Berkeley’s account of action words, then his account of
the attribution of actions to spirits.
Philonous tells Hylas, “I have no notion of any action distinct from volition” (DHP,
239). When this assertion is interpreted in light of Berkeley’s notebooks, it becomes
27. For discussion, see R. M. Adams 1973, 58-59; Woozley 1976; Flage 1985; Winkler 1989, 279-282.
28. Thus we do not need to hold, with R. M. Adams 1973, 63, that Berkeley went through an early
phase of development in which he thought the mind was unknowable, and we certainly do not need to
hold, with Tipton 1974, 72-74, that Berkeley failed to update the opening section of the Principles to
reflect his ‘new’ view that the mind is knowable. Berkeley’s consistent view, stated explicitly in PHK,
x142, is that there are two distinct senses of ‘know,’ and in one of these senses it is correct to say ‘spirits
are known’ and in the other it is not.
29. See R. M. Adams 1973, 59; Tipton 1974, 270; Woozley 1976; Winkler 1989, 281. Berkeley likely
got this usage of ‘notion’ from Sergeant. See Sergeant 1697, Prefx21 and p. 33. For discussion of
Sergeant’s theory of notions, see above,x1.4.
234
clear that Berkeley means to be identifying action with volition (N,xx621, 635, 644; cf.
Bettcher 2007, 72-74). The attempt to separate acting (or causing) from volition, even
in thought, is, in Berkeley’s view, an illicit abstraction (Roberts 2007, 117-120).
V olition is contrasted by Berkeley with perception (PHK,x27). The latter is in some
sense, or to some degree, passive, insofar as we typically do not have voluntary control
over which particular ideas we perceive (PHK,x28; DHP, 195-197); nevertheless, only
active things (minds or spirits) can perceive (PHK,xx2-3, 6-7, 138). I will not here at-
tempt to untangle Berkeley’s uses of ‘active,’ ‘passive,’ and related vocabulary.
30
Suffice
it to say that perceiving, whether it is to be regarded as active or passive, or as active in
some respects and passive in others, is nevertheless something which must be done by
an active being, and is closely enough related to activity to preclude our having any idea
of it (PHK,x27; Cummins 2007,x5). For this reason, despite the lack of clarity about
whether perception is ‘active,’ I will persist in calling it an ‘action’ and treat it together
with volition.
According to Berkeley, actions and active beings are to be understood by ‘reflexion’
on one’s own mind (PHK,x140; DHP, 231-233). In giving an account of my ability to
conceive of God, Philonous says, “I have . . . in myself some sort of an active thinking
image of the Deity” (DHP, 232). In the Principles, Berkeley says that just as we con-
ceive of ideas not actually perceived by us by having ideas which resemble them, “so
we know other spirits by means of our own soul, which in that sense is the image or idea
of them” (PHK,x140). We will come to the spirits themselves momentarily; as for the
names of the actions attributed to spirits, this suggests that they can be understood by
30. On this, see McKim 1989; Migely 2007; Bettcher 2007, ch. 5; Winkler 2011, 241-249.
235
analogy to the names of sensible qualities. As sensible qualities are known in the sens-
ing of them, so actions are known in the doing of them.
31
We have ‘reflective awareness’
of our own actions, and just as we can use words to label ideas we have, so we can use
them to label actions we are aware of performing. By this means, action words become
genuine referring expressions.
Note that in both the case of ideas perceived by other minds and the case of actions
done by other minds, reference depends on resemblance. Berkeley is not here lapsing
back into the general theory of representation by resemblance; rather, his view is that I
am following a certain rule in determining whether to label ideas as ‘red’ or actions as
‘imagining.’ The distinction drawn by the rule I follow cannot apply to things utterly
unlike the exemplars by means of which the rule is learned. A person blind from birth
could distinguish soft from hard, and she could use the words ‘soft’ and ‘hard’ in such
a way that every idea she perceived was either one or the other. (Her usage would, of
course, be non-standard.) Nevertheless, despite the fact that the classification is, from
her perspective, exhaustive, there would be no sense in asking whether red was soft or
hard. Red is simply too different from the objects which she has learned to classify for
it to have a place in here classification scheme (cf. NTV,x128).The blind person can
still describe red with some of her most general words, like ‘thing,’ ‘idea,’ and ‘sensible
quality,’ and she can use the word ‘red’ in a way that is parasitic on the labeling use by
sighted members of her linguistic community. What she cannot do is follow rules which
directly distinguish among visual ideas.
Our situation with respect to ideas had by other minds and actions performed by
other minds is not like the case of the blind person. Even though, since these ideas and
actions exist in other minds, I cannot actually apply my rule to them, there is a fact of
31. Cf. Tipton 1974, 267: “Doubting, to take one case of an act of mind, is not an idea because it is not
primarily an object for mind but rather something we are aware of through doing it.”
236
the matter about what my rule says about them, which is to say, about whether my words
apply to them (cf. Craig 1982, 556). That there is such a fact of the matter is due to their
resemblance to my own actions and ideas. Thus the word ‘imagining’ originally gets its
meaning from acts I am reflectively aware of myself performing, but can nevertheless
genuinely refer to acts of other minds.
Note further that I do not need to have a red idea or perform an act of imagining in
order to think about red or think about imagining (see Alc,x7.7). Once I acquire the rule
whereby ‘red’ is used as a label for red ideas, and ‘imagining’ is used as a label for acts
of imagining, I can use these words to think about red or about imagining even when I
am not presently perceiving a red idea or performing an act of imagining.
32
Other ideas,
besides words, can also be used as signs in this way.
In this section, we have been concerned only with how it is possible for me to “ap-
prehend the possibility of the existence of other spirits and ideas,” that is, of actions not
done by me and ideas not perceived by me (DHP, 232); how my belief that there actually
are such actions and ideas can be justified will be discussed below (x8.6). Berkeley’s
view is that action words initially get their meaning by individuals’ use of them to label
their own particular actions, but that individuals can coherently suppose them to apply
to actions done by others provided those actions are similar enough to the individual’s
own actions that the rule has application to them.
32. If I am thinking silently, I will be imagining the word ‘imagining’ (either visually or auditorially),
so strictly speaking there must be an act of imagining going on in this case. But suppose I am ‘thinking
out loud’ or ‘thinking along with’ a written text; then the word ‘imagining’ will be sensed rather than
imagined.
237
6.5 Spirits
So far, we have seen that the names of sensible qualities and actions are genuine refer-
ring expressions, while the names of occult qualities and bodies are not. In this section,
I defend the claim that ‘spirit’ and related terms are, on Berkeley’s view, genuine re-
ferring expressions. I begin with a careful examination of certain problematic aspects
of Berkeley’s thought which might cast doubt on my thesis, then proceed to show how
these doubts can be dispelled.
At the beginning of the Principles, contrasting spirits with ideas, Berkeley tells us
that by the words ‘mind’, ‘spirit’, ‘soul’, or ‘my self’, “I do not denote any one of
my ideas, but a thing entirely distinct from them” (PHK,x2). Later in the Principles,
Berkeley reiterates: “the words ‘will’, ‘soul’, ‘spirit’ do not stand for different ideas or,
in truth, for any idea at all, but for something which is very different from ideas, and
which being an agent cannot be like unto, or represented by, any idea whatsoever” (x27
[1734 ed.]).
33
An even stronger statement occurs near the end of the Principles, where
Berkeley says that ‘soul’, ‘spirit’, and ‘substance’ “signify a real thing, which is neither
an idea nor like an idea, but that which perceives ideas and wills and reasons about
them” (x139).
Although Berkeley’s talk of these words ‘denoting’ or ‘standing for’ something fa-
vors the interpretation of ‘spirit’ (etc.) as a genuine referring expression, it is not con-
clusive, for I have suggested that Berkeley has no reason to object to the use of words
like ‘refer’ or ‘mark out’ or ‘denote’ in connection with quasi-referring terms. Indeed,
as I noted above, Berkeley says that the word ‘apple’ ‘signifies’ the quasi-entity we con-
struct out of certain sensible ideas (x1). Furthermore, we cannot read too much into the
33. The 1710 edition differs only in having a longer list of examples of spirit words.
238
use of the word ‘real,’ since, as I will argue below (x6.6), mere quasi-entities can, on
Berkeley’s view, be correctly described as ‘real things.’
Another factor which favors the interpretation of ‘spirit’ and related terms as genuine
referring expressions is Berkeley’s defense of the status of spirits as genuine substances.
Berkeley is content to allow spirits, ideas, and bodies all to be called ‘things’ (PHK,
xx38-39, 89), but spirits alone are properly termed ‘substances’ (xx7, 135), and the word
‘substance’ (and its equivalents in other languages) connotes metaphysical fundamental-
ity in both the Aristotelian and Cartesian traditions (Aristotle Cat. 5 1b11-19; Met., Z1
1028a10-15; CSM, 1:210-211). This is further supported by a passage in Siris, where
Berkeley says that “mind, soul, or spirit truly and really exists [while] bodies exist only
in a secondary and dependent sense” (Siris,x266; cf. N,x24; for discussion, see Roberts
2007, 23-24).
This line of thought is, however, again inconclusive. Berkeley’s central claim is that,
if ‘substance’ is to be intelligible it must be synonymous with ‘spirit.’ The claim is that
spirit is the only intelligible concept ‘in the neighborhood’ of what has traditionally been
called ‘substance.’ The claim is not that spirit corresponds precisely with the traditional
notion of substance.
In defending the claim that ‘substance’ should be reinterpreted as a synonym of
‘spirit,’ Berkeley does not appeal to fundamentality, but instead to two other elements of
the traditional notion of substance. The first, and most important, is the notion of sup-
port, that is, the relation of a substance to (in Aristotelian language) its accidents, or (in
Cartesian language) its principal attribute and various modifications (see the previously
cited references). It is this element of the tradition that plays the largest role in Locke
(EHU,xx2.23.2-3), and it was Locke’s discussion that likely had the most significant in-
fluence on Berkeley. Berkeley argues that the only intelligible notion of support (in this
239
context) is perception, and spirit is, by definition, that which perceives (PHK,xx16-17,
49, 135; DHP, 197-200, 234; Cummins 2007).
The second element of the traditional notion of substance which Berkeley takes over
is the view that a substance is a thing which can have multiple distinct accidents and
persist through a change in its accidents (Aristotle Cat. 5 4a10-20). In Berkeley’s view, a
spirit, like a traditional substance, can perform various actions and perceive various ideas
and persist through a change in the actions it is performing and ideas it is perceiving
(DHP, 234). Since Berkeley is, in certain respects, a critic of the traditional doctrine
of substance, we should not simply assume that Berkeley associates these traits with
fundamentality, as the tradition does.
Furthermore, the quoted text from Siris is the only place in his published works
where Berkeley speaks of spirit existing in a more primary or fundamental sense than
body, and that passage is an exposition of the views of “the Pythagoreans and Platonists”
(Siris,x266). Now, Berkeley is clearly sympathetic to the view he is expositing; he says
that these philosophers “had a notion of the true system of the world.” Indeed, John
Russell Roberts argues, with considerable plausibility, that “one of [Siris’s] central aims
is to show that Berkeley’s metaphysics has its roots in an ancient and venerable tradition
that shares his fundamental metaphysical thesis: it places minds first in the order of
beings” (Roberts 2007, 24; cf. Siris,x263). However, Berkeley himself warns us that
the “hoary maxims . . . scattered in [Siris], are not proposed as principles, but barely as
hints to awaken and exercise the inquisitive reader” (Siris,x350). It is only in light
of independent argument, based on Berkeley’s other writings, that we can have any
confidence in identifying those among Siris’s ‘hoary maxims’ which Berkeley himself
endorses as approximately or entirely correct. As a result, it does not seem that there is
a simple, direct textual argument to be made for the claim that ‘spirit’ and related terms
240
must be genuine referring expressions on account of the metaphysical fundamentality
of spirit.
There are, however, two indirect arguments which provide strong evidence for the
metaphysical fundamentality of spirit. The first is the way in which perception replaces
inherence as the preferred notion of ‘support’ in Berkeley’s theory. The notion of in-
herence played two roles in the tradition: it was held that, necessarily, every property
instance must inhere in some substance, and it was held that every true predication is
true because the property denoted by the predicate inheres in the entity denoted by the
subject (e.g., ‘snow is white’ is true because whiteness inheres in snow). Descartes com-
bines both elements when he writes, “if we perceive the presence of some attribute, we
can infer that there must also be present an existing thing or substance to which it may
be attributed” (CSM, 1:210, emphasis added; cf. EHU,x2.23.1).
34
Berkeley accepts the
claim about the modal metaphysics of properties, but rejects the claim about predication
(cf. Cummins 2007, 134-138). Similarly, according to Aristotle, non-substances “are
said to be because they are, some of them, quantities of [substance], others qualities
of it, others affections of it, and others some other determination” (Aristotle Met., Z1
1028a18-20); thus, for instance, it is Aristotle’s view that red exists in virtue of some
substance’s being red. According to Berkeley, “the existence of an idea consists in being
perceived” (PHK,x2); that is, red exists in virtue of some substance’s (mind’s) perceiv-
ing red. It is in this sense that perception, for Berkeley, plays the role inherence played
in the tradition. However, Berkeley explicitly rejects the claim that it follows from this
that red should be predicated of any substance (mind) (x49). The key point here is that
34. The Latin of the last clause is in fact rather more emphatic than the translation indicates. It reads,
“cui illud tribui possit, necessario etiam adesse” – more literally, “to which it [sc. the attribute] may
be attributed, and [which it] is even necessary that it be present in.” I take it ‘etiam’ is here used as a
conjunction “to introduce a stronger statement” (C. T. Lewis 1890, s.v. ‘etiam’, def. II). Thus the second
half of the clause is intended as a strengthening of the first half: not only may the attribute be attributed
to the substance, it must necessarily be in the substance, otherwise the attribute could not exist at all.
241
the part of the substance tradition Berkeley accepts is the claim that necessarily every
non-substance depends on some substance for its existence. This certainly seems to
make spirits (substances) metaphysically fundamental.
Second, Berkeley argues that only substances can be genuine efficient causes (PHK,
x26). This, again, seems to imply fundamentality.
Non-substances are of two kinds, ideas and actions. Ideas exist by being perceived
by spirits, while actions exist by being done by spirits. This certainly seems to provide
the most straightforward interpretation of Berkeley’s text. If this is the case, however, it
seems that ‘spirit’ and related terms must be genuine referring expressions, since it does
not seem that mere quasi-entities could play this role. The next question we must ask is,
therefore, how do we succeed in using these terms to label the entities in question?
In the Dialogues, Philonous gives the following account of the matter:
I . . . know that I, who am a spirit or thinking substance, exist as certainly
as I know my ideas exist. Farther, I know what I mean by the terms ‘I’
and ‘myself’; and I know this immediately, or intuitively, though I do not
perceive it as I perceive a triangle, a colour, or a sound. The mind, spirit, or
soul, is that indivisible unextended thing, which thinks, acts, and perceives
. . . My own mind and my own ideas I have an immediate knowledge of; and
by the help of these, do mediately apprehend the possibility of the existence
of other spirits and ideas (DHP, 231-232).
In the expanded text of the 1734 edition, Philonous reiterates the point twice, in slightly
different words:
[T]he being of my self, that is, my own soul, mind, or thinking principle, I
evidently know by reflection (233 [1734 ed.]).
242
I know or am conscious of my own being; and that I myself am not my ideas,
but somewhat else, a thinking active principle that perceives, knows, wills,
and operates about ideas. I know that I, one and the same self, perceive
both colours and sounds; that a colour cannot perceive a sound, nor a sound
a colour; that I am therefore one individual principle, distinct from colour
and sound and, for the same reason, from all other sensible things and inert
ideas (DHP, 233-234 [1734 ed.]).
35
It is clear from these texts that our ‘spirit’ talk must originally derive its meaning-
fulness from its use to label the self. Berkeley’s emphasis on the immediacy of my
knowledge of myself might seem to indicate that I can simply apply the label ‘spirit’
directly to myself, in the way I apply ‘red’ to an idea or ‘imagining’ to an action. How-
ever, if this were the case then it would appear that I could have a notion of myself which
was utterly independent of my ideas and actions, and Berkeley explicitly denies that this
is possible: “by the word ‘spirit’ we mean only that which thinks, wills, and perceives”
(PHK,x138). Furthermore, Berkeley denies that it is possible to “abstract the existence
of a spirit from its cogitation” (x98). This is in line with Berkeley’s famous comment
in his notebooks that “existere is percipi or percipere”(N,x429; cf. N,x646). Berkeley
later added to this entry the words: “or velle i.e. agere.” Spirits cannot be separated,
even in thought, from their perceiving, willing, and acting, any more than ideas can be
separated from their being perceived. Thus it seems that we know even our own spirit
through our own activity.
36
35. The use of the word ‘principle’ in these passages also suggests fundamentality. See Siris,x335.
However, this is again not conclusive by itself, since Berkeley explicitly recognizes several other meanings
of the word ‘principle’ (Alc,x3.1).
36. This difficulty seems to be part of what led Daniel Flage to the conclusion that we know spirits
only by relative notions, which he connects with Russellian ‘knowledge by description’ (Flage 1985).
However, Berkeley quite clearly claims that my knowledge of myself (one particular spirit) is immediate.
243
This does not violate Berkeley’s claims of immediacy, for Berkeley certainly holds
that we know our ideas immediately, or intuitively, yet we surely know them only by
perceiving them. The act of perception itself is not the kind of intermediary that Berke-
ley is concerned to rule out here. Thus Berkeley can hold, consistently with his claims
of immediacy, that I know myself in my activity. Perceiving essentially has subject-
verb-object structure (cf. Cummins 2007, 139-140). In my perceiving I have immediate
knowledge of myself, the subject, and my idea, the object. (Remember, however, that in
the preceding sentence the word ‘knowledge’ must, according to Berkeley, be equivocal:
the relation I bear to myself is not, and indeed cannot be, the same as the relation I bear
to my idea.)
This provides me with a kind of de re knowledge of my self, to which I can now
attach a label, ‘I’. This will be a genuine referring expression; it is used to label the
perceiver. This does not mean that I can label myself in a way that is independent of my
perceiving. Even though sensible qualities can be immediately labeled, they cannot be
coherently separated from their being perceived.
I now have a label for myself, one individual spirit. What I must do next is generalize
my concept of self into a concept of spirit. In order to do this, we simply introduce the
practice of attributing ideas not perceived by me and actions not done by me to some
other spirit.
Here is a more careful account of how this works. We said above that my rule for the
application of ‘red’ may give a determinate answer to the question of whether a certain
idea should be called ‘red’ despite the fact that I am not in a position to apply my rule
to that idea, since I am not the perceiver of that idea. The same applies to action words
such as ‘imagining.’ My rule for recognizing instances of imagining gives an answer to
My interpretation below is able to account for these data while still taking Berkeley’s claims of immediacy
with the utmost seriousness, rather than trying to explain them away, as Flage does.
244
the question of whether someone else’s act is to count as an act of imagining, but I am
not in a position to apply my rule to that act directly, in the way I apply it to my own
imaginings. Now what we have introduced is the practice of attributing perceptions and
other actions to an enduring subject, the self. This rule can equally well be used in the
case of any perception at all, though only, of course, by the one who has that perception.
Knowing, however, that any perception must necessarily be self-attributable, since the
rules for the use of ‘I’ and ‘self’ permit this, I can begin to speak of other ‘selfs,’ by
attributing to them actions not done by me. Assuming that the action so attributed really
occurred, this will be a genuine referring expression, though it operates, in Russellian
terms, by description rather than by acquaintance. My knowledge of other spirits, unlike
my knowledge of myself, is indirect.
There is a difficult question here about personal identity; that is, about how
Philonous is sure that “one and the same self, perceive[s] both colours and sounds”
(DHP, 234 [1734 ed.]). This will be dealt with below, along with the various other uses
of the words ‘same’ and ‘one’ (x6.6). A second concern, which I already mentioned
above (x5.4), is that it is not clear what justification Berkeley has for his claim that every
action must be done by someone. The short answer to this objection is, I believe, that
we are acquainted with particular instances of action from our own case, and we are
able to think about actions in general by the adoption of conventional rules for the use
of signs. But we are unable to abstract the action from the mind that does it. Whenever I
am immediately aware of any action, I am aware of myself doing it, and I cannot adopt
any rule which would apply the word ‘action’ beyond the scope of those actions which
are done by agents.
245
When this is combined with the impossibility of abstracting causation from action,
then we have Berkeley’s conclusion that there is an agent behind every instance of cau-
sation. However, it is still not clear how we are entitled to the claim that every change
must have a cause, a principle which Berkeley uses implicitly in arguing for the exis-
tence of God in the Principles (PHK,x26). We are frequently aware of changes without
being aware of causation/volition/action. Indeed, this is a central premise of Berkeley’s
argument here: because I am not immediately aware of the volition, I must attribute it to
some other agent. However, since I can separate awareness of the change from aware-
ness of a cause of that change, it seems that I can coherently conceive a change without
a cause.
Berkeley’s use of the causal principle was unlikely to be challenged by any of his
contemporaries, and this is perhaps why he does not even so much as state it, let alone
defend it (cf. Tipton 1974, 218-219).
37
Perhaps, indeed, it was so widely accepted as to
constitute one of the conventional rules for the proper use of such words as ‘cause’ and
‘change.’ However, as I already indicated (x5.4), it is possible for a mismatch with the
world to be built into the conventions of language, so this would not render Berkeley’s
assumption immune from criticism.
This much, at least, can be said on Berkeley’s behalf: if it is really true that we
cannot separate cause from volition, then the only possible genuine explanation is one
in terms of the actions of an agent (PHK,x107; Siris,x237). Thus, given Berkeley’s
analysis of ‘cause,’ we are faced with a stark choice: either we say that the observed
changes are the actions of some mind, or we say that they happen for no reason at
all. ‘Physical’ or ‘mechanical’ explanation will, of course, be left intact, for this is
37. Of course, the principle is challenged in Hume’s Treatise of Human Nature (THN,x1.3.3), and this
work was published in 1739-1740, which places it before the third edition of Alciphron, and also before
Siris. Berkeley, however, does not seem to have read it.
246
merely a matter of subsuming the phenomena under general laws, but there will be
no explanation of why the phenomena are orderly in the first place, nor will anything
at all make them happen. Thus if inference to the best efficient causal explanation is
justified, then Berkeley’s implicit reliance on the causal principle is justified, though the
causal principle would here not be a metaphysical law or analytic truth, but merely a
methodological assumption.
6.6 Some Ontological Vocabulary
In Berkeley’s notebooks, he writes, “’tis on the Discovering of the nature & meaning &
import of Existence that I chiefly insist.” In the same entry, Berkeley claims that it is
due to an error about the meaning of ‘exists’ that “many of the Ancient Philosophers run
into so great absurditys as even to deny the existence of motion and those other things
they perceiv’d actually by their senses.” Furthermore, Berkeley claims, it is his view
about the meaning of ‘exists’ that “puts a wide difference betwixt the Sceptics &c” and
himself (N,x491; cf. N,xx593, 604; PHK,x89). Similar remarks apply to words such as
‘real,’ ‘thing,’ and ‘being.’ In Berkeley’s view, one cannot intelligibly wonder whether
the things actually perceived by the senses exist; any doubt about this is symptomatic of
a confusion about the proper use of ‘exists’ in plain language (Pappas 2002, 56). To this
Berkeley would add that philosophers have not succeeded in assigning any intelligible
meaning to ‘exists’ other than the one they inherited from plain language.
38
38. Whether there might be a special ‘ontological’ sense of ‘exists,’ distinct from the plain language
sense, is one of the principal matters of dispute among the authors of the essays in Chalmers, Manley,
and Wasserman 2009. However, the main issue in that dispute is whether a wedge can be driven between
the truth (or correctness) of plain language statements of the form ‘there areF s’ and genuine ontological
commitment toF s. Here, Berkeley is clearly on the side of the ‘ontological realists:’ he holds that, among
the things which can be correctly said, in English, to ‘exist,’ some exist in an ontologically weightier
sense than others. What Berkeley denies is not the existence of an ontologically heavyweight notion of
existence, but rather the existence of an abstract philosophical sense of ‘exists’ which can be used to drive
247
We shall have more to say about this general strategy when we address Berkeley’s
response to skepticism in chapter 9. Our task for the present is to understand how,
according to Berkeley, the ontological terms ‘exists,’ ‘thing,’ ‘being,’ ‘real,’ and ‘same’
get to be meaningful, and how Berkeley’s account of these linguistic facts supports his
immaterialism.
As we have already observed, Berkeley asserts explicitly that ‘exists’ is equivocal as
applied to minds and ideas (PHK,x142). In another passage he makes similar remarks
about ‘thing’ and ‘being’ (x89). I will show that, more generally, Berkeley is committed
to the claim that all of this ontological language is equivocal across the categories we
have so far described; that is, ‘exists,’ ‘same,’ and so forth take different meanings as
applied to sensible qualities, bodies, actions, and spirits.
39
I begin with ‘exists.’
6.6.1 ‘Exists’
Sensible Things
At the beginning of the Principles, Berkeley says that “an intuitive knowledge may be
obtained of [immaterialism], by any one that shall attend to what is meant by the term
‘exist’ when applied to sensible things” (x3, emphasis added). The word ‘exists’ takes
a distinctive meaning when applied to sensible things, i.e. sensible qualities and bodies.
Berkeley treats these two cases together because in both of them existence is inextricably
linked to being perceived; there are, however, differences between them.
Lumping ‘exists’ together with the existential use of ‘is’ (see DHP, 234), Berkeley
draws our attention to the following expressions: “There was an odour, that is, it was
a wedge between the existence of ideas and their being perceived, or the existence of actions and their
being done, or the existence of minds and their acting.
39. I leave out occult qualities because Berkeley does not discuss the application of these words to them,
and because they do not play the same important role in his metaphysics as the other categories.
248
smelled; there was a sound, that is to say, it was heard; a colour or figure, and it was
perceived by sight or touch” (PHK,x3). The attempt to separate the being or existence
of a sensible quality from its being perceived involves an illicit abstraction. Berkeley
writes, “as it is impossible for me to see or feel any thing without an actual sensation of
that thing, so is it impossible for me to conceive in my thoughts any sensible thing or
object distinct from the sensation or perception of it” (x5).
40
Berkeley’s reason for supposing that it is impossible to imagine an unconceived
sensible quality has been discussed above (x2.2.2). Berkeley’s view is that ideas are
intrinsically well-suited to represent only what resembles them exactly, but every idea
that I could use to represent something is an idea I conceive, hence all ideas represent
their objects as conceived, and, indeed represent them as conceived by me. However,
I have been arguing that, on Berkeley’s view, the introduction of conventional rules for
the use of ideas as signs can expand our powers of representation. Such an expansion
is needed, as we saw above, if I am to represent ideas in other minds. If, however, I
can introduce rules whereby I can represent sensible qualities as existing in other minds,
then perhaps it is possible to introduce rules for the use of signs whereby we would
represent sensible qualities as existing unperceived.
No such rule can be introduced. As we saw above (x5.1), it is essential to Berkeley’s
notion of a rule that it be followed by a mind, and there is no rule which could possibly be
followed by any mind such that some unperceived sensible qualities would be labeled as
‘existing’ (or labeled at all). In order for a mind to apply a label to a quality, the quality
must be perceived by that mind, so there cannot be a rule which instructs anyone to label
40. Muehlmann 1992, 19 claims that this text is ambiguous between an interpretation on which ‘so’ is
an ‘inference indicator’ and one on which ‘so’ is an ‘analogy indicator.’ Muehlmann favors the former.
He is, however, mistaken. The word ‘as’ at the beginning of the quotation clearly and unambiguously
indicates that an analogy is being drawn. Also, the use of ‘so is it’ rather than ‘so it is’ would sound odd
if ‘so’ were taken as an ‘inference indicator.’
249
what is unperceived. Thus any sensible quality that exists is perceived. But it is a truism
that the labels ‘is’ and ‘exists’ can be applied to just anything, so any sensible quality
that is perceived exists. Thus ‘perceived’ and ‘exists,’ as applied to sensible qualities,
label all the same things, namely, every sensible quality. “Their esse is percipi” (PHK,
x3).
In the very same passage of the Principles, Berkeley gives a different account of
the existence of his table: “The table I write on, I say, exists, that is, I see and feel
it; and if I were out of my study I should say it existed, meaning thereby that if I was
in my study I might perceive it, or that some other spirit actually does perceive it”
(x3). Berkeley lists three distinct conditions, each of which is sufficient for the proper
attribution of existence to the table. No such disjunction is applied to odors, sounds,
colors, or figures. Contrary to Ian Tipton, it really is the case, on Berkeley’s view, that
‘there was an odor’ can only ever mean ‘it was smelled’ (Tipton 1974, 101). An odor is
an idea (sensible quality), and not a body. The account of the application of ‘exists’ to
tables does not contradict the account of the application of ‘exists’ to sensible qualities,
for bodies and sensible qualities are different sorts of things, and ‘exists’ is equivocal as
applied to them.
41
Berkeley’s aim in this passage is to show that even for bodies existence cannot be
utterly separated from perception, although the dependence of existence on perception
is more complicated in the case of bodies than in the case of sensible qualities. This is
in line with the intention he stated in his notebooks, “to say the things . . . themselves
to really exist even w
n
not actually seen perceiv’d but still with relation to perception”
(N,x802, emphasis added). The first difference between bodies and sensible qualities
is, then, that bodies can exist when not actually perceived. A second issue is that we
41. One scholar who explicitly recognizes the need to separate the account of sensible qualities from
the account of bodies is Muehlmann 1992, 13-15, et passim.
250
do not ordinarily apply the predicate ‘exists’ to bodies which are not real,
42
though
philosophers, at least, are (or were in Berkeley’s day) in the habit of saying that dreamed,
imagined, or hallucinated ideas exist. Berkeley may be recognizing this point when he
uses the word ‘really’ in the notebook entry just quoted. Elsewhere in the notebooks,
Berkeley explicitly recognizes that ‘exists’ is not usually applied to imagined bodies,
such as chimeras, though he notes that there is no problem with a philosopher, such as
himself, extending the use of the word ‘exists’ to say of bodies, as of ideas, that some
which are not real (e.g., those that are imagined) nevertheless exist (N,x473).
43
To say that a body exists is to say that it satisfies one of three conditions: either it
is perceived by me, or it would be perceived by me were I appropriately situated, or it
is perceived by someone else. If I am not presently perceiving the table (I am not in
the study), then the first disjunct is ruled out, and so my utterance of ‘the table exists’
in this context can only mean that either I would perceive the table were I appropriately
situated or someone else actually perceives it. This is not a change in the meaning of
the statement; it is merely that one of the ways in which the statement could be true is
clearly ruled out by the context. This, I take it, is what Berkeley has in mind when he
talks about what he would mean by the utterance if he were out of his study.
There are three rather obvious problems with this interpretation of Berkeley, two
philosophical and one textual. First, what makes the subjunctive conditional in the sec-
ond disjunct true (when it is true)? Second, how can we know the conditional to be true?
Third, how are we to reconcile this reading with those texts in which Berkeley speaks
of bodies as ideas in the mind of God (e.g. DHP, 212, 230-231, 248, 254)? In my view,
42. This is the point behind Tipton’s complaint that Berkeley’s “assumption that whatever I conceive
must exist” commits him to the existence of “mermaids and chimeras” (Tipton 1974, 160).
43. Tipton recognizes the availability of a reply similar to this one, and indeed quotes this notebook
entry (162-164). However, he does not recognize that there is a contrast between ideas and bodies here,
and he also does not draw the connection to Berkeley’s remarks on the meaning of ‘real.’
251
the answers to all three of these questions depend on Berkeley’s theory of the perceived
world as a language in which God speaks to us. This theory will be examined in detail
in chapter 8, and in chapter 9 the results of that study will be applied to these questions.
Spirits and Actions
The existence of an action consists in its being done, and the existence of a spirit consists
in its acting (N,x429). That the existence of an action consists in its being done is guar-
anteed by the same reasoning which guaranteed that the existence of a sensible quality
consists in its being perceived. We know actions in the doing of them, and it is this that
gives us the opportunity to label them. As a result there is no possible rule by which the
label ‘exists’ could be given a broader scope, among actions, than the label ‘done.’ (Of
course, the non-existence of unperformed actions is relatively uncontroversial, whereas
the non-existence of unperceived sensible qualities is a distinctively Berkeleian thesis.)
The matter is somewhat more complicated for spirits, in much the same way it is
complicated for bodies. It follows from Berkeley’s theory of time that the mind always
thinks (PHK,x98). However, the ‘always’ here is with reference to that mind’s subjec-
tive time – i.e., to the “the succession of ideas” in that mind. Because of the orderliness
of the perceived world, we are able to construct an inter-subjective time order, and in
this order it may sometimes be correct to assert that there is a time when a certain mind
252
exists but does not think.
44
Berkeley mentions to Samuel Johnson that his early re-
flections about time led him into the ‘paradox’ “that the Resurrection follows the next
moment to death” (BW, 2:293; cf. N,x590).
45
If this were true, then there would be an
important sense in which Berkeley himself, for instance, presently (in 2013) exists but is
not presently thinking, for in Berkeley’s succession of ideas there is no idea which can
be said to be simultaneous, in the constructed inter-subjective time order, with any of
the 2013 ideas. Nevertheless, there are ideas of Berkeley’s which are inter-subjectively
before and (if there is to be a resurrection) after the ideas I am having now. It seems,
therefore, at least to be a possible and coherent use of the word ‘exists’ to say that
Berkeley presently exists on account of the fact that there are ideas in his subjective
time which come both before and after the inter-subjective present. Thus just as a body
need not be presently perceived in order to exist, a spirit need not be presently acting
in order to exist. This, however, depends on a use of ‘presently’ which is parasitic on a
more basic notion of time, namely the succession of ideas in an individual mind. In the
more basic sense of ‘presently,’ minds exist when and only when they are acting (and,
indeed, there is no time when a mind is not acting).
44. Berkeley certainly sees the need for an inter-subjective time order: “Bid your servant meet you at
such a time, in such a place, and he shall never stay to deliberate on the meaning of those words; in
conceiving that particular time and place, or the motion by which he is to get thither, he finds not the
least difficulty” (PHK,x97). This would make no sense if the only notion of time was the subjective time
of each individual. Unfortunately, Berkeley says little about how the inter-subjective time order is to be
constructed from the “succession of ideas.” Perhaps it is Berkeley’s silence on this point which leads
Tipton to say that Berkeley’s view of time is “surely unsatisfactory” (Tipton 1974, 275; cf. 278, 280-282).
45. I do not see anything paradoxical about this position, if we take it to refer to the individual’s subjec-
tive time; Berkeley’s real reason for hesitating is probably that the theological orthodoxy of this view is
questionable. On the orthodox view of the afterlife which Berkeley wished to defend, see Pearce, forth-
coming(b),x7. Alternatively, it may be that the view seemed paradoxical to Berkeley because he did not
see the need to distinguish between subjective and inter-subjective time with sufficient clarity.
253
6.6.2 ‘Thing’ and ‘Being’
We move now to the words ‘thing’ and ‘being.’ Berkeley writes, “‘thing’ or ‘being’ is
the most general name of all; it comprehends under it two kinds entirely distinct and
heterogeneous, and which have nothing in common but the name, to wit, spirits and
ideas” (PHK,x89; cf. N,x643). These words can be used to label just anything and,
indeed, can also be used to quasi-refer to whatever quasi-entities our sign systems have
introduced. This, of course, is unsurprising, since ‘being’ is just a noun form of the
existential ‘is.’
Berkeley also recognizes two narrower uses of ‘thing,’ a perfectly good plain lan-
guage use, and a nonsensical philosophical use. Concerning the plain language use,
Berkeley writes that “in common discourse . . . the several combinations of sensible qual-
ities . . . are called ‘things’” (PHK,x38; cf. DHP, 251). In this use, ‘thing’ is just a
synonym for ‘body.’ It is this use that Philonous has in mind when he says, “I am not
for changing things into ideas, but rather ideas into things” (DHP, 244). There is a cor-
responding narrow sense of ‘idea’ on which things and ideas are contrasted: namely, a
use on which ‘idea’ refers solely to ideas of the imagination (PHK,x33; DHP, 251).
There is, on the other hand, a narrow philosophical use of ‘thing’ “to denote some-
what existing without the mind” (PHK,x39). Here, the word ‘thing’ has an incoherent
usage equivalent to the usage of ‘matter’ to be discussed below. This usage also con-
trasts ‘thing’ with ‘idea,’ and it is this contrast, Berkeley says, that is the reason for
his use of the word ‘idea:’ ideas are the immediate objects of perception and depend
for their existence on being perceived. Berkeley denies that it is coherent to suppose a
contrast class of unthinking things which are not ideas. However, in the broad sense of
‘thing,’ there are things which are not ideas, namely, spirits (PHK,x39; cf. N,x872).
254
6.6.3 ‘Real’
Among sensible things, that is, sensible qualities and bodies, some are said to be ‘real.’
In plain language, this word gets its meaning from a contrast with those things that are
imaginary. Thus Berkeley says, “the sun I see by day is the real sun, and that which I
imagine by night is the idea [in the narrow plain language sense] of the former” (PHK,
x36). This is an empirical distinction. When Hylas asks how Philonous can distin-
guish “between real things and chimeras formed by the imagination or the visions of a
dream,” Philonous responds, in part, that “by whatever method you distinguish things
from chimeras on your own scheme, the same, it is evident, will hold also upon mine.
For it must be, I presume, by some perceived difference, and I am not for depriving you
of any one thing that you perceive” (DHP, 235).
The words ‘real’ and ‘imaginary’ are used, in plain language, to label two differ-
ent varieties of perceptions. As it turns out, Berkeley holds that ‘reality’ comes in
degrees: “The ideas of sense are allowed to have more reality in them, that is, to be
more strong, orderly, and coherent than the creatures of the mind” (PHK,x33, emphasis
added). However, Berkeley thinks that, as a matter of empirical fact, “there is no danger
of confounding” the real ideas with the imaginary ones (DHP, 235), for there are few if
any borderline cases. In our actual experience, there is a wide gulf between the “vivid
and clear” ideas “imprinted on the mind by a spirit distinct from us” and independent of
our will, and the “faint and indistinct” ideas of the imagination, or the “dim, irregular and
confused” ideas in dreams (235). Thus although, strictly speaking, reality, for Berkeley,
is a complex notion combining several criteria, each of which comes in degrees, we use
the terms ‘real’ and ‘imaginary’ as though they marked a sharp all-or-nothing contrast
because the ideas we actually experience tend to score either very high or very low on
all of the criteria together, rather than to give mixed or borderline results. This is a case,
255
again, where the conventions governing our sign system build in assumptions about the
world, and there is a conceivable course of experience which would cause, not merely
the falsity of our utterances, but the breakdown of the sign system (see above,x5.4).
In addition to sensible qualities, bodies are described as ‘real’ or ‘imaginary.’ When
we perceive a real body as a body, the sensible qualities we perceive suggest to us certain
other qualities, and these other qualities are imagined. We thus attribute to real bodies
both real and imagined qualities. It seems that if any of the qualities attributed to the
body are real, then the body itself is to be described as real. Note, however, that among
the criteria for reality of sensible qualities (ideas) are orderliness and coherence (PHK,
x33). Thus it is due, in part, to our success in using the language of body to talk about
these qualities that we say they are real. This can help to explain the unreality of after-
images: although these may be ‘lively’ or ‘vivid,’ they are not related to other ideas in
such a way as to allow us to make correct predictions using the technology of ‘body’
talk. If we attempt to attribute them to bodies, we necessarily go wrong. Hence we write
them off as ‘unreal.’
Berkeley does not apply the real/imaginary distinction to actions or spirits. Since
Berkeley does not discuss the matter, it is impossible to be certain about his reasons,
but a likely hypothesis is this: on Berkeley’s view, imaginary ideas and bodies may
coherently be said to exist. They exist, that is, in the mind of the one who imagines them.
Thus the real/imaginary distinction is a distinction between two kinds of existing things
(although, as I said above, the word ‘exist’ is not usually applied to imaginary bodies;
my point here is simply that it may coherently be used in this way). However, Berkeley
would not, I think, admit that such a distinction can be drawn between minds. Fictional
characters, for instance, are not existing things with ‘less reality’ than us; neither they
256
nor their actions exist at all. Thus ‘real’ as applied to spirits and actions can only mean
‘existing.’
6.6.4 ‘Same’
We now turn, finally, to the meaning of ‘same’ and the related term ‘one.’ In the Dia-
logues, Philonous distinguishes between two senses of ‘same.’ “Strictly speaking,” he
says, “we do not see the same object that we feel; neither is the same object perceived by
the microscope which was by the naked eye” (DHP, 245). Immediately after this remark,
Philonous introduces the notion, discussed above, that these qualities are attributed to
a body in virtue of their lawful connections. Then, he talks about “examin[ing] by my
other senses a thing I have seen” (245). This commits him to the claim that there is
an intelligible (though non-strict) sense in which we may “see the same object that we
feel:” the thing he saw before is the thing he will now feel. Clearly what it means to
see and feel the same thing in this non-strict sense is just to have a visible idea and a
tangible idea which are correctly attributed to the same body. Sameness of bodies is a
rather tricky matter, and there is danger of being caught up in purely verbal disputes, es-
pecially if one is under the illusion that there exists an ‘abstracted idea of identity’ (248).
The fundamental facts are just facts about which ideas are connected with which other
ideas. We then adopt conventions whereby we attribute these ideas to bodies and call
these bodies the same or different from one another. Similarly, there are verbal disputes
among philosophers about the application of ‘same’ to ideas. On one interpretation, two
people may be said to have the same idea if they are having sufficiently qualitatively
similar experiences; on another interpretation it is impossible, by definition, for two
people to have the same idea (DHP, 247; cf. N,x568). There is, according to Berkeley,
257
no substantive fact at issue.
46
This is merely a dispute about the meaning of ‘same.’
Since we are here talking about a technical usage of ‘same,’ the problem is to be solved
by stipulation.
47
Berkeley’s views on sameness of bodies can be further clarified by considering his
remarks on the closely related terms ‘one’ and ‘unit.’ Berkeley writes,
number . . . is entirely the creature of the mind, considering either an idea
by itself, or in any combination of ideas to which it gives one name, and
so makes it pass for an unit. According as the mind variously combines its
ideas, the unit varies . . . We call a window one, a chimney one, and yet a
house in which there are many windows and many chimneys has an equal
right to be called one . . . Whatever, therefore, the mind considers as one,
that is an unit. Every combination of ideas is considered as one thing by
the mind, and in token thereof is marked by one name. Now, this naming
and combining together of ideas is perfectly arbitrary, and done by the mind
in such sort as experience shows it to be most convenient: without which
our ideas had never been collected into such sundry distinct combinations
as they now are (NTV,x109; cf. PHK,x12; Siris,xx288, 346-347).
In this passage note first how closely Berkeley connects ‘considering’ with ‘calling.’
On the interpretation I have been developing, this close connection is quite understand-
able: in order to consider the ideas as one thing (body), I need to engage in the linguistic
practice of attributing those ideas to a body, and I can do this only if I have some way
46. For a detailed defense of this line of interpretation of Berkeley’s view about sameness of ideas, see
Bettcher 2007, 89-95.
47. At one point in his notebooks, Berkeley also warns that we are in danger of purely verbal disputes
about sameness of ‘Will’ (N,x714); ‘will’ is, of course, just another name for spirit (PHK,x27). Berke-
ley’s views about sameness of spirits are quite difficult to reconstruct. See below.
258
of talking about that body, or, in other words, only if that body has been introduced
as a quasi-entity in some sign system. The introduction of such a quasi-entity requires
a name. (Of course, we can use the very general names ‘thing’ or ‘body’ rather than
having a particular name.)
In Berkeley’s view, it is by the conventions for attributing qualities to bodies that
the various ideas come to constitute one and the same thing. These conventions are
conceptually relative: that is, they depend on what kind of body we are talking about.
48
Like the other conventions of plain language, these are adapted to practical uses and
hence not perfectly precise (MI,x19; DHP, 245-246). This is why there is danger of
purely verbal disputes about the sameness, for instance, of houses over time (DHP,
248).
Berkeley says little about the sameness of spirits. As a result, it is impossible to be
certain what his view was. There are, however, comments in Siris which lead to an in-
triguing speculation. There Berkeley says that “number . . . is an act of the mind” (Siris,
x288, emphasis added), and he contrasts spirits with bodies, saying that “the person or
mind of all created beings seemeth alone indivisible, and to partake most of unity. But
sensible things are rather considered as one than truly so” (x347). Berkeley also ap-
provingly cites the dictum of Aristotle that “it is the mind that maketh each thing to be
one” (Siris,x356; quoting Aristotle De Anima, G6 430b5-6), and at one point describes
the human person as “the self-thinking individual” (Siris,x257). The rather speculative
conclusion I would draw from this is that Berkeley’s claim that unity, and hence num-
ber, stems from an act of the mind is an absolutely unrestricted claim, applying even
to spirits. Spirits, on this interpretation, are ‘self-thinking individuals’ who necessarily
unify themselves by the reflective awareness in which the self is considered as one. This
48. Here Berkeley is no doubt following Locke (EHU, ch. 2.27).
259
is why spirits are more truly one than bodies: a spirit is literally (to use the Scholastic
phrase) ‘unum per se,’ one through or by itself, whereas a body is ‘unum per accidens,’
or, more accurately, ‘unum per aliud’ – one through or by another. Being ‘unum per
se’ is another traditional mark of substance, hence this also fits nicely with Berkeley’s
desire to see minds as substances. Furthermore, this reading would line up with Berke-
ley’s remark that “it should seem that personality is the indivisible center of the soul or
mind, which is a monad so far forth as she is a person” (Siris,x346), for it would take
consciousness of oneself as author of an action, Locke’s criterion of personal identity
(EHU,x2.27.9), as the metaphysical basis for identity of substances.
A similar approach to Berkeley’s theory of mind has been developed in great detail
by John Russell Roberts (Roberts 2007, ch. 4). However, I take issue with Roberts’
account on two points. First, on the account under consideration Berkeley is co-opting
Locke’s criterion of personal identity to do a great deal of metaphysical work. Per-
sons are fundamental to Berkeley’s ontology, and they are the fundamental explainers in
metaphysics, which is where all genuine explanation takes place. Roberts has Berkeley
doing this while at the same time taking seriously Locke’s claim that “Person . . . is a
Forensick Term appropriating Actions and their Merit” (EHU,x2.27.26; Roberts 2007,
100-102). Due to his understanding of Berkeley’s meta-ethics, this tight connection
between the word ‘person’ and moral discourse leads Roberts to attribute to Berkeley
the view that ‘person,’ and hence its synonym ‘spirit,’ is a non-cognitive term (Roberts
2007, 81-82). I have already argued that Berkeley recognizes no sharp distinction be-
tween the cognitive and the non-cognitive, and also that he takes ‘spirit’ to be a genuine
referring term. Hence I cannot accept this conclusion. Besides this, to let ethics drive
metaphysics in this way leads, as Roberts explicitly recognizes (96-97), to a quite strong
form of pragmatism. I will argue below (x7.2.2) that, although considerations of utility
260
enter into Berkeley’s understanding of truth, Berkeley sees a wider gap between utility
and truth than would be permitted by this sort of pragmatism. Thus although I think
Locke meant his claim about ‘person’ being a ‘forensic’ term quite seriously, it seems
to me that if Berkeley uses Locke’s account in the way Roberts suggests, he must be
seen as taking Locke’s account out of that ‘forensic’ context, and making it properly
metaphysical rather than merely moral.
The second point on which I disagree with Roberts is simply that, as I have already
said, I think this account must be regarded as quite speculative. Roberts apparently be-
lieves that, on the basis of these texts from Siris and other elements of Berkeley’s thought
we can reconstruct with some confidence the account of spirits that would have appeared
in the projected Part II of the Principles (see, e.g., Roberts 2007, 118). However, as
I have already noted, it is extremely difficult to determine which of the claims in Siris
Berkeley actually endorses. The claims about unity in Siris,xx346-347, for instance,
are prefaced with “According to the Platonic philosophy. . . ” Furthermore, in Alciphron
Berkeley seems to think that there are unresolved problems with Locke’s account of per-
sonal identity (Alc,x7.11), and there is not even the slightest hint of a solution to these
problems anywhere in Berkeley’s writings. On this basis, I would make a further claim:
not only is this account of Berkeley’s view quite speculative, but if he held this view at
all, he held it tentatively.
49
What is certainly true is that Berkeley did not give any other
account of ‘same’ as applied to spirits.
49. Many previous commentators have suggested that Berkeley was never fully satisfied with his ac-
count of spirit, and that this may indeed be one of the reasons why Part II of the Principles was never
completed. See, e.g., Tipton 1974, 258-262; Muehlmann 1992, 246; Stoneham 2002, 212-213. How-
ever, these commentators have not typically looked carefully at the material in Alciphron and Siris which
bears on the question, and in this respect, although Roberts overstates his case, his work on the subject
represents a significant advance.
261
6.7 Matter
According to Berkeley, the words ‘absolute existence of sensible objects in themselves,
or without the mind,’ “make out either a direct contradiction, or else nothing at all”
(PHK,x24). This, however, is precisely what ‘matter’ is supposed to be (PHK,x76;
DHP, 216, 225, 261). Thus, Philonous concludes that in using the word ‘matter’ one
“mean[s] nothing at all, [and] employ[s] words to no manner of purpose, without any
design or signification whatsoever” (DHP, 223). Given the interpretation of Berkeley’s
general philosophy of language I have been defending, we may read the second clause
as a gloss on the first: to ‘mean nothing at all’ just is to ‘employ words to no manner
of purpose.’ Similarly, Philonous charges that in his various attempts to find a referent
for the word ‘matter,’ Hylas “ha[s] been only supposing [he] know[s] not what, for no
manner of reason and to no kind of use” (220). This is radically different from the quasi-
entities stipulated in physics; those have a very important purpose and use, but ‘matter’
has none. Although ‘matter’ appears grammatically to be a referring term, it is in fact
neither a genuine referring term nor a quasi-referring term; it is simply nonsense.
The reason for this is quite simple. Matter is supposed to be the unthinking substra-
tum of perceived qualities. However, as we have seen, the claim ‘sensible qualities exist
unperceived,’ is either contradictory (if ‘exists’ is used as a synonym of ‘perceived’) or
meaningless (since no other meaning can here be assigned to ‘exists’). Therefore, no
unthinking thing can preserve sensible qualities in existence. Furthermore, the notion of
‘supporting’ qualities – the key notion in the concept of ‘substratum’ – can only possi-
bly be understood as perceiving them. Hence either ‘substratum’ is meaningless, or it is
the perceiving spirit, not the alleged unthinking thing, that is the substratum. Hence we
cannot introduce the word ‘matter’ by means of the description ‘unthinking substratum,’
for that description is incoherent.
262
Note that the point here is not just that ‘matter’ could not possibly be used to label
entities (i.e., that it is not a genuine referring term); the point is that the word ‘matter,’ as
it is here defined, cannot be used to do anything (except confuse poor Hylas). Either the
definition is contradictory, and so we are given instructions for the word’s application
which cannot be followed, or else the definition contains terms which are themselves
meaningless, and so we are not given any instructions for its use at all.
No help can be got by making matter the cause of the sensible qualities rather than
their support, for ‘cause’ can be given no meaning apart from volition (DHP, 216, 239-
240). Nor is it any help to make matter the occasion of the sensible qualities, for ‘occa-
sion’ can be given no meaning, unless an occasion is merely an idea in the mind of an
agent (PHK,xx69-75; DHP, 219-220).
In the end, the only way to give meaning to ‘matter’ is to make it a synonym of
‘body’ (DHP, 261). There is, however, no need for this, since we already have a perfectly
good word for this purpose. Furthermore, ‘matter’ carries metaphysical baggage which
we would be better off without. The word ‘matter’ is therefore best rejected.
6.8 Conclusion: Avoiding Immaterialism?
In this chapter, I have shown how Berkeley’s (in)famous immaterialist ontology follows
from his theory of mental and linguistic representation. Now, most philosophers find
Berkeley’s immaterialism quite incredible.
50
Indeed, no less a mind than David Hume
found Berkeley’s view literally incredible, i.e. psychologically impossible to believe:
50. Recent exceptions include Foster 1982, 2008; and R. M. Adams 2007.
263
Berkeley’s arguments, Hume famously wrote, “admit no answer and produce no con-
viction” (Enq,x12.15n).
51
Even for those of us who do not begin from a prior conviction
that Berkeley simply must have gone wrong somewhere, it will be useful to review the
steps by which we have arrived at Berkeley’s conclusion, in order to inventory the places
at which one might get off the boat, so to speak.
In chapter 2, I argued that Berkeley’s critique of abstract ideas in the Introduction
to the Principles is of broad application, showing that the ‘meanings’ posited by the
Theory of Meanings are metaphysically suspicious entities which we would be better
off to do without, if we can. I further argued that the elimination of meanings requires
the adoption of some form of ‘use theory’ of linguistic and mental representation on
which both words and mental states get their representational content from the way
in which agents use them according to rules. Inx5.2, I articulated and motivated a
principle I called ‘Berkeley’s Internalist Requirement,’ according to which the rules of
language must be such that the agent has some independent grasp of the conditions for
the rule’s application. In this chapter, I developed an anti-Quinean meta-ontology, on
which ontological commitment is not tied to quantification but instead to the use of
words as labels. Berkeley’s Internalist Requirement, combined with his sparse theory
of mind, guarantees that only minds, actions, and ideas can be successfully labeled. As
a result, only minds, ideas, and actions should be admitted to our ontology.
There are two rather obvious ways of escaping this conclusion. The first way of es-
cape would be to adopt a different conception of the mind and its representative powers
51. Hume is likely echoing Hylas here: “To deal frankly with you, Philonous, your arguments seem
themselves unanswerable, but they have not so great an effect on me as to produce that entire conviction,
that hearty acquiescence which attend demonstration” (DHP, 223; see Raynor 1990, 234; Bettcher 2007,
109). Alciphron makes a similar remark about ‘metaphysical’ arguments for the existence of God (Alc,
x4.2).
264
from Berkeley’s.
52
If there are ‘meanings,’ such as Platonic forms, Aristotelian univer-
sals, or Fregean senses, grasped by the mind, then Berkeley’s entire approach to the
philosophy of language is mistaken, and the argument we have been examining never
gets off the ground.
53
If, on the other hand, the mind somehow makes direct contact with
mind-independent external objects, then most of Berkeley’s arguments about language
can be accepted without drawing his immaterialist conclusion, since we would then be
able to label these mind-independent external objects (cf. Craig 1982, 542-544).
The second way of escape would be to accept most of Berkeley’s theses about mental
representation while rejecting the Internalist Requirement. This would allow us to label
external objects despite having only indirect contact with them.
It might be thought that there is a third way of escape, namely the rejection of the sig-
nificance of the contrast between labeling and non-labeling uses of words I have drawn
on Berkeley’s behalf. In other words, one could insist on a Quinean quantificational
meta-ontology. This would immediately secure ontological commitment to both bodies
and the theoretical entities of physics, for I have argued that Berkeley does not believe
we should eschew quantification over such entities. However, it is not clear, on this
strategy, how the mind-independence of bodies and forces, or the existence of material
substratum (or anything like it) is to be secured. Quantificational meta-ontology, as a
thesis about how to determine ‘what there is,’ tells us nothing about relations of onto-
logical dependence or relative fundamentality. Furthermore, Berkeley denies that there
52. Rickless 2013, 188 similarly emphasizes that “[f]rom the perspective of the twenty-first century”
the most obvious way of avoiding Berkeley’s conclusion is to reject the whole ‘Way of Ideas.’
53. I do not, however, mean to suggest that Berkeley has no other considerations to offer in favor of
immaterialism if his views on mental and linguistic representation are rejected. For instance, Berkeley
shows that issues about perceptual relativity cause problems for certain varieties of materialism, and these
arguments are more or less independent of the issues we have been considering.
265
is meaningful quantification over ‘substrata.’ In other words, the anti-Quinean meta-
ontology I attributed to Berkeley is not in fact at the heart of his system; it explains his
treatment of bodies and of theoretical entities from physics, but it is not strictly nec-
essary to his basic view that all of fundamental reality is mind-dependent. Thus the
rejection of this meta-ontology is not by itself a strategy for escaping immaterialism.
One would still need to show how we can coherently separate sensible objects from
their being sensed, and to provide a coherent meaning for ‘material substratum.’ It is far
from clear that this can be done by adjustments in meta-ontology alone without broader
changes to the philosophy of mind and language, such as the two strategies discussed
above.
Some, perhaps most, philosophers will be willing to incur the cost of some mysteri-
ous metaphysics or the counter-intuitive rejection of Berkeley’s Internalist Requirement
simply on the basis of what they see as the intrinsic implausibility of immaterialism. In
my view, the alleged intrinsic implausibility of immaterialism has never been made to
appear. What is of greater concern, however, is the theory’s extrinsic implausibility, that
is, its problematic consequences. The most problematic of these consequences have to
do with structure: how do we get from fleeting ideas to the complex structure of the
world of commonsense and of natural science?
We have already made some headway on this topic. We have seen that ‘body’ talk
and the technical jargon of physics can be understood as technologies whereby we rep-
resent to ourselves complex patterns in our sense experience. Language, in other words,
allows us humans to structure our experience in useful ways which give us correct ex-
pectations about future sensory experience and thereby allow us to order our actions
for our benefit and the benefit of others. To say this, however, is only to explain how
we are able to think of the world as structured; it does not preserve the claim that the
266
world is structured. To stop here and deny that the world really is structured would be
problematic for two reasons. First, it would be a radical affront to commonsense, whose
defender Berkeley claims to be. Second, it would render the practical success of our
linguistic technology a mystery (cf. Putnam 1979, 72-73).
What we want to say, then, is that when, in plain language or in physics, we attribute
structure to the world, we are giving a true description of the way the world really is.
If Berkeley could not defend this conclusion, this would certainly be sufficient cause
to reevaluate the plausibility of the various ways of escape from Berkeley’s immateri-
alism. However, I will argue in the coming chapters that Berkeley can get the correct
conclusion here. In chapter 7, I will examine Berkeley’s views on assent and truth in
order to get clear on the sort of objective, metaphysical basis that is needed to ground the
truth of our structure attributions. In chapter 8, I will argue that the requisite underlying
metaphysical reality – the objective structure of the world which is correctly reflected
in our language – can be provided by Berkeley’s theory of the physical universe as “a
most coherent, entertaining, and instructive Discourse” (Siris,x254) ‘spoken’ by God.
Finally, in chapter 9, I will show how this approach can be used to solve the most notori-
ously difficult problem for Berkeley’s attempts to accommodate commonsense, namely,
the question of how he can maintain that the (philosophically untutored) gardener is
justifiably certain of the existence of his cherry tree when no human perceives it.
267
Chapter 7
Assent and Truth
In 1733, Peter Browne, formerly Provost of Trinity College during Berkeley’s time there,
and at this time Bishop of Cork and Ross (neighboring the episcopal see to which Berke-
ley would be appointed one year hence),
1
appended to his Divine Analogy a scathing
critique of the account of religious language found in Berkeley’s Alciphron. Browne’s
critique is long and rambling and sometimes confused, and much of it is in response to
Berkeley’s rejection of the doctrine of analogy in Alciphron IV . Nevertheless, Browne
succeeds in putting his finger on a crucial problem for the theory of language enun-
ciated in Alciphron VII. According to Browne, it is Berkeley’s view “that Words may
be Significant, tho’ they signify Nothing” (Browne 1733, 534). On this view, Browne
alleges, “Faith . . . is no other than believing in certain Sounds and Syllables,” so that
“Your believing a God may be no more than Faith in a Monosyllable” and “all Faith
would terminate in the Ear” (539). This is the basis for Browne’s charge that “in the
particular Instance of divine Grace, [Berkeley] in effect gives up the whole cause of
Revelation and Mystery” (508).
Browne has, I say, put his finger on a crucial problem. For both Berkeley and
Browne, the issue here is made especially pressing by religious concerns: Browne takes
Berkeley to be committed to what we would now describe as a form of theological
anti-realism, endorsing theological language without taking it to describe an objective
reality. Some interpreters would have Berkeley plead ‘guilty as charged’ (see especially
1. For an intellectual biography of Browne, see Winnett 1974.
268
Berman 1994, 146-148).
2
However, if the conclusions I defended in chapter 4 are cor-
rect, not only would Berkeley find this unpalatable, religiously speaking, but it would
have a devastating effect on his philosophy as a whole. I argued that Alciphron VII
is concerned to give a general theory of language, and that Berkeley does not accept
any sharp demarcation between cognitive and non-cognitive discourse. On my reading,
Berkeley does not regard theological language as a special case. As a result, if Berkeley
cannot answer Browne’s charges, he will be brought to the conclusion that not only faith
in religious mysteries, but assent of every kind ‘terminates in the ear.’
The difficulty Browne has identified stems from Berkeley’s elimination of the Lock-
ean ‘mental proposition’ (Browne 1733, 510-512; cf. Bordner 2011, 322-323n42).
3
In
the Theory of Meanings, the mental proposition (or its analog) does important theoret-
ical work. Locke’s mental propositions are mental states which represent the world as
being a certain way. The having of the relevant mental proposition is what distinguishes
sincere from insincere assertions. It is also what distinguishes understanding a sentence
from merely hearing the sounds. Finally, verbal propositions (sentences) inherit their
representational content from the associated mental proposition. Thus it is only by be-
ing associated with a certain mental proposition that a sentence, which, considered in
itself, is merely a string of sounds or characters, can come to represent the world as be-
ing a certain way. One cannot just delete mental propositions from this picture without
providing some alternative account of the phenomena mental propositions were meant
to explain.
2. Belfrage 1986b is even more explicit than Berman in attributing to Berkeley the view that our talk of
religious mysteries is not descriptive of an objective reality. However, Belfrage’s paper is concerned only
with the Manuscript Introduction and not with Alciphron. More recently, Belfrage has said that Berkeley
was dissatisfied with this consequence of his early view and “spent the rest of his life trying to satisfy
himself on this point” (Belfrage 2007, 51).
3. Although the term ‘mental proposition’ is Locke’s, analogous entities were posited by other theories
current in Berkeley’s lifetime. See chapter 1.
269
Although the issue raised by Browne is an extremely serious one, Browne does not
acknowledge that it is also one that Berkeley addresses explicitly and at length in Alci-
phron VII. The aim of this chapter is to understand Berkeley’s solution to this problem.
This will prepare the way for the last two chapters of this dissertation, in which I will
defend the claim that it is in virtue of the status of the perceived world as “a most co-
herent, entertaining, and instructive Discourse” (Siris,x254) ‘spoken’ by God that plain
language assertions about bodies are true.
The first section of this chapter examines Berkeley’s account of assent. We shall
be especially concerned to understand Berkeley’s claim that faith, which is or involves
a kind of assent (Alc,xx7.1, 7.11-14, 7.17), is “placed in the will and affections rather
than in the understanding, and producing holy lives, rather than subtle theories” (x7.13).
I will argue that Berkeley endorses a theory of belief similar to one recently defended by
Eric Schwitzgebel under the name ‘phenomenal dispositionalism’ (Schwitzgebel 2002).
On this view, belief consists in a certain complex of dispositions; however, unlike many
20th century dispositional theories, Schwitzgebel’s phenomenal dispositionalism allows
that some of the dispositions may be dispositions to experience private mental episodes,
such as images in the imagination. On the view I attribute to Berkeley, assent involves
a wide variety of dispositions, including dispositions to have ideas, feel emotions, and
perform linguistic and non-linguistic actions.
Equipped with our understanding of the nature of assent, we will be prepared to
examine the conditions under which one’s assent is correct, that is, the conditions in
which the sentence to which one assents is true. Here we will find that Berkeley endorses
a form of semantic holism more radical than Quine’s. For instance, with respect to
the word ‘grace,’ Berkeley holds not only that it can be meaningful only within the
context of a theological theory, but that it can be meaningful only in the context of the
270
religious life. In this respect, Berkeley anticipates the Wittgensteinian tradition in the
philosophy of religion. However, in Berkeley’s view, one particular ‘form of life’ (as
Wittgenstein 1953,xx1.19, 241 calls it) can better fit objective reality than another. Thus
correspondence and objectivity are preserved in Berkeley’s theory of truth. Berkeley’s
view can be summed up in the slogan: realism about truth, holism about meaning.
7.1 The Nature of Assent
To assent to a sentence is to agree with it, or take it to be true. Berkeley’s account of this
phenomenon is developed in conscious opposition to Locke, and in particular the use of
Locke’s theory in John Toland’s notorious Christianity Not Mysterious (Toland 1696).
In Locke’s usage, ‘assent’ is a genus which is divided into two species. Locke calls the
first of these ‘knowledge’ and the second ‘belief,’ ‘judgment,’ or ‘opinion.’ Religious
faith belongs to the latter category (EHU,xx4.18.4-6). Both varieties of assent involve
the joining or separating of ideas by the mind (x4.14.4). As a result, it is impossible
for there to be any form of assent in the absence of ideas (EHU,xx4.1.1, 4.3.1, 4.18.3;
Alc,x7.3). According to Toland, this renders belief in religious mysteries, such as the
doctrines of various Christian denominations regarding the Eucharist and the Trinity,
impossible since we have no corresponding ideas (Toland 1696, 26-27).
Berkeley’s strategy in Alciphron VII is, first, to show that assent in the absence of
ideas is a widespread phenomenon, not only in religion, but in science and in everyday
life, and then to provide an account of the nature and possibility of such assent.
That Berkeley is committed to the first claim should by now be abundantly clear;
nevertheless it will be helpful briefly to review the examples Euphranor cites in its de-
fense. First, he argues that, since there are no abstract general ideas, there is no one
idea which corresponds to any general term. Furthermore, as Berkeley had pointed out
271
as early as the Manuscript Introduction, in ordinary predications like ‘Melampus is an
animal,’ the general term ‘animal’ cannot stand for some one of the items in its exten-
sion: “if it be made stand for another [animal] different from that [which] is marked by
the Name Melampus, the Proposition is false and includes a Contradiction. And if it be
made [to] signify the very same Iindividual that Melampus doth, it is a Tautology” (MI,
x35). But in fact the sentence is neither a falsehood nor a tautology (see above,x3.1).
Hence ‘animal’ must here not stand for any idea at all (Alc,x7.7). Additionally, no idea
can possibly represent (except as a conventional sign) any action or spirit, yet we can
speak meaningfully of actions and spirits (x7.8).
Thus far, however, Berkeley’s departure from the Theory of Meanings is not radical.
‘Melampus’ and ‘animal’ are both names or labels of Melampus, and ‘spirit’ is a name
for spirits. Perhaps, one might think, what happens in the Melampus case, is that one
considers Melampus under the name ‘Melampus,’ and considers Melampus under the
name ‘animal,’ and recognizes the identity of the object considered in the two cases.
This would be a kind of mental “joining or separating of Signs,” and it would be correct
(true) precisely because “the Things signified by them, do agree . . . one with another”
(EHU,x4.5.2). Furthermore, this would involve one of the four kinds of agreement
Locke enumerates, namely, identity (xx4.1.3-4). On this reading, it is precisely because
Melampus is considered under two different names that the proposition is non-trivial;
hence to take away the names and try to form a mental proposition is to turn it into a
tautology. Nevertheless, the sentence satisfies Locke’s definition of truth.
As we will see in our discussion of truth, below, there is something correct about
this line of thought: Berkeley does hold that some basic sentences are true because
they join terms which label the same object or objects. However, even in the simplest
cases, assent, for Berkeley, does not consist in considering an object under two guises
272
and recognizing the identity of the object. This is because it is possible to understand
and assent sincerely to a sentence without having any of the relevant ideas at all (Alc,
x7.7). Hence assent cannot amount to recognition of sameness of idea. Furthermore,
it is only in simple cases that such ideas even exist. Thus in Euphranor’s concluding
summary we read that the ends of language may be achieved “not only although the
ideas marked [by the words] are not offered to the mind, but even although there should
be no possibility of offering or exhibiting any such idea to the mind” (x7.17). Euphranor
gives as examples the use of ‘force’ as a technical term of Newtonian physics (xx7.9-10),
and “the algebraic mark, which denotes the root of a negative square” (x7.17).
These examples are sufficient to show the falsehood of the general principle that as-
sent requires ideas, and hence to undermine the argument against the possibility of faith
in religious mysteries. Much of the negative argument of Alciphron VII is, of course,
just the attack on meanings from the Introduction to the Principles, which we discussed
at length in chapter 2. However, Alciphron goes significantly farther in attempting to
provide a replacement theory.
Some commentators (e.g. Bennett 1971,x10; Berman 1994, ch. 6) would have us
believe that the account Berkeley provides is an account specifically of religious faith,
and certainly religious faith is Berkeley’s chief concern here. However, as I have already
shown in chapter 4, this narrow interpretation is directly contrary to the text. Euphranor
says, “Science [i.e. scientific knowledge] and faith agree in this, that they both imply an
assent of the mind: and, as the nature of the first is most clear and evident, it should be
first considered in order to cast a light on the other” (Alc,x7.14). Faith implies assent,
and if we want to understand it, Euphranor tells us, we must examine other forms of
assent, such as scientific knowledge, and see what they all have in common. He therefore
proceeds to a discussion of the nature of scientific knowledge, from which he concludes
273
“that the true end of speech, reason, science, faith, assent, in all its different degrees,
is not merely, or principally, or always the imparting or acquiring of ideas, but rather
something of an active, operative nature, tending to a conceived good” (Alc,x7.17,
emphasis added).
4
Berkeley’s aim is to provide a general account of assent, which takes
all assent to be like religious faith in involving the will (x7.13).
To understand what Berkeley has in mind here, we must understand what he means
by ‘will’ and how it contrasts with ‘understanding.’ In the Principles, we read, “A spirit
is one simple, undivided, active being. As it perceives ideas, it is called the ‘understand-
ing’, and as it produces or otherwise operates about them, it is called the ‘will’” (PHK,
x27). This way of contrasting will with understanding makes good sense of Berkeley’s
account of the origin of science:
the human mind, naturally furnished with the ideas of things particular and
concrete, and being designed, not for the bare intuition of ideas, but for
action and operation about them, and pursuing her own happiness therein,
stands in need of certain general rules or theorems to direct her operations
in this pursuit: the supplying which want is the true, original, reasonable
end of studying the arts and science (Alc,x7.14).
Scientific knowledge consists in facility with these general rules; it is a kind of know-
how. In arithmetic, for instance, “the mind is immediately occupied about the signs or
notes, by mediation of which it is directed to act about things” (x7.15, emphasis added).
The point is general: “If I mistake not, all sciences, so far as they are universal and
demonstrable by human reason, will be found conversant about signs as their immediate
object, though these in the application are referred to things” (x7.16).
4. The talk of ‘degrees’ of assent is likely a reference to EHU, ch. 4.16.
274
Scientific knowledge is a kind of assent, and it consists (to repeat) in facility with
certain general rules. These rules are of two kinds: rules for ‘computing in signs,’ and
rules for applying the results to things in the world. This is what Berkeley means by
saying that assent resides primarily in the will rather than the understanding: assent is
not a matter of what ideas one has (“bare intuition of ideas”), but of what one does (or
is disposed to do) with them.
Now, for Locke, dispositional assent might very well be described this way, since
joining two ideas is something the mind does. Thus I might be said to assent disposi-
tionally to the proposition that snow is white in virtue of having a disposition to join
the idea of white to the idea of snow. The act of (occurrent) assent, for Locke, is just
this act of joining. One way we could put the contrast is this: for Locke, particular acts
of (occurrent) assent are fundamental, and dispositional assent must be defined in terms
of these. For Berkeley, assent is fundamentally a (dispositional) state of the will. We
may, if we so choose, describe certain actualizations of this dispositional state as ‘acts
of assent,’ but Berkeley’s concept of assent is the concept of the state. Furthermore, acts
of assent, so defined, will be a radically heterogeneous class, including linguistic and
non-linguistic actions, as well as feelings and imaginings.
A believer in a verbal proposition (sentence) is a follower of certain rules. In or-
dinary cases, these include rules for the manipulation of the signs which make up the
proposition, and rules for action. The rules for manipulation of the signs will of course
include the rules for inferring other propositions from the propositions one believes. We
discussed these rules at some length above (x5.4). This manipulation can be internal
or external: one can reason ‘in one’s head,’ on paper, or in speech. Another use of the
signs in question which the rules will authorize is the assertion of the proposition one
275
believes, and the use of signs to indicate agreement with utterances of that proposition
by others.
5
When one makes an assertion, one hopes that one’s audience will assent to it (see
Alc,x6.8). In the Introduction to the Principles Berkeley famously claimed that, besides
“the communicating of ideas” language has “other ends, as the raising of some passion,
the exciting to, or deterring from an action, [or] the putting the mind in some particular
disposition” (PHK, Introx20). In light of Alciphron VII we can see that, unlike paradig-
matic non-cognitive theories of ethical or religious discourse (e.g. Ayer 1952, ch. 6),
Berkeley’s theory does not claim that when we utter certain declarative sentences we
are not making genuine assertions or aiming to gain the assent of our hearers. Rather,
he is claiming that assent need not involve ideas. Assent involves emotions, actions, or,
more generally, dispositions of the mind.
One way in which religious discourse and moral discourse contrast with scientific
discourse is that they are strongly emotionally valenced. Euphranor contrasts the case in
which words “influenc[e] our conduct and actions . . . by forming rules for us to act by”
with that in which the influence proceeds “by raising certain passions, dispositions, or
5. I reiterate that by ‘proposition’ Berkeley always means a sentence. Neither Berkeley nor Locke
believes in the Fregean ‘third realm’ propositions widely accepted by analytic philosophers today. I
retain Berkeley’s use of ‘proposition’ for two reasons. The first reason is simply to make my exposition
terminologically consistent with Berkeley’s text. The second reason is that propositions were introduced
in philosophy to play certain theoretical roles – for instance, to be the primary bearers of truth and falsity
and to be the immediate objects of the so-called ‘propositional attitudes’ (belief, hope, conjecture, doubt,
etc.) – and it is Berkeley’s view that sentences play these roles. Nevertheless, because there is very real
danger of terminological confusion, I will continue to remind the reader frequently that ‘proposition’ and
‘sentence’ are being used equivalently.
In order to avoid serious difficulties relating to vagueness and ambiguity, Berkeley needs to hold that
there can be distinct sentences which are homonymous. Thus the sequence of characters, ‘she is going
to the bank’ may be either of two (or more) distinct sentences, depending on whether ‘bank’ means river
bank or financial institution. Insofar as a sentence is not a sequence markings (or sounds), but rather a
sequence of signs, this is a perfectly natural thing to say, since a sign is not just a symbol, but a symbol
associated with certain rules of usage. A particular instance of the sound or inscription ‘bank’ may thus be
an instance of either of two different signs, depending on which rules are being followed on the particular
occasion in question.
276
emotions in our minds” (Alc,x7.8). It seems that scientific claims belong primarily or
exclusively to the former category, whereas religious and moral claims contain a gener-
ous intermixture of the latter. Thus Euphranor says of the doctrine of the Trinity, that one
may be said to believe it “provided that this doctrine . . . makes proper impressions on his
mind, producing therein, love, hope, gratitude, and obedience” (x7.11). Similarly, Crito
says of the doctrine of original sin, that “the belief thereof may produce in [one’s] mind
a salutary sense of [one’s] own unworthiness, and the goodness of [one’s] Redeemer”
(x7.13). The belief (assent) is the dispositional state; the outcomes it ‘produces’ include
these feelings of unworthiness.
We can see, then, that assent in general, and religious faith in particular, do not, for
Berkeley, “terminate in the ear”; rather, just as he says, they ‘terminate’ in the will. In
the specific case of faith in religious mysteries, Crito says that assent is “placed in the
will and affections rather than in the understanding, and producing holy lives, rather
than subtle theories” (x7.13, emphasis added). In these cases, there are no ideas to be
exhibited, so the difference between merely hearing a sound and assenting to a sentence
is constituted entirely by the affective and volitional response. The understanding –
that is, the mind considered merely as a perceiver of ideas – has no role to play. This
contrasts with sentences such as ‘snow is white,’ where acceptance of the sentence is
constituted in part by a disposition to form mental images of white snow under appro-
priate circumstances. (Perhaps the disposition is rather: when one imagines snow, to
imagine it as white. Of course the disposition may be overridden in particular cases.)
Even in the straightforward ‘snow is white’ case, assent to the sentence is not constituted
by a particular instance of imagining white snow. Assenting to the sentence involves a
277
certain disposition to imagine, which disposition may or may not be actualized on a par-
ticular occasion when one hears or utters the sentence. An act of imagination is neither
necessary nor sufficient for belief.
It will be helpful at this point to review some of the linguistic rules we have al-
ready encountered, and consider their bearing on what precisely is involved in assent to
different sorts of propositions.
Let us begin with ‘Melampus is an animal.’ In the Manuscript Introduction, Berkeley
said that this sentence signified nothing but “That the particular [creature]
6
thing I call
Melampus has a right to be called by the Name Animal” (MI,x34). Inx3.1, I argued
that Berkeley need not, and should not, regard this as (necessarily) involving an explicit
meta-linguistic belief. That is, the speaker need not be disposed to make such meta-
linguistic claims as ‘Melampus is rightly called ‘animal.” Rather, the speaker has, and
seeks to create in the hearer, a disposition to first-order linguistic behavior; specifically,
the use of the word ‘animal’ to describe Melampus. Assent to the sentence ‘Melampus
is an animal’ consists chiefly in this linguistic disposition.
7
However, this belief also
involves the disposition to use ‘Melampus is an animal’ as a premise in reasoning, and
by the rules for the use of the word ‘animal,’ the individual can get from here to the
claim that, for instance, Melampus is a living thing.
Next, let us proceed to ‘reward’ talk. Suppose Melampus is the lost puppy dis-
cussed inx3.2, and consider the declarative sentence: ‘the return of Melampus will be
rewarded.’ To assent to this sentence involves a variety of dispositions including, im-
portantly, the disposition to describe the thing given to whoever returns Melampus as
a ‘reward,’ and some degree of motivation (however small) to return Melampus. The
6. Brackets original.
7. The problem of ascribing the belief that Melampus is an animal to non-English speakers will be
addressed below.
278
motivation may be negligible in force, and may be overridden by countervailing moti-
vations. Alternatively one may of course understand the sentence without assenting to
it. Also, finding out what the thing to be given to the returner is may lead one to stop
assenting to the sentence, perhaps because of an overriding conviction that certificates
of appreciation are not really rewards.
The kind of disposition to act which is generated by assent to reward sentences is not
a raw, ungrounded disposition. As I emphasized inx3.2, promising a reward is not the
same as issuing a command. In the latter case, a rule of action is given explicitly, and if
it is accepted it guides conduct directly, due to the agent having the rule in mind and fol-
lowing it intentionally. Reward talk, however, produces cheerful obedience, because it
has the effect of attaching positive expectations to the action in question (MI,x37). That
is, assent to ‘the return of Melampus will be rewarded’ attaches a positively valenced
emotion to the prospect of returning Melampus. Threat of punishment, of course, works
in precisely the reverse fashion.
Faith in the promise of an inconceivable heavenly reward need not, on this theory,
function any differently than ordinary beliefs about rewards. The other religious mys-
teries are more complicated, but the point is general: religious faith attaches positively
valenced emotions to some thoughts and actions, and negatively valenced emotions to
others, thereby altering our motivational structure, and becoming “a principle destruc-
tive of evil habits and productive of good ones” (Alc,x7.10).
8
Assent to these sorts of
sentences also involves the same sorts of inferential dispositions as any other syntacti-
cally similar sentences.
8. The literal syntax of the sentence from which this quotation is taken strongly suggests that the
‘principle’ in question is grace itself, rather than belief in grace. However, this would make the statement
irrelevant to the surrounding discussion. See Pearce, forthcoming(b),x8.
279
Assent to a mathematical ‘sentence’ such as ‘2 + 2 = 4’ likewise involves both
inferential and practical dispositions. The inferential dispositions are defined by the
conventions governing the notation. The practical dispositions include, for instance, the
expectation that combining two groups of two things will result in a group of four things.
Claims in the natural sciences are to be handled similarly.
Finally, let us briefly revisit moral beliefs. As I indicated inx3.2, Berkeley’s is
severely underdeveloped, and he seems to have been quite hesitant about it. What is
clear, however, is that Berkeley was consistent in holding that “morality consist[s] in the
V olition chiefly” (N,x669; cf. N,x149), and he clearly takes this to be a meta-ethical
and not merely normative claim. That is, he holds that assent to moral propositions,
like assent to religious mysteries, has little or nothing to do with the understanding
(i.e. the perceiving of ideas). According to the Manuscript Introduction, the aim of
describing an action as “Honourable and vertuous” is to “excite in the Mind of the
Hearer an esteem of that particular Action and stirr him up to the performance of it”
(MI,x41). This suggests that assenting to the proposition that a particular action is
‘honorable and virtuous’ consists in attaching a positively valenced emotion to it, an
emotion of a sort which is apt to “stirr [one] up to the performance of it.” As I pointed
out inx3.2, Berkeley’s later markings on this passage suggest that he came to think that
this account was inadequate to cover all of the uses of these sentences, and so that there
is more going on in moral discourse than this. Passive Obedience appears to assume
some sort of definitional connection between goodness and pleasure, but Berkeley never
spells out the details. The central point, though, is that assent to moral propositions, like
assent to religious mysteries, necessarily involves the attachment of certain emotions to
certain actions.
280
This, I hope, suffices to make clear the broad outline of Berkeley’s understanding
of assent. Unfortunately, Berkeley has given us little more than the broad outline. In
asking whether this outline can be filled in in a philosophically satisfying way, a number
of questions arise. The first, and most important is this. We began by saying that to
assent to a sentence is to agree with it or take it to be true. However, it is not at all clear
what assent, on Berkeley’s view, has to do with truth. It is not obvious, on the view
developed, that in assenting to sentences we take the world to be a certain way. This
issue will be the topic ofx7.2. There are several other issues, however, which it will be
well to pause and consider here.
The theory I have attributed to Berkeley takes assent to consist in a variety of dispo-
sitions to speech, thought, and action. However, insofar as we are not perfect reasoners,
or our linguistic competence is imperfect, or perhaps if we are akratic or otherwise prac-
tically irrational, we will not have all of the dispositions in question. Thus, for instance,
no human is really disposed to assent to all of the logical consequences of sentences to
which she assents. Moreover, we use English to ascribe beliefs to non-English speakers,
who do not have the relevant linguistic dispositions at all. Thus the view in question
appears to demand far too much of believers.
It is at this point that the account of belief recently developed by Eric Schwitzgebel
can be of assistance to us. According to Schwitzgebel, the possibility of ‘mixed cases,’
in which the agent has some, but not all, of the relevant dispositions is in fact a key
selling point of the phenomenal, dispositional account.
It is widely recognized that belief is not an all-or-nothing affair. This is, for instance,
explicitly recognized by Locke in his discussion of ‘degrees of assent’ (EHU, ch. 4.16),
and it is central to the Bayesian approach to epistemology. However, what has not, ac-
cording to Schwitzgebel, been adequately recognized is that belief with low confidence
281
is not the only kind of ‘in-between belief’ (Schwitzgebel 2001). For instance, one might
be able to produce a fact in some contexts but not others, or one might be disposed to
affirm in one context or state of mind what one is disposed to deny in another, despite
not having any mental reservations or intention to mislead. In these sorts of cases, one
does not exactly believe the proposition, nor does one exactly fail to believe it; one is
rather in an ‘in-between’ state. Yet this in-between state is not to be explained in terms
of a lack of confidence in one’s (dis)belief, or in terms of changing one’s mind. The
agent’s mental state is the same throughout.
This can be made clearer by consideration of an example of Schwitzgebel’s, which
would be of special interest to Berkeley:
In certain moods and in certain contexts, Antonio feels quite sure that
the universe is guided by a benevolent deity. In other moods and contexts,
he finds himself inclined to think of talk about God as ‘a beautiful metaphor’
or even, sometimes, ‘a crock of hooey’. When his atheistic buddies at work
mock religious belief, he does not join in, but neither does he feel an im-
pulse to defend belief in God; at such moments, especially if it is mid-week,
the whole God business seems rather silly. When Antonio goes to church
with his wife, he is not inclined to believe everything the pastor says, but,
particularly if the pastor waxes poetic about the magnificence of creation,
he may feel that there must be a divine force guiding the world. At the birth
of a child or the death of a friend, he feels certain God is involved; when the
church gossip group has invaded his house, the idea of taking literally talk
about the existence of a benevolent deity strikes him as foolish.
Does Antonio believe that God exists? A simple yes or no answer to
this question would be misleading (78).
282
There is nothing particularly unusual about Antonio’s case. Yet little effort has been
made to understand how, on standard accounts which take believing that God exists to
involve some relatively stable mental state representing God’s existence, such a case is
possible. The ‘in-betweenness’ of Antonio’s belief is neither a function of its content,
nor of his level of confidence, but these are all of the resources standard accounts give
us for distinguishing between kinds of belief.
Schwitzgebel suggests that a dispositional account can deal with the case as follows.
Our concept of belief that God exists can be understood as a ‘dispositional stereotype’
– that is, as a cluster of dispositions which we take to go together (Schwitzgebel 2002,
250). Schwitzgebel offers as an example, being hot-tempered. “This stereotype,” he
says, “will include the disposition to respond angrily to minor provocations, the disposi-
tion to be slow in cooling off after a fight, the disposition to feel and express frustration
quickly when one’s will is thwarted, and so forth” (251). Beliefs can be understood
analogously. The unambiguous cases of belief in a particular proposition are those in
which the agent has (almost) all of the dispositions included in the stereotype. The ‘in-
between’ cases are cases in which there is a partial match between the agent and the
stereotype. This provides a straightforward account of Antonio’s case: he has some, but
not all, of the dispositions in the stereotype associated with believing that God exists.
Filling in the Berkeleian outline with a theory along the lines of Schwitzgebel’s has
both philosophical and textual benefits. Philosophically, it provides a plausible account
which will do the work Berkeley needs it to do. Textually, it makes good sense of a
particular instance of ‘in-between belief’ which is crucial to the project of the Principles,
namely, the pernicious sort of practical belief in matter which Berkeley admits to be
common to the philosophers and the mob (PHK,x54).
283
Taking the philosophical issue first, I have argued that, for Berkeley, to be a be-
liever in a proposition (sentence) is to be a follower of certain rules, and that it is the
conventional association of the proposition with these rules that makes the proposition
meaningful. We have already discussed in some detail the question of implicit belief that
someone else follows a certain rule (see above,x5.3). What is involved here is the ex-
pectation that the other will act in certain ways in certain circumstances. This includes
being disposed to imagine the other acting in a particular way when I imagine her in
the relevant circumstance. Thus, adapting Schwitzgebel’s view to Berkeley, we would
hold that our concept of belief is used to lump together certain rules and thus organize
our thinking about the actions of others. The ‘in-between’ cases are cases in which the
individual is a follower of some of the relevant rules but not others, or in which the indi-
vidual is an inconsistent or otherwise imperfect follower of the rules. Berkeley is able to
agree with Schwitzgebel that neither of these things is necessarily connected with lack
of confidence.
This approach also helps to make sense of Berkeley’s discussion of belief in matter
at PHK,x54. He writes:
upon a narrow inquiry, it will not perhaps be found so many as imagined do
really believe the existence of matter or things without the mind. Strictly
speaking, to believe that which involves a contradiction, or has no meaning
in it, is impossible . . . In one sense, indeed, men may be said to believe that
matter exists, that is, they act as if the immediate cause of their sensations
. . . were some senseless unthinking being. But that they should clearly ap-
prehend any meaning marked by those words, and form thereof a settled
speculative opinion, is what I am not able to conceive. This is not the only
instance wherein men impose upon themselves, by imagining they believe
284
those propositions they have often heard, though at bottom they have no
meaning in them.
Berkeley has argued that, depending on how it is interpreted, the sentence ‘matter exists’
is either contradictory or meaningless (see above,x6.7). Here, he makes three claims.
First, he claims that it follows from the fact that ‘matter exists’ is contradictory or mean-
ingless that it is ‘strictly speaking’ impossible to believe that matter exists. Second, he
allows that there is a looser sense in which it is correct to ascribe belief in matter to peo-
ple. In that looser sense, to believe in matter just is to “act as if the immediate cause of
[one’s] sensations . . . [is] some senseless unthinking being.” Third and finally, Berkeley
holds that people have a false second-order belief, the belief that they believe in matter.
The proposed interpretation can make sense of all three of these claims. First, it is
impossible strictly speaking to believe what is contradictory or meaningless, because to
believe a contradictory proposition would be to have inconsistent dispositions – that is,
dispositions to inconsistent courses of action in the same circumstances. Since two in-
consistent dispositions cannot both be actualized on the same occasion, it is impossible
reliably to follow the instructions given by a contradictory proposition; hence there is no
such thing as a ‘stereotypical’ believer in such a proposition. A meaningless proposition,
on the other hand, gives no instructions, or at least problematically incomplete instruc-
tions. Accordingly, meaningless propositions likewise fail to give directions which can
be followed. However, a kind of ‘in-between belief’ is possible here, and this is what
is at issue in Berkeley’s second claim. The practical directions involved in the belief
that matter exists can be followed, at least in part. As Berkeley makes clear in the
closing sections of the Principles, his immaterialist arguments aim to secure belief in
“a spirit who is intimately present to our minds, producing in them all that variety of
ideas and sensations, which continually affect us, on whom we have an absolute and
285
entire dependence” (PHK,x149). According to Berkeley, “it is downright impossible
that a soul pierced and enlightened with a thoroughgoing sense of the omnipresence,
holiness, and justice of that Almighty Spirit should persist in a remorseless violation of
his laws” (x155). Immaterialism brings God near, and the belief in the nearness of God
is partly constituted by feelings of obligation to live up to God’s law, and remorse if one
fails. Materialism is, on the other hand, the “foundation [on which] have been raised all
the impious schemes of Atheism and Irreligion” (x92). Even if the materialist does not
fall into outright atheism, materialism puts God ‘at a distance,’ as it were, and thereby
lessens the felt urgency of obedience to God’s commands. This explains how Berkeley
can hold, at the same time, that belief in materialism is impossible, and that belief in
materialism is pernicious: full belief, ‘settled speculative opinion,’ is impossible, but
there is a widespread sort of ‘in-between belief’ in materialism which is pernicious.
Finally, people believe of themselves and of others that they believe in matter. This
belief consists in linguistic dispositions and behavioral expectations. This belief must
be to some degree defective, since the behavioral expectations it should involve are not
entirely consistent. What Berkeley should say about this is that we have an inconsistent
stereotype and are unaware of its inconsistency. Since some departure from the stereo-
type is consistent with belief, we continue to think that people conform well enough to
the stereotype, despite the fact that rigorous conformity is impossible.
There are additional problems associated with the attribution of beliefs to individuals
who do not speak the language (or are simply not familiar with the vocabulary) in which
the belief is described. These too are neatly dealt with by the phenomenal dispositional
theory.
286
Consider, first, the most straightforward case. Pierre, a monolingual French speaker,
is disposed to give external signs of assent to the sentence ‘Londres est jolie.’
9
We
ought, it seems, to be able to describe Pierre in English as believing that London is
pretty, despite the fact that Pierre has no disposition to make external signs of assent to
the English sentence ‘London is pretty.’
Although linguistic dispositions are central to our stereotypes, and Pierre fails to
exhibit one of the relevant linguistic dispositions, we nevertheless have no hesitation,
in this case, in ascribing the belief to Pierre. This is because he exhibits the other
relevant dispositions (e.g., the disposition to aesthetic appreciation of postcards from
London), and because his failure to exhibit the relevant linguistic disposition is a failure
of a well-known and accepted sort. The rules corresponding to a particular instance of
assent require that the believer assent likewise to paraphrases (in the same language)
and translations (in other languages) of the sentence in question, but it is understood
that the believer will fail to recognize translations into languages she does not speak,
and will therefore fail to follow the rule in these cases. Thus Pierre is not precisely a
stereotypical believer in ‘London is pretty,’ but it is nevertheless unambiguously correct
to describe him as a believer, for he has the relevant dispositions of feeling, imagination,
and action, as well as the disposition to make outward signs of assent to translations of
‘London is pretty’ into languages he speaks.
This, of course, raises the question, in virtue of what does one sentence count as a
translation or paraphrase of another? What is it about ‘London is pretty’ and ‘Londres
est jolie’ that makes them translations of one another?
The Theory of Meanings gives a simple answer to this question: the two sentences
correspond to the same mental proposition. Having eliminated mental propositions, and
9. This example, together with the various modifications of it considered below, is drawn from Saul
Kripke’s influential discussion of this issue (Kripke 1979).
287
‘meanings’ more generally, Berkeley is not in a position to give this answer. Neverthe-
less, Berkeley has significant resources. To believe a proposition is to be a follower of
certain rules. Hence Berkeley can appeal to sameness (or similarity) of rules. In the
case of genuine referring expressions, this is quite simple: two words which use the
same rule to pick out which things to label are at least partial synonyms. If they are
associated with other rules, then they may not be precisely synonymous (consider the
difference between an affectively neutral racial label and a slur); however, it is simply
a fact that it is only in the simplest cases, if ever, that exact sameness of meaning can
be achieved between two sentences of the same or different languages. Similarity of
meaning is good enough.
In addition to the similarity of the rules, when there is a community of bilingual
speakers (as in the case of English and French) there may well be conventional rules
of translation, and these rules are just as eligible as any others to help constitute the
meaning of a word. Thus, for instance, one of the rules associated with ‘London is
pretty’ may be to assent to it only if one also assents to ‘Londres est jolie.’ I said
above that the rules which the believer follows include the rule of likewise assenting
to any paraphrase or translation of the proposition believed; what I am claiming now
is that, when there is a bilingual community, the conventional rules which make the
sentence meaningful may actually include a partial enumeration of these paraphrases
and translations, or some rules for generating them.
In addition to similarity of meaning, a necessary condition for being a translation or
paraphrase is, of course, having the same truth conditions. We will return to the issue of
truth in the next section.
288
Having dealt with the simple case, let us move on to the more difficult one, one
in which a bilingual speaker has contradictory linguistic dispositions when speaking
different languages. Kripke describes the case like this:
Later, Pierre . . . moves to England, in fact to London itself, though to an
unattractive part of the city with fairly uneducated inhabitants. He, like
most of his neighbors, rarely ever leaves this part of the city. None of his
neighbors know any French, so he must learn English by ‘direct method,’
without using any translation of English into French: by talking and mix-
ing with the people he eventually beings to pick up English. In particular,
everyone speaks of the city, ‘London,’ where they all live (Kripke 1979,
255).
Pierre, we are to understand, never learns that ‘London’ and ‘Londres’ name the same
city, and he comes to be inclined to assert, and make outward signs of assent to, ‘London
is not pretty,’ despite continuing to be inclined to assert, and make outward signs of
assent to, ‘Londres est jolie.’ The puzzle is, what does Pierre believe?
Schwitzgebel, rightly I think, classifies this as a kind of ‘in-between belief’
(Schwitzgebel 2002, 265). Pierre exhibits some of the dispositions (follows some of
the rules) of each of the pair of contradictory beliefs. However, insofar as the beliefs
are contradictory and so give contradictory instructions, it is impossible that he should
fully believe both of them at once. It is not that he believes one or the other. Rather, in
light of his contradictory dispositions, it is not quite accurate to describe him as either
having or lacking either belief. In these kinds of cases, the language of belief or assent
is simply too blunt an instrument to describe the rules governing Pierre’s behavior.
289
Berkeley holds that assent consists in the following of certain rules, which rules
govern not only overt (linguistic and non-linguistic) behavior but also the having of
ideas, feelings, and so forth. This schematic approach is conveniently fleshed out by
means of Schwitzgebel’s notion of a dispositional stereotype associated with assent to
a given proposition. An individual unambiguously believes a proposition when she
assents to enough of the central elements in the stereotype. However, there are also
‘in-between’ cases, in which the assumptions behind our practice of belief ascription –
namely, the assumption that certain dispositions ‘go together’ – fail. Berkeley’s account
of practical belief in matter is naturally seen as such a case. The central problem for this
sort of account is how it can make sense of what is arguably the most salient fact about
assent (belief): that it represents the world as being a certain way or, in other words, that
beliefs can be either true or false, depending on whether they ‘get things right’ about
how the world is. This is closely connected with the more general problem of how a use
theory of meaning can provide any account of truth. It is to this issue that we now turn.
7.2 The Nature of Truth
It has often been thought that Berkeley was an anti-realist about certain domains of dis-
course, such as natural science (Tipton 1982, 166-167; Newton-Smith 1985; Downing
1995b; Stoneham 2002, 101-106; Hight 2010; Kail 2010), pure mathematics (Brook
1973, 146-155; Hight 2010), ethics (Belfrage 1986a), or revealed religion (Belfrage
290
1986a, 1986b, 2007; Berman 1994, ch. 6).
10
By ‘anti-realism about a domain of dis-
course’ I mean the view that speakers of the discourse in question are not or should not
be attempting to make true assertions. Anti-realism about a domain of discourse should
be distinguished from anti-realism about a putative class of entities: the former has to
do with truth, the latter with reference. To be an anti-realist about talk of forces (a do-
main of discourse) is to hold that utterances of declarative sentences containing the word
‘force’ are not or should not be attempts to state truths. To be an anti-realist about forces
(a putative class of entities) is to hold that users of the word ‘force’ do not or should not
aim (genuinely) to refer to an object.
Anti-realism about a domain of discourse can take either of two forms, fictionalism
or non-cognitivism. According to fictionalism, the sentences of the target domain have
truth-values but, as with the declarative sentences uttered in the course of telling a fic-
tional story, the truth-value of a given sentence is irrelevant to the propriety of uttering
that sentence. According to non-cogntivism, the sentences do not have truth-values at
all. Each of these views comes in two forms, which are known as ‘revolutionary’ and
‘hermeneutic’ (see Kalderon 2005a, 5-7). The revolutionary varieties make the norma-
tive claim that we should change our linguistic practice and aim at something other than
truth (e.g., utility or empirical adequacy) in our assertions in these domains, while the
‘hermeneutic’ forms make the descriptive claim that ordinary speakers do in fact aim at
something other than truth.
11
10. Belfrage’s thesis is restricted to the Manuscript Introduction (see above, p. 269, note 2), and Berman
takes Berkeley’s position in Alciphron to be somewhat ambiguous (Berman 1994, 161-163), but both
suggest that Berkeley may not have held religious mysteries to be genuinely (descriptively) true.
11. The hermeneutic anti-realist typically also holds that we ought to keep on using language in this
way. There is, however, room for at least two other views: one could hold that we do use language in the
anti-realist way but shouldn’t, or that we should reject the discourse under discussion altogether.
291
Various anti-realist views have been attributed to Berkeley, and there is indeed some
reason for thinking that Berkeley does hold a view of this general sort about certain
kinds of talk. For instance, he writes, “[i]n the ordinary affairs of life, any phrases may
be retained so long as they excite in us proper sentiments, or dispositions to act in such
a manner as is necessary for our well-being, how false soever they may be if taken in a
strict and speculative sense” (PHK,x52).
The context of this remark is a discussion of physical causation. P. J. E. Kail con-
cludes from this text and its context that Berkeley is a fictionalist about plain language
causal talk (Kail 2010, 33-36).
Although Kail’s conclusion has some plausibility, it is not forced upon us, for Berke-
ley does not say that these useful phrases are false simpliciter, but that they are false if
taken in a strict and speculative sense. What this suggests is that these ways of speaking
have an interpretation – one which is tempting to philosophers – on which they are false.
Berkeley goes on to say that “even in the most rigid philosophic reasonings” it is im-
possible entirely to avoid these “inaccurate modes of speech,” since they are part of the
very structure of our language. Hence the philosophical reader must “collect the sense,
from the scope and tenor and connexion of a discourse” (PHK,x52). This obviously
includes the Principles itself. Thus, Berkeley seems to think that properly interpreted
these expressions are true.
This interpretation is confirmed by Alciphron. There, Berkeley clearly holds that,
in the language of mathematics, physics, ethics, and revealed religion there are many
words which do not stand for ideas. Nevertheless, he repeatedly says that there are true
propositions in these domains (Alc,xx1.14-16, 2.6, 5.9-10, 6.28, 7.10, 7.18).
The correct conclusion to draw, then, is that Berkeley is, in fact, a realist about
these crucial domains of discourse. That is, he holds that statements in mathematics,
292
physics, ethics, and revealed religion can be genuinely true, despite the fact that they will
typically contain words which do not stand for ideas. But how can Berkeley maintain
this?
The interpretation of Berkeley’s philosophy of language which we have so far de-
veloped provides reason to associate Berkeley with any of three distinct well-known
approaches to theorizing about truth: deflationism, pragmatism, and Carnapian rela-
tivism. In what follows, I examine each of these interpretations in turn and find that
each captures an aspect of Berkeley’s thought, but none is ultimately satisfactory. After
this, I develop what I take to be the correct account of Berkeley’s thought on the subject:
the combination of a realist (correspondence) theory of truth with a very strong version
of holism about meaning.
7.2.1 Berkeley and Deflationism
In light of the interpretation of Berkeley’s philosophy of language I have so far de-
fended, the most obvious account of truth to attribute to Berkeley would be some form
of deflationism. Given that, as I have argued, it is Berkeley’s view that words get to be
meaningful by being associated with rules of use, it seems that in order to answer the
question at hand we ought to attend to the rules of use associated with the word ‘true.’
There is little difficulty here. The most important rule for the use of the word ‘true’ (at
least as applied to sentences) is surely: for any sentence ‘p,’ ‘p’ is true if and only ifp
(Tarski [1935] 1983, 159; 1969, 64).
12
Deflationism, as I understand it here, is the view
that understanding the rules of this sort which govern the use of ‘true’ is all there is to
12. This formulation quantifies into a quote-name, which gives rise to certain well-known technical dif-
ficulties (Tarski [1935] 1983, 159-163). However, these difficulties do not prevent us from understanding
and applying the rule in non-pathological cases. The philosophical issues in which we are here interested
will be clearer if we stick with this simple, though admittedly technically defective, formulation.
293
understanding truth (cf. Burgess and Burgess 2011, 33-34). In the language of chapter
6, this would be to make truth a quasi-entity.
The Theory of Meanings combined a correspondence (realist) theory of truth with
semantic atomism – the view that individual words are the fundamental bearers of mean-
ing. A true sentence is one where things in the world are arranged in a way that corre-
sponds to the arrangement of the words in the sentence. Thus, for instance, on Locke’s
theory, the verbal proposition ‘Melampus is an animal’ corresponds to a certain men-
tal proposition which is an act of joining the idea of Melampus with the abstract idea
of animal. The idea of Melampus corresponds to Melampus, and the abstract idea of
animal corresponds to any animal. The verbal proposition ‘Melampus is an animal’
and its corresponding mental proposition are true only if the thing in the world which
is represented by the idea of Melampus is properly related to one of the things in the
world which are represented by the idea of animal. In this case the relation is identity:
Melampus is identical to one of the animals.
As was observed above, Berkeley could say something like this about the Melampus
example, despite his elimination of mental propositions, by relating words directly to
objects without going through ideas. Thus one might say, for instance, that on Berke-
ley’s view ‘Melampus is an animal’ is true because the extension of ‘Melampus’ (i.e.
the singletonfMelampusg) is a subset of the extension of ‘animal’ (i.e. the set of all
animals).
13
We may assume for the sake of argument that this can be given a nominalistically
acceptable paraphrase. Nevertheless, Berkeley cannot accept anything like this as a
general account of truth, since it can handle only genuine referring expressions. Of
13. Although he discusses only the meaning of terms, and not whole sentences, Daniel Flage’s claim
that Berkeley holds an ‘extensional theory of meaning’ strongly suggests that he would read Berkeley this
way (Flage 1987, ch. 3).
294
course we can talk about the extension of a quasi-referring expression like ‘force,’ but
since the quasi-entities in the extension have no language-independent existence, they
do not seem to be able to do the work set out for them here.
Deflationism neatly solves this problem by absolving Berkeley of the need to find
anything in common between all the truths. The rules governing ‘Melampus is an an-
imal’ are of one sort, and different sorts of rules govern ‘force is proportional to the
product of mass and acceleration,’ and still different sorts of rules govern ‘murder is
wrong.’ On the deflationist proposal, all of these can be true sentences despite the fact
that no relevant, substantive similarity between them can be found. In each case, we
are prepared to assert the sentence, and ought therefore to be prepared to assert that the
sentence is true, and that is all there is to say about the matter.
The main problem with a deflationist interpretation of Berkeley is that, unless some-
how supplemented, it fails to secure any sort of objectivity in the discourse of ethics and
revealed religion. Although Berkeley is willing to accept a kind of relativism or conven-
tionalism about physics (DM,x67), he is surely unwilling to accept any such relativism
about ethics or revealed religion. However, given that the rules of these domains of dis-
course are primarily volitional and affective, it is hard to see how, absent a substantive
characterization of truth, there could be an objectively right set of objects of assent. This
is, of course, the well-known problem of how deflationism can secure the intuitive claim
that truth is a norm of belief and assertion (Wright 1992, 15-21; Price 1998). Why, one
might well ask, should we assent to one sentence rather than another? This is a question
a theory of truth should answer, but deflationism does not answer it. For this reason,
I do not find it plausible to attribute an unadorned deflationism to Berkeley. A more
substantive notion of truth is needed.
295
7.2.2 Berkeley and Pragmatism
Why believe one thing rather than another? At least in the realm of ethics and religion,
there is an answer to this question which is both interpretively and philosophically at-
tractive. Given that Berkeley ties assent so closely to action, it might be thought that
assent is to be judged by the norms of practical rationality, that is, that we ought to
regulate our assent in such a way as to achieve beneficial outcomes. This fits nicely
with Berkeley’s constant emphasis on the usefulness of the philosophical, religious, and
moral doctrines he favors.
This kind of pragmatist or instrumentalist conception of truth has been attributed to
Berkeley by a number of scholars. Unfortunately, most of these scholars have neglected
to explain what they mean by ‘pragmatism.’
Paul Olscamp characterizes Berkeley as holding that “the truth of a sentence . . . is
a function of its effects when interpreted and followed as a rule or direction” (Olscamp
1970b, 45). Although this characterization is rather vague,
14
its central idea fits well
with the account of assent developed above: since assenting to a proposition just is
having some set of dispositions, the conduit by which assent makes contact with the
world would seem to be by the actualization of those dispositions, that is, by our taking
certain actions pursuant to that proposition. It would seem that for the proposition to
make contact with the world in the correct way is just for assent to that proposition
to be useful, i.e. for the effects of actualizing the dispositions involved in the assent to
be beneficial. The pragmatist interpretation of Berkeley attributes to him the view that
usefulness is somehow constitutive of truth.
14. Olscamp admits as much, and says that this is unavoidable because there is “no body of statements
in Berkeley’s writings which can plausibly be taken to constitute a theory of truth” (Olscamp 1970b, 145).
Olscamp is, of course, correct that Berkeley never addresses the issue directly. I think, however, that given
Berkeley’s account of meaning – something he does address directly – and his other commitments we can
come to a rather less vague and schematic Berkeleian view about truth.
296
Claire Schwartz is another scholar who asserts that Berkeley adopts an “instrumen-
talist conception of truth” (C. Schwartz 2010, 46). However, it is not clear that she
actually endorses the pragmatist interpretation as I have defined it, for her language is
more often epistemic:
“what allows us to know whether a proposition is true or false is its utility”
(46)
“Berkeley implies that the utility is also what makes us recognize the truth
of a physical proposition” (46)
“[The] utility [of mathematical propositions] . . . is the proof of their truth”
(54)
We must be careful to distinguish the metaphysical from the epistemological issue.
What we are at present concerned with is the question of what constitutes the truth of
a proposition, not what counts as evidence of its truth. As we shall see, Schwartz is
correct that, for Berkeley the utility of a proposition counts toward its truth; however,
her claim that Berkeley has an instrumentalist conception of truth strongly suggests that
she holds that utility is conclusive proof of truth because truth is entirely constituted by
utility.
It cannot be seriously maintained that Berkeley identifies truth with usefulness. The
very structure of Alciphron speaks against such identification: dialogues 2 and 3 are
concerned with the usefulness of natural religion, and 4 with its truth. Dialogue 5 is
concerned with the usefulness of Christianity and 6 with its truth. The characters in
the dialogue, on both sides of the dispute, explicitly and repeatedly recognize these as
distinct questions (e.g. Alc,xx2.25, 6.1). Furthermore, there are explicit recognitions,
again on both sides, that utility and truth can come apart:
297
CRITO. . . . Errors and nonsense as such are of small concern in the eye of
the public, which consider not the metaphysical truth of notions, so much
as the tendency they have to produce good or evil (Alc,x2.24)
EUPHRANOR. . . . you should not, therefore, if you would appear con-
sistent with yourself, think it necessary or wise to publish hurtful truths
(x3.16).
ALCIPHRON. . . . all proofs drawn from utility or convenience are foreign
to the purpose [of proving the existence of God]. They may prove indeed
the usefulness of the notion, but not the existence of the thing. Whatever
legislators or statesmen may think, truth and convenience are very different
things to the rigorous eyes of a philosopher (x4.1).
15
Finally, it is repeatedly emphasized in the early dialogues that “truth is constant and
uniform” (Alc,x1.15; cf. Alc,x3.15), or, in Alciphron’s more emphatic statement, that
truth is “of a stable, permanent, and uniform nature; not various and changeable like
modes or fashions, and things depending on fancy” (Alc,x1.8). However, as Alciphron
points out, “a thing may be useful to one man’s views, and not to another’s” (x1.16).
Thus it is common ground, in Alciphron, between the freethinkers and the Christians,
not only that utility and truth are conceptually distinct, but that they can come apart. This
implies that usefulness is not even conclusive proof of truth. Admittedly, there are some
texts in which Euphranor and Crito seem to suggest the contrary. The most emphatic is
found inx2.6, where Crito says, “Can anything be more inconsistent, than to condemn in
practice what is approved in speculation? Truth is one and the same; it being impossible
a thing should be practically wrong, and speculatively right.” However, careful attention
to these texts will show that they are all restricted to ‘practical’ truths. Thus Euphranor
15. Also see An,x10.
298
asks rhetorically, “is not the general good of mankind to be regarded as a rule or measure
of moral truths, of all such truths as direct or influence the moral actions of men?” (Alc,
x1.16), and Crito says,
the one great mark of the truth of Christianity is . . . its tendency to do good,
which seems the north star to conduct our judgment in moral matters, and in
all things of a practic nature; moral or practical truths being ever connected
with universal benefit (x5.4).
The conclusion, then is that truth and utility are distinct, but utility is evidence of truth,
and that for explicitly practical claims, that is, claims about what we ought to do, it is
conclusive evidence.
That truth is not to be identified with utility does not entail the falsity of the prag-
matist interpretation, for truth may be constituted by utility in some more complicated
fashion than simple identity. William James, a classical exemplar of pragmatism, stren-
uously objects to the characterization of pragmatism as “calling everything true which,
if it were true, would be pleasant” (James 1907, 234). It is not sufficient for truth, ac-
cording to classical pragmatism, that a belief should bring us a favorable balance of
pleasure over pain. Rather, classical pragmatism, as enunciated by James, holds “that
truth is one species of good” (75). More precisely:
ideas (which themselves are but parts of our experience) become true just
in so far as they help us get into satisfactory relation with other parts of our
experience . . . Any idea that will carry us prosperously from any one part
of our experience to any other part, linking things satisfactorily, working
securely, simplifying, saving labor; is true for just so much, true so far forth,
true instrumentally (58).
299
True beliefs, for James, are beliefs which are good for getting along in the world,
beliefs we can successfully do something with. Beliefs that give us a warm, fuzzy feeling
do not thereby get to be true. The ‘utility’ involved is not merely an optimal balance of
pleasure over pain.
According to Robert Schwartz, this is precisely the sort of view Berkeley adopts.
On this interpretation, although Berkeley accepts that there is a sense in which truth
involves correspondence to reality, this ‘correspondence,’ in the end, involves only cor-
relation between our ideas. The true sentences are those which “enable us to navigate
our cognitive and behavioral environment successfully” (R. Schwartz n.d., 10).
Although the interpretation of Berkeley as a sophisticated pragmatist of this sort en-
joys considerable textual and philosophical support, it does not open the gap between
utility and truth wide enough to account for the textual data from Alciphron. The cases
where Berkeley sees utility and truth coming apart are not just cases where the utility
consists in the belief itself being pleasant. The case seems rather to be something like
this. When it comes to truths about what we ought to do, there can be no separation of
truth from utility, since believing that one ought to do something cannot be separated
from being at least somewhat motivated to do it. Hence correct normative beliefs nec-
essarily, though not infallibly, tend to produce correct actions, and are for this reason
useful. However, all beliefs have some tendency to effect action, since beliefs consist
in dispositions. Other beliefs, which are not explicitly normative, may be true, but nev-
ertheless have a pernicious effect in action. Berkeley does not say when or how this
happens, but the fact that the possibility is left open by the discussion in Alciphron, and
that theism in general and Christianity in particular must, separately, be shown both use-
ful and true, shows that Berkeley cannot hold a general theory of truth which joins truth
to utility so closely as classical pragmatism does.
300
7.2.3 Berkeley and Carnap
Utility is, for Berkeley, not a guarantee of truth. Nevertheless he clearly takes utility as
a reason for belief. Berkeley particularly emphasizes the utility of certain sign systems,
such as Newtonian physics (Alc,x7.10). This might suggest the Carnapian thesis that we
can make sense of truth only within a linguistic framework (sign system) and the choice
between such frameworks is a pragmatic one (Carnap 1950). This is further supported
by the holistic tendency of Berkeley’s theory of quasi-entities: because forces (and, as I
argued in chapter 6, bodies) exist only as artifacts of a particular sign system, the word
‘force’ cannot be assigned a meaning apart from that system. Since the word ‘force’
cannot be meaningful apart from the system, surely no sentence containing ‘force’ can
be true apart from the system.
It is in fact somewhat difficult to arrive at an understanding of Carnap’s claim which
does not make it trivial. It is certainly true that if one makes a sound, or draws a series of
symbols, then whether what one says or writes is true depends on what language one is
speaking or writing. It is further true that, for any false sentence of English, it is possible
to speak a language in which that string of sounds or characters expresses a truth. Thus
one may very well speak a language in which ‘twice two is equal to seven’ means that
snow is white, and in that case when one uttered ‘twice two is equal to seven’ what
one said would be true (cf. PHK,x79). Finally, it is surely correct that decisions about
whether to make linguistic reforms are (often, if not always) to be made on pragmatic
grounds.
These obvious facts, however, answer none of our questions about truth. All that
this tells us is that whether a given sentence is true depends on what it means, and what
it means depends on the conventions of the language being spoken. We have made no
progress on the question of how language makes contact with the world.
301
Matti Eklund has defended the claim that these obvious points are all that Carnap
has in mind in his distinction between ‘internal’ and ‘external’ questions (Eklund 2009).
However, Carnap is often taken to be proposing something more radical. According to
Scott Soames, Carnap’s central claims are two: first, that analytic sentences are true (or
false) only by the conventions of a particular linguistic framework and hence make no
claims about the world, and, second, that many ontological claims, such as ‘there are
numbers,’ are analytic (Soames 2009, 428-435). Thus Carnap holds that we must simply
decide whether to speak a language in which Platonism is trivially true or a language
in which it is trivially false by adopting the relevant convention regarding the truth or
falsity of ‘there are numbers’ and related sentences. There is no substantive fact under
dispute, only the question of which language we ought to speak, and that is a practical
question.
This view does have something to say about our present question for, as mentioned
above (p. 196), on this view the conventions which make certain sentences assertable
would be sufficient to make them true. Thus the conventions governing ‘force’ would
guarantee that ‘there are forces’ (or, at least, ‘where there are accelerations, there are
forces’) is true. On Berkeley’s theory of quasi-entities, this much is clearly correct.
Furthermore, Carnap is quite explicit that “the introduction of the new ways of speaking
[e.g. talk of abstract objects] does not need any theoretical justification because it does
not imply any assertion of reality” (Carnap 1950,x3). This is in line with the anti-
Quinean meta-ontology I attributed to Berkeley in chapter 6: quasi-entities exist, not as
a matter of objective, metaphysical fact, but simply because we have introduced quasi-
referring expressions into our language. The claim that such objects exist gets to be
true in virtue of the introduction of conventions authorizing its assertion; no cooperation
from the world is required.
302
Carnap’s aim in introducing his notion of linguistic frameworks, and the distinction
between ‘theoretical’ framework-internal questions and ‘pragmatic’ framework-external
questions, was to trivialize metaphysical disputes such as the dispute between nominal-
ists and Platonists. It was part of the positivist program of the elimination of meta-
physics. Now, Berkeley can actually be seen as having some degree of sympathy with
this program. For instance, in his notebooks he reminds himself “To be eternally banish-
ing Metaphysics & recalling Men to Common Sense” (N,x751). However, Berkeley’s
contrast between ‘metaphysics’ and ‘commonsense’ is certainly not the same as Car-
nap’s distinction between ‘metaphysics’ and ‘science.’ By Carnap’s lights, Berkeley is
surely to be classed as a metaphysician and, what is worse, a theologian.
16
In “Empiricism, Semantics, and Ontology,” Carnap seems to want to deflate all onto-
logical questions. This would seem to suggest that Carnap’s anti-metaphysical position
is that all entities are (in our terms) mere quasi-entities. However, it is not clear that this
extreme doctrine is tenable, or consistently held by Carnap. When Carnap talks about
the decision to adopt the ‘thing language,’ he treats this as a question of whether to add
talk of ‘things’ to a pre-existing “language of sense data and other ‘phenomenal’ enti-
ties” (Carnap 1950,x2); the ‘erlebs’ which were the fundamental entities in Carnap’s
Aufbau were of this sort.
17
Furthermore, he describes the pragmatic question of whether
to adopt a certain linguistic framework as the question, “Are our experiences such that
the use of the linguistic forms in question will be expedient and fruitful?” (x2). Carnap
thus seems to recognize the existence of phenomenal experience as prior to our decision
16. A. J. Ayer, rather oddly, denies that it is “fair to regard Berkeley as a metaphysician,” and then
goes on to complain about Berkeley’s understanding of God and the self (Ayer 1952, 53-54). These very
doctrines clearly show that, despite certain similarities, Berkeley is not an anti-metaphysical thinker in
anything like the way the positivists were.
17. For a brief, but rigorous, critical exposition of Carnap’s system in the Aufbau, see Goodman 1977,
ch. 5. ‘Erleb’ is Goodman’s abbreviation for Carnap’s German neologism, ‘elementarerlebnisse’ (112).
303
of which linguistic frameworks to adopt. This appears to commit him to holding that
sense data, and sense data alone, are (in our terms) genuine entities. If this is his view,
then he was perhaps not so anti-metaphysical as he intended to be (see Carnap 1950,
x4).
Leaving aside the question of whether Carnap succeeds in his anti-metaphysical pre-
tensions, let us return to the question of truth. Carnap’s view is that some sentences (the
analytic truths) are true by convention alone, whereas others (the synthetic truths) are
true due to a combination of convention with experience. The conventions of language
authorize the assertion of analytic truths regardless of what the experienced world is like;
by contrast they authorize the assertion of synthetic sentences only if the experienced
world is a certain way. That the assertion is duly authorized (warranted) is sufficient for
its truth.
Carnap’s general approach is easily severed from his verificationism. All that is
needed is to give a different account of the circumstances in which the assertion of
synthetic sentences is authorized. These circumstances will, of course, be different for
different sentences. However, when this separation is combined with Berkeley’s Inter-
nalist Requirement (above,x5.2), difficulties emerge. The Internalist Requirement is
what prevented us from having a linguistic rule like say ‘there is a table here’ only in
the presence of mind-independent, material tables (ch. 6). The application conditions
of the linguistic rules we follow must be accessible to us prior to learning the rule.
However, we cannot follow the rule say ‘there are other minds’ only if there are other
minds, because our ability to think about other minds depends on our ability to general-
ize from our own experience, and generalization is only possible by means of language
(see above,x6.5). We cannot recognize the presence of other minds in a way that is
independent of language, yet other minds are not mere quasi-entities; ‘there are other
304
minds’ is not analytic. Thus the conditions in which the assertion of ‘there are other
minds’ is warranted must come apart from the conditions in which it is true.
Furthermore, if we did allow externalistic rules for assertion, we would no longer
have a distinctive theory of truth, for at that point we might as well say that the rule is
assert a sentence only if it is true. It is the internalistic nature of the rules of assertion
that make this a distinctive conception of truth.
Additionally, Berkeley does not seem to distinguish ‘pragmatic’ and ‘theoretical’
questions so sharply as Carnap does. As we have seen, although Alciphron distinguishes
the question of usefulness from the question of truth, both together are taken to be rele-
vant to the question of whether to endorse Christianity (see especially Alc,x7.32). Now,
either Christianity counts as a linguistic framework or not. If it does then the question
whether to endorse Christianity is an external question, and thus, on Carnap’s view, only
the pragmatic considerations ought to count; if Christianity does not count as a linguistic
framework, then this is an internal question and only the evidential considerations ought
to count. But for Berkeley both kinds of considerations matter. Carnap does admit that
decisions about accepting linguistic frameworks, “although . . . not of a cognitive nature,
will nevertheless by influenced by theoretical knowledge” (Carnap 1950,x2), but the
role of theoretical knowledge here is only to help us determine what will be useful.
Finally, Carnap’s whole aim is to defend a sort of pluralism or relativism according
to which there is no question of one language being better than another in any terms
other than practical utility. Although Berkeley does make a move of this sort with re-
spect to the dispute between Newton and Torricelli about the nature of force (DM,x67),
he does not want to take this kind of approach across the board. According to Berkeley,
the ontologies of the materialist and the atheist are not only pernicious but also factually
305
false. The materialist does not merely adopt an alternative set of conventions; she mis-
describes the world. If Berkeley is to preserve this claim, he must adopt an account of
truth which takes ‘fit with objective reality’ more seriously.
Berkeley has real affinities with Carnap. He holds, for instance, that we may intro-
duce new entities simply by adopting conventions for the use of certain words, with-
out committing ourselves to the language-independent existence of those entities. He
holds that more than one system of such entities is possible, and that we ought to adopt
whichever such system proves most useful. Furthermore, he holds that once such a sys-
tem exists, we may state truths within it, despite the fact that the entities these truths
appear to be about have no existence or nature independent of the system (see above,
x6.2). Nevertheless, the Carnapian theory simply does not capture all of Berkeley’s com-
mitments. Berkeley was a metaphysician and, as such, was in need of a more robust,
objective conception of truth.
7.2.4 Realism about Truth, Holism about Meaning
Each of the theories of truth we have so far discussed captures an aspect of Berkeley’s
thought, but none of them can be Berkeley’s ultimate position. Berkeley needs to be
able to account for the meaning of the word ‘true’ within his use theory, and surely the
disquotation schema will play a large role in that account, as on the deflationist theory,
but Berkeley is in need of a more ‘inflated’ account of truth than this. Berkeley sees a
strong connection between utility and truth, yet also leaves a gap between them which
is wider than that allowed by classical pragmatism. Finally, Berkeley recognizes that
different sign systems, involving different quasi-entities, are possible, and that the state-
ments within each of these systems may be true. He further holds that we should choose
306
among these sign systems on a pragmatic basis. Yet Berkeley intermingles considera-
tions of utility and considerations of truth far more thoroughly than Carnap does, and he
wants more objectivity than Carnap allows. What kind of theory of truth can we then
attribute to Berkeley?
To begin to answer this question, let us return to Berkeley’s discussion of ‘speaking
with the vulgar.’ In the passage already quoted, Berkeley tells us that “[i]n the ordinary
affairs of life, any phrases may be retained so long as they excite in us proper sentiments,
or dispositions to act in such a manner as is necessary for our well-being, how false
soever they may be if taken in a strict and speculative sense” (PHK,x52). Note that
Berkeley speaks of how false the ‘phrases’ are, implying that falsity comes in degrees.
It might be though that this is merely rhetorical. However, Berkeley goes on to say that
“propriety being regulated by custom, language is suited to the received opinions, which
are not always the truest” (x52, emphasis added). Philosophers may, to some small
degree “alter the bent and genius of the tongue we speak,” but some “inaccurate modes
of speech” will always remain, because we can only incrementally alter the language we
have received from custom. Thus the phrases used, even in Berkeley’s own works, are to
some degree false or inaccurate, but if we consider “the scope and tenor and connexion”
of the work, we will be able to “collect the sense” Berkeley intends (x52).
This is not the only place Berkeley speaks of degrees of truth and falsity.
18
In re-
sponse to Alciphron’s insistence that Pagan religion brought about virtue in the Greco-
Roman world, Crito says: “we [Christians] freely own [that Pagan religions] produced
good effects on some people: but then these good effects were owing to truths contained
in those false religions: the truer, therefore, the more useful” (Alc,x5.10). Here the total
religious system of Greco-Roman Paganism is first described as false, but then admitted
18. Cf. also NTV,x90: “this account [of the inverted retinal image] to me does not seem in any degree
true.”
307
to be true to some degree (though, Crito insists, to a much lesser degree than Christian-
ity), insofar as it contains individual truths. Similarly, back in the Principles, plain lan-
guage contains falsehoods. Berkeley compares this to the way in which ordinary modes
of speech assume heliocentrism, and argues that just as Copernicanism can be adopted
without disturbing these modes of speech, so immaterialism can be adopted while still
speaking, for instance, of physical causation. What is necessary is a reinterpretation of
these modes of speech.
As I mentioned above, P. J. E. Kail finds evidence of a form of fictionalism here
(Kail 2010, 33-36). Kail takes Berkeley’s discussion of Copernicanism as evidence that
Berkeley endorses a form of revolutionary fictionalism – that is, that Berkeley is arguing
that we ought to change how we use language, and begin to regard talk of physical
causation as a fiction.
I argued above that this text does not provide adequate support for the attribution
of fictionalism to Berkeley, since Berkeley seems to hold that these ‘inaccurate expres-
sions’ are true when correctly interpreted. Nevertheless, there is a degree of truth (!) in
Kail’s proposal. Most importantly: Berkeley is proposing linguistic reform. What he
is not doing is suggesting that these claims, as ordinarily interpreted in plain language,
are entirely false, and that the linguistic reform we should implement is to regard them
as mere figures of speech. What he is suggesting instead is that by taking a holistic ap-
proach (attending to “the scope and tenor and connexion of a discourse”) we can come
to recognize the intermixture of truth and falsehood in our ordinary ways of speaking
and, by incremental reform, come to speak more truly.
The incremental nature of this process is even more explicit in the Theory of Vision
Vindicated:
308
The work of science and speculation is to unravel our prejudices and mis-
takes, untwisting the closest connexions, distinguishing things that are dif-
ferent, instead of confused and perplexed, giving us distinct views, gradu-
ally correcting our judgment, and reducing it to a philosophical exactness.
And, as this is the work of time, and done by degrees, it is extremely dif-
ficult, if at all possible, to escape the snares of popular language, and the
being betrayed thereby to say things strictly speaking neither true nor con-
sistent . . . For, language being accommodated to the prænotions of men and
use of life, it is difficult to express therein the precise truth of things, which
is so distant from their use, and so contrary to our prænotions (TVV,x35;
cf. Siris,x296).
One consequence of this approach is that when Crito admits that Pagan religion
contained some useful truths, he need not be taken as claiming that there are a few indi-
vidual sentences to which Pagans assented which were true, while the majority of their
sentences were false. Rather, it may well be that one and the same sentence contains
both truth and falsehood, and it may perhaps be the case that it is not until after some
linguistic reform that we are in a position to untangle the truth from the falsehood.
To see how this works, let us consider three cases, one historical case, and two
thought experiments from recent philosophical literature.
Phlogiston. Joseph Priestley hypothesized the existence of a substance
called ‘phlogiston’ which was released in combustion. He was thus inclined
to assert ‘phlogiston is being released’ whenever combustion was going on.
As it turns out, there is no such thing as phlogiston.
Bligers. When the first settlers arrived in the hitherto unpeopled land of
Pluralia, they observed (always from a fair distance) what appeared to be
309
black tigers, and they coined the name ‘bliger’ for them . . . Occasionally,
Pluralians would make idle remarks along the lines of ‘There’s a bliger
crossing the field’, and that was the extent of their interaction with bligers.
A few centuries after the settlement of Pluralia, however, a foreign zoolog-
ical expedition discovered that, in a way, there were no bligers. “A bliger
(Quasi-Tigris Multiplex Pluralianus),” their report read, “is really six ani-
mals. Its ‘legs’ are four monkey-like creatures, its ‘trunk’ a sort of sloth,
and its ‘head’ a species of owl. Any six animals of the proper species can
combine temporarily to form a bliger” (Van Inwagen 1990, 104).
Spirits. Imagine a girl who has inherited a primitive conception of person
from her linguistic community. Say that she takes persons, including her-
self, to be spirits, and that she takes spirits to be parcels of a subtle and
ethereal fluid that permeates bodies. What is the content of her assertion
“I am in pain” made after a visit to the local medicine man? (Gupta 2006,
140)
What these three cases have in common is that the assertion in question is reliably
correlated with some real fact about the world. Priestley asserts ‘phlogiston is being
released’ only when combustion is going on; the Pluralians asserts ‘a bliger is crossing
the field’ only when six animals of certain species are crossing the field in a certain
arrangement; and the girl in Spirits asserts ‘I am in pain’ only when she is in pain.
Yet in each case a false presupposition about the world is so deeply embedded among
the speaker community that it may plausibly be regarded as part of the meaning of
the assertion (cf. our previous remarks about ‘phlogiston,’ p. 197). Furthermore, in a
case like Spirits it may be impossible, prior to linguistic reform, to express the truth
contained within the sentence without presupposing the falsehood. We can say, in our
310
actual language, “you are indeed in pain, but there is no such thing as a subtle, ethereal
fluid,” but the girl in the story may not have the linguistic apparatus to say that she is in
pain in a way that does not presuppose the existence of subtle, ethereal fluids.
Berkeley’s claim in PHK,x52 is that all human language is like this to a greater
or lesser degree. We cannot entirely escape from our linguistic inheritance to express
absolute truth, entirely untainted by falsehood. As Berkeley puts it in a similar passage
in the New Theory of Vision, “language being accommodated to the common notions
and prejudices of men, it is scarce possible to deliver the naked and precise truth without
great circumlocution, impropriety, and (to an unwary reader) seeming contradictions”
(NTV,x120).
What should we say about these cases? Do the speakers in our three stories speak
the truth? According to Anil Gupta “it [is] best not to give a straight answer to these
questions” (Gupta 2006, 140). Berkeley agrees. These assertions are not true or false
simpliciter, but rather to some degree true and to some degree false.
As I have said, Berkeley does not think that we can free ourselves from falsehood
entirely, but he does think that we can make incremental progress by linguistic and
conceptual reform. His discussion of causation – the context of the remark we have
been explicating – is instructive here. Berkeley’s claim is that the sentence ‘fire heats’
(PHK,x51) is false to a high degree “if taken in a strict and speculative sense,” but may
be retained insofar as it is associated with dispositions to feeling and action “necessary
for our well-being” (x52). The key truth in the claim that fire heats is that a good way
to make something hot is to put it near a fire. Assent to the sentence ‘fire heats’ does
involve a disposition to follow this piece of advice. This useful part of the belief is
unaffected when ‘fire heats’ is reinterpreted as the claim that fire signifies heat in the
language of nature (see below,x8.4) and thereby freed from the pernicious half-belief in
311
matter with which it is now connected (see above,x7.1). Contrary to Kail, this does not
involve regarding ‘fire heats’ as literally false and continuing to use it in some non-literal
sense (e.g.using it to mean: ‘according to the fiction of physical causation, fire heats’).
Instead, it involves recognizing that ‘fire heats,’ interpreted literally, contains a degree
of truth and a degree of falsity and then adjusting the literal meaning of the phrase to
get more truth and less falsity.
As we have observed, it is not always possible, prior to linguistic reform, to state
separately the truth and the falsehood contained in a partially true statement. This is
why Berkeley insists that a holistic approach must be taken. Not all of the needed
linguistic reforms can be recommended explicitly by stating the new rules in our existing
language. Thus Berkeley must model the new language he is recommending. We must
pay attention to the whole of Berkeley’s writings, and begin speaking and thinking in
his language, before we can understand what he is after.
Of course, Berkeley tries, as best he can, to help us along the way by making explicit
remarks about the meanings and various ambiguities of key words like ‘exist.’ What he
tells us in PHK,x52, though, is that it is not possible for him to do this so thoroughly
“as never to give a handle for cavillers to pretend difficulties and inconsistencies.” For
the girl in Spirits, to say ‘you are indeed in pain, but there are no subtle, ethereal fluids’
is to assert a contradiction. Yet how can we bring about the needed linguistic reform
except to begin asserting what is now a contradiction? It is by this very change in our
linguistic practice that we sever mental language from commitment to subtle ethereal
fluids. Indeed, to use an example discussed by Berkeley, van Inwagen, and Gupta, we
are still inclined to say such things as ‘it was cooler in the garden after the sun had
moved behind the elms’ (Van Inwagen 1990, 101; cf. PHK,x51; Gupta 2006, 140),
though we may now go on to say ‘of course the sun didn’t really move at all; the earth
312
did.’ If Copernicus or Galileo had said one of these things right after the other, he
would have contradicted himself. But asserting such a contradiction, and thereby forcing
reinterpretation of old patterns of speech, is a way of introducing reform. By continuing
to insist that fire heats, and that the fire is the physical cause of (e.g.) the boiling of the
water while at the same time denying that fire could possibly be the efficient cause of
anything, Berkeley forces reinterpretation of our ordinary talk about causation in the
physical world. In this case, Berkeley is, of course, able to give some account of what
physical causation amounts to, and so to say quite a bit about the reform he is advocating
(again, seex8.4, below), but his concern in PHK,x52 is that he cannot possibly say
enough about this sort of reform to avoid all “difficulties and inconsistencies.” Thus it
is primarily by personally adopting this new way of speaking that Berkeley pushes his
reformist agenda.
We are now in a position to give a rough intuitive characterization of what I take
to be Berkeley’s understanding of truth. A large collection of sentences which embody
a certain set of linguistic rules (including rules for non-linguistic actions) can together
constitute a sort of portrait of the world, and as a portrait can be a more or less accurate
likeness of its subject, so such a collection of sentences can, as a whole, be more or
less true. To attribute truth or falsity to an individual sentence is like saying that one
brush stroke is a good or poor likeness of some aspect of the subject. This may be a
perfectly sensible thing to say in the context of the portrait as a whole; on the other
hand, the single brush stroke, by itself, does not represent or portray anything at all. It
is the context of the portrait that makes the brush stroke a representation.
This general approach to truth can be captured by the slogan: ‘realism about truth,
holism about meaning.’ Berkeley is a realist about truth insofar as he holds that truth
does involve some sort of genuine matching or correspondence between what we say
313
and how the world is. He is a holist about meaning insofar as he holds that the match or
mismatch is primarily between a total way of speaking and the world, and only deriva-
tively between individual sentences and the world. Indeed, I will go on to argue that, due
to the strong pragmatic elements in Berkeley’s understanding of meaning and truth, it is
in fact more accurate to say that Berkeley’s theory of truth involves a matching between
a form of life and the world.
19
This kind of matching comes in degrees, and sentences
inherit a degree of match or mismatch from the form of life in which they are embedded.
This account is, as I said, rough and intuitive. Furthermore, the previous two para-
graphs have glossed over the most difficult question for a theory of truth, namely, in
what this matching with the world consists. Additionally, the only reason I have so far
given for attributing this theory to Berkeley is that it makes good sense of one particular
difficult text, PHK,x52. In order to put some flesh on the bones of this theory, and to
defend its attribution to Berkeley more fully, I now proceed to show that, in combination
with the account of assent I attributed to Berkeley inx7.1, it can account for those as-
pects of Berkeley’s thought which rendered each of the three interpretations dismissed
above initially attractive, and that it does not share their faults.
19. I use this Wittgensteinian phrase intentionally. The ‘holist’ aspect of my interpretation, as applied
to Christianity, sees Berkeley as anticipating Wittgensteinian approaches to philosophy of religion. See,
e.g., Phillips 1966, (1970) 1993. The relationship of the ‘realist’ aspect to Wittgenstein and his followers
is more complex. D. Z. Phillips takes himself to be following Wittgenstein in holding that the debate
between realists and anti-realists is fundamentally confused (Phillips [1991] 1993). On the other hand,
Iakovos Vasiliou takes Wittgenstein to endorse a thesis Vasiliou calls ‘the factuality of religious belief’
(Vasiliou 2001, 30-31), which certainly sounds like a realist thesis, though it is not clear how this is
to be reconciled with Vasiliou’s uncertainty about whether Wittgenstein endorsed a form of relativism
(47-48). At the opposite extreme from Vasiliou, several scholars take Wittgenstein to endorse a more or
less straightforward form of anti-realism (see, e.g., Bailey 2001; Nielsen 2001; Arrington 2001). This is
obviously not the place for extended interpretation of Wittgenstein and his followers. I have said this much
only to indicate that there is one relatively clear element of the Wittgensteinian point of view which, on my
interpretation, Berkeley anticipates, and this is the view that religious propositions (or at least propositions
pertaining to religious mysteries) are intelligible only within the context of the religious form of life in
which they have their use. Beyond this, I take no position on the interpretation of Wittgesntein, and
therefore can take no position on the extent to which Berkeley’s views are similar to Wittgenstein’s.
314
The attribution of a deflationist account of truth to Berkeley was suggested by Berke-
ley’s use theory of meaning, but was found unable to secure the degree of objectivity
Berkeley wants. On the theory of truth under consideration, the disquotation schema
turns out not to be the most important rule for the use of ‘true.’ This is because the
all-or-nothing attribution of truth to sentences is not the most basic use of ‘true.’ Rather,
a Christian, immaterialist form of life is true to a large degree, while a freethinking, ma-
terialist form of life is false to a large degree. Within such a form of life, sentences can
be evaluated as true or false. However, the truth of (e.g.) the doctrine of the Trinity is
not a matter of its being true according to Christianity. It is rather a matter of Trinitarian
modes of speech making a positive contribution to the truth of Christianity. ‘The Son
is of one substance with the Father’ is one of the good brush strokes, which makes the
portrait a good likeness.
Strictly speaking, Christianity, immaterialism, freethinking, and the like are not
themselves forms of life (Phillips 1986, 79), but can be significant aspects of a form
of life, with widespread consequences. The reason we do not talk as though it is an
entire form of life we are evaluating for truth is that, if the form of life in question were
totally alien compared to our own, we would be unable to engage in any evaluation in
the first place. What we can do is consider forms of life which are possible revisions
of our own form of life. Thus Alciphron can consider the form of life he would live if
he became a Christian, and Euphranor can consider the form of life that he would live
if he became a freethinker, and these can be evaluated as true or false to some degree.
This is what one is doing when one says such things as ‘Christianity is true:’ one is
comparing the Christian forms of life one might adopt with the non-Christian forms of
life one might adopt.
315
Since these sorts of evaluations are the fundamental ones in Berkeley’s understand-
ing of truth, we need to begin by understanding the rules for these evaluations. First,
there is a disquotation-like rule. It is disingenuous to attribute truth to a form of life
which one does not adopt. To believe in Christianity is to have a complex set of disposi-
tions to thought, speech, feeling, and action, all of which, according to Berkeley, center
around “the love of God and man” (Alc,x5.15). So the disquotation-like principle for
this use of ‘true’ is that the attribution of truth to a form of life is correct if and only if
the adoption of that form of life is correct.
This rule is disquotation-like in the following way. According to the disquotation
rule, it is a linguistic mistake to assert “p’ is true’ when one is not prepared to assert ‘p.’
Christianity, or immaterialism, or Marxism, or any form of life one chooses, is consti-
tuted by linguistic and non-linguistic behavior. Thus the Christian is committed, on the
one hand, to asserting ‘God exists’ and, on the other hand, to refraining from swindling
the poor. To assert ‘Christianity is true’ when one is unprepared to assert ‘God exists’ or
to refrain from swindling the poor is to make a mistake. It is to contradict oneself. Now,
of course, people do contradict themselves, and that with some frequency. The claim
is not that to assert ‘Christianity is true’ and fail consistently to adopt Christianity as a
form of life is impossible; the claim is, rather, that to do this is to commit an error, an
error akin to asserting “p’ is true’ while being unprepared to assert ‘p.’ This is because
to say that Christianity is true is to evaluate it as worthy of adoption, just as to say “p’ is
true’ is to evaluate ‘p’ as worthy of assertion.
The correctness of adopting a form of life is not an all or nothing matter. Whether to
adopt a form of life is a practical question and, as a result, practical utility must surely
be one of the factors to be weighed. But practical utility clearly comes in degrees.
20
So,
20. For the same reason, William James also sometimes speaks of truth as coming in degrees (James
1907, 63-64).
316
to put matters more correctly, a form of life is true to the degree that it is worthy of our
adoption, and the belief that a form of life is worthy of adoption is constituted in part by
an inclination to adopt it.
21
This disquotation-like rule cannot be all there is to the truth of a form of life. If we
are to have the kind of substantive conception of truth which I have argued Berkeley
needs, then we need to say something about what makes a form of life worthy of adop-
tion. We have already indicated that utility plays a role here but since, as I argued in
x7.2.2, Berkeley permits a wider gap between utility and truth than classical pragmatism,
utility cannot be the only desideratum.
The text where Berkeley comes closest to addressing our present question – what
makes a form of life worthy of adoption? – directly is in Crito’s speech on how we
ought to choose among religions. Crito says:
one of more spirit and a juster way of thinking [than the freethinkers], makes
[the plurality of religions] a step whence he looks about, and proceeds to ex-
amine, and compare the differing institutions of religion. He will observe,
which of these is the most sublime and rational in its doctrines, most vener-
able in its mysteries, most useful in its precepts, most decent in its worship?
Which creates the noblest hopes, and most worthy views? He will consider
their rise and progress: which owes least to human arts or arms? Which flat-
ters the senses and gross inclinations of men? Which adorns and improves
the most excellent part of our nature? Which has been propagated in the
most wonderful manner? Which has surmounted the greatest difficulties or
21. This does not imply the Socratic thesis that we are always inclined to adopt a form of life just to the
degree that we judge it worthy of adoption, for we may have other inclinations besides those which help
to constitute our evaluations. Furthermore, Berkeley likely holds the libertarian thesis that we are able
to act contrary to our inclinations. Each of these considerations is individually sufficient to preserve the
possibility of akrasia.
317
showed the most disinterested zeal and sincerity in its professors? He will
inquire, which best accords with nature and history? He will consider, what
savours of the world, and what looks like wisdom from above? (Alc,x7.33).
The criteria identified by Crito can be reduced to five general headings. We ought,
according to Crito, to adopt a religion which: (1) exhibits evidence of supernatural
origin; (2) has doctrines which can be rationally believed; (3) has practices which can
rationally and morally be followed; (4) produces useful emotions; and (5) exhibits pos-
itive effects on human character.
The last three of these are pragmatic in nature: based on what we already know or
believe, we must evaluate whether the form of life in question will be useful to us. This
includes the question of what the rules therein actually tell us to do, as well as the likely
effects of any emotions that way of life would tend to make us feel. We can confirm the
helpful or harmful effects of adopting a form of life by observation of those who have
adopted it.
The first two criteria, on the other hand, do not appear to be pragmatic. Here we
are told to look for evidence that the doctrines of the contemplated religion are true.
This includes examining both evidence that bears directly on their truth and evidence
that bears on the question of whether they are actually revealed by God.
22
(If they
are revealed by God, then they must of course be true.) What this means is that even
though these individual sentences cannot be evaluated for truth apart from the form of
life in which they are embedded, we must be able to evaluate them for truth without
first adopting the form of life. We must be able to ask, counterfactually, if I adopted
Christianity, would I go around saying a lot of false things?
23
Here it is important to
22. On the question of how these two types of evidence fit together, see Pearce, forthcoming(a).
23. This is in agreement with Kai Nielsen: “to understand religious discourse one must have a partic-
ipant’s understanding of it. However, this certainly does not entail that one is actually a participant, that
318
remember that being true within the context of Christianity is not the same as being true
according to Christianity. Obviously, the religious utterances of Christians are, for the
most part, true according to Christianity,
24
but this is not the question Crito recommends
that we ask. He recommends that we ask whether these things would be true in the
context in which they would be uttered, namely, the context of a Christian form of life.
This brings us, then, to the question of the attribution of truth to individual sentences.
Here the disquotation schema applies, though again it must be modified to accommo-
date Berkeley’s recognition of degrees of truth. Thus we can say
25
that for any sentence
p, ‘p is true’ exhibits the same degree of truth or falsity as p. This, however, does not
get us very far toward an account of truth: it only tells us when, and to what degree,
truth is properly attributed to sentences which themselves attribute truth to other sen-
tences, where an independent evaluation of the degree of truth of those other sentences
is possible. Again, a more substantive conception is needed.
one accepts or believes in the religion in question . . . [thus] philosophy [can] relevantly criticise religions
or forms of life” (Nielsen 1967, 193). Nielsen frames this claim as a criticism of the Wittgensteinian
tradition in philosophy of religion, but D. Z. Phillips has vigorously disputed the notion that Wittgen-
steinianism was ever committed to the kind of fideism Nielsen attempts to foist on it (Phillips 1986, ch.
1; 2003).
24. Or perhaps not; many churches are characterized by shocking levels of theological ignorance. There
is an interesting theoretical question here about the relationship between Christian beliefs and the beliefs
of Christians. Could it be the case that there is some Christian belief which is not believed by most
Christians (e.g., because only sophisticated theologians understand it)? There is a strong populist ten-
dency in Berkeley’s religious thought, which would no doubt incline him to answer ‘no’ (N,x405; Alc,
x7.13; TVV,x6; see Pearce, forthcoming[b],x4); on the other hand, although Berkeley criticizes Ro-
man Catholicism for demanding excessive deference to the clergy, he does believe in a special role for the
clergy as authoritative teachers of religious doctrine: “I grant it is meet the Law of Christ should like other
laws have magistrates [i.e., the clergy] to explain and apply it” (CGB, 425). In any event, someone who
adopted Christianity after undertaking the kind of thorough investigation recommended by Crito would
certainly then go around saying things which are true according to Christianity.
25. Subject to the caveats about quantifying into quote-names mentioned above (p. 293, note 12).
319
To make progress toward such a substantive conception, recall that, on Berkeley’s
view, sentences provide rules of thought, feeling, speech, and action. What we need is
to identify a sense in which such rules can be said to match the world or not.
There is an obvious sort of match or mismatch which can be attributed to certain
sentences containing genuine referring expressions. Thus consider the sentence ‘all cir-
cles are red.’ One of the instructions this gives us is, when imagining a circle, to imagine
it as red. Thus if we are trying to imagine what a situation we might get ourselves into
would be like, we will imagine that any circles in that situation will be red. This will
result in a feeling of disappointed expectations in case we should find a non-red circle.
These sorts of cases are the fundamental ones in terms of the contact we make with the
world.
Thus far, Berkeley will sound like a verificationist. However, we should recall that
on Berkeley’s view there will be sentences of this sort which are in principle unverifiable
by me. Such sentences include ‘other minds exist’ and ‘there is shame not felt by me’
(see below,x8.6). These contain only genuine referring expressions, that is, the words
in these sentences are used as labels; however, I cannot personally get into a position to
label those things directly.
Adapting a term from Quine, let us call atomic sentences containing no categore-
matic terms other than genuine referring expressions ‘observation sentences’ (Quine
1960). Although such sentences cannot, even in principle, always be connected to ob-
servation by me, they can always in principle be connected quite directly to observation
by someone.
26
The truth or falsity of observation sentences is partly constituted by the
26. This connection will not necessarily involve the possibility of absolute verification or falsification;
for instance no one (other than God) can falsify ‘there is shame [felt by someone]’ or verify ‘there is no
shame [felt by anyone].’
320
simple referential function of those expressions, and this is as independent of the broader
framework in which they would embedded as one could like.
This referential ‘matching’ is, however, only one contributor to the truth or falsity of
observation sentences. As we have discussed (x5.4), these sentences will be connected,
within a framework, with inference rules, whereby we can get from one sentence to
another. If the rules associated with the words in the sentences allow us to infer false-
hoods from them, then the sentences in question will have a degree of falsity themselves.
Furthermore, the sentences may have emotional or pragmatic associations, which can
constitute part of their meaning, which give bad instructions for getting around in the
world. For instance, one can imagine a word, call it ‘bircle,’ which is a genuine referring
expression picking out blue visual circles, but has a strong negative emotional valence.
It may be that all of the other observation sentences entailed by the ‘bircle’ sentences
are true. Nevertheless, ‘there is a bircle’ will be in some degree false if the blue visual
circle in question is a blueberry and the emotional association in question prevents me
from eating it. To take a real-world case: the same sort of falsity would be exhibited by
a sentence which used a racial slur to make a true descriptive statement about the mem-
bers of the race in question. Since we have neutral racial labels in our language, we can
separate the true from the false in such a sentence by saying, e.g., ‘it is true that many
Irish-Americans have red hair, but there is no need to call them names.’ However, one
can imagine a language in which the only label for a particular group was the negatively
valenced one. There might be a language in which ‘he’s an Irish-American, but he’s not
a bad person’ was a contradiction. The best way to change this state of affairs might very
well be to go around asserting such contradictions, thereby forcing reinterpretation, as
discussed above. Indeed, things like this seem to happen with some frequency, and this
is the reason that the preferred labels for negatively stereotyped groups change so often:
321
the negative stereotypes about the group tend over time to get built in to the meaning
of the label for that group. Thus, for instance, at some point in the history of English
‘she’s dumb but not stupid’ became a contradiction (it hadn’t been one before), and this
unfortunate state of affairs was corrected by the introduction of a new word, ‘mute,’ for
people who can’t speak.
27
Observation sentences are the simplest case, and we can see that even here things are
quite complicated. My key point at present is that even for observation sentences, there
is a very strong holism about meaning, and there are significant pragmatic elements of
meaning. Where genuine reference is absent – which is to say, in most of our language –
the emotional, practical, and inferential connections are all there is to the truth or falsity
of a sentence, and since sentences have a variety of such conventional connections, they
will be likely to contain a mixture of truth and falsity. As Quine said, the ‘fabric’ of
our language “impinges on experience only along the edges” (Quine 1951, 39), yet it
is in terms of these edges that the whole fabric must be evaluated. Berkeley endorses a
stronger pragmatism than Quine or Carnap (though weaker than James); for him feeling
and action are as much a part of the interface between the linguistic fabric and the world
as observation. This is why I have said that Berkeley is a more radical holist even
than Quine: the ‘whole’ relative to which the individual statements must be evaluated
includes feeling and action. It is a form of life and not just a theory.
We return now to our question of the conditions under which we ought to adopt a
form of life. We have said that we ought to consider evidence of the usefulness of that
form of life as a whole, but also evidence of the truth of the individual sentences to
which we would assent were we to adopt it. We have now seen that the truth of some
individual sentences (namely, the observation sentences) does include a straightforward
27. Objection: This whole discussion runs roughshod over the distinction between semantics and prag-
matics. Reply: Berkeley doesn’t have that distinction.
322
sort of piece-by-piece correspondence, but, even in this case, this correspondence is not
the whole story. We must ask, in the context of the form of life in which it is embedded,
does assent to this sentence lead to assent to other sentences which are false (to a high
degree)? We must also ask, does assent to this sentence (in that context) lead to helpful
or unhelpful emotions and actions?
Many of the historical claims involved in Christianity (and Crito explicitly mentions
history) do imply observation sentences about past occurrences: Christianity claims that
certain people saw and heard certain things. Furthermore, there are claims about what
people will experience in the future (the heavenly reward). In these cases there is a ques-
tion of descriptive adequacy.
28
However, there are also, in Christianity, specific practical
instructions. Furthermore, words and phrases have certain emotional associations (the
widespread adoption of Christianity, for instance, radically altered the emotional asso-
ciations of the word ‘cross’ and its equivalents in other languages). We need to ask
about the correctness of the individual practical and emotional associations in the indi-
vidual statements Christianity requires us to affirm, as well as the general tendency of
the system as a whole.
The deflationist is right about the close connection – indeed, identity – between the
question whetherp and the question whether ‘p’ is true. However, Berkeley has a more
substantive account of how we are to evaluate the question whetherp. The pragmatist
is right about the close connection between truth and utility. However, because of the
correspondence which is needed for truth in the case of observation sentences, truth and
utility can come apart to a greater degree than James recognized. This will be the case
particularly when the true sentence is embedded in a false sign system. The sentence
may be more true than false, in virtue of entailing the right observation sentences, and
28. This may be a point of disagreement with Wittgenstein (see Wittgenstein 1966, 55-59).
323
yet be deeply problematic pragmatically due to bad emotional and practical associations.
Finally, Carnap is right about our freedom to construct alternative sign systems and the
need to take account of pragmatic considerations in deciding between them. However,
Berkeley, unlike Carnap, sees pragmatic considerations at work even in so-called ‘in-
ternal questions,’ and he sees pragmatic considerations as contributing to the truth of
answers to both internal and external questions.
We have now seen how Berkeley answers Browne. Faith does not ‘terminate in
the ear,’ but rather in the will. Assent, including religious faith, is a matter of having
certain dispositions to speech, thought, feeling, and action. It is this that distinguishes
sincere from insincere professions. Furthermore, these dispositions serve to connect our
thought and action to the world, and they can do so well or badly. To the degree that they
connect us well, the beliefs are true. This does lead to a certain degree of pragmatism and
relativism, but it stops short of radical theories such as those of James and (the traditional
interpretation of) Carnap by insisting on an objective notion of correspondence. The
element of Berkeley’s view which is truly radical is his extreme holism which holds
that individual sentences can be evaluated for truth only in the context of the total form
of life in which they are embedded. The question which shall occupy us in the final
two chapters of this dissertation is how Berkeley can defend the claim that our ordinary
ways of speaking – our actual form of life – does correspond in the way it must for our
assertions to exhibit a high degree of truth. Berkeley’s answer to this question is to be
found in his theory of a divine language of nature.
324
Chapter 8
The Linguistic Structure of Berkeley’s
World
“Qualities,” Berkeley says, “are nothing else but sensations or ideas, which exist only in
a mind perceiving them” (PHK,x78). Bodies, furthermore, are nothing over and above
these sensible qualities. Nevertheless, Berkeley insists, it is not the case that “by [his]
principles, all that is real and substantial is banished out of the world, and instead thereof
a chimerical scheme of ideas takes place.”
1
Rather, “[t]here is a rerum natura” (x34).
This ‘rerum natura’ is meant to be an independent, objective reality to which our system
of body talk ought to conform. The reality in question consists in God’s systematic and
orderly activity in causing our sensory ideas (PHK,xx29-31, 33, 36; Siris,x293). It is the
understanding of this system of divine activity, especially insofar as such understanding
leads to successful prediction, that Berkeley calls ‘knowledge of nature’ (PHK,x59;
DHP, 245; Siris,xx252-253). This knowledge is embodied to some degree in ordinary
body talk, but in a much higher degree in Newtonian physics (PHK,xx105, 110). Thus
we can say that, despite some metaphysical confusions, plain language body talk is true
to a high degree, but in many contexts the body talk contained in Newtonian physics is
true to a higher degree (see above,x7.2.4).
The central question immediately raised by this picture is: what is God up to? (See
PHK,x64.) Why, that is, is God engaged in this systematic and orderly causing of my
1. The wording of this objection makes it clear that Berkeley has Edward Stillingfleet’s criticism of
Locke in mind (Stillingfleet 1697, 234). For discussion, see Pearce, forthcoming(a).
325
ideas? When Berkeley faces up to this problem, he typically begins by making a simple
observation: if the ideas God caused were not systematic and orderly, I would be unable
to get around in the world (PHK,x62). The system of rules is “necessary to assist, not
the Governor, but the governed” (Siris,x160). This answer is, however, not particularly
satisfactory. We may grant that, assuming God is going to give me a series of ideas, it
would be inconsistent with divine benevolence that those ideas should be “any how and
at random produced” (PHK,x64), but this does not answer the question of why God is
giving me ideas in the first place. It is simply not clear what Berkeley’s ‘world’ is or
why it is here.
To answer this objection, Berkeley makes a surprising and transformative use of a
very old metaphor, the ‘book of nature:’ “the reason why ideas are formed into ma-
chines, that is artificial and regular combinations, is the same with that for combining
letters into words” (x65). The structure of Berkeley’s world is a linguistic structure.
Berkeley himself traces the ‘book of nature’ metaphor back to Plotinus (Siris,x252).
Closer to Berkeley’s time, the metaphor had appeared in the (Dutch Reformed) Belgic
Confession (article 2), and, famously, in Galileo.
There was, then, nothing surprising, in Berkeley’s day, about speaking of the world
as a ‘book’ or ‘discourse’ or talking about the ‘characters’ in which the book was writ-
ten. However, Berkeley’s specific understanding of the language of nature is original,
even radical, in two respects. First, Berkeley’s theory represents a bold synthesis of the
old Aristotelian use of this metaphor with the new Galilean use. For the Aristotelians,
the ‘characters’ of the book of nature were sensible qualities, while for Galileo the book
of nature “is written in mathematical language, and its characters are triangles, circles,
and other geometrical figures” (Galilei [1623] 2008, 183; see Redondi 1987, 52). In
Berkeley’s philosophy, the language of nature is a mathematical language insofar as
326
its grammar is provided by Newtonian mathematical physics; nevertheless, the charac-
ters of that language are sensible qualities, including the so-called ‘secondary qualities’
which Galileo shunned.
Berkeley’s second transformation of this traditional talk of a ‘book of nature’ is more
radical still. In Berkeley’s hands, the metaphor of a language in which “the Author
of Nature constantly speaks to the eyes of all mankind” (Alc,x4.11; cf. NTV,x152)
ceases to be a metaphor. In Berkeley’s view, it is literally the case that “the phenomena
of nature . . . form . . . a most coherent, entertaining, and instructive Discourse” (Siris,
x254). This is Berkeley’s answer to our earlier question about the nature and purpose of
the system of ideas God impresses upon us: that system is a discourse, and its structure
is, literally, a linguistic structure.
The aim of this chapter will be to show that it is by attributing this linguistic struc-
ture to the world that Berkeley aims to preserve, within his immaterialist metaphysics,
the truth of claims about structure in both plain language and Newtonian physics. We
shall be particularly concerned with two types of structure: the co-instantiation of mul-
tiple qualities in a single, enduring object, as represented by our practice of predication
(x8.3), and the relation of physical causation (x8.4). Additionally, we shall be concerned
with the question of the content, the semantics, of this ‘discourse,’ and I will show that,
contrary to a number of commentators, it is not the case that the ideas in Berkeley’s lan-
guage of nature signify only other ideas. Rather, it is by the interpretation of the divine
discourse that we learn about other minds (x8.6). First, however, it will be necessary to
establish the fundamental premise of this investigation, namely, that Berkeley’s divine
language hypothesis is intended literally. It is to this task that we now proceed.
327
8.1 A Literal Language of Nature
My aim in this section is to understand Berkeley’s talk of a ‘language of nature.’ I begin
by arguing that Berkeley held that vision is literally a language – that is, that it satisfies
the same definition of ‘language’ which is satisfied, e.g., by English – then proceed to
examine Berkeley’s understanding of the other sense modalities. I will not be concerned
with the much-discussed questions of the extent to which Berkeley’s theory of vision
can be severed from his immaterialism, whether there are inconsistencies between the
New Theory and the Principles, or whether the theory of vision provides (or is taken by
Berkeley to provide) part of a positive argument for immaterialism.
2
Rather, as I have
indicated above, my concern in this chapter is to show that Berkeley uses the theory
of a language of nature to answer a certain family of objections to immaterialism. The
task of this section is, therefore, to answer the question, what is the theory to which
Berkeley appeals in answering these objections? I will be concerned with Berkeley’s
direct arguments for this theory only insofar as these arguments shed light on the na-
ture of the theory.
3
The theory to which Berkeley appeals, I argue, is one which states
that our visual ideas, and possibly also our tactile ideas, literally form a “a most coher-
ent, entertaining, and instructive Discourse” (Siris,x254) ‘spoken’ by God,
4
while the
other sense modalities provide what we may call ‘adjunct signs,’ contributing to what is
communicated by the language without entering into its syntax.
2. On these questions see, e.g., Armstrong 1960; Atherton 1990, 9-15, 215-229; Yaffe 2003; Atherton
2005.
3. Of course, if Berkeley has compelling arguments for immaterialism, and the theory of a language
of nature answers objections to immaterialism better than any rival proposal, this will provide an indirect
argument for a language of nature.
4. As the scare-quotes around ‘spoken’ indicate, the discourse is obviously not literally spoken, insofar
as it is not produced by the use of a vocal apparatus to create sound waves. More on this below.
328
Even the best scholars of Berkeley’s theory of vision often fall into the trap of speak-
ing of a ‘language metaphor,’ ‘language analogy,’ or ‘language model’ (see, e.g., Ather-
ton 1990, ch. 11, et passim). There is a very good reason for this: in the New Theory
of Vision, Berkeley proceeds by making a series of analogies or comparisons between
vision and language before finally stating his conclusion “that the proper objects of vi-
sion constitute an universal language of the Author of nature” (NTV,x147 [1732 ed.]).
In each of these previous passages, Berkeley says that vision is like a language in one
respect or another (xx17, 32, 51, 66, 73, 140, 143). It is, therefore, natural to read the as-
sertion inx147 metaphorically, and take Berkeley’s assertion that vision is (constitutes)
a language as a rhetorical flourish emphasizing how far the similarities go.
This interpretation cannot, however, make sense of Berkeley’s later writings. In the
Theory of Vision Vindicated, Berkeley explains that, in the earlier New Theory of Vision,
he had “observed a certain known method, wherein, from false and popular suppositions,
men do often arrive at truth” (TVV,x38). This was known as the ‘analytic’ method, for
it involved beginning with complex preexisting beliefs about the phenomena and ‘tak-
ing them apart’ (the etymological meaning of ‘analysis’) to discover underlying true
principles. Thus in the New Theory of Vision, Berkeley examines a number of well-
known optical phenomena and attempts to show, by this examination, that vision must
be a language. In the Theory of Vision Vindicated, however, Berkeley explicitly adopted
what was known as the ‘synthetic’ method, beginning from the first principles of his
theory and ‘putting them together’ (the etymological meaning of ‘synthesis’) to “de-
duc[e] theorems and solutions of phenomena, and explain[] the nature of visible things
and the visive faculty” (x38). Berkeley had likewise, though less explicitly, observed
329
this ‘synthetic’ method in Alciphron IV . In both texts, Berkeley begins with a defini-
tion of ‘language’ and proceeds to argue that each element of the definition is (literally)
satisfied by vision (Olscamp 1970b, 36-37; Creery 1972; Stoneham 2013, 219-220).
Alciphron defines ‘language’ as:
the arbitrary use of sensible signs, which have no similitude or necessary
connexion with the things signified; so as by the apposite management of
them, to suggest and exhibit to [the] mind an endless variety of things,
differing in nature, time, and place: thereby informing [one], entertaining
[one], and directing [one] how to act, not only with regard to things near
and present, but also with regard to things distant and future. No matter,
whether these signs are pronounced or written, whether they enter by the
eye or the ear (Alc,x4.7).
In a later speech, this is explicitly called a ‘definition’ (x4.12). In a passage added to the
1752 edition, Euphranor adds an important clarification:
all signs are not language . . . It is the articulation, combination, variety, co-
piousness, extensive and general use and easy application of signs (all which
are commonly found in vision) that constitute the true nature of language
(x4.12 [1752 ed.]).
In the Theory of Vision Vindicated, a briefer definition is given: “A great number of
arbitrary signs, various and apposite, do constitute a language” (TVV,x40).
A ‘sign’ in Berkeley’s usage, is an idea which is used according to a rule. However,
in order to constitute a language, there must be a great many signs, organized into a
system, and these signs must be (1) arbitrary, (2) various, and (3) apposite. As is well-
known, Berkeley’s central focus in his writings on vision is to show that visual ideas are
330
arbitrary signs of tangible ideas. What he means by this is that the connection between
visible and tangible ideas is “neither from necessary deduction to the judgment, nor
from similitude to the fancy, but purely and solely from experience, custom, and habit”
(Alc,x4.10).
5
The key here is that there is nothing in the intrinsic nature of the ideas
which ‘qualifies’ one to suggest the other (TVV,x68). Instead, the rules connecting
visible ideas with tangible ideas are arbitrarily instituted by God.
The ‘variety’ which is required is indicated by the definition from Alciphron: the
signs making up a language must “suggest and exhibit to [the] mind an endless variety
of things, differing in nature, time and place: thereby informing [one], entertaining
[one], and directing [one] how to act, not only with regard to things near and present,
but also with regard to things distant and future” (Alc,x4.7). Here it is important that
the definition is put in Alciphron’s mouth, and Euphranor has not yet expounded his
general theory of language. Alciphron is assuming that language accomplishes all of its
ends by “suggest[ing] and exhibit[ing] to the mind” various things, an assumption which
Euphranor will challenge in the seventh dialogue. Removing this false assumption, we
may say that, on Berkeley’s view, a genuine language must be able to talk about a wide
variety of subjects and serve a wide variety of ends. A language must not, in other
words, be a special purpose sign system like the notation used to record the moves in
chess games.
Finally, the use of signs in a language must be ‘apposite.’ The interpretation of this
requirement is somewhat difficult. In Alciphron’s definition, it is by “apposite manage-
ment” that the signs which constitute a language are able to achieve the wide variety
of ends at which they aim. The claim here seems to be that which signs are ‘apposite’
(suitable) varies widely from one circumstance to another and it is the ability to make
5. On the nature of this threefold classification, see above, pp. 166-170.
331
these judgments that Alciphron takes to be a sign of intelligent agency. If this is Alci-
phron’s point, then he is likely following Descartes, who claimed that, although it would
be possible to make a talking machine, yet “it is not conceivable that such a machine
should produce different arrangements of words so as to give an appropriately mean-
ingful answer to whatever is said in its presence, as the dullest of men can do” (CSM,
1:140).
6
However, there is more than this involved in the requirement that the signs be ‘appo-
site.’ In the passage added in 1752, Euphranor refers to “the articulation, combination,
variety, copiousness, extensive and general use and easy application of signs.” The talk
about ‘copiousness [and] extensive and general use’ is probably intended as an elucida-
tion of the ‘variety’ requirement, but the rest of this passage may be taken as an explana-
tion of the requirement that the signs be ‘apposite.’ The passage is explicitly contrasting
language, properly so-called, with other “significant sounds, such as the natural cries of
animals, or the inarticulate sounds and interjections of men” (Alc,x4.12 [1752 ed.]). A
dog’s bark or a human’s scream may indicate that he is hurt or in danger, yet these are
not language, and part of the reason they are not language is that they are not ‘articu-
late.’ This is, I take it, a matter of phonology (or, in written language, orthography). The
‘phonology’ of the language of nature will be the topic of the next section. The general
idea, however, is that there are a number of basic signs which are distinguished from
one another. These play the role of, for instance, the alphabet of written English.
The next requirement is ‘combination.’ These basic signs are put together into com-
plex signs in a systematic way, and different combinations differ in meaning.
The final requirement, “extensive and general use and easy application,” could be
taken either as part of being ‘apposite’ or as part of being ‘various.’ In any event, the
6. On the possible influence of this discussion on Alciphron IV , see Kline 1987.
332
point is that the signs are organized into a system in such a way that it is easy to con-
struct the complex signs appropriate to any given situation. Alciphron refers to “an
endless variety of things;” the basic signs that make up a language can be recombined
in indefinitely many ways, to describe unforeseen and unforeseeable situations. Thus,
for instance, bizarre sentences like ‘the space aliens are infested with yetis’ are no more
difficult to construct than commonplace sentences like ‘the tomatoes are infested with
aphids.’
The requirement that the signs be ‘apposite’ is closely related to the requirement
that they be ‘various.’ The latter says that there must be many different signs with a
wide variety of different meanings; the former says that the signs must be ‘managed’
in such a way as to say the appropriate things given the relevant circumstances and
purposes. This entails a very strong variety requirement, for it must be possible to have
something ‘apposite’ to say in situations unforeseen when the language was developing.
This flexibility is to be obtained by ‘articulation’ and ‘combination’ – in other words,
by the compositionality of language. The basic signs must be put together to compose
complex signs so as to be capable of dealing with indefinitely many situations.
In Alciphron and the Theory of Vision Vindicated, Berkeley argues specifically that
vision satisfies each of these conditions, and hence is literally a language. In each case,
he takes himself to be summarizing his earlier argument from the New Theory of Vision,
and his contention is supported by specific texts from the earlier work. Thus it is perhaps
not incorrect to say, with Paul Olscamp, that “[t]here is a development in [Berkeley’s]
writings from a position where one may reasonably interpret his view of [a language of]
nature as a metaphor, to a position where this is no longer reasonable” (Olscamp 1970a,
333
39).
7
However, there is no good reason to suppose that this is a substantive change of
view rather than merely (as Berkeley himself claims) a change in the mode of exposition.
The bulk of Berkeley’s work on vision is devoted to the defense of the claim that
vision consists of arbitrary signs. This has been explored at great length in the liter-
ature (see especially Atherton 1990), and there is no need to revisit it here. Berkeley
spends fewer words on the defense of the other two requirements, and there has been
correspondingly less attention to them in the literature. However, that Berkeley takes
vision to have the last two features is abundantly clear. For instance, in Alciphron the
visual language is said to involve “[t]he instantaneous production and reproduction of
so many signs . . . adapted to such an endless variety of purposes, ever shifting with the
occasions, and suited to them” (Alc,x4.14). In the New Theory of Vision, Berkeley had
also emphasized the variety of purposes served by the visual language, claiming that,
by it, “we are instructed how to regulate our actions in order to attain those things that
are necessary to the preservation and well-being of our bodies, as also to avoid what-
ever may be hurtful and destructive of them” (NTV,x147). Alciphron, however, goes
farther in identifying diverse purposes served by the visual language: “while it informs,
it amuses and entertains the mind with such singular pleasure and delight” (Alc,x4.15).
Further, as Euphranor explicitly claims, the visual language “answers so apposite
to the uses and necessities of mankind, informing us more distinctly of those objects,
whose nearness and magnitude qualify them to be of greatest detriment or benefit to
our bodies” (x4.15, emphasis added). Berkeley further emphasizes that the signs are
“combined, dissolved, transposed, [and] diversified” (x4.14) – in other words, that they
exhibit the kind of complex contextual variation discussed above. In the New Theory,
Berkeley gives a specific example of this:
7. Cf. Olscamp 1970b, 36: “By the middle of the Principles it is evident that Berkeley is no longer
considering the language metaphor as a metaphor.”
334
Faintness, as well as all other ideas or perceptions which suggest mag-
nitude or distance, does it in the same way that words suggest the notions to
which they are annexed. Now, it is known a word pronounced with certain
circumstances, or in a certain context, with other words, has not always the
same import and signification that it has when pronounced in some other
circumstances or different context of words.
The very same visible appearance as to faintness and all other respects,
if placed on high, shall not suggest the same magnitude that it would if it
were seen at an equal distance on a level with the eye (NTV,x73 [2nd 1709
ed.]).
The context of this discussion is the moon illusion. Berkeley’s claim is that the very
same visual stimulus, when it occurs while looking to the horizon, will be interpreted
differently than when it occurs while looking directly overhead. The complexity of the
contextual variation is supposed to explain why we misinterpret the ‘horizontal moon.’
Whether the particular example is correct need not concern us; the point is that Berkeley
takes visual signs to vary in their meaning with their context.
A second example appears in all three of Berkeley’s expositions of his theory of
vision:
what [a person] sees only suggests to his understanding that, after having
passed a certain distance, to be measured by the motion of his body, which
is perceivable by touch, he shall come to perceive such and such tangible
ideas which have been usually connected with such and such visible ideas.
But that one might be deceived by these suggestions of sense, and that there
is no necessary connexion between visible and tangible ideas suggested by
335
them, we need go no farther than the next looking-glass or picture to be
convinced (NTV,x45; cf. Alc,x4.12; TVV,x25).
Especially given the context of the Alciphron version of this claim, it might be thought to
be an answer to the objection that vision is unlike a language insofar as it is impossible
for vision to say something false.
8
However, this objection is not even mentioned in
either NTV or TVV, and even in Alciphron the claim that the visual language cannot
represent falsely is actually only a premise in an argument for the conclusion that “the
connexion [between vision and touch is] necessary” (Alc,x4.12).
9
This is the view
Berkeley is arguing against. His argument is that it cannot possibly be the case that,
for instance, there is something about the intrinsic nature of a certain kind of visual
experience in virtue of which it is necessarily connected with an object’s being (e.g.) five
feet in front of me because in the context of a mirror the very same visual experience
might represent the object as being behind me. We know that it is by these contextual
factors that we interpret our visual experience because if I fail to recognize that I am
looking into a mirror I will systematically misinterpret the data I am receiving. That
the signs involved in the visual language vary with their context in these complex, yet
systematic, ways is part of what it means for the signs to be ‘apposite.’
It is, then, Berkeley’s view that vision is literally made up of a “great number of
arbitrary signs, various and apposite” (TVV,x40) and that this is sufficient for vision
literally to be a language. Of course Berkeley is speaking metaphorically when he says
that “the voice of the Author of nature . . . speaks to our eyes” (NTV,x152; cf. Alc,
x4.11), for God does not literally ‘speak’ – that is, he does not agitate air molecules
8. Alciphron has just objected that human languages “do not always suggest real matters of fact.
Whereas this natural Language, as you call it, or these visible signs . . . have the same constant regular
connexion with matters of fact” (Alc,x4.12).
9. The objection that the visual language cannot represent falsely will be discussed below (x8.6).
336
by means of a physical vocal apparatus so as to convey signals to us – and the eyes
do not literally ‘hear’ – that is, they do not receive auditory ideas. It is for this reason
that, when Alciphron first says that he knows the existence of human persons by their
speaking to him, Berkeley has Euphranor ask, “Is then the impression of sound so much
more evident than that of the other senses?” and Alciphron respond, “Alas! you mistake
the point. What I mean is not the sound of speech merely as such, but the arbitrary use of
sensible signs . . . No matter whether these signs are pronounced or written, whether they
enter by the eye or the ear” (Alc,x4.7). The fact that the divine language is not literally
‘spoken’ does not prevent it from being, literally, a language. Written languages and
sign languages are languages in the same sense as spoken languages.
Berkeley’s claim, then, is that vision is literally a language. Below, I will argue that
Berkeley takes the language of nature to have a structure very similar to the structure
of human languages: it has a ‘phonology’ or ‘orthography’ (x8.2), a lexicon (x8.3), a
syntax (x8.4), and a semantics (x8.6). Furthermore, in chapter 9, I will argue that certain
difficult problems in Berkeley’s metaphysics and epistemology can be solved by taking
the language of nature, like other languages, to be a matter of public social convention,
and will show that this can be held consistently with recognizing the privileged status
of God as ‘speaker’ of the great ‘discourse’ of nature (Siris,x254). Before proceeding,
however, we should pause to examine the status of the other sense modalities in the
language of nature.
Famously, in the New Theory of Vision, Berkeley’s view appears to be that visual
stimuli signify an extra-linguistic, and extra-mental, tangible reality. This, however, is
dismissed in the Principles as a ‘vulgar error’ (PHK,x44). One therefore wonders what
has become of the other sense modalities.
337
Berkeley addresses this issue explicitly only once, in the 1752 edition of Alciphron.
In that text, Alciphron objects that “every other sense may as well be deemed a language
as that of vision. Smells and tastes, for instance, are signs that inform us of other qual-
ities to which they have neither likeness nor necessary connexion.” It is not clear why
this is supposed to be problematic. In any event, Euphranor responds, “That they are
signs is certain . . . But it is as certain that all signs are not language . . . Other senses may
indeed furnish signs; and yet those signs have no more right than inarticulate sounds to
be thought a language” (Alc,x4.12 [1752 ed.]).
Berkeley’s position, then, is that smells and tastes are what we may call ‘adjunct
signs.’ They are significant, but they lack “the articulation, combination, variety, co-
piousness, extensive and general use and easy application” which are characteristic of
language (x4.12 [1752 ed.]). These signs form part of the total apparatus by which God
‘speaks’ to us, but they do not fit into the syntax of the visual language. For this reason,
I have elsewhere compared them to stoplights (Pearce 2008, 251). Stoplights are no part
of the English language, and the reason they do not form a part of it is that they do not fit
into the syntax of the language (a green light cannot be one of the words of a sentence).
Nevertheless, they are part of the total apparatus by which English speakers customarily
communicate. Smell and taste are, in fact, far richer than stoplights, so that we may
perhaps also compare them to ‘body language:’ one’s facial expression, or stance, or
tone of voice, can change the meanings of one’s words, and at least some of the features
that have this effect have it by convention only, yet none of these things is actually part
of the English language.
The fact that the discussion in Alciphron mentions only ‘smells and tastes’ is signif-
icant. Berkeley nowhere discusses the role of hearing in the language of nature, and his
remarks about touch are prima facie inconsistent. In the New Theory of Vision, Berkeley
338
wrote that “visible figures represent tangible figures much after the same manner that
written words do sounds” (NTV,x143). Colin Turbayne has defended an extremely lit-
eral reading of this remark (Turbayne 1970b). However, this is apparently contradicted
by a remark in Alciphron: “light and colours, with their several shades and degrees
. . . suggest and exhibit to us the distances, figures, situations, dimensions, and various
qualities of tangible objects . . . just as words suggest the things signified by them” (Alc,
x4.10; cf. NTV,x77). The relation between a written word and its spoken equivalent
is not at all the same as the relationship between a written word and its referent in the
world. We will return to this issue inx8.6.1, below, where I will show that following
Turbayne in taking visual ideas to signify tangible stimuli in the way written words sig-
nify spoken words, rather than in the way words signify their referents, will help to solve
a number of difficult problems about the meaningfulness of the language of nature.
10
Conventionally, we speak of one language, English, being both spoken and written,
despite the radical differences between spoken and written English (cf. Turbayne 1970b,
7-12). Furthermore, we conventionally say that two people are ‘speaking English’ even
if a good deal of their communication is non-verbal. It is for this reason that, beginning
in the Principles, and especially in Siris, Berkeley is able to move beyond his theory of
vision as language and speak of a total language of nature which “forms a sort of rational
discourse” (Siris,x254; cf. PHK,xx65-66).
11
This ‘discourse’ consists in a system of
signs – chiefly, but not exclusively, visible and tangible ideas – which are ‘various and
apposite,’ exhibiting the same kind of flexibility and structure which is exhibited by the
signs in human languages. We now proceed to a thorough examination of this structure.
10. Here I depart from my own previous treatment of this issue (Pearce 2008, 250-251), where I classed
tactile ideas among what I am now calling the ‘adjunct signs,’ along with smell and taste.
11. For discussion see Turbayne 1970b; Olscamp 1970b, 19-20, 30-31; Pearce 2008, 251.
339
8.2 Phonology: Sensible Qualities
Written languages have an orthography, and spoken languages a phonology. Sign lan-
guages likewise have certain basic units, analogous to letters, of which they are com-
posed. In this section I speak broadly of ‘phonology’ as covering all of these categories,
as well as the corresponding items in the language of nature.
In describing the phonology of a language, linguists classify individual sounds into
phonemes. Roughly speaking, the phoneme is the class of individual symbols between
which no distinction is drawn within the language. Thus, for instance, in written En-
glish no distinction is drawn between ‘a’ and ‘a,’ while ‘a’ and ‘d’ are distinguished.
One reason it is necessary to allow this kind of variation is that there are limits on the ac-
curacy with which we are able to produce symbols matching the paradigms from which
we learned. Thus some degree of so-called ‘free variation’ must be permitted (Gleason
1955, 163; Quine 1960, 88-90). Additionally, in actual languages there are systematic
variations in the pronunciation of a given phoneme based on its context (Gleason 1955,
164-170).
These details were not known in Berkeley’s day, and thus he was unable to provide
a detailed treatment of the phonology of his language of nature. However, it was well-
known, and Berkeley is clearly aware (see, e.g., PHK,x65), that a language must start
from an ‘alphabet’ of basic symbols (see Arnauld and Lancelot [1660] 1975, part 1).
Furthermore, Berkeley would certainly have been familiar with at least two instances in
which the ‘same’ letter is written differently in different contexts: the characters ‘s’ and
‘s’ in 18th century English, and ‘sv’ and ‘c’ in the miniscule script for ancient Greek. So
Berkeley was aware that pronunciation or orthography for the ‘same’ letter could vary
with context.
340
If we are to assign a phonology to the language of nature, we must first identify the
basic symbols out of which the language of nature is constructed, and then show how
they are grouped into phonemes. The first step is not difficult, and Berkeley discusses it
fairly explicitly. The second step, however, poses grave difficulties which Berkeley does
not seem to recognize, yet it seems that some such grouping is needed for his theory
to work. These difficulties stretch Berkeley’s comparison of the language of nature to
human languages but, I will argue, not to the breaking point. Berkeley’s language of
nature must clearly be a language of a quite unusual sort, but this does not prevent it
from satisfying Berkeley’s definition of ‘language.’
There are two texts in which Berkeley explicitly discusses the phonology of the
language of nature:
[T]he reason why ideas are formed into machines, that is, artificial and reg-
ular combinations, is the same with that for combining letters into words.
That a few original ideas may be made to signify a great number of effects
and actions, it is necessary they be variously combined together: and to the
end their use be permanent and universal, these combinations must be made
by rule, and with wise contrivance. By this means abundance of informa-
tion is conveyed unto us, concerning what to expect from such and such
actions, and what methods are proper to be taken for the exciting of such
and such ideas; which in effect is all that I conceive to be distinctly meant
when it is said that, by discerning the figure, texture, and mechanism of the
inward parts of bodies, whether natural or artificial, we may attain to know
the several uses and properties depending thereon or the nature of the thing
(PHK,x65).
341
[L]ights, shades, and colours, variously combined, answer to the several
articulations of sound in language (Alc,x4.10).
The first quotation is about nature in general; the latter is speaking solely of vision. Quite
generally, then, the ‘characters’ of the language of nature will be individual sensible
qualities.
It may be tempting to identify these characters with Lockean simple ideas. However,
there are strong reasons for supposing that Berkeley rejects the existence of simple ideas
(Winkler 1989, ch. 3). Most importantly, Berkeley denies the possibility of abstracting
visible extension from color or color from visible extension (PHK,x10). We have seen,
though, that Berkeley is nevertheless able to group ideas according to different respects
of resemblance between them. Thus he can meaningfully talk of the color and the visible
figure of ideas as distinct aspects of them, because he can recognize that, for instance,
two ideas may resemble in color but not in figure. The visual language consists, es-
sentially, of visual colored regions which sometimes occur together in a single visual
experience, and other times succeed one another temporally. A colored region can only
be experienced (or imagined) in a total visual field. Nevertheless, one can recognize the
same color and visual figure recurring in another context.
Similar remarks will apply to touch. One has, at any moment, a total ‘tangible
experience.’ This experience may have identifiable components which can recur in other
contexts, but nevertheless cannot occur alone. Thus one cannot feel a shape without
feeling it as having some degree of roughness or smoothness and feeling it as having
some temperature. As in vision, so in touch there is variation both in co-occurring
components and in succeeding components.
I suspect that Berkeley thought that what set vision and touch apart from the other
senses was the complexity of the co-occurring qualities in vision and touch. That is, in
342
vision and touch one has, in an instant, a complex experience with distinct identifiable
components. Berkeley may have thought that this was not the case with the other senses.
If he thought this, then it is not clear whether he was correct. There is a certain
amount of intuitive pull to the idea that, for instance, my gustatory experience at any
given moment is ‘one simple uniform sensation;’
12
on the other hand, the experienced
wine taster may be able to identify several distinct components of the experience of
a particular wine. Likewise, the expert musician may be able separately to identify
each instrument in the orchestra, and even the less experienced listener may be able to
distinguish a few.
In any event, these basic components or aspects, sensible qualities, are the ‘char-
acters’ of Berkeley’s language of nature. If the language of nature is similar to human
languages, then we ought to be able to group these characters into phonemes. Here, how-
ever, we observe a curious phenomenon. Like speakers who are familiar with a writing
system, we native ‘speakers’ of the language of nature have names for what we seem to
take as its phonemes, yet there is radical variation among different communities about
what these alleged phonemes are. I am referring, of course, to the notoriously wide
cross-linguistic variation in categorization of colors and other similar qualities (Quine
1960, 41).
This in itself is not surprising. As the Port-Royalists knew, the letters of written
languages typically do not have a precise correspondence with the sounds of spoken
languages (Arnauld and Lancelot [1660] 1975,x1.5). This makes it plausible that two
12. I borrow this phrase from see DHP, 176, but Berkeley is there talking about the impossibility of
distinguishing “the most vehement and intense degree of heat” from “a very great pain.”
343
groups who shared a spoken language might have different written languages, with dif-
ferent phonetic alphabets, where there was not a direct correspondence between the
letters of one alphabet and those of the other.
13
Nevertheless, it might be thought, it must be the case that, whatever labels we have in
English as opposed to Greek or Mandarin, if nature is a language and we all understand
that language, we must be, implicitly, grouping qualities in the same way. Yet it does
not seem that we do this. We seem, rather, all to group qualities in the way prescribed
by our native (human) language.
It is, however, not at all clear that, in order to understand the language, we must
group the phonemes in the same way. The linguist’s phonemes are not to be identified
with any of the particular psychological, behavioral, or acoustic phenomena which con-
stitute the language; they are, rather, elements of the linguist’s model of the language
(Gleason 1955, 170-171). In other words, the notion of a phoneme comes in, not in the
course of trying to speak a language, but in the course of trying to gain explicit, artic-
ulable understanding of how that language works. Thus the grouping into phonemes is
to be justified by the role that it plays in a systematic exposition of the working of the
language.
Here, though, is where the real trouble arises. As it turns out, the ‘grammarians’ of
nature – that is, physicists (PHK,x108; see below,x8.4) – have not found it useful to
divide colors into discrete categories. In discussing physical optics, they instead discuss
the wavelength of the light, which is treated as a continuous variable. Similarly, in phys-
iological optics, we can talk about the degree of stimulation the light produces in each
13. James Van Cleve suggested to me that Hindi and Urdu may be a real-world case like this. The spoken
languages are mutually intelligible, but in written form Urdu has 39 letters (all representing consonants)
and 10 vowels marks (BBC 2014b), while Hindi has 35 consonants and 11 vowels (BBC 2014a).
344
kind of receptor in the eye, giving rise to (typically) three continuous variables.
14
Simi-
larly, temperature is continuous, and while Berkeley denies that distance and magnitude
(whether tangible or visible) are truly continuous, these are clearly treated as continu-
ous in physics. Thus whereas, in the course of giving a systematic treatment of a human
language, we lump the symbols together into discrete ‘bins,’ in the course of giving a
systematic treatment of the language of nature, we treat the symbols as continuous. The
language of nature does not have a ‘phonology’ of the sort human languages have.
There is, however, a ready answer to be given on Berkeley’s behalf. As Quine once
noted (Quine 1960, 87-88), we can imagine the use of a continuous variable as a medium
of communication, and this particularly makes sense when the thing we want to repre-
sent is itself continuous. Thus Quine envisions a complex mapping of sounds onto
colors. The key reason this sort of device is not used in human communication is that
we are unable either to produce or to perceive arbitrarily fine gradations in, e.g., sound.
This, however, is no limitation for God. Thus the fact that the divine language exhibits
continuous phonology (as we may call it) is not a bug, it’s a feature. The divine language
is infinitely richer and more sophisticated than human languages. This is just the result
Berkeley would want.
One who understands the language of nature must be able to respond to such con-
tinuous variation. This requires that the components of our experience be arranged in
‘quality spaces’ (see Quine 1960, 83-85; Goodman 1977, chs. 9 and 10). This is, indeed,
a feature of our experience: for instance, some colors seem to us to be ‘closer together’
than others, and this apparent closeness is, as it turns out, connected with various fea-
tures of the grammar of nature, namely, the science of optics.
14. On the sort of variation that occurs between human individuals at the interface between physical
optics and physiological optics, and the difficulties this causes for theories of color, see Block 1999.
345
The continuous phonology exhibited by the language of nature is the underlying re-
ality which our talk of qualities, both in plain language and in science, aims to capture.
Each does indeed capture it. The more or less arbitrary division of color space into dis-
tinct categories exhibited by human languages is a useful tool for getting around in the
world in ordinary circumstances, and thus a statement using ‘red’ will in many circum-
stances be ‘truer’ than a statement about the precise wavelength of light being reflected.
Yet this does not mean that the categories of one language are ‘truer’ than those of an-
other. Furthermore, it is not the case that plain language is universally to be preferred
over physics. Berkeley, to be sure, will not allow physics to displace plain language
and render our secondary quality attributions false, as Galileo and the Cartesians do (see
below,x8.5), but this does not prevent him from recognizing that the explanation given
in physics is more precise and hence will be ‘truer’ where such precision is appropriate.
Here is another place where “one and the same thing can be explained in different ways”
(DM,x67).
8.3 Lexicography: Co-Instantiation
The basic symbols of a language are put together into words. This also occurs in the
language of nature: “the reason ideas are formed into machines, that is, artificial and reg-
ular combinations, is the same with that for combining letters into words” (PHK,x65).
This passage occurs in Berkeley’s answer to the objection that, on his view, no purpose
can be given for “the curious organizations of plants, and the admirable mechanisms of
the parts of animals” (x60). These ‘mechanisms’ are ‘artificial’ in the sense that they
are the product of artifice, that is, they have evidently (Berkeley thinks) been arranged
for a purpose by a mind. But, it is alleged, if Berkeley’s view is correct, no purpose can
346
be assigned to them.
15
Berkeley’s answer is that the ideas are indeed orderly, and they
do indeed have a purpose, but it is unnecessary to suppose that their organization and
purpose are causal. (Indeed, Berkeley argues, it is incoherent to suppose this; see above,
x6.4.) Rather, the ideas are grammatically structured and serve a communicative pur-
pose. To serve this purpose, ideas must “be variously combined together . . . [and] these
combinations must be made by rule, and with wise contrivance” (PHK,x65). These
combinations of ideas are the words of the language of nature.
This provides an alternative avenue of approach to a question discussed above (x6.3),
namely, how ideas are combined into bodies. In the previous discussion, I emphasized
the role of the human perceiver in combining the ideas. However, there is a well-known
tension in Berkeley’s thought regarding the respective roles of the human perceiver and
the divine cause of sensible ideas in the construction of bodies (Glauser 2007; Hight
2007b; Atherton 2008a). In a number of passages Berkeley says quite explicitly that it
is we humans who combine ideas together into objects. For instance:
According as the mind variously combines its ideas, the unit varies . . . We
call a window one, a chimney one, and yet a house in which there are many
windows and many chimneys has an equal right to be called one (NTV,
x109).
[M]en combine together several ideas, apprehended by divers senses, or by
the same sense at different times or in different circumstances . . . all which
they refer to one name and consider as one thing (DHP, 245)
15. There is some question as to how the ‘combinations’ could appear purposive without appearing to
have some particular purpose. The objection seems to presuppose that this is possible, though Berkeley’s
reply carries no such presupposition, since Berkeley answers the question of what purpose they appear to
have: their purpose is to communicate information in order to guide or actions.
347
[A] cherry, I say, is nothing but a congeries of sensible impressions or ideas
perceived by various senses; which ideas are united into one thing (or have
one name given them) by the mind (DHP, 249)
In these passages, it is clear that, when we talk about familiar objects such as win-
dows, chimneys, houses, and cherries, we combine ideas. Furthermore, Berkeley often
emphasizes that we have a great deal of freedom in deciding which combinations to
construct (see, e.g., NTV,x109; PHK,x12). However, as Richard Glauser points out
(Glauser 2007, 54-55), the very passage from the Principles which we have been dis-
cussing causes rather serious problems for this interpretation, for it is surely God, and
not humans, who has combined ideas into ‘machines’ as described in PHK,xx60-66.
The standard way of solving this problem is to appeal to Berkeley’s frequent remarks
to the effect that we combine ideas which we have “observed . . . to have some connection
in nature” (DHP, 245), that is, that our combining of ideas is done according to the
connections which we call ‘laws of nature’ (PHK,x30; see, e.g., Muehlmann 1992, 219-
223, et passim; Atherton 2008a, 93). This approach can allow for significant freedom in
how we combine ideas, while still holding that there are divinely instituted constraints
on our choice (Stoneham 2002,x8.3.1), thus giving God an important role in ‘forming
ideas into machines.’
Glauser advocates an alternative interpretation on which particular objects consist
of ideas combined together by God, while sorts of objects are made by humans (Glauser
2007). This interpretation sees Berkeley as following Locke’s view “That General
and Universal, belong not to the real existence of Things; but are the Inventions and
Creatures of the Understanding” (EHU,x3.3.11). On Glauser’s reading, Berkeley, like
Locke, would take this claim to mark a contrast between the general and the particular,
348
with the particular existing in rerum natura, while the general exists only in the finite
minds that construct general signs.
Neither of these readings is fully satisfactory as it stands. The standard interpreta-
tion does not adequately account for the kind of unity God is taken to bestow on objects
in PHK,xx60-66, for not all ideas which, according to the laws of nature, ‘go together,’
thereby become constituents of a common object. As a result, if the standard interpreta-
tion is to solve the problem, it must be supplemented by an account of the kind of ‘going
together’ which constitutes being ‘formed into a machine’ and explain the role of this
phenomenon in our combining activity.
Glauser’s position, on the other hand, fails to recognize just how radical Berkeley’s
views on identity are. I have argued above (x6.6) that, in Berkeley’s view, there is no
such thing as unity independent of being ‘considered as one’ by some mind. I have also
argued that, since humans cannot have all of the constituent ideas of a body at once, we
can consider these bodies as one only by establishing conventions for the use of words
(x6.3). Thus, if my earlier conclusions are correct, individuals, like sorts, must be ‘the
workmanship of the understanding.’ If I am right, though, then how are we to account
for those passages which, Glauser rightly points out, seem clearly to imply that God
combines ideas into bodies? What are the respective roles of the divine and the human
in the assembling of ideas into bodies?
Margaret Atherton has suggested that Berkeley’s language of nature theory provides
a way forward. On Atherton’s reading, our grouping of ideas into objects is part of our
process of interpreting the language of nature. “What ties the entire collection [of ideas]
together [into a physical object] are the relations of sign to signification” (Atherton
2008a, 95). As Atherton understands this process of interpretation, it is primarily a
process of deriving “a set of expectations or predictions” (95) from some particular
349
group of (e.g.) cherry perceptions, by understanding the lawful connections between
one idea and another (cf. Siris,xx252-253).
Atherton has, I believe, identified the most important point that must be recognized
if we are to make progress on the problem of divine and human roles in object construc-
tion: our ideas are signs, used according to rules adopted by God, which we call ‘laws of
nature’ (PHK,x30), and our collecting or combining of those ideas is part of the process
of interpretation of those signs. However, as Atherton explicitly recognizes (Atherton
2008a, 99-100), this is not yet a full solution to the problem.
Atherton discusses one reason why the problem is not yet solved. If we are to sup-
pose that God, by ‘speaking’ the language of nature to us, ‘forms ideas into machines’
by means of the laws of nature God has established, then it seems that which ideas ‘go
together’ is determined by God’s intentions. However, if this were so then it would
seem to raise the possibility that we might misinterpret God’s intentions while never-
theless making correct predictions. Yet Berkeley seems to think that as long as our
expectations are being fulfilled, we are succeeding at our interpretive task (97-100).
Beyond this, Atherton has not yet solved the general difficulty for lawfulness ac-
counts which I mentioned above. That is, she has not shown how to distinguish between
those lawful connections which bind ideas together into objects and those which do
not. To put the matter another way: she has not shown how to distinguish between
co-instantiation and physical causation. Both of these relations must involve lawful
connections between ideas and, if Berkeley’s world is to exhibit the kind of structure
which commonsense and Newtonian physics take the world to exhibit, then these two
relations must be distinct.
Atherton’s divine language approach can, however, be extended to solve these diffi-
culties, and the needed extensions are both philosophically and textually well-motivated
350
in terms of Berkeley’s theory of language and his pronouncements about how to under-
stand the structure of the perceived world in linguistic terms.
We can begin to respond to the problem of possible misinterpretation by observing
that a language can be correctly described in more than one way. The rules followed by
speakers of a language are not themselves linguistic entities and may be given more than
one equally correct linguistic expression. Furthermore, as the phrase ‘equally correct’
suggests, and as I have argued inx7.2.4, this sort of ‘correctness,’ for Berkeley, comes
in degrees and is largely constituted by correctness of prediction. To know what rule
is being followed is to be able to ‘see’ what the rule will command in real or imagined
circumstances; a linguistic formulation of the rule is merely an aid to such knowledge
(see ch. 5).
In terms of the metaphor used inx7.2.4, we may say that what Berkeley means in
claiming that we are correct in our ordinary claims about bodies is that our body talk
is painting a highly, but not perfectly, accurate portrait of God’s orderly, rule-governed
productive activity. There are many different ways of painting a highly accurate por-
trait. (Paintings in different styles, different media, from different angles, or in different
lighting may all be equally accurate representations of the subject.) This accounts for
Berkeley’s pronouncements about our freedom in grouping ideas. Nevertheless, some
groupings ‘fit’ better than others, and these degrees of fit are due to the rules God fol-
lows. Hence we can say that these ideas are ‘put together’ or ‘formed into machines’ by
God. We, however, do not perceive all of the grouped ideas at once. Since we perceive
them separately we must put them together as part of our interpretive activity. In this
activity of reassembly, there are many different ways of getting things (mostly) right.
There is an important objection to this approach. Part of what the correctness of
my understanding amounts to is my being able to imagine what ideas I would have in
351
various counterfactual situations. It seems, then, that a Cartesian demon could deceive
me by explicitly formulating some bizarre rule which diverges radically from my ex-
pectations in counterfactual situations and forming the intention to follow this rule in
causing my ideas. Berkeley’s anti-skeptical project requires him to eliminate this appar-
ent possibility. In chapter 9, I will argue that Berkeley’s recognition of the essentially
public nature of language can and should be applied to the language of nature and that,
by this means, this kind of skeptical scenario can indeed be ruled out. For now, however,
we will set this difficulty aside and assume that God follows rules which do not strike
us as bizarre, and hence that our inductively formed beliefs about what counterfactual
circumstances would be like are roughly accurate.
16
The second problem, the problem of distinguishing physical causation from co-
instantiation, can be solved by taking seriously Berkeley’s remarks about the role of
physical objects in the language of nature, already mentioned above. According to
Berkeley, “ideas are formed into machines” for the same reason letters are combined
into words (PHK,x65). By ‘machines,’ Berkeley means complex, highly organized
bodies, but presumably similar remarks apply to simpler bodies, like rocks. Bodies are
the words of the language of nature. Thus, although Atherton is correct that grouping
ideas into objects is part of the process of interpretation, it is important to recognize that
it is only the beginning of this process. This combining activity is part of parsing the
language of nature, which is merely a preliminary to interpretation proper. It is true, as
Atherton says, that this ‘parsing’ is sufficient to allow us to make certain correct predic-
tions, yet Berkeley emphasizes that “fram[ing] general rules from the phenomena, and
afterward deriv[ing] the phenomena from those rules” is not yet interpreting nature, but
merely describing its grammar (x108). To interpret nature is
16. Berkeley explicitly recognizes the need for such an assumption. See PHK,x107; DM,x34.
352
to recreate and exalt the mind, with a prospect of the beauty, order extent,
and variety of natural things; hence by proper inferences to enlarge our
notions of the grandeur, wisdom, and beneficence of the Creator; and lastly,
to make the several parts of the Creation, so far as in us lies, subservient
to the ends they were designs for, God’s glory, and the sustentation and
comfort of our selves and fellow creatures (PHK,x109).
The grouping with which we are presently concerned is part of the grammar of the
language, and not part of its semantics. In particular, the activity of grouping ideas
into objects is one of identifying the word boundaries of the language and, more gen-
erally, the identification and classification of physical objects is the lexicography of the
language of nature.
There are two kinds of grouping involved in lexicography. The first is the group-
ing of sounds or characters in a single utterance or inscription into a single word. The
second is the grouping of words in various utterances and inscriptions, which are not
qualitatively identical, into lexemes. A dictionary aims to capture, in each entry, a sin-
gle lexeme. This single lexeme will typically involve many distinct words which, in
many languages (including all languages familiar to Berkeley) will vary systematically
according to patterns of conjugation or declension. Thus, for instance, ‘run’ and ‘runs’
are different words belonging to the same lexeme.
This can provide a Berkeleian solution to the problem of the bent visual perception
of the straight oar half-submersed in water. Philonous says of this case,
[The perceiver] is not mistaken with respect to the ideas he actually per-
ceives; but in the inferences he makes from his present perceptions. Thus
in the case of the oar, what he immediately perceives by sight is certainly
crooked; and so far he is in the right. But if he thence conclude, that upon
353
taking the oar out of the water he shall perceive the same crookedness; or
that it would affect his touch as crooked things are wont to do: in that he is
mistaken (DHP, 238).
One can compare the person confused about the oar to a beginning Greek student
encountering the word ‘êlusa’ and looking under ‘e’ in a lexicon. The student has not
yet learned the proper conjugation patterns, and, in particular, the rule whereby past
indicative forms of ancient Greek verbs receive receive an initial ‘e’ as an ‘augment,’
so that the relevant dictionary headword will in fact be found under ‘l.’ Similarly, the
perceiver who is confused by the oar has gotten as far as correctly recognizing the visual
perception of the oar as a single word, but has not understood the ‘conjugation’ pattern
for this word by means of which she ought to have predicted that most of the other
appearances of the oar would be straight.
As with phonology, so lexicography is part of a model constructed by linguists to
aid us in representing and understanding the patterns of linguistic phenomena. The lex-
icographer will adopt a convention whereby some particular verb form is used as the
dictionary headword. (In our actual conventions, usually the first person present indica-
tive active.) This form need not, however, be in any way privileged in the psychological
processes of native speakers. For instance, if it were to turn out that children learning
English natively typically first thought of the form ‘runs’ and then dropped the ‘s’ to
form the first person, it would not therefore follow that the lexicographer was making
an error in using ‘run’ as the dictionary headword. What is more important from the
lexicographer’s perspective is that the selection should make things as easy as possible
for those who are learning the language ‘by rule’ (seex5.1), and that it should make the
rules of the language as easy to state as possible. It is to achieve this kind of ease in
stating and understanding the rules that we think of the oar – the lexeme – as straight,
354
despite the fact that in appropriate circumstances it will appear bent, just as we think of
the lexeme ‘luw’ as beginning with ‘l,’ despite the fact that that, under certain circum-
stances, it is spelled with an initial ‘e.’
The grouping of ideas into objects is analogous to the grouping of sounds or letters
into words, and words into lexemes, in human languages. We have so far been implicitly
restricting our attention to visual appearances. What shall we say of the other senses?
As we have discussed, smell and taste, at least, are held by Berkeley to lack the sort
of complex structure exhibited by language proper, so it is to be expected that they will
lack a lexicography. Of course one can distinguish one smell or taste from another, and
could thereby construct a sort of ‘dictionary’ of smells and tastes, but these signs would
not be words, insofar as they would not exhibit morphology, and would not combine
syntactically into larger units. We therefore need not discuss these cases further. We
will also continue to follow Berkeley in neglecting hearing.
What remains is to discuss touch. The relation between sight and touch is, of course,
a major object of Berkeley’s concern. Inx8.1, I endorsed Turbayne’s suggestion that we
take Berkeley seriously when he says that the relation of vision to touch is like the
relation of written to spoken English (NTV,x143; Turbayne 1970b). ‘Commonsense’
objects can be both seen and touched but there is, according to Berkeley, something
sloppy about the claim that we “see the same object that we feel” (DHP, 245; cf. NTV,
x49). The reason we do not draw this distinction is that “the combination of visible
ideas has constantly the same name with the combination of tangible ideas wherewith it
is connected, which of necessity arises from the use and end of language” (NTV,x49).
Later in the New Theory, Berkeley explains how this occurs:
It is customary to call written words and the things they signify by the same
name: for words not being regarded in their own nature, or otherwise than
355
as they are marks of things, it had been superfluous, and beside the design
of language, to have given them names distinct from those of the things
marked by them (NTV,x140).
The spoken word, of course, likewise shares its name with the written word and the
referent.
The observation Berkeley is making is that it is only for purposes of linguistics,
philosophy, and the like intellectual endeavors that a careful distinction between written
words, spoken words, and their objects is needed. In plain language contexts there is
typically no danger of confusion, and so it is more efficient to avoid the multiplication
of vocabulary. The same is true regarding the distinction between visible and tangible
objects. It is necessary in certain scientific and philosophical contexts, but it would be
counter-productive to try to introduce it into plain language.
The conventions of plain English physical object talk ‘tie together’ multi-modal col-
lections of ideas. What Berkeley argues in the New Theory is that we can achieve a more
precise, explicit understanding of the structure of the language of nature by consistently
distinguishing the tangible words from the visible words. However, just as constant
explicit attention to the specific mode of communication (spoken, written, signed) is
rarely helpful to speaking and understanding human language, so likewise the distinc-
tion which serves us well in analyzing the language of nature would serve us poorly in
‘reading’ that language most of the time. This is an instance of a more general obser-
vation Berkeley makes. On the one hand, Newtonian science is extremely practically
useful in a wide variety of contexts: it has led to “many inventions in Mechanics” so
that we can “frame engines by means of which things difficult and otherwise impossible
are performed” (Alc,x7.10). Yet, on the other hand, just as rough-and-ready intuitive
356
grammatical judgments are often more useful than explicit rules in speaking and under-
standing a language, so also our intuitive ‘folk physics’ is often more useful in getting
around in the world than explicit judgments arrived at by calculation from the Newto-
nian laws (PHK,xx108-109). This is a theme to which we shall frequently return in the
remainder of this chapter.
8.4 Syntax: Causation and Laws
Our study thus far has already touched on the topic of physical causes and laws at a
number of points. It will be helpful briefly to review these before addressing the topic
of how physical causes and laws fit into the language of nature.
Inx7.2, I argued that Berkeley holds that plain language causal talk – including
sentences like ‘fire warms’ (PHK,x31) which do not include the word ‘cause’ or its
cognates – is mostly true or true to a high degree. There are two closely related aspects of
this kind of talk which are responsible for this high degree of truth: it gives us accurate
expectations and provides useful instructions. There is, however, a temptation to think
that such claims tell us what really makes things happen, and this is an error, for only
spirits can genuinely make things happen. Thus this kind of talk contains a mixture of
truth and falsehood, and reform is needed in order that we may speak more truly. This
reform involves distinguishing physical causation from genuine efficient causation (see
PHK,xx51-52; Siris,xx154-155).
The reason the expectations and instructions provided by our causal talk map cor-
rectly onto the world is that, in causing our perceptions, God is following certain rules,
and it is these rules which we call ‘the laws of nature’ (see above, pp.175-178). Physics
aims to improve on our plain language causal talk by the explicit formulation of rules
which allow us to use a ‘mathematical method’ (DM,x38) to arrive at predictions with
357
greater precision and accuracy (see above,xx3.3 and 6.2). In the context of this en-
deavor, we may use the term ‘law’ in a narrower sense to refer to the fundamental laws,
or principles, “in which are grounded and contained the whole discipline” of physics
(DM,x36). Berkeley likely has Newton’s three laws of motion in mind here.
Berkeley is quite explicit in connecting this talk of laws and causes with the language
of nature. Yet there is an apparent inconsistency in Berkeley’s remarks. On the one hand,
Berkeley says repeatedly that the relation of cause to effect is the relation of sign to thing
signified (PHK,xx65-66; TVV,x13). On the other hand, he also repeatedly says that
the laws of physics provide a grammar for nature (PHK,xx108-110; Siris,xx252-254).
The first claim suggests that the relation of cause to effect is a semantic relation, while
the second claim suggests that it is a syntactic relation.
This may become clearer by the following illustration. Berkeley’s claim that the
relation of cause to effect is sign to thing signified suggests that it is like the relationship
of ‘rock’ to rocks. On the other hand, Berkeley’s claim that laws are grammatical rules
suggests that a law can be understood on analogy to syntactic rules of English, like the
rule that a transitive verb must be followed by its direct object. But the relationship of
(physical) cause to effect is one in which a law requires that the effect follows the cause.
Thus it would seem that the relationship of cause to effect is like the relationship of a
transitive verb to its direct object. But this is not at all a relation of sign to thing signified
(cf. Pearce 2008, 257-259).
17
One obvious approach to solving this problem would proceed by pointing out that
Berkeley describes visual ideas as ‘marks and prognostics’ of tactile ideas (PHK,x44),
and likewise asserts that “The fire which I see is not the cause of the pain I suffer upon
my approaching it, but the mark that forewarns me of it” (x65). The approach I have in
17. The response to this problem I provide here differs from the one I suggested in Pearce 2008, 264.
358
mind would take this to be all Berkeley means in describing the cause as the sign of the
effect: there is not a real semantic relation, it is merely that one may predict the effect
on the basis of the cause.
I have two reasons for rejecting this approach. First, it would make Berkeley’s se-
riousness about regarding the world as a ‘discourse’ fluctuate wildly, sometimes using
‘sign’ in a very strong sense for a conventionally meaningful item such as a word in a
language, and at other times using it in this very weak sense. Second, and more impor-
tantly, Berkeley insists that providing successful predictions and useful instructions is
part of what makes the ‘discourse’ of nature meaningful (PHK,x109; Alc,xx4.14-15;
see below,x8.6). Hence it seems that physical causation should indeed be taken as part
of the semantics of the language.
A better solution begins by pointing out that we have been making a simplistic and
uncritical assumption about the relation of causes to laws. The two claims which are
the root of our difficulty here are: (1) that causes signify their effects, and (2) that laws
of nature are grammar rules. These two claims produce contradictory results only given
the further assumption that A causes B if and only if B follows from A according to
the laws. This assumption about the relationship of causes to laws makes it appear that
causes and laws should be either both syntactic or both semantic. However, this further
assumption is not something Berkeley ever asserts. If Berkeley rejected this assumption,
that would explain why he didn’t see any problem with the combination of a semantic
view of causes with a syntactic view of laws.
Let us briefly reflect on the relation of syntax to semantics, and see whether this
can clarify Berkeley’s view. The syntax of English is heavily dependent on word or-
der. Thus, in a simple subject-verb-object sentence like ‘John loves Mary,’ the subject is
whichever noun comes before the verb, and the object is whichever noun comes after. As
359
a result, switching the word order to ‘Mary loves John’ produces a sentence with a very
different meaning. The syntactic rules determine the grammatical relations between the
words, and this in turn determines how those words, together, produce a meaningful
sentence. This can be observed vividly in the case of sentences which are structurally
ambiguous, such as ‘I saw the person with the telescope.’ Here (as in many English ex-
amples) the structural ambiguity is due to uncertainty as to what the prepositional phrase
modifies (it can modify either the verb ‘saw’ or the noun phrase ‘the person’). There are
also cases in which an individual word will have different meanings depending on the
syntax of the sentence. Thus, for instance, the sentence ‘fruit flies like a banana’ can be
(perversely) parsed with ‘flies’ as the main verb and ‘like’ as a preposition, yielding an
interpretation radically different from the obviously intended one.
Now let us consider Newton’s laws of motion. As we have observed above (x6.2),
these laws provide instructions both for the attribution of forces to bodies and for the
generation of further predictions from such attributions. If we consider, for instance,
the elliptical motion of the earth, the first law tells us that some force is acting on the
earth, since elliptical motion is non-inertial. The second law allows us to calculate the
magnitude and direction of that force. The third law tells us that the force must be
exerted by a body on which the earth exerts an equal and opposite force. If we assume
that we are dealing with a single gravitational force, then we may bring in the Law of
Universal Gravitation. Given sufficiently complete information about the earth’s motion,
this will allow us to compute the mass of this other body and the location of its center
of gravity.
The earth and the sun are bodies and bodies, I have argued, are the words of the
language of nature. What we have done by considering the laws of physics is to give an
account of the dynamic relationship between these two bodies. The relationships with
360
which syntax are concerned are, again, purely part of the linguist’s model and not part of
her data. In English, syntactic relations happen to be tied heavily to word order, but there
are other languages in which the same syntactical relation (e.g., of verb to direct object)
occurs in which the word order is consistently reversed, and there are even languages in
which the words can bear the same syntactic relation regardless of which order they are
in, because these syntactic relations are marked by some means other than word order
(typically morphology). Similarly, the force between the earth and the sun is a feature
not of the perceived earth and sun, but of the model we are given in Newton’s ‘grammar
of nature’ (see PHK,xx108-110).
In parsing the sun-Earth system, we must first group the sun ideas into one object
and the Earth ideas into another, recognizing the word boundaries. Then, we need to
understand the grammatical (i.e., dynamic) relationship between these two words. Once
we have done this, we are in a position to interpret the system by seeing what it means
in terms of prediction.
The sun-Earth system is far enough removed from our ordinary experience that we
would not parse it correctly without the aid of explicitly formulated laws of physics. Of
course, once one gains some experience in physics, or simply learns some new rules
of thumb discovered by physicists, one may be able to do this without explicitly stat-
ing what the forces at work in the system are, but one cannot do this with untutored
commonsense.
However, untutored commonsense is a perfectly good interpreter of nature in most
everyday cases (x108). Physics is an explicit account of the grammar of nature. The
reason having such an account is of practical, rather than merely academic, interest is
that it expands the domain within which our interpretations are accurate, and it increases
361
the precision of our interpretations. It also gives us an articulable basis for our predic-
tions (rather than a mere ‘knack’ or ‘hunch’ or ‘intuition’) which improves our ability
to communicate about the interpretation of nature. (Similarly, even among native En-
glish speakers, the explicit use of grammatical categories can be useful for resolving
interpretive disputes about English texts, as historians of philosophy well know.)
In the language of nature, perhaps even more so than in English, the syntax is rarely
ambiguous in context. However, in the absence of context there can be extreme ambigu-
ity. Imagine, for instance, being shown a photograph of a tabletop from above and being
asked to describe the Newtonian forces in operation. One would be expected to describe
the balanced forces of gravity pulling the tabletop down on its legs, and the legs pushing
up on the tabletop. However, the legs are out of view in the photo, so for all one knows
from the photo alone (without background assumptions about tables) there may be no
legs there at all: the tabletop could be in free fall.
These are two different ways of interpreting the photo, which yield different beliefs
about what happened immediately before and after the photo was taken. The different
interpretations arise from different understandings of the syntax of the physical system
in view, just as in cases of structural ambiguity in English.
Let us return now to causal talk. Berkeley writes, “The fire which I see is not the
cause of the pain I suffer on my approaching it, but the mark that forewarns me of it.”
(PHK,x65). When one has the visual orange sensations of the fire, one engages in an
activity of parsing this experience. Thus one groups the fire sensations together into one
object (word), and one implicitly takes it to be related to the other objects around it in
a certain way. For instance, one takes the fire to be burning the wood. This relation
between the fire and the wood is a syntactic one. One does not perceive the burning.
Rather, one perceives the fire and the wood, in close proximity to one another, each
362
undergoing certain transformations. In this way the notion of burning is like the notion
of force. It is a notion one uses to construe the objects of one’s experience as related
in certain ways, and this construal is part of the process of interpretation which leads
to prediction. Parsing the situation properly, so that we regard the fire as burning the
wood, leads us to predict that the wood will turn to ash, and for this reason we say
that the fire causes the wood to turn to ash. The commonsense ‘laws’ at work here
are generalizations like fire needs fuel and wood is combustible. When we see the fire,
before we can interpret the situation and make our predictions, we need to find some
combustible object which can be assigned the grammatical role of fuel for the fire. My
body, like the wood, is combustible, and as there are certain ideas that let me know that
the wood is burning, so pain is one of the ideas that lets me know my body is burning.
Thus once I parse the situation as fire burning wood, I am in a position to interpret the
situation by predicting that I will experience pain if I get too close.
In non-linguistic terms, we can put Berkeley’s view of the relationship between laws
and causes like this: the laws are the general principles needed to conceptualize a given
situation in a way that makes accurate prediction possible. The causes are the items
which, in that conceptualization, are thought of as leading to the effect. In linguistic
terms, the laws are the syntactic rules that allow us to conceive of the relations between
words (bodies) in the way we need to in order to interpret our experience.
8.5 Excursus on Commonsense and Natural Science
It is often said that Berkeley differs from most other early modern philosophers, and
especially from his fellow idealist, Leibniz, in that his philosophy aims to vindicate the
‘manifest image’ at the expense of the ‘scientific image’ (Wilson 1987; R. M. Adams
1994, 226; Roberts 2007, ch. 6). This assertion is correct as far as it goes, but if the
363
interpretation of Berkeley’s metaphysics and philosophy of science which I have so far
developed is correct, then Berkeley’s vindication is far less ‘expensive’ than is often
supposed. In this section, we take a brief detour from our main line of argument to
examine Berkeley’s attitude to the ‘new science’ and the various non-commonsensical
entities posited by it.
One of the leading ideas of the new science was its emphasis on the use of instru-
ments. The proponents of this view held, against the Aristotelians, that it was by sys-
tematic experimentation with the aid of instruments such as the telescope, microscope,
or thermometer, rather than by the unaided senses that the ‘true natures’ of objects were
revealed (see, e.g., Bacon [1620] 2000, 17-18, 125-126, 170-172; Galilei [1632] 2008,
244-248). Berkeley’s defense of ‘commonsense’ is in large part a defense of trust in
the deliverances of the unaided senses (see below,x9.1), and to this extent he can be
seen as an opponent of the new science. Berkeley sees the new science as having the
consequence that “something there is in every drop of water, every grain of sand, which
it is beyond the power of human understanding to fathom or comprehend,” and regards
this conclusion as objectionably skeptical. In accepting this skeptical result, Berkeley
says, “we are influenced by false principles to that degree as to mistrust our senses,
and think we know nothing of those things which we perfectly comprehend” (PHK,
x101). Philonous puts Berkeley’s position more bluntly: “What [the materialist] call[s]
the empty forms and outside of things seems to me the very things themselves” (DHP,
244). It is by the unaided senses that we discover the true natures of objects.
In addition to insisting that real knowledge is gained by the unaided senses, Berkeley
demotes many of the theoretical entities of physics – including, most notably, forces – to
the status of mere quasi-entities, whose ‘natures’ and existence conditions are whatever
the physicist defines them to be (DM,x67). Finally, Berkeley is committed to the claim
364
that nothing inanimate can be a genuine cause, or have any existence independent of
perception. However, the whole scientific (natural philosophical) enterprise, according
to many of its early modern practitioners, is concerned with the discovery of the hidden
(i.e. unperceived) causes of the qualities and behavior of objects. Berkeley clearly and
explicitly rejects this approach to natural philosophy (PHK,xx50, 107; DM,xx35-37;
Siris,xx231, 247, 249, 251).
On the other hand, Berkeley’s overall attitude toward Newton, the leading prac-
titioner of the ‘new science’ in his day, is overwhelmingly positive (see, e.g., PHK,
x110), and Berkeley says that his remarks about scientific methodology are directed to-
ward rendering an already useful discipline all the more useful by eliminating pointless
disputes within it (DHP, subtitle [1713 ed.]; DM,x66; Alc,xx7.9-10). Thus Berkeley
must, in some sense, affirm the ‘new science.’
In order to solve this puzzle, we need first to get clear on the exact nature of Berke-
ley’s criticism of the natural philosophy of his day. When Berkeley objects to the view
that our unaided senses do not reveal to us the ‘true natures’ of water and sand and other
familiar things, he is not objecting to the view that such things have microscopic parts.
Berkeley clearly and explicitly accepts the existence of “the clockwork of nature, [a]
18
great part whereof is so wonderfully fine and subtle as scarce to be discerned by the best
microscope” (PHK,x60; cf. DHP, 245).
19
What Berkeley rejects is rather “the current
opinion that every thing includes within it self the cause of its properties, or that there
is in each object an inward essence, which is the source whence its discernible qualities
flow and whereon they depend” (PHK,x102). When Berkeley writes in his notebooks,
18. Clarke’s brackets.
19. For discussion, see Tipton 1982, 168-169; Garber 1982, 182-184; Wilson 1985, 134-138; Winkler
1989, 265-275; Atherton 1991, 52-53.
365
“My Doctrine affects the Essences of the Corpuscularians” (N,x234), it is the essences
and not the corpuscles he is rejecting.
As Margaret Atherton has shown in her careful and detailed study (Atherton 1991),
the chief target here is the Cartesian notion of scientific explanation, of which Berkeley
provides an admirably clear account in the Principles passage just quoted. The Carte-
sians thought that scientific explanation depended on ‘grasping’ an ‘inward essence’ and
showing how the observed qualities and behavior of the object follow from this essence.
Locke argued that such a ‘grasp’ was impossible and, hence, that the prospects for sci-
entific explanation – or, as he preferred to put it, for making natural philosophy a science
(EHU,x4.12.10) – were dim (xx4.3.25-26, 4.6.15, 4.12.9-10).
20
Berkeley’s response is
to deny that this is what natural philosophy is about. Natural philosophy cannot be about
this, for a physical object does not have an “inward essence . . . whence its discernible
qualities flow.” The ‘real nature’ of a body is nothing but its sensible appearance.
If the senses grasp the ‘real natures’ of things, then what role is there for science,
with its use of experiments, instruments, and mathematical reasoning? Berkeley writes,
If we . . . consider the difference there is betwixt natural philosophers and
other men with regard to their knowledge of the phenomena, we shall find
it consists . . . only in a greater largeness of comprehension, whereby analo-
gies, harmonies, and agreements are discovered in the works of nature, and
the particular effects explained, that is, reduced to general rules . . . [This
knowledge] enable[s] us to make very probable conjectures, touching things
that may have happened at very great distances of time and place, as well as
to predict things to come (PHK,x105)
20. See Garber 1982, 175-178; McCann 1994,x3; Downing 1998, 399-412.
366
This is in line with Philonous’s remark in the Dialogues that the use of microscopes
is “only to know what ideas are connected together” and that this is all we mean by
knowledge “of the nature of things” (DHP, 245).
The methods of natural science – including careful, systematic observation with the
aid of instruments, as well as the intellectual activity whereby the observed rules of na-
ture “are refined by reason and rendered universal” (DM,x36) – are simply methods
for arriving at a more precise and comprehensive knowledge of the relations between
ideas. The fact that, prior to engaging in natural science we are ignorant of many of
these connections does not threaten Berkeley’s anti-skeptical project. Rather than un-
dermining commonsense, knowledge of natural science merely supplements it (Garber
1982, 189-191).
The introduction of ‘mathematical hypotheses’ is, as we have already seen (above,
x6.2), an essential part of the ‘univeralizing’ task of natural science. The introduc-
tion of these entities is necessary to “make notions or, at least, propositions, universal”
(DM,x7). Just as the linguist, in order to give a general account of the grammar of
English must conceptualize the auditory phenomena of the language by means of cer-
tain specialized grammatical concepts which cannot be straightforwardly mapped onto
particular disturbances of the air, the physicist, in order to give a general account of the
grammar of nature, must introduce entities which she does not directly observe.
Berkeley is demoting the reality of the physicist’s theoretical entities below that of
commonsense objects. Certainly he is rejecting the claim that the entities of fundamental
physics are more real, or more fundamental, than the entities of commonsense. But, as
I argued in chapter 6, the gap between bodies and forces is not so wide as is commonly
supposed, and this result generalizes. In linguistic terms, the point can be made like
367
this: commonsense concepts like word and letter are part of a constructed model of lan-
guage, just like the linguist’s more precise concepts, lexeme and phoneme. Indeed, the
typical speaker, even one who has never studied linguistics, will have a certain amount
of explicit grammatical knowledge, and this knowledge must necessarily be given in
terms of such a model. The linguist’s model is more precise, accurate, and general, but
for just this reason must be further removed from the concrete phenomena of language.
To repeat an example used above, a direct object cannot be defined in terms of sounds
or marks alone, for no such definition could possibly be valid cross-linguistically. Thus
the status of direct-objecthood is something like a mathematical hypothesis: this status
is not represented in sense experience, but must be introduced in order to give a suit-
ably general account of the phenomena of language. For Berkeley, this movement away
from the senses is part and parcel of generality, and hence is essential to all ‘scientific’
knowledge whatsoever (Alc,xx7.14-16).
Plain language body talk is part of a commonsense grammar of the language of
nature. Like the commonsense grammar of English, it involves only concepts that are
fairly closely tied to concrete sense experience. Yet it ultimately differs from the more
general and technical grammar of the physicist only in degree. In this way, Berkeley
preserves the ability of natural science to teach us things we didn’t know before, and
defends the legitimacy of its introduction of new entities unknown to commonsense,
while denying to natural science the right to take away the knowledge the person on the
street has had all along.
8.6 Semantics
Perhaps the most central problem with Berkeley’s divine language hypothesis is the
difficulty of giving an account of the semantics of the divine language (Mabbott 1931,
368
26-28; Olscamp 1970b, 32-46; Creery 1972, 218-221; Roberts 2007, 65). If the Author
of Nature is speaking to our eyes (NTV,x152), what is he saying?
The texts in which Berkeley deals with this issue can be divided into three categories.
In the first set of texts, Berkeley speaks of sensory ideas conveying information about
other ideas. In these texts, Berkeley typically emphasizes the role of the language of
nature both in providing beliefs or expectations about what ideas we will have in the
future (often conditionally on our own actions), and in giving us instructions about what
course of action to take. In the second set of texts, Berkeley says that sensory ideas
signify other finite minds and their states and activities. In the third set of texts, Berkeley
says that the language of nature reveals to us God and his attributes.
Most existing treatments of Berkeley’s language of nature focus on one or the other
of the first and third sets of texts. The second set of texts is almost uniformly ignored.
This is quite odd, since – as I have argued before (Pearce 2008), and will now argue
again – unless information about other minds is provided to us by the language of nature,
Berkeley will be stuck in solipsism.
Given Berkeley’s emphasis on the flexibility of language (including the language
of nature) and the wide variety of purposes it can serve (see above,x8.1), there is no
reason why Berkeley should not hold that the language of nature does all three of these
things, with sensory ideas giving us information about other ideas, God, and other finite
minds. In this section, I will address each set of texts in turn, and then show how these
three types of meaning to be found in the language of nature fit together into a coherent
overall picture of the structure and significance of Berkeley’s world.
369
8.6.1 Informing and Instructing about Ideas
The passages in which Berkeley apparently takes the language of nature to be about our
sensory ideas (i.e., about itself) are numerous, and may be subdivided into four (over-
lapping) categories. First, there are those in which he says that visual ideas “suggest
and exhibit to us the . . . various qualities of tangible objects . . . just as words suggest the
things signified by them” (Alc,x4.10; cf. NTV,xx17, 32, 51, 77, 140).
21
Second, there
is NTV,x143, where Berkeley says that “visible figures represent tangible figures much
after the same manner that written words do sounds.” Third, there are those passages in
which Berkeley emphasizes that the interpretation of the language of nature allows us
to predict our future perceptions (NTV,x45; PHK,x65; Alc,x4.7). Fourth and finally,
there is the cluster of texts, closely related to and overlapping the third group, in which
Berkeley emphasizes that the language of nature gives us practical instructions about
how to conduct ourselves in the world of sense experience (NTV,x147; PHK,xx65,
109; Alc,x4.7).
Let us begin with the previously promised reconciliation of the first two groups of
texts. As indicated above (pp. 338-339), there is an apparent conflict here between
Berkeley’s claim that the relation of vision to touch is that of sign to thing signified, and
his claim that it is that of written word to sound. This conflict can, however, be resolved
by calling attention to a systematic ambiguity which Berkeley says is necessary to “the
design of language.” This is the fact that “It is customary to call written words and the
things they signify by the same name.” That is, to use Berkeley’s own example, “a tan-
gible square, and the monosyllable consisting of six letters whereby it is marked” are
21. Strictly speaking, not all of the suggesting ideas in these passages are visual. For instance, in NTV,
x17, one of the ideas is “the sensation arising from the different turn [the mind] gives the eyes, in order to
bring the pupils nearer or farther asunder.”
370
both named by a single sound (NTV,x140). Note also that the relation here is symmet-
ric: the inscription ‘square’ names both squares and the sound in question. Berkeley’s
view, then, would seem to be that visual ideas do indeed name tangible ideas, in the way
that spoken words name their written counterparts and vice versa. The use of sounds to
name inscriptions and vice versa is, however, not the primary use of words in language.
Rather, the primary use occurs at the ‘object level,’ where both ‘square’ and its corre-
sponding sound are used to name squares, and for a variety of other related purposes. In
philosophical discourse (including this paragraph), we use quote-names to disambiguate
some of these usages, but this convention was not well-established in Berkeley’s day,
22
and is still not followed consistently in non-philosophical contexts.
If this is what Berkeley has in mind, then he is not committed to the problematic
claim, sometimes attributed to him (Mabbott 1931, 26-27; Olscamp 1970b, 32-33; Win-
kler 1989, 21), that the words of the language of nature signify only other words. Every
word of spoken English signifies its written counterpart, and vice versa, but this does
not mean that English is only about itself.
Nevertheless, there is a sense in which at least part of the language of nature clearly
is about itself, for Berkeley clearly holds that a crucial aspect of the meaningfulness
of the language of nature is the way in which “we are instructed how to regulate our
actions in order to attain those things that are necessary to the preservation and well-
being of our bodies, as also to avoid whatever may be hurtful and destructive to them”
(NTV,x147). Berkeley adamantly insists that these things are only ideas and hence that
they are part of the language of nature. In this passage, and in most of those where
Berkeley addresses this issue, the language of nature is seen as both giving instructions
22. Berkeley often marks the use-mention distinction by means of italics, but he is not fully consistent
about this, and also sometimes uses italics for other purposes. In Clarke’s edition, from which I quote in
this dissertation, italicized words are replaced with quote-names where appropriate.
371
and providing ‘information.’ Yet both the instructions and the information are apparently
about the language itself.
If there was no more to the language of nature than this, then the whole thing would
be rather pointless. However, this is not all there is, for the language of nature is also,
and in fact primarily, about minds, as I will now proceed to show. Once we have un-
derstood God’s talk about minds, we will be in a position to return to the puzzling
self-referential nature of the rest of the divine discourse.
8.6.2 Informing about Other Finite Minds
Although some scholars (e.g. Olscamp 1970b, 32; Winkler 1989, 21) have held that all
of the ideas in Berkeley’s language of nature signify only other ideas, Berkeley explicitly
and repeatedly claims that our ideas of sense signify other minds. Three times in the New
Theory of Vision, Berkeley illustrates the nature of the suggestion relation by discussing
the way in which the appearance of a person’s face suggests to us the ‘passions’ he or
she is experiencing (NTV,xx9-10, 23, 65), and hence provides information about this
other mind. Berkeley insists that these suggestion relations are of the same sort as the
suggestion relations between visual ideas and tactile ideas.
The role of these kinds of signs in the language of nature is developed in more detail
in the discussion of other minds at the end of the Principles:
A human spirit or person is not perceived by sense, as not being an idea.
When therefore we see the colour, size, figure, and motions of a man, we
perceive only certain sensations or ideas excited in our own minds; and
these being exhibited to our view in sundry distinct collections, serve to
mark out unto us the existence of finite and created spirits like our selves
(PHK,x148, emphasis added).
372
‘Mark out’ is, for Berkeley, a synonym for ‘signify’ (see, e.g., NTV,x140). As if this
were not explicit enough, Berkeley goes on to write, in the same section, “one finite
and narrow assemblage of ideas denotes a particular human mind” (PHK,x148, empha-
sis added; cf. Alc,x4.5; Pearce 2008, 252-253). This is framed as an explanation of
Berkeley’s immediately preceding claim that God is better known, and more immedi-
ately ‘seen’ than any human person. According to Berkeley, “He [that is, God] alone
it is who ‘upholding all things by the Word of his Power’, maintains that intercourse
between spirits whereby they are able to perceive the existence of each other” (PHK,
x147). Whatever knowledge (or reasonable belief) about other minds we have, we have
on the basis of God’s testimony through the language of nature.
However, as Denis Grey observed long ago (Grey 1954), there is an interesting ap-
parent conflict between the discussion in (PHK,xx147-148) and an assertion Berkeley
makes not long before it. In this earlier passage, Berkeley says that “we cannot know
the existence of other spirits, otherwise than by their operations, or the ideas by them
excited in us” (x145). This suggests knowledge, not by arbitrary signification, but rather
by causal inference. As Grey likewise notes, this parallels two ways Berkeley talks
about our knowledge of God. These have come to be known as the ‘passivity argument’
and the ‘divine language argument’ (see Pearce, forthcoming[b],xx1-2).
We shall have more to say about the way in which the divine language informs us
about God shortly. However, we should pause here to discuss the relationship between
the passivity and divine language arguments in order to show how the apparent conflict
identified by Grey is to be resolved.
Berkeley’s passivity argument, upon which we have already had occasion to com-
ment (see above, pp. 167-168, 197-199, and 245-247), argues from the fact that ideas
of sense are passively received to the existence of a foreign agent who is their cause.
373
We must note, however, just how little this proves. Berkeley’s stated conclusion is only
“[t]here is therefore some other will or spirit which produces” my ideas of sense (PHK,
x29). Berkeley then argues that the patterns to be found in my ideas of sense demon-
strate “the goodness and wisdom” of the mind that causes them (x32). At the end of the
Principles, Berkeley makes the stronger claim that
if we attentively consider the constant regularity, order, and concatenation
of natural things, the surprising magnificence, beauty, and perfection of the
larger, and the exquisite contrivance of the smaller parts of creation, to-
gether with the exact harmony and correspondence of the whole, but above
all the never enough admired laws of pain and pleasure, and the instincts or
natural inclinations, appetites, and passions of animals; I say if we consider
all these things, and at the same time attend to the meaning and import of the
attributes, one, eternal, infinitely wise, good, and perfect, we shall clearly
perceive that they belong to the aforesaid spirit, ‘who works all in all’, and
‘by whom all things consist’ (x146).
As Ekaterina Ksenjek and Daniel Flage have noted, this description of God is far
more religiously adequate than the earlier one, and is in fact closely related to the de-
scription found at the beginning of the Thirty-Nine Articles, the doctrinal standard of
Berkeley’s church (Ksenjek and Flage 2012, 291-292; cf. Pearce, forthcoming[b],x2).
All of the features Berkeley here cites are clearly aspects of the divine language, and
in fact Berkeley had earlier listed “enlarg[ing] our notions of the grandeur, wisdom,
and beneficence of the Creator” as one of the aims we ought to have in ‘reading’ the
language of nature (PHK,x109).
In the Principles, Berkeley first argues by means of the passivity argument that some
mind is the cause of our sensory perceptions (PHK,xx25-30). We then ask the general
374
question of this chapter, what is that mind up to? The answer is that that mind is ‘speak-
ing’ to us (xx31-33, 65, 108-110). In the course of ‘reading’ the discourse which is
presented to us, we draw conclusions about the nature of that mind, and see that its
attributes match those of the traditional God (xx31-32, 109, 146-150). In Berkeley’s op-
tical writings, he argues that vision is a language without first defending immaterialism,
and hence without rejecting ‘second causes.’
23
Berkeley thus evidently believes that
the deliverances of the senses can be seen to be a language without the need for prior
knowledge of their efficient cause. Nevertheless, in Alciphron IV , where Berkeley spells
out the inference from his view of the world as a ‘discourse’ to the existence and nature
of God, the emphasis is again on the divine attributes, rather than on mere existence.
We will come to this last inference, from the language of nature to the divine at-
tributes shortly. The point I want to make at present is that, contrary to Grey, there is no
conflict between the “causal theory of God” and the model of “God as the Divine Inter-
locutor” (Grey 1954, 29-30), for, after all, a speaker is typically (always?) the cause of
her utterance. What is going on with the two arguments of the Principles is that we first
infer a speaker, and then get to know something about him by interpreting his utterances.
However, Grey’s claim that Berkeley is inconsistent is on much stronger ground in
the case of other finite minds. In this case, Berkeley first asserts that we know other
minds by means of “the ideas by them excited in us” (PHK,x145) and then, only two
sections later, asserts that “in affecting other persons, the will of man has no other ob-
ject than barely the motion of the limbs of his body; but that such a motion should be
attended by, or excite any idea in the mind of another, depends wholly on the will of the
Creator” (PHK,x147). The first passage apparently asserts, and the second apparently
denies, that human beings can cause ideas in other human beings. There is clearly a
23. On the rejection of ‘second causes,’ see PHK,x32.
375
deep tension here. On the other hand, it is deeply implausible to suppose that Berke-
ley would not notice an explicit contradiction between two texts so close together. It
would be better by far if an explanation could be given of how Berkeley understood the
two passages just quoted on which they are at least not obviously contradictory to one
another.
The difficulty we find here is symptomatic of a deep ambivalence in Berkeley’s
philosophy: his attitude toward occasionalism. Berkeley wants to insist, against Male-
branche, that “We move our Legs our Selves” (N,x548). He also wants to defend the
claim that the whole sensible world is the product of God’s willing. Finally, he wants,
of course, to defend the claim that legs are sensible.
We can attribute a consistent view on this matter to Berkeley if we attend to the
qualifications he adds to those passages where he insists that we humans are the agents
in our bodily movements. In the notebooks passage just quoted, Berkeley goes on to
remark, as if by way of explanation, “’tis we that will their [i.e., our legs’] movement”
(x548). At PHK,x145, Berkeley says that perception of human bodies in motion informs
me of “certain particular agents like myself, which accompany them and concur in their
production.” Thus my willing the motion of my legs constitutes my concurrence in
God’s production of that motion, and the motion is perceived by others. In the next
section, Berkeley confirms this by saying, vaguely, that “there be some things which
convince us, human agents are concerned in producing them” (x146, emphasis added).
Humans do not produce these actions on their own, but by concurrence with God.
Concurrentism about divine and human action is rather a slippery view in early mod-
ern philosophy and theology. Generally speaking, this view involves holding that the
divine and human wills cooperate in the production of the effect. But how exactly does
this ‘cooperation’ work, in Berkeley’s view? In PHK,x147, Berkeley says that “the
376
will of man has no other object than barely the motion of the limbs of the body.” Given
Berkeley’s general theory of action (see above,x6.4), this would seem to mean that what
the man does is merely move his limbs. This is contrasted with “excit[ing an] idea in the
mind of another.” This latter event “depends wholly on the will of the Creator.” What
this suggests is that my moving my legs can be distinguished from anyone else having
any ideas.
What I want to suggest is that my moving my legs consists in a certain kinesthetic
sensation experienced solely by me, and that I am the efficient cause of this sensation.
What God does is to cause corresponding sensory ideas both in me and in others. Since
God causes these ideas in response to my volition to move my leg, I can be said to
concur with God in the production of these ideas.
This hypothesis helps to answer our central question about the semantics of the lan-
guage of nature: the visual perception of the motion of my leg means or signifies my
action of moving my leg. Actions, for Berkeley, just are volitions (see above,x6.4), and
volitions cannot be perceived, since they are not ideas nor are they like ideas. Neverthe-
less, just as “We often see shame or fear in the looks of a man, by perceiving changes
of his countenance to red or pale” (NTV,x9), so we often see actions by seeing human
bodies in motion. In both cases, this seeing is mediated by the language of nature. This
is how Berkeley can insist that God “alone . . . maintains that intercourse between spirits,
whereby they are able to perceive the existence of each other” (PHK,x147).
In observing human bodies, I immediately perceive signs which signify actions and
passions. Yet, I am aware that these actions and passions are not my own. As a result, if I
trust the author of nature, I will conclude that these actions are done, and these passions
experienced, by other finite minds like myself. I learn the meaning of these signs from
my own case, and extrapolate to others.
377
We must now briefly address a difficulty in the interpretation of nature about which
Berkeley says far too little. This is the status of animals in his system (cf. Pearce 2008,
265-266). The higher animals, at least, exhibit many of the features which lead us to
posit minds in the case of humans. Furthermore, Berkeley seems in several places to
recognize that animals perceive the world in much the way that we do (NTV,xx24, 59;
DHP, 188; Siris,x254). S´ ebastien Charles has suggested that Berkeley held, as against
the Cartesians, that the difference between humans and animals is a matter only of de-
gree (Charles 2010). This is supported by Berkeley’s apparent endorsement, throughout
Siris, of the Neoplatonic Great Chain of Being (cf. Moked 1988, 120-121). However,
we should tread carefully here, for there is in Berkeley’s philosophy, as in the Cartesian
system, a fundamental contrast between active mind and passive body. Berkeley differs
from the Cartesians in analyzing body into ideas. The contrast between mind and ideas,
however, is clearly, for Berkeley, one of kind rather than merely degree. Thus if Berke-
ley does endorse the view Charles attributes to him, he will be committed to classifying
animals as minds. Furthermore, Berkeley’s argument for the natural immortality of the
human soul (PHK,x141) would then apparently apply to animals, a result with which
he would probably not be pleased (cf. Charles 2010, 190-191, 197-199).
Oddly, despite the fact that Charles’ thesis of continuity derives its strongest support
from the presence of the Great Chain of Being doctrine in Siris, Charles sees Berkeley
as relapsing into the Cartesian view of animals as machines in that work. In Siris,x257,
Berkeley suggests that the intelligence which moves “the systole and diastole of the
heart,” as well as the diaphragm, may well also be the agent in the habitual actions of
humans, and, further, that this intelligence is “the same perhaps which governs bees and
spiders, and moves the limbs of those who walk in their sleep.” Charles sees this as a
return to a Cartesian understanding of animals (Charles 2010, 197).
378
A few sections earlier, Berkeley drew another contrast between humans and animals:
with respect to the language of nature, “A beast is like a man who hears a strange tongue
but understands nothing” (Siris,x254).
Note, however, that in both of these passages Berkeley is drawing attention both
to similarities as well as differences between humans and animals. Bees and spiders
are like humans in engaging in involuntary actions which are nevertheless (in some
sense) attributable to them (the sleep walker is the one who walks, although the action is
involuntary), but unlike humans in that they do not (apparently) perform truly voluntary
actions. ‘Beasts’ more generally are like humans in that they perceive the signs which
make up the language of nature, but unlike humans in that they do not understand it.
These passages certainly seem to attribute genuine mental lives to animals and thus,
given Berkeley’s general view, to take them to be minds. From the perspective of the
language of nature, it is not hard to see why this is so. Much of what we observe
in animals seems to be the same as what, in humans, signifies actions and passions.
Berkeley’s God seems to be telling us that animals too perform actions and experience
passions, though Berkeley claims that lower animals, such as bees and spiders, are not
genuine agents with respect to their complex patterns of instinctive behavior.
Does this imply the immortality of animal souls? Not necessarily. Berkeley’s ar-
gument for the natural immortality of the soul in PHK,x141 is simply an argument
for the conclusion that the soul “is not liable to be broken or dissolved by the ordinary
laws of nature or motion.” It is not an argument that the soul actually persists beyond
death (see Pearce, forthcoming[b],x3). Berkeley does elsewhere argue that the actual
immortality of human souls is probable independent of revelation (BW, 7:73, 114-115,
181-184; Alc,x6.11).
24
These arguments, however, do not appear to apply to animals.
24. For discussion, see Hight 2007a.
379
Thus Berkeley should be seen as holding that animals are genuine minds, and hence im-
material and possibly immortal. However, for animals, unlike humans, we do not have
probable reasons either from nature or from revelation, to suppose actual immortality.
8.6.3 Informing about God
In addition to informing us about ideas and other finite minds, Berkeley frequently says
that the language of nature informs us about God. In the Principles, Berkeley says,
As in reading other books, a wise man will choose to fix his thoughts on
the sense and apply it to use . . . so in perusing the volume of nature . . . We
should propose to our selves nobler views, such as to recreate and exalt the
mind, with a prospect of the beauty, order, extent, and variety of natural
things; hence by proper inferences, to enlarge our notions of the grandeur,
wisdom, and beneficence of the Creator (PHK,x109).
Similarly, in Alciphron we read,
this visual Language proves, not a Creator merely, but a provident Governor
actually and intimately present, and attentive to our interests and motions:
who watches over our conduct, and takes care of our minutest actions and
designs, throughout the whole course of our lives, informing, admonish-
ing, and directing incessantly, in a most evident and sensible manner (Alc,
x4.14).
In both the Principles and Alciphron, Berkeley contrasts the way the language of
nature informs us about God with the way it informs us about other finite minds:
[W]hereas some one finite and narrow assemblage of ideas denotes a partic-
ular human mind, withersoever we direct our view, we do at all times and in
380
all places perceive manifest tokens of the divinity, every thing we see, hear,
feel, or any wise perceive by sense, being a sign or effect of the power of
God, as is our perception of those very motions which are produced by men
(PHK,x148; cf. Alc,x4.5).
Everything in nature is a ‘sign or effect’ of God insofar as these ideas may suggest their
inferred efficient cause (see above, pp. 167-168). This contrasts with the human case
in which there is some specific item in the divine discourse which is used by God to
inform us about the human being. In the discussion of the interpretation of the divine
discourse in PHK,x109, Berkeley instructs us to look at the discourse holistically in
order to ‘enlarge our notions’ of God. Berkeley specifically mentions the ‘beauty, order,
extent, and variety’ which is to be observed in God’s discourse. Similarly, Euphranor
says,
Something there is of divine and admirable in this Language, addressed to
our eyes, that may well awaken the mind, and deserves its utmost attention:
it is learned with so little pains: it expresses the differences of things so
clearly and aptly: it instructs with such facility and dispatch, by one glance
of the eye conveying a greater variety of advices, and a more distinct knowl-
edge of things, than could be got by a discourse of several hours. And while
it informs, it amuses and entertains the mind with such singular pleasure
and delight (Alc,x4.15).
Berkeley emphasizes the overall character of the discourse as allowing us to make
inferences about God. What this suggests is that this information about God may not be
part of the semantic content of the discourse at all. Rather, we gain knowledge of God in
the way that one may gain knowledge of an author by reading her novel, even if no part
of the novel is autobiographical. What a speaker or writer says or writes conveys a great
381
deal of information about the speaker or writer, even if he is not explicitly talking or
writing about himself.
25
This is the case to an even greater degree with God’s “coherent,
entertaining, and instructive Discourse” (Siris,x254), for God is not only the author of
the discourse, but also the designer of the language in which it takes place.
8.6.4 The Interpretation of the Discourse of Nature
What is God’s discourse about, and what is its purpose? The greater part of God’s dis-
course lacks direct extra-linguistic reference. Using either commonsense or Newtonian
laws, I parse my visual impression of fire, and this parsing allows me to interpret it in or-
der to make a variety of predictions. These predictions constitute the meaning or content
conveyed by this ‘sentence’ of the language. Like most of English, most of the divine
language consists in pure ‘inference tickets’ which do not make direct contact with any-
thing extra-linguistic. This is what the allegedly problematic s