University of Southern California Dissertations and Theses

Representation, truth, and the metaphysics of propositions

(USC Thesis Other)

Representation, truth, and the metaphysics of propositions

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Representation, Truth, and the Metaphysics of Propositions by Greg Ackerman 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) May 2016 Copyright 2016 Greg Ackerman 2 Contents Acknowledgements...............................................................................................................................4 Introduction.............................................................................................................................................5 Chapter 1: The Way Propositions Need to Be............................................................................11 1. Introduction..........................................................................................................................................................12 2. The Meaning Route............................................................................................................................................16 3. The Embedded That-Clause Route..............................................................................................................21 4. Why Minor Cognitive Contact is Enough..................................................................................................24 5. Artifactualism – A Minimal Theory of Propositions............................................................................30 6. Competing Views.................................................................................................................................................35 7. Concluding Remarks..........................................................................................................................................39 Chapter 2: Bridging the Gap between Representation and Truth......................................40 1. Introduction..........................................................................................................................................................41 2. The General Demarcation Problem............................................................................................................43 3. The Link between Representation, Accuracy, and Truth...................................................................44 4. The Naïve Representational Theory of Truth-Aptitude.....................................................................46 5. The Unity of the Sentence................................................................................................................................50 6. The Special Demarcation Problem..............................................................................................................52 7. The Functional Theory of Truth-Aptitude................................................................................................53 8. Consequences for the Nature of Propositions.........................................................................................60 9. Concluding Remarks..........................................................................................................................................64 Chapter 3: Living without Representational Mental States..................................................66 1. Introduction..........................................................................................................................................................66 2. Background and Motivation..........................................................................................................................68 3. Features of Representation............................................................................................................................70 4. Denying that there are Representational Mental States...................................................................72 5. Accommodating the Truisms of Folk-Psychology.................................................................................75 6. Options for Accounting for the Nature of Mental States...................................................................77 7. Solving the Problem of Intentionality........................................................................................................80 8. Consequences for the Current Debate on the Nature of Propositions.........................................83 9. Consequences for the Relationalism/Representationalism Debate..............................................84 10. Concluding Remarks.......................................................................................................................................90 Chapter 4: What Complex Sentences and Propositions Represent....................................91 1. All Propositions Represent Things as Being Ways...............................................................................92 2. Intuitions about Aboutness Provide Constraints..................................................................................95 3. Complex Types of Sentences/Propositions...............................................................................................97 3.1 Multi-Place Relations...........................................................................................................................97 3.2 Sentential Operators............................................................................................................................98 3.3 Quantifiers.............................................................................................................................................104 3.4 Remaining Types................................................................................................................................107 4. Application to Poorly Understood Semantic Phenomena..............................................................107 4.1 Multi-Clausal Sentences..................................................................................................................107 3 4.2 Descriptions.........................................................................................................................................108 4.3 Lists, Paragraphs, Speeches, and Discourses.........................................................................109 4.4 Plurals.....................................................................................................................................................110 4.5 Tense and Aspect...............................................................................................................................111 4.6 Non-Referring Demonstratives....................................................................................................112 4.7 Definite Descriptions........................................................................................................................112 4.8 Phenomena Often Categorized as “Convential Implicature”...........................................114 5. Concluding Remarks.......................................................................................................................................115 Chapter 5 : Artifactualism and the Puzzles of Belief.............................................................116 1. Introduction.......................................................................................................................................................116 2. For Propositionalist Solutions....................................................................................................................118 3. Actionism.............................................................................................................................................................120 3.1 Details and Motivation.....................................................................................................................121 3.2 How it can Solve the Puzzles.........................................................................................................122 3.3 Actionism’s Problem with Truth.................................................................................................126 4. How Artifactualism Solves the Puzzles...................................................................................................127 5. The Nature of Attitude Reports in General...........................................................................................131 5.1 The Range of Actionist Options....................................................................................................131 5.2 An Option Exclusive to Artifactualism......................................................................................133 6. Concluding Remarks.......................................................................................................................................136 Bibliography.......................................................................................................................................137 4 Acknowledgements I would like to thank the members of my committee: Scott Soames, Robin Jeshion, Gabriel Uzquiano-Cruz, and Barry Schein. In addition to the members of my committee, John Hawthorne, Janet Levin, Andrew Bacon, and Mark Schroeder offered helpful feedback on various chapters of this dissertation. Finally, I would very much like to thank Keri Smith for her support throughout this process. 5 Introduction This dissertation is in an investigation into the nature of propositions and propositional attitudes. It is inspired by the growing optimism that returning to a serious investigation into the nature of propositions is a promising way to make progress on long standing problems in the philosophy of language and the philosophy of mind. The view of propositions I offer here, Artifactualism, is the view that propositions are kinds of artifacts. Propositions are kinds. Their instances are particular representational artifacts. The proposition that snow is white is the kind of artifact that has the function of representing snow as being white and has no other functions involving representing things as being ways. Instances of this kind include tokens of the English sentence ‘Snow is white’, tokens of the Spanish sentence ‘La nieve es blanca’, and many other particular linguistic artifacts that have the function, i.e. serve the purpose, of representing snow as being white. In general, propositions are representational kinds/types whose instances all share a common representing function. This account of propositions has two important advantages over other recent views that also take seriously the question of what propositions are and how they could have all the properties that they are alleged to have. The first advantage is that the account requires only minimal theoretical assumptions. We only need to grant the obvious truth that there are particular artifacts that have the function of representing things as being ways and grant that there are various kinds of these artifacts. We need not introduce radically new entities into our ontology. We need not take on bold assumptions about the nature of the mind. The second advantage is that we are not forced to accept any revisionary results about what kinds of things are true or false. The heart of our inquiry into the nature of propositions consists in a careful investigation into what way an entity needs to be in order to be truth-apt, i.e. have any truth-conditions at all. Chapter 1 examines the theoretical reasons for positing propositions. First, it argues contra Jeff King, Scott Soames, and Jeff Speaks, that propositions have to meet relatively minimal constraints in order to fulfill our theoretical needs. It turns out that King, Soames, and Speaks are right that propositions cannot be tools created by the theorist for modeling purposes. However, I show that when we evaluate the various arguments King, Soames, and Speaks put forward to establish this and see what our theoretical needs actually require, it turns out that their respective 6 accounts of propositions take on unnecessary commitments. 1 They take our theoretical needs to require more of propositions than what is actually required of them. Consequently, they overlook options for accounting for the nature of propositions that are plausible and have minimal theoretical commitments. The chapter ends by introducing Artifactualism and arguing that it is well positioned to meet our theoretical needs without incurring the substantive and unnecessary theoretical commitments taken on by its competitors. Chapter 2 gives a positive argument to show that Artifactualism is indeed the correct account of propositions. The argument proceeds by first arguing for a demanding account of truth- aptitude. An account of truth-aptitude must answer two questions. First, what way does an entity need to be in order to be truth-apt, i.e. have any truth-conditions at all. Second, how are the truth- conditions of a truth-apt entity determined by its other properties? A careful investigation into the nature of truth-aptitude reveals that a truth-apt entity must have the function, i.e. serve the purpose, of representing. I argue that this result rules out nearly every metaphysically reductive theory of propositions that has been proposed so far. Only Artifactualism is consistent with the demanding theory of truth-aptitude. Chapter 3’s main concern is the nature of representation and whether there are representational mental states. The success of Artifactualism does not strictly depend on the results of chapter 3, however the results of chapter 3 allow Artifactualism to avoid making strong theoretical commitments about the nature of the mind. Importantly, this chapter also serves to shed light on the familiar but often misunderstood notion of representation – a notion that is at the core of defining and arguing for Artifactualism. The first half of the chapter argues, contrary to widespread opinion, that there are no good pre-theoretical or theoretical reasons for holding that there are representational mental states. We can deny there are representational mental states while still accounting for the truisms of folk-psychology and without severely limiting our options when it comes to giving account of what propositional attitudes are. The second half of the chapter draws out the argument’s consequences for the nature of propositional attitudes, the problem of intentionality, the nature of propositions, and the nature of perception. The results of this chapter support a view of propositional attitudes where modern agents use propositions to help pick out mental states that share certain features in common with propositions. Importantly, being representational is not one of those features. Furthermore, the results of this chapter support a view of propositions where propositions are not mental entities. 1 This is not to claim that an account of propositions can ignore accommodating our pre-theoretical judgments. The focus here is just limited to theoretical reasons for positing propositions. 7 Chapter 4 addresses some complications that arise for Artifactualism when it comes to complex propositions, i.e. non-atomic propositions. Our demanding account of truth-aptitude in Chapter 2 commits us to the result that all propositions represent things as being ways. So we must give an account where, for example, sentences with quantifiers and sentential operators represent things as being ways. Chapter 2 went over some options open to us for complex sentences and propositions, but Chapter 4 argues for the preferred way of proceeding. This chapter ends by showing that an additional benefit of giving our account of complex sentences and propositions is that it helps us make progress on semantic phenomena that we are currently puzzled by. Such phenomena include multi-clausal sentences, definite descriptions, tense, and plurals. In Chapter 5 we see how Artifactualism can give an elegant solution to long-standing semantic puzzles such as Frege’s puzzle. The solution is Fregean in one sense because it posits a proposition distinct from that Venus is identical to Venus that the ancients came to know. The basic idea is that some artifacts are specialized representers. Specifically, some artifacts have the function of representing Venus as being identical to Venus, but other artifacts have the more specialized function of representing Venus as being identical to Venus by referring to Venus with ‘Hesperus’ in one instance and referring to Venus with ‘Phosphorus’ in the other instance. These specialized token tools are instances of a specialized proposition that is representationally identical to the plain proposition that Venus = Venus. Importantly, this specialized proposition is what the ancients came to know when they learned that Hesperus is Phosphorus. I argue that although Scott Soames’ and Peter Hanks’ theory of propositions can give a similar solution the semantic puzzles, Artifactualism gives the most plausible and straightforward account of how tokens of sentences like ‘The ancients learned that Hesperus is Phosphorus’ can be true. Before diving into the arguments, it will be instructive to both go over some general reasons for positing propositions as well as cover the impetus behind the recent revival in investigating the nature of propositions. There are both pre-theoretical and theoretical reasons for positing propositions. One pre- theoretical reason concerns the fact that we say things like ‘Anne believes that Oceans once existed on Mars and this is the exact same thing that Bill was asserting yesterday. Moreover, we have evidence to support its truth, although Cathy thinks that this is false.’ So we appear to talk and judge and as if there is some sort of entity that is believed, that is asserted, and that itself can be true or false. We will say that any entity that is the sort of thing that is believed, asserted, and 8 is itself possibly true or false, is an entity that plays the traditional role required of propositions. 2 Positing propositions that play their traditional roles allows us to easily account for our pre- theoretic talk and judgment concerning what agents believe, assert, and so on. Furthermore, positing such entities allows us to easily account for the validity of certain inferences. It allows us to easily account for the validity of the inference from ‘Anne believes that snow is white’ and ‘It is true that snow is white’ to the conclusion ‘There is something that Anne believes that is true’. If the embedded-that clause refers to an entity that plays the traditional role required of propositions, the inference is valid for the same reason that any existential generalization is valid. The theoretical reasons for positing the existence of propositions include the following. One might think that a satisfactory theory of meaning is going to need to appeal to propositions. The idea would be that the theorems of a satisfactory theory of meaning must tell us, for each truth-apt sentence, what proposition that sentence expresses. Another potential theoretical reason for positing propositions concerns their role in modeling the truth-conditions of sentences. When using possible world semantics to model the truth-conditions of sentences, the models give us theorems that tell us which worlds w a sentence S is true-at. Soames (2010) argues that the our understanding of what it is for a sentence being true-at-w must be ultimately analyzed in terms of our understanding of what it is for a proposition to be true simpliciter. Some creative philosophers have tried to make do without positing entities that play the traditional role required of propositions. Such creative philosophers have tried to show that our theoretical needs merely require us to think of propositions as tools that the theorist creates and uses to model linguistic and mental phenomena. According to this way of thinking about propositions, the theorist first finds some suitable mathematical entities like sets of possible worlds or ordered pairs consisting of objects and properties. Next, the theorist endows the suitable mathematical entities with truth-conditions and other intentional properties. After that, the theorist calls these entities ‘propositions’ and goes on to use these entities to model phenomena such as agents making assertions and agents having beliefs. As for the pre-theoretical reasons for positing propositions that play their traditional role, such creative philosophers will either ignore our pre-theoretical judgments or attempt to give a revisionary account of our talk so that we are not committed to entities that are things believed, asserted, and so on. 2 I do not include being the primary bearers of truth in the traditional role because I do not think our pre- theoretic judgments include that the truth-conditions all truth-apt entities, in some way, are determined the relations they stand in to proposition. Chapter 1, however, will cover theoretical reasons for taking propositions to be the primary bearers of truth. 9 For example, those that use propositions as mere tools for modeling the truth-conditions of assertive utterances often adopt a modified version of the account in Stalnaker (1976) where the expressing-relation is defined to hold between an utterance and a theoretical tool that has the same truth-conditions as the utterance. Or if we take propositions to be the interpreted logical forms of, for example, Larson and Ludlow (1993), we could model what a given utterance represents by stipulating that an utterance expresses a given logical form when the utterance and interpreted logical form are identical in what they represent. The recent revival in giving a metaphysics of propositions, as seen in King (2014b), Soames (2014b), Speaks (2014b), and Hanks (2011), is motivated by the shared belief that in order for propositions to meet our theoretical needs, propositions must also play their traditional role. According to the revival, propositions need to be more than mere tools created by the theorist in order to meet our theoretical needs. They need to play their traditional roles of being things believed, asserted, and so on in order to meet our theoretical needs. The revival is also motivated by the shared belief that all previous accounts of propositions, where propositions play their traditional role, are unsatisfactory. King, Soames, and Speaks argue that if propositions play their traditional role, then they also end up having troubling properties. 3 They appear to have their truth-conditions essentially and appear to have these essential truth- conditions without agents intentionally making them this way. But one might wonder why we cannot just rest satisfied with a theory where propositions are sui generis entities that have these properties. The following is one reason to worry about such a move. If we plausibly suppose that the truth-conditions that an entity e has depends on what e represents, the assumptions of the revival get us the further result that a proposition must necessarily represent what it represents. The assumptions also get us that a proposition necessarily represents what it represents not in virtue of the direct decisions of agents. Many representational things we bump into everyday represent what they do in virtue of the intentions of agents. Other things perhaps represent what they represent given causal facts. 4 Maybe it is even possible for us to decide to create an entity that has its representational properties necessarily. But we are left wondering how there could be an entity that necessarily represents what it represents and does so without us intentionally making it that way. 3 Hanks motivation is different than this. His motivation is to solve the long-standing puzzles of belief like Frege’s puzzle. However, this is also a motivation in Soames (2015). 4 Fodor (1990), Stalnaker (1987), and Dretske (1988) are some prominent examples of the common assumption that mental states are representational in virtue of causal facts. 10 Views such as Frege (1997c), Bealer (1998), Schiffer (2003), McGlone (2012), and Merricks (2015) simply take it as primitive that some abstract object essentially has its truth-conditions. A view like the one in Stalnaker (1976) identifies propositions with sets of possible worlds, but we are given no reason for thinking that a set of worlds represents anything without the help of a theorist. We ought to do better than this if we can. One reason for resisting positing entities that play the traditional role required of propositions is that it is hard to see how any entity could have the requisite properties to play this role. Again, how could any entity be representational without agents or contingent events making it that way? If we can solve the mystery of what propositions are, we can stop jumping through revisionary hoops to avoid positing them at all costs. 11 Chapter 1: The Way Propositions Need to Be Abstract Recent work by Jeff King, Scott Soames, and Jeff Speaks seeks to establish that we need propositions in our ontology. The guiding thought is that propositions are needed to play certain theoretical roles and that there are significant constraints on how propositions must be if they are to play these roles. This chapter attempts to achieve two main results. First, it argues, contra King, Soames, and Speaks, that propositions have to meet relatively minimal constraints in order to fulfill our theoretical needs. Second, it introduces and argues for a novel theory of propositions called Artifactualism that is well positioned to meet our theoretical needs without incurring the substantive and unnecessary theoretical commitments taken on by its competitors. Recent work by Jeff King, Scott Soames, and Jeff Speaks seeks to establish that we need propositions in our ontology. 5 The guiding thought is that propositions are needed to play certain theoretical roles and that there are significant constraints on how propositions must be if they are to play these roles. Specifically, we are told that propositions cannot be mere tools created by the theorist to model linguistic and mental phenomena. They argue that it is not enough for the theorist to find suitable mathematical entities like sets of possible worlds, to endow these suitable entities with truth-conditions, to call these entities ‘propositions’, and to use these entities to model features of assertions and beliefs. Our theoretical needs require more of propositions. To put a phrase to the way that King, Soames, and Speaks allege propositions to be: a conception of propositions is metaphysically robust iff it is one where propositions are not mere tools created by the theorist. After arguing that we need to posit metaphysically robust propositions, King, Soames, and Speaks go on to each give their own account of the underlying nature of propositions. I intend to show that when we evaluate the various arguments King, Soames, and Speaks put forward to establish the need for metaphysically robust propositions and see what our theoretical needs actually require, it turns out that their respective accounts of propositions take on 5 C.f. King (2007), King (2014a), Soames (1992), Soames (2002), Soames (2010), Soames (2014a), Speaks(2006), and Speaks(2014a). 12 unnecessary commitments. 6 King, Soames, and Speaks take our theoretical needs to require more of propositions than what is actually required of them. Consequently, they overlook options for accounting for the nature of propositions that are plausible and have minimal theoretical commitments. This chapter attempts to achieve two main results. First, it argues that propositions have to meet relatively minimal constraints in order to fulfill our theoretical needs. Second, it introduces and argues for a novel theory of propositions called Artifactualism that is well positioned to meet our theoretical needs without taking on unnecessary commitments. 1. Introduction There are a variety of views when it comes to what properties propositions need to have. The divergence in views mainly stems from disagreement as to what theoretical roles propositions are supposed to play, assuming that propositions are needed to play any theoretical roles at all. Propositions are often assumed to be the things believed, the things asserted, bearers of truth and falsity, and the meaning of indicative sentences. They are often posited as designata of certain terms in order to make sense of statements like ‘Anne believes that oceans once existed on Mars and this is exactly what Bill was asserting yesterday and we have evidence to support its truth’. On the other hand, there are ways of thinking of propositions where their theoretical roles are simply to be tools for modeling linguistic and mental phenomena like assertions and beliefs. For our purposes, there are three areas of interest concerning what properties propositions need to have in order to meet our theoretical needs. First, there are questions concerning the explanatory relationship between propositions and truth-apt entities. Call these questions of primality. Second, there are questions concerning what type of cognitive contact normal agents, i.e. non-theorists, have with propositions. Call these questions of cognitive contact. Third, there are questions concerning whether propositions themselves have truth-conditions, and, if so, what the modal status of these truth-conditions are. Call these questions of truth-conditions. Within each area of interest, there are various options distinguished by the level of theoretical commitment incurred. 6 This is not to claim that an account of propositions can ignore accommodating our pre-theoretical judgments. The focus here is just limited to theoretical reasons for positing propositions. 13 Understandably, many philosophers try to take on the fewest possible commitments in the three areas of interest. They look for creative ways of making do without metaphysically robust propositions and take ‘propositions’ to refer to theoretical tools that we use for modeling purposes. 7 To this end, such creative philosophers accept: No Primality: An explanation for why a given truth-apt entity e has truth-conditions need not appeal to e standing in some relationship to a proposition. No Cognitive Contact: Agents only began having cognitive access to propositions when theorists began modeling linguistic and mental phenomena. Non-theorist agents never designate or think about propositions. The creative philosopher accepts No Primality because if propositions are simply created for modeling purposes, they are at most used to model features of truth-apt entities. They do not explain why the truth-apt entities have certain properties. No Cognitive Contact is accepted for a similar reason. The notion of cognitive access that No Cognitive Contact appeals to will become more clear later in this chapter. The basic idea, though, is that having cognitive access to a proposition p amounts to p being “around” such that some agent, with the right abilities and in favorable circumstances, could think about p. The dialectical purposes of using this notion will soon become apparent. As for the question of truth-conditions, the creative philosopher can accept either: No Truth-Conditions: Propositions are not the kinds of entities that have truth- conditions. Contingent Truth-Conditions: Propositions have their truth-conditions contingently. Necessary Truth-Conditions: Propositions have their truth-conditions necessarily. Since the creative philosopher takes propositions to be mere theoretical tools, the question of truth-conditions is just a question of which theoretical tools get called ‘propositions’. If we stipulate that a set or some other mathematical entity is to have certain truth-conditions and call this entity a ‘proposition’, we will accept Contingent Truth-Conditions. If we stipulate that a set or some other mathematical entity is to represent the circumstances or conditions under which a 7 Davison (1968) and the projects it inspired are examples of trying to make do without metaphysically robust propositions. 14 sentence is true and call this entity a ‘proposition’, we will accept No Truth-Conditions because we will be treating propositions as the conditions under which something is true rather than as a thing that itself has truth-conditions. As for creating theoretical tools with necessary truth- conditions and calling them ‘propositions’, Soames (2010) contains a good account of how this might go. If our theoretical needs rule out either No Primality or No Cognitive Contact, then our theoretical needs require metaphysically robust propositions. Here is why. As we have just seen, if propositions are mere tools created by the theorist, then No Primality and No Cognitive Contact are true. And by definition, a conception of propositions is metaphysically robust iff it is one where proposition are not mere tools created by theorist for modeling purposes. It will be instructive to now go over the options that have the highest level of commitment within the three areas of interest. They are as follows: Immodest Primality: The truth-conditions of all truth-apt entities are explained by the relations these entities stand in to propositions. Major Cognitive Contact: By the time agents first started having propositional attitudes, agents must have had cognitive access to propositions. Necessary Truth-Conditions: Propositions have their truth-conditions necessarily. These three options make the task of finding suitable entities to play the role of propositions extremely difficult. Immodest Primality greatly constrains how we explain why a given proposition has the truth-conditions it has. The explanation cannot appeal to other truth-apt entities. Major Cognitive Contact requires that propositions were around and available as objects of thought at least by the time that there were agents with propositional attitudes. The respective accounts of propositions given by King, Soames, and Speaks are designed to satisfy the options that take on the highest level of commitment. They each give us an account that satisfies Necessary Truth-Conditions and Major Cognitive Contact. The accounts of Soames and Speaks also satisfy Immodest Primality. How King’s theory relates to questions of primality is complicated and need not concern us here. 8 8 Part of this complication comes from the fact that King (2014b) claims to give us an account where agents endow propositions with truth-conditions in virtue of the things agents do with sentence-types. 15 There is remaining middle ground between the options with the greatest and fewest commitments. For the questions of primality and cognitive contact, two relevant options are: Modest Primality: The truth-conditions that propositions have cannot be explained by the contingent relations they stand in to any other truth-apt entities. Minor Cognitive Contact: By the time agents first started making propositional attitude reports and using embedded that-clauses, 9 normal non-theorist agents must have had cognitive access to propositions. Unlike Immodest Primality, Modest Primality places a more moderate constraint on how we account for the truth-conditions of propositions rather than having propositions force their way into every explanation concerning truth-aptitude. Minor Cognitive Contact is the sort of thing we will be lead to accept if we think that making propositional attitude reports and using embedded that-clauses requires agents to think about and designate propositions. With the relevant options concerning the nature of propositions on the table, here is a more detailed account of the two main goals of this chapter. First, this chapter argues that our theoretical needs only require an account of propositions that satisfies Necessary Truth- Conditions and Minor Cognitive Contact. Our theoretical needs place no direct constraints on issues of primality. Second, this chapter introduces and argues for a novel of account of propositions, Artifactualism, that is well positioned to accommodate these theoretical needs without taking on unnecessary commitments. The first part of this chapter evaluates King, Soames, and Speaks’ attempts to establish the need for metaphysically robust propositions. I will assume that their attempts exhaust the possible ways of establishing the theoretical need for propositions. In light of this assumption, evaluating which of their attempts are successful will tell us what theoretical roles propositions must play and will thereby indicate the way propositions need to be. There are two main routes that King, Soames, and Speaks take to establish the theoretical need for metaphysically robust propositions. The first route concerns our theoretical needs in giving a satisfactory theory of meaning. The second route concerns accounting for phenomena 9 To keep things simple I am only talking about embedded that-clauses. But the same considerations apply to suitable pronouns and wh-interrogatives used in relative clauses. A sentence like ‘What Anne believes is the same thing that Bill asserted, and we have evidence to support its truth’ illustrates the relevant phenomena. 16 that embedded-that clauses exhibit. These routes are respectively referred to as the meaning route and the embedded-that clause route. 10 Section 2 argues that the meaning route is unsuccessful and places no constrains on how propositions must be. Section 3 argues that the embedded that- clause route does establish the need for an account of propositions that at least satisfies Minor Cognitive Contact and Necessary Truth-Conditions. Section 4 argues that, contra Soames, the embedded that-clause route does not establish the need for an account of propositions that satisfies Major Cognitive Contact. The second part of this chapter presents a novel theory of propositions that is positioned to meet our theoretical needs without taking on unnecessary commitments. Section 5 introduces Artifactualsim and shows how it satisfies Necessary Truth-Conditions and Minor Cognitive Contact. We see how Artifactualsim can be easily modified if independent considerations about the nature of propositional attitudes end up forcing us to accept Major Cognitive Contact. Section 6 argues that Artifactualism is preferable to competing theories of propositions. 2. The Meaning Route Soames and Speaks attempt to show that, when it comes to giving a satisfactory theory of meaning, our theoretical needs require an account of propositions that satisfies Immodest Primality. In saying that we want propositions to “play the role of being the information encoded by sentences relative to contexts and the things that determine whether a sentence relative to a context is true or false”, King also appears to think that our theoretical needs support Immodest Primality (King 2014a: p.5) But the main arguments for establishing Immodest Primality come from Soames and Speaks. The basic idea is that a satisfactory theory of meaning must have theorems that pair sentences with propositions that satisfy Immodest Primality. The propositions the sentences are paired with must explain why the truth-apt sentences have the truth-conditions that they have. I argue in this section that the meaning route is unsuccessful. Considerations about giving a satisfactory theory of meaning place no constraints on primality. 10 There is a third route that Soames (2014a) takes that basically claims we need propositions in order to explain how we get information from models that use the ‘true at’ operator to give the truth-conditions of sentences. However, it turns out that this route does not establish anything more than the embedded-that clause route and that the embedded-that clause route is successful – so we need not cover this route. 17 The following is a prima facie reason for doubting that we need an account of propositions that satisfies Immodest Primality. It might be the case, that as a matter of fact, we indicate the truth-conditions of a sentence relative to a context by saying what proposition it expresses. But which proposition is expressed by a sentence relative to a context is still ultimately going to be due to intentions, actions, beliefs, and conventions of agents. So, in other words, the intentions, actions, beliefs, and conventions of agents will ultimately be responsible for a sentence’s properties that determine which proposition it expresses relative to a context and will thereby determine the sentence’s truth-conditions relative to a context. If this is so, what theoretical benefit do we gain from supposing that there is a type of entity e for which the truth-conditions of all other entities asymmetrically depend on the truth-conditions of e? An entity that satisfies Immodest Primality would appear to be a theoretically dispensable middleman if a sentence’s truth-conditions relative to a context are still ultimately determined by the intentions, actions, beliefs, and conventions of agents. But Speaks and Soames seek to show that there is something flawed with this way of thinking. The meaning route begins by making some plausible assumptions about what a satisfactory theory of meaning must be like. A satisfactory theory of meaning θ for a language L must have two important properties. First, θ should obviously give us information about the meaning of the linguistic items that belong to L, including its words and sentences. Second, θ should be such that knowledge of θ is sufficient for understanding, i.e. being semantically competent, with L. This is because the theory should tell us the meaning of all the linguistic items in the language and so knowing the theory should suffice for knowing the meaning of all the items of the language. And knowing the meaning of the all the linguistic items in a language certainly suffices for understanding the language. Speaks (2014a), Speaks (2006), and Soames (1992) contain arguments to show that θ cannot have these two properties if either the theorems of θ merely give the truth-conditions of the sentences of L or if the theorems of θ merely amount to claims that linguistic items of L have the same meaning as the linguistic items of another language. Speaks and Soames argue that if θ is either of these ways, then θ will both give us insufficient information about the meaning of the linguistic items of L to count as a theory of meaning and will not be such that understanding θ suffices for understanding L. Soames and Speaks then reach the conclusion that propositions must satisfy Immodest Primality in the following way. Since they assume that a theory of meaning must do more than 18 give truth-conditions, Speaks and Soames argue that a satisfactory theory of meaning must give us theorems of the form S means in L that p where that p designates a proposition. They then argue that these propositions cannot be such that they were endowed with truth-conditions by agents; otherwise, the theorems will merely provide translations into a technical artificial language. Accordingly, they think these propositions must have their truth-conditions for different reasons than why truth-apt sentences have truth-conditions and that the truth-apt sentences get their truth-conditions by standing in the expressing relation to propositions. So, they think that we need propositions and propositions must satisfy Immodest Primality. I will now argue that we do not need propositions that satisfy Immodest Primality in order to give a satisfactory of meaning. I am first going argue that Soames and Speaks are right to claim that we need a theory of meaning that does more than give the truth-conditions of the sentences L and does more than pairing up the linguistic items of L with linguistic items, from another language, that have the same meaning. However, I will then show that it is possible to give a satisfactory theory of meaning that does more than this without positing propositions that satisfy Immodest Primality. A theory of meaning for L must do more than give the truth-conditions of the sentences of L. There is quite a bit of literature meant to establish this. 11 The most important problem comes down to the simple fact that you can know that (the English sentence ‘Snow is white’ is true iff snow is white) without knowing what the sentence means. Even if you know that (if we hold fixed the actual semantic properties of ‘Snow is white’, (then the sentence is true at a world w iff snow is white in w)), you could still not know what the sentence means. In fact, no matter how much you know about when ‘Snow is white’ is true or false, you could be mistaken about what it means. A theory of meaning that merely gives the truth-conditions of sentences does not tell us what an agent knows when she knows the meaning of a linguistic item of L. Such a theory gives us insufficient information about the meanings of the linguistic items of L to count as a satisfactory theory of meaning. A theory of meaning for L must do more than pair the linguistic items of L with items of another language L* such that the paired items have the same meaning. You can know that, for example, two sentences of a different language have the same meaning without knowing what either of the sentences mean. Such “Translation-manual” theories of meaning fail to present us with what an agent knows when she knows the meaning of a linguistic item of L. 11 See, for example, Speaks (2006), Soames (1992), and Foster (1976). 19 Now there have been many attempts to show we can modify a meaning theory that merely gives truth-conditions or a translational theory of meaning so that it is satisfactory. 12 But it is reasonable to grant that if the theorems of θ merely give the truth-conditions of the sentences of L or state that the linguistic items of L have the same meaning of linguistics items from other languages, then θ is not a satisfactory theory of meaning. We can know the truth-conditions of a sentence of S of L and know that it has the same meaning as another sentence S’ from L’ and still not know what S means. What we should not grant is that positing propositions that satisfy Immodest Primality is the only way to get a satisfactory theory of meaning. Here is a plausible alternative to positing propositions that satisfy Immodest Primality. Imagine that we learned more about what basic properties a sentence needs to have in order to count as expressing the proposition it expresses and having the truth-conditions it has. For example, imagine we learned that if S has the truth-conditions of being true iff o is F, then this is because S represents o as being F. Not only is this a plausible claim about the nature of how truth- conditions are determined, it is something that Soames and King appear to endorse. 13 Now, assuming it is true that S’s truth-conditions are determined by the way it represents things as being, imagine we make a theory of meaning θ where θ’s theorems tells us what each sentence of L represents as being what way. For English, we would get the theorem that ‘Snow is white’ represents snow as being white. This alternative avoids the problems of truth-conditional and translational theories of meaning. Could we know the English sentence ‘Snow is white’ represents snow as being white yet still not know its meaning? No argument has been given to show this. When we are just dealing with truth-conditions, it was possible for someone to know the truth-conditions of ‘Snow is white’ but think that it meant snow is white and arithmetic is incomplete. But if we know that ‘Snow is white’ does not represent arithmetic as being any way but does represent snow as being white, we know it does not mean that snow is white and arithmetic is incomplete. Now this alternative might require adjustment. It could very well turn out that it is not quite right that a sentence’s truth-conditions are determined by what things it represents as being what way. Or it could turn out that we can present a case where knowing the way an indicative sentence relative to a context represents things as being is compatible with not knowing its meanings. However, and this is important, once we have discovered the correct theory as to how 12 See, for example, Higginbotham (1992). 13 C.f. King (2007: p. 59) and Soames (2014c: p.235). 20 an entity’s truth-conditions and other semantic properties are explained by its more basic properties, our theory of meaning just needs to tell us the nature of linguistic items in terms of these more basic properties and show us how these more basic properties of the linguistic items account for why the linguistic items have their semantic properties. 14 So it appears that there will always be some plausible alternative to positing propositions that satisfy Immodest Primality. As long as we plausibly suppose that which proposition is expressed by a sentence is ultimately explained by the intentions, actions, beliefs, and conventions of agents, then indicating which proposition a (complete) 15 sentence expresses will only being doing middleman-work in a theory of meaning. The sentence expressing a proposition will amount to the sentence having certain semantic properties that include its truth-conditions and other intentional properties. That the sentence has these semantic properties will be explained in terms of its other more basic properties. It will have these more basic properties in virtue of the intentions, actions, beliefs, and conventions of agents. A theory of meaning θ for L just needs to tell us for a given linguistic item i of L what i’s relevant basic properties are. As we said earlier, if one of these basic properties is the way i represents things as being, the theory of meaning will tell us what i represents. And although it might be only roughly correct that the truth-conditions of an entity are determined by the way the entity represents things as being, we can already see that a theory of meaning that told us what each sentence of L represents would do a pretty good job giving us information about the meaning of the sentences of L. This is not to deny that there are certain sub-sentential items for which a theory of meaning is going to ultimately require positing propositions with certain properties. We will cover this in the next section. Specifically, there ends up being good theoretical reasons for thinking that a sentence like ‘Anne believes that snow is white’ represents Anne as standing in relation to a proposition. But we have not seen any theoretical reason for thinking that a satisfactory theory of meaning must have theorems that pair complete sentences with propositions that satisfy Immodest Primality. The meaning route places no constraints on primality because a successful theory of meaning does not require theorems that pair up complete sentences and propositions. One might worry that a satisfactory theory of meaning will necessarily appeal to the beliefs and intentions of agents and thereby put constraints on how propositions must be. Eventually we 14 Thanks to Scott Soames for pointing out to me in conversation that it would not be enough for the basic properties to merely determine the semantic properties, if the basic properties still failed to provide an explanation for why the linguistic items have their semantic properties. 15 Why this qualification is necessary will become clear momentarily. 21 are going to need to address the fact that independent considerations about the nature of propositional attitudes might lead one to take various positions on cognitive contact. For example, if we make the very strong assumption that an agent A believing that S requires A to be able to think about the proposition expressed by ⎡S⎤, 16 this alone entails that Major Cognitive Contact is true. This would entail that any theory that appeals to the propositional attitudes of agents in giving an explanation, requires admitting propositions into our ontology that satisfy Major Cognitive Contact. We are going to put aside such issues for now but we will return to them shortly. In summary, we have no reason to think that a satisfactory theory of meaning must contain theorems that pair sentences with propositions and so the meaning route does not place any constraints on primality. Speaks and Soames attempted to show that a satisfactory theory of meaning must have theorems that pair sentences with propositions that satisfy Immodest Primality. However, we have been given no reason to doubt that the sentence a proposition expresses will always be explained by its more basic properties, and so the truth-conditions and other intentional properties of a sentence will always be explained by more basic properties. As long as our theory of meaning for L explains the nature of L’s linguistic items in terms of these more basic properties, it will be at least as satisfactory as a theory of meaning that pairs sentences with propositions. Consequently, a satisfactory theory of meaning need not have theorems that pair sentences with propositions, and so a satisfactory theory of meaning places no constraints on how propositions need to be. 3. The Embedded That-Clause Route In this section, we see that, given certain phenomena that embedded that-clauses exhibit, there is a theoretical need for metaphysically robust propositions. There is a theoretical need for an account of propositions that satisfies Necessary Truth-Conditions and Minor Cognitive Contact. Speaks (2014b) argues that we need propositions to account for the validity of arguments such as: P1) Violet believes that the sky is blue. 16 Here I use ‘S’ as a schematic letter for a sentence although it was previously used as a variable that ranges over sentences. Throughout the rest of the chapter, it should be clear whether ‘S’ is being used as a schematic letter or variable. 22 P2) It is true that the sky is blue C) Therefore Violet believes something true. The logical form of the argument appears to be: P1) R(o, o*) P2) F(o*) C) ∃x(R(o, x) & F(x)) And what else can o* be besides a proposition? Although one could try to give a revisionary account of the logical form of this argument, Speaks shows that all attempts at a revisionary account are highly problematic. 17 So barring accepting a highly problematic and revisionary account, we need to posit propositions as the designata of embedded-that clauses and this is if the first step towards establishing the need for metaphysically robust propositions. Perhaps the most promising way to block the first step in establishing the need for metaphysically robust propositions is to claim that a sentence like ‘Violet believes the sky is blue’ does not represent Violet as standing in a relation to a proposition, but rather represents Violet as being such that she is in a belief state that has the same truth-conditions and meaning (i.e. content) as some other entity that means that the sky is blue. However, it cannot be right that the belief report represents Violet as being in a belief state that has the same truth-conditions as the sentence ‘The sky is blue’. The sentence could have meant that the sky is red, but in no possible world is Violet’s belief that the sky is blue true iff the sky is red. To get around this problem, Speaks notes that one could give a different revisionary account where ‘Violet believes the sky is blue’ represents Violet as being in a belief state that has same the truth-conditions/content as the sentence ‘The sky is blue’ has in the actual world @. But as Soames (2002) argues, this has untenable consequences for sentences like ‘Amelia knows that Violet believes that the sky is blue’. This sentence could be true at a world w where Amelia is not able to have any thoughts about @. If the sentence is true at w, then, in w, Amelia believes-that- Violet-believes-that-the-sky-is-blue. The revisionary account we are considering entails that Amelia, in w, is therefore in a belief state that is about @. But, in w, Amelia cannot think about @, so this revisionary account does not work. 17 C.f. Speaks (2014b: pp.11-19). 23 There are other ways to try to make do without positing propositions as the designata of embedded that-clauses but they also come with substantial, revisionary, and highly problematic commitments. 18 This shows that there is indeed a theoretical need for positing propositions that can at least be the designata of embedded that-clauses. It does not show that it is our only option. It just shows that our theoretical needs would be well, if not be best, served by propositions. The next task is to show that if we need propositions to be the designata of embedded that-clauses, then propositions must be metaphysically robust. I will now argue that if propositions are the designata of embedded that-clauses, then an account of propositions must satisfy Necessary Truth-Conditions and Minor Cognitive Contact. Starting with truth-conditions, if propositions are designated by embedded that-clauses, then ‘John believes that snow is white’ relates John to a proposition – namely John’s belief. John truly believes that snow is white only if John’s belief (the proposition he believes) is true. John’s belief is true iff snow is white. The proposition has the property of being-an-x-such-that-x-is-true-iff- snow-is-white. Importantly, the proposition has this property necessarily. If the proposition did not have this property necessarily, then: 1) There would be some world w such that an agent A in w truly believes that snow is white 2) A’s belief (the proposition that snow is white) is true in w 3) Snow is not white in w. This is absurd. There is no world where both someone truly believes that snow is white and snow is not white. Consequently, if propositions are designated by embedded that-clauses, they have their truth-conditions necessarily and we need an account of propositions that satisfies Necessary Truth-Conditions. As for Minor Cognitive Contact, if propositions are the designata of embedded that-clauses, then agents designate propositions when they use embedded that-clauses. This happens for example, when agents give propositional attitude-reports. To establish the need for Minor Cognitive Contact I need to argue that agents designating propositions requires that they have cognitive access to propositions. A creative non-committal philosopher might push back here and try to give a deflationary account of what it is to designate a proposition. But I claim that this will not work. Long before there were any theorists around, agents correctly made valid inferences involving existential generalizations with embedded that-clauses. For example, they thought that John believes that snow is white and then made the valid inference to the thought that there is something that John believes. If embedded-that clauses designate propositions, the reason these 18 C.f. Speaks (2014a). 24 inferences were valid is just the same reason that any existential generalization is valid. But it is not enough to be able to give a model where these inferences share a common structure with other existential generalizations. Non theorist-agents knew that these existential generalizations with embedded that-clauses were valid. How did they know? They recognized that these inferences shared something in common with inferences like: the Sun is yellow, therefore something is yellow. They recognized that the embedded that-clause designates some thing, and they had to be able to recognize when two embedded that-clauses did or did not designate one and the same thing. This entails that these agents could think about the designata of embedded that-clauses. And since we are assuming that such designata are propositions, this entails that these non-theorist agents could think about propositions. If they could think about propositions, then they had cognitive access to propositions. So we need an account of propositions that satisfies Minor Cognitive Contact. In summary, there is a theoretical need to posit propositions as the designata of embedded that-clauses. Attempts to make do without assuming this have very unwelcome results. Once we assume that propositions are the designata of embedded that-clauses we need to posit metaphysically robust propositions. We need to posit metaphysically robust propositions because the success of the embedded that-clause route requires propositions to satisfy Minor Cognitive Contact. 4. Why Minor Cognitive Contact is Enough This section argues that the successes of the embedded-that clause route does not require us to take on a stronger cognitive contact option than Minor Cognitive Contact. I will first explain why independent considerations about the nature of propositional attitudes along with the embedded that-clause route might lead one to conclude we need something stronger than Minor Cognitive Contact. In doing so, I lay out some options for accounting for the nature of propositional attitudes that are distinguished by their various levels of theoretical commitment. I then address and respond to particular arguments in Somes (2010) that are supposed to show that we need more than Minor Cognitive Contact. What reasons could there be for thinking that the embedded-that clause route entails that propositions must satisfy Major Cognitive Contact? Well, before going down the embedded-that clause route, there were at least two extreme options open to us concerning the nature of propositional attitudes. For the following options, let ‘φ’ be a schematic letter for a propositional 25 attitude verb like ‘believes’, ‘hopes’, and so on. Let ‘S’ be a schematic letter for a suitable indicative sentence: Inflationary Attitude Relation: The attitude report ⎡A φ’s that S⎤ represents A as standing in a relation R to the proposition p that is expressed by ⎡S⎤. For the attitude report to be true, A must stand in R to p, and standing in this relation requires that A have cognitive access to p. No Attitude Relation: The attitude report ⎡A φ’s that S⎤ does not represent A standing in a relation to a proposition. The success of the embedded that-clause route rules out No Attitude Relation because the route entails that propositions are designated by embedded that-clauses, so a sentence like ‘Violet believes that the sky is blue’ represents Violet as standing in a relation to a proposition. If Inflationary Attitude Relation were the only other option on the table, then the embedded that- clause route would entail Inflationary Attitude Relation. And, in turn, Inflationary Attitude Relation would entail that we need an account of propositions that satisfies Major Cognitive Contact since agents merely having propositional attitudes would require agents to have cognitive access to propositions. But once again there is a middle path. The basic idea is that having a propositional attitude does consist in standing in a relation to a proposition but the relevant relation between agent and proposition is deflationary: Deflationary Attitude Relation: The attitude report ⎡A φ’s that S⎤ represents A as standing in a relation R to the proposition p that is expressed by ⎡S⎤. For the attitude report to be true, A must stand in R to p, but standing in this relation does not require that A have cognitive access to p. Accounts of propositional attitudes that accept Deflationary Attitude Relation include Stalnaker (1987) and Loar (1981). On these accounts there is a strong analogy between numbers and propositions. In Stalnker’s words: The analogy between propositions and numbers suggests that there may be further alternative strategies for explaining how a person can be related to a proposition… What is it about such physical properties as having a certain height 26 or weight that makes it correct to represent them as relations between the thing to which the property is ascribed and a number? The reason we can understand such properties––physical quantities––in this way is that they belong to families of properties which have a structure in common with the real numbers… That, I think, is all there is to the fact that weights and other physical quantities, are, or can be understood as, relations between physical objects and numbers. (Stalnaker 1987: p. 9). When we think of propositional attitude reports this way, we are lead to a view where an agent A can believe that o is F without standing in any immediate causal or cognitive contact with the proposition that o is F. A just needs to be in a mental state that has some sort of similarities to the proposition we use to pick out that mental state. It is important to see that the success of the embedded that-clause route does not immediately threaten this way of thinking about propositional attitudes. We need not think of propositions as mere theoretical tools in order to think that having a propositional attitude does not require standing in a very robust relation to a proposition. Under this deflationary conception, propositions are tools that normal, non-theorist, agents use to talk and think about mental phenomena. However, the argument in Soames (2010: pp. 69-98) is meant to show that even if we start out accepting something like Deflationary Attitude Relation, other plausible assumptions about propositional attitudes will still lead us to conclude that we need an account of propositions that satisfies Major Cognitive Contact. Soames argues as follows. He claims that someone that accepts a view like Deflationary Attitude Relation must also accept that: the truth of ⎡A φ’s that o is F⎤ is to be explained in terms of A (mentally) representing o as being F (Soames 2010: pp.81- 83). For example, if ‘Violet believes that the sky is blue’ is true, this requires Violet being in some state that has to do with Violet representing the sky as being blue. Perhaps it requires her to be disposed to do this. The idea behind Soames’ assumption is that the common properties shared between propositions and attitudes/mental states must be representational properties. Soames then goes on to argue that complex sentences like ‘The sky is blue and grass is green’ represent propositions as being ways (Soames 2010: pp. 83-86). So the truth of ‘Violet believes that the sky is blue and grass is green’ needs to be explained in terms of Violet representing the propositions that the sky is blue and that grass is green as being a certain way. In order for an 27 agent to represent or be disposed to represent these propositions as being some way, an agent must have cognitive access to these propositions. This entails that, at least by the time agents started having propositional attitudes where complex propositions were the object of the attitude, agents had cognitive access to propositions. This result alone does not entail that we need an account of propositions that satisfies Major Cognitive Contact. But it is hard to see how agents could begin to form complex attitudes if we accept Soames’ assumptions. At one point, agents only had beliefs as complex as believing that the sky is blue. Then they went on to have more complex beliefs like believing that the sky is blue and grass is green. Having these complex beliefs requires being able to represent propositions as being ways. In order to do this, agents must have had cognitive access to propositions. So we have good reason to think that by the time agents started having any propositional attitudes they had cognitive access to propositions even if they did not yet have the cognitive means to think about propositions. Once they had the cognitive means to think about propositions, they went on to have complex propositional attitudes. Other complex propositional attitudes that Soames appeals to include attitudes that are about the propositional attitudes of others (Soames 2010: pp. 86-88). For example, suppose Amelia believes that Violet believes that the sky is blue. Soames’ assumptions along with the embedded that-clause route get us that Amelia must be able to represent Violet as standing in a relation to the proposition that the sky is blue. So Amelia believing what she believes requires Amelia to be able to represent a proposition. And this therefore requires her to have cognitive access to a proposition. To respond to Soames’ argument, we should first examine his assumption that the only way to accept Deflationary Attitude Relation is to accept that the truth of ⎡A φ’s that o is F⎤ requires A to be able to represent o as being F. This basically amounts to assuming that the commonality between propositions and the mental states/attitudes we pick out with propositions must be representational. But this is not the place to argue against this assumption, and the assumption is stronger than what Soames actually needs. Soames just needs to assume that, extreme cases aside, 19 whenever an agent A believes that S, A must have cognitive access to the things that ⎡S⎤ is 19 The caveat is to allow for extreme cases such as where Aristotle believes all philosophers are egomaniacs but does not know of a particular philosopher Phil. We might be still lead to accept, in some cases, that Aristotle believes of Phil that Phil is an egomaniac. Hawthorne and Manley (2012: 50-53) gives a similar example and some reasons for being sympathetic to such a view. Whatever we say about these deviant cases, it is plausible that in most cases where A has a propositional attitude, and that o is F is the object of 28 about. So, whenever an agent believes that the sky is blue, the agent must have cognitive access to the sky and the property of being-blue. This assumption seems quite plausible and we will proceed by taking it to be true. Soames’ other assumptions still appear to get the result that we need an account of propositions that satisfies Major Cognitive Contact. How else could agents start having complex propositional attitudes? If ‘Snow is white and grass is green’ represents propositions as being ways, then this sentence is about propositions. So we still appear to get the result that believing snow is white and grass is green requires having cognitive access to propositions. To show that we do not need to assume something stronger than Minor Cognitive Contact, I need to show that before there were agents that used embedded that-clauses, agents did not need to have cognitive access propositions. So I need to show that we can explain how pre-that-clause- using-agents had all the propositional attitudes we think they had without supposing that this required pre-that-clause-using-agents to have cognitive access to propositions. For sake of simplicity, I will call all pre-that-clause-using-agents, primitive agents. To argue against Soames and show that primitive agents did not need cognitive access to propositions, we could try to deny his assumption that ‘Snow is white and grass is green’ represents propositions as being some way. I think there are good reasons to deny this, 20 however such a move will not help us for sentences like ‘Amelia believes that Violet believes that the sky is blue’. The embedded-that clause route established that such sentences do represent propositions as being ways. ‘Violet believes that the sky is blue’ represents Violet as standing in a relation to a proposition. So, our assumptions entail that Amelia must have cognitive access to a proposition. We cannot deny that we modern humans use our cognitive access to propositions to think about the attitudes/mental states of others. But perhaps more primitive agents had other ways of thinking about the attitudes/mental states of other agents. The following is an account of how primitive agents can have all the propositional attitudes we think they can have without having cognitive access to propositions. Suppose that primitive humans did not think of the attitudes/mental states of their peers in terms of their peers standing in a relation to a proposition. We modern humans often pick out these attitudes/states in terms of the attitude, A must stand in some important cognitive relationship with o and being F in order to have the attitude. This is not to assume that A must be acquainted with o or stand in a causal relation with o in all of these cases. We can allow that complex agents can think thoughts about, and generally cognize o, merely by thinking of some o as the φ. 20 These reasons are covered in chapter 4. 29 them standing in relations to propositions, but early humans did not do this, and we in fact do not always do this. I can think of your belief as the reason or thing that is making you act strangely. For example, a mom might fear that a stranger across the street will abduct her child. The child can start to have an attitude towards her mom’s propositional attitude by thinking of her mom’s propositional attitude as the thing that is making her mom act strangely. 21 We are able to recognize different types/kinds of attitudes/mental states that others exhibit by their overt behavior. Perhaps we can even de re cognize the underlying attitude/mental state by picking it out as the thing responsible for certain overt behavior. It is plausible that pre-linguistic humans thought about the attitudes of others in this way that does not involve cognizing propositions. Imagine a scenario where we are primitive hunters. Why did the lead hunter of our tribe raise his hand? Perhaps he wants us to be quiet. We can cognize this attitude as the thing that made the lead hunter raise his hand. Why did the lead hunter just shiver? Perhaps he believes a deadly boar is near. We can cognize this attitude as the thing that made the lead hunter shiver. If pre-linguistic humans thought about the attitudes of others this way, then the fact that pre-linguistic humans had propositional attitudes about the attitudes of others does not force us to give up on Minor Cognitive Contact. Turning back to other complex attitudes, it is plausible that there is more than one way to think complex thoughts such as thoughts we pick out with sentential operators. Suppose my dog believes that either the ball is under the armchair or the ball is still in my hand. If possible, it would be good to avoid claiming that my dog can think about propositions. Perhaps my dog is thinking of the ball, rather than two propositions, as being a certain way. He is thinking of the ball as either-being-under-the-armchair-or-still-in-my-hand. My dog does not have the concept of what is to be an armchair. But perhaps he can still cognize this complex property by cognizing it some other way. If primitive agents can cognize such complex properties, then they can have such complex attitudes without being able to cognize propositions. It is hard to see how taking primitive agents to be able to cognize complex properties is more commitment laden than taking primitive agents to be able to cognize propositions. Of course we could also simply deny that primitive agents had such complex attitudes, but this should be a last resort. So, with some plausible assumptions, we can explain how primitive agents can have all the propositional attitudes we attribute to them without being committed to positing that such 21 In a similar spirit, Butterfill and Apperly (2013, p. 607) claims that agents with limited conceptual sophistication can track the belief states of others without being able to cognize “propositional attitudes as such.” 30 primitive agents have cognitive access to propositions. Now a lot of details still need to be worked out if we are to be sure of the assumptions that get us this result. And these details will be worked in chapter 3. But the point at hand is that Major Cognitive Contact is no sure thing. Soames gave us an argument that the success of the embedded-that clause route ultimately requires that we accept Major Cognitive Contact. It seemed that even if we accept Deflationary Attitude Relation, we were still forced to accept Major Cognitive Contact. Phenomena involving primitive agents having complex attitudes seemed to require that primitive agents have cognitive access to propositions. But as we just saw, there are plausible ways of explaining how primitive agents could have all the complex attitudes we think that they have without assuming that they have cognitive access to propositions. This means that our theoretical needs alone do not entail that we need anything stronger than Minor Cognitive Contact. Now it could very well turn out that accounting for the nature of propositional attitudes still requires something stronger than Minor Cognitive Contact. But it would be clearly worthwhile to try to give a minimal account of propositions that satisfied nothing stronger than Minor Cognitive Contact. 5. Artifactualism – A Minimal Theory of Propositions In the last three sections we saw that our theoretical needs require an account of propositions that satisfies Necessary Truth-Conditions and Minor Cognitive Contact. Our theoretical needs place no constraints on primality. Our theoretical needs alone are not enough to establish the need for Major Cognitive Contact. This section introduces and argues for Aritfactualism. After presenting and motivating the view, we see how it stacks up against some competing views. Artifactualism takes a distinctive stance on the relationship between sentences and propositions. Consider the following two token sentences written on this page: Snow is white | Snow is white. The first token is written in the font Courier and the second is written in Comic Sans. Despite their orthographic differences, it is uncontroversial that the two sentences share common properties and belong to a common type – namely the sentence-type ‘Snow is white’. Now consider the following two sentence tokens: La nieve es blanca | Schnee ist weiß. It is also uncontroversial that these tokens also have something in common despite their orthographic differences. Those sympathetic to positing propositions normally explain this by claiming that the two tokens share something in common because they both express the proposition that snow is white. But Artifactualism is different. Artifactualism claims that the proposition that snow is white just is the type that is common to these sentence tokens. The 31 proposition that snow is white is not some extra entity that serves to explain what these tokens have in common. Rather, the proposition is the type common to all sentence tokens that mean that snow is white. So Artifactualism presents a radically different view of the relationship between sentences and propositions. Standing in the expressing relationship to the proposition that snow is white does not explain what is common to all sentence tokens that mean that snow is white. As a result, Artifactualism owes us an independent account of what all these sentence tokens have in common. And this account is how Artifactualism comes by its name. What these sentence tokens all have in common is their unique representing function. Tokens of ‘Snow is white’ and ‘La nieve es blanca’ all have the function of representing snow as being white. The proposition that snow is white just is the type that is common to all the things that have this unique representing function. In general, Artifactualism claims that propositions are types whose instances all share a unique representing function. In most instances, with a possible caveat that I will consider shortly, the things that have representing functions are artifacts like sentences, and this is how Artifactualism earns its name. Propositions end up being types that have representational artifacts, like sentence-tokens, as instances. Artifactualism only needs to make the incontrovertible assumption that there are indeed artifacts that have the function of representing things as being ways and that there are types/kinds that these things belong to. No other metaphysical baggage is necessary. And admitting that there are functional-types is really not such a shock when we recognize that we already do this in using sentence-types and word-types in our linguistic theorizing. 22 Some linguistic types are distinguished by their phonetic and morphological properties but others are distinguished by their function. The conjunction is a linguistic type whose instances all have a common function but need not have common phonetic or morphological properties. Artifactualism simply claims that there are functional-types whose instances all share a common function of representing but need not share common phonetic or morphological properties. 22 We often attribute properties to linguistic-types in our semantic theorizing without seriously investigating the nature of these types/kinds. Kaplan (1990) and Wetzel (2009) are two places where these problems are actually addressed. Ali Khalidi (2015), Thomasson (2003), Searle (1995), and Boyd (1991) argue for various views of kinds that allow for kinds of artifacts. Hawthorne and Lepore (2011: 480-485) present some worries for these inclusive approaches. 32 To show that Artifactualism satisfies our theoretical needs, we need to first show that it satisfies Minor Cognitive Contact. We can plausibly assume that the functional-types that are propositions at least exist whenever their instances exist. 23 So whenever agents started using sentences that had the function of representing things as being ways, corresponding propositions came into existence. Whenever agents used an embedded-that clause, the proposition that this embedded-that clause is an instance of was around for agents to designate and think about. So, Artifactualism satisfies Minor Cognitive Contact. If it turns out that independent considerations about the nature of propositional attitudes require us to accept Major Cognitive Contact, we can easily modify Artifactualism. We can assume that the functional-types that are propositions do not just have representational tools/artifacts as instances. They also have biological/mental states as instances. That snow is white would have instances that include, not only sentence tokens, but also token biological/mental states that have the function representing snow as being white. 24 And one could take a view of propositional attitudes where to believe that snow is white partially consists in being in a mental state that is an instance of the proposition that snow is white. Turning to Necessary Truth-Conditions, it might seem a bit strange at first to think that a functional-type could have any truth-conditions. On the other hand, we already take sentence- types to have truth-conditions. Without going into all the details, here are some reasons why it is plausible to suppose that the functional-types that are propositions both have truth-conditions and have them necessarily. According to Artifacualism, the proposition that snow is white is the functional-type whose instances all have the common function of representing snow as being white. In fact, in order for anything to be an instance of this type, that thing must have this function. Functional-types have membership criteria that require their instances to have certain functions. And it is because of these membership criteria that the type counts as having certain properties. The letter ‘S’ is a tool-type whose instances must all be curved. This is why we say that the letter ‘S’ is curved even though the reason for which it counts as being curved is not the same as the reason that tokens of the letter ‘S’ are curved in the normal sense. Similarly, all instances of 23 However, we do not need to assume that the proposition exists only if its instances exist. We could adopt some combinatorial principle such that if an instance of one proposition exists, many other propositions exist. 24 Millikan (1987) is an example of a prominent view that holds that certain mental states have the function of representing. 33 that snow is white must be such that they are true iff snow is white. This is why the functional- type that is that snow is white counts as being true iff snow is white. The membership criteria of this functional-type gives it this property. Since functional-types have their membership criteria necessarily, propositions have their truth-conditions necessarily. There are certainly worries for this way of explaining why Artifactualism satisfies Necessary Truth-Conditions. One worry is that propositions (and sentence-types too) seem to end up having truth-conditions in a derivative sense whereas sentence tokens have truth-conditions in the normal sense. I take this to be an acceptable result, especially because our theoretical needs place no constraints on primality. But there is still the worry of how exactly these functional-types count as having truth- conditions in a derivative sense. This worry is an instance of a more general worry concerning the properties that types and kinds have. The dog is an animal that has four legs, but does the Dog, a kind of animal, really have four legs? For the cases where it seems implausible to claim that the way in which a kind is F is one and the same as the way in which its instances are F, there are a range of plausible accounts. One option is to claim that there is one property of having-four-legs, but particular dogs with four legs stand in a different relation to this property than the kind dog. 25 Another option is to claim that there is a derivative property that we also call ‘having four legs’ and the kind dog has this property whereas particular dogs do not. 26 For sake of simplicity, we will assume the first option is the correct one. We will say that the kind dog counts as having four legs and mean by this that it stands in a different relation to the property of having-four-legs than the relation that particular dogs stand in to this property. Accordingly, the musical note standard A counts as having a frequency of 440hz. It essentially counts as having this frequency. 25 The ontology of Lowe (2006) is one way to spell out what such a distinction might amount to. In Lowe’s ontology, the proposition, as a kind, would stand in a relation to the universal of being-true, whereas instances of the proposition would stand in the same relation to instances of the universal being-true. All we need to assert for this option, though, is that sentences of generic form represent the denotata of the noun-phrase and verb-phrase, respectively, as standing in a different relation than non-generic sentences. 26 Yet another option would be to try to analyze our claims such as ‘Standard A has a frequency of 440hz’ as not making any reference to a kind. See Carlson and Pelletier (1995) for a collection of articles on the semantics of generics that share the common view that we ought to treat such locutions, when used this way, as attributing properties to a kind/type. There might be reasons for analyzing these sentences as having a more complicated logical form, but no potentially successfully proposal has been given that completely does away with an appeal to kinds/types. We would have to be highly revisionary in order to analyze away all references to types/kinds, and so we will assume that some of our judgments are indeed about the properties of kinds/types. 34 Although explaining why kinds essentially have the property of being such that their instances are all a certain way is beyond our reach for now, we can explain why kinds essentially have properties like counting-as-having-a-frequency-of-440hz or counting-as-having-four-legs. No contingent fact about the actual world explains this. Rather, the musical note standard A essentially counts-as-having-a-frequency-of-440hz because all instances of this kind must be this way in order to be an instance of this kind. As a first pass, our claim in general would be that: A kind k essentially counts-as-being-F iff k is essentially such that, for all x, x must be F in order to be an instance of k. Now Soames (2014c) and Soames (2015) appear to assume something similar in trying to show that act-types essentially count as representing things as being ways. Contra such views, Caplan et al. (2013) argues that if this general claim about the properties of kinds were true, then every kind k would count as having the property of being an instance of the kind k. For example, every particular dog is such that it is an instance of the kind dog and must be this way to be an instance of the kind dog. And so our general claim about kinds entails that the Dog is a kind of animal that counts as being an instance of the kind dog. This is absurd. To get things right we need to modify our general claim. But before we do that, it is worth pointing out that we are not trying to find a way to make an ad hoc claim about the nature of types/kinds true. We are trying accommodate some pre-theoretic data. We say that the dog is an animal, that the triangle is a shape, that the letter ‘S’ is curved, and so on. We can all agree that the reason the dog counts as being animal, in the sense of ‘counting’ we have in mind, is roughly because all instances of this kind must be this way. We are just trying to be more precise on how exactly a kind must be in order to count-as-being-F. With that in mind: A kind k counts-as-being-F iff k is essentially such that, for every one of its instances o, o must be F in order to be an instance of k and the fact that o is F (partially) explains why o is an instance of k. Apples being an animal partially explains why Apples is an instance of the kind Horse. However, Apples being an instance of the kind horse does not explain why Apples is an instance of the kind horse. The fact that S does not explain the fact that S. Furthermore, since Apples being a horse is likely no different from the fact that Apples is an instance of the kind horse, Apples being a horse does not explain why Apples is an instance of the kind horse. All of this is why the Horse counts 35 as being a kind of animal but neither counts as being an instance of the kind horse nor counts as being a horse. The functional-type that is the proposition that snow is white essentially and necessarily counts as having the truth-conditions it has for a similar reason. Its instances must all represent snow as being white and have the property of being-true-iff-snow-is-white. The instances being this way explain why they are instances of the proposition that snow is white and not instances of say, the proposition that grass is green. In general the functional-types that are proposition necessarily count as having truth-conditions and this is how Artifuactualism satisfies Necessary Truth-Conditions, So we have good reason to conclude that Artifactualism satisfies both Minor Cognitive Contact and Necessary Truth-Conditions. Furthermore, it can also be easily modified to satisfy Major Cognitive Contact. Finally, it comes with little metaphysical baggage and few theoretical commitments. The competitors to Artifactualism are not this way. 6. Competing Views This section argues that the competitors to Artifactualism are worse off when it comes to the theoretical commitments they incur. The competitors include all accounts of propositions where propositions are metaphysically robust, not taken to be sui generis entities, and there is some explanation as to why propositions have the properties that they have. Stalnaker (1976), for example, does not count since there is no explanation as to why propositions have the truth- conditions that they have. So, we will compare Artifactualism to the views in Speaks (2014b), King (2014b), and Soames (2014b). The view in Hanks (2011) is close enough to Soames’ view that considerations against Soames’ view apply mutatis mutandis to Hanks’. To be clear, these are not going to be knock down arguments against these views, and we have not altogether ruled out views like Stalnaker (1976). Chapter 2 contains positive arguments to show that only Artifactualism is a suitable account of propositions. The argument here is to show that Artifactualism has theoretical advantage over competing views. According to Speaks (2014b), propositions are properties. The proposition that snow is white is the property of being-such-that-snow-is-white. According to King (2014b), propositions are facts, i.e. states of affairs. The proposition that snow is white is, roughly, the existential state of affairs of there being a syntactic structure that encodes whiteness being ascribed to snow. For 36 Soames (2014b), propositions are types of actions. The proposition that snow is white is the act- type of predicating whiteness of snow. These three accounts of propositions have serious theoretical commitments. The theoretical commitments become apparent when it comes to how each account is supposed to satisfy Necessary Truth-Conditions. For Speaks, the truth-conditions of the properties that are allegedly propositions are determined by their instantiation conditions. So, being-such-that-snow-is-white is true at w whenever it is instantiated by everything in w. But as Speaks (2014c: pp.89-90) admits, there is a lack of explanation as for why, according to him, some properties, like being-such-that-snow-is- white, have truth-conditions and are propositions whereas others, such as being-white, lack truth- conditions. Speaks furthermore admits that it looks very unlikely that such an explanation could be given. It is not enough to point out that the properties he claims are propositions are necessarily either instantiated by everything or nothing in a given world. The property of being self-identical is also this way, but is not plausibly such that it has truth-conditions. So his view of propositions is likely to engender the theoretical cost of positing a primitive distinction between properties that have truth-conditions and are propositions and properties that are not propositions. King’s view also faces a serious problem when it comes to satisfying Necessary Truth- Conditions. Although King claims that propositions are facts, by ‘fact’ he means things such as “an object having a property, n objects standing in an n-place relation”, and so on (King 2014b: p. 50). So for King, propositions are what are traditionally called ‘states of affairs’. According to King, that snow is white is the following existential state of affairs: There is a context c, assignment g, and language L such that for some lexical items a and b of L, snow is the semantic value of a relative to g and c and the property of being-white is the semantic value of b relative to g and c, and a occurs at the left terminal node of syntactic relation R that in L encodes ascription and b occurs at R’s right terminal node. Call this state of affairs SNOW. Now the question is why on Earth should SNOW, this complicated state of affairs, be such that it is true iff snow is white? We normally do not take states of affairs to have truth-conditions and King will of course grant that most states of affairs lack truth-conditions. So we need to know what makes SNOW special such that it has truth- conditions and can therefore be a proposition. 37 The way King attempts to show this is by first claiming that sentence-types are states of affairs because “it seems plausible that word-types are properties and hence that sentences are properties standing in sentential relations” (King 2014b: p.53). So According to King, the English sentence-type ‘Snow is white’ is also a state of affairs. Call this state of affairs that is allegedly a sentence, ENGLISH-SNOW. King roughly claims that a state of affairs like SNOW ends up having truth-conditions in virtue of the things agents do with states of affairs like ENGLISH- SNOW (King 2014b: pp. 187-190). So what agents do with sentences end up giving truth- conditions to the states of affairs that are propositions. Whether this account works is beyond the scope of what we are up to here, but chapter 2 will give us very good reason to reject this claim. What should be clear now is that the only reason King can plausibly claim that states of affairs can have truth-conditions is by giving his revisionary account of sentence-types where sentence- types are states of affairs. Sentence-types have truth-conditions, and if we somehow are lead to believe that sentence-types are states of affairs, then we ought to accept that states of affairs can have truth-conditions. King’s reason for taking sentence-types to be states of affairs is not good. If we assume that word-types are properties, as King does, why would we not also conclude that sentence-types are properties? King is thinking that sentence-types, in some sense, consist of word-types standing in a syntactic relation to one and another, so they must be states of affairs. But word-types, in some sense, also consist of other entities standing in relation to each other. The word-type ‘snow’, among other things, consists of the letter-type ‘S’ and the letter-type ‘N’ standing in a certain relation. Consequently, King has not given us a reason for thinking that sentence-types are states of affairs unless we also think that words-types are states of affairs. But then what things are not states of affairs? Almost every object, in some sense, consists of other things standing in a relation to each other. 27 Furthermore, what is it for these types, that are allegedly states of affairs, to have instances? So King’s account can only satisfy Necessary Truth-Conditions by making the problematic and unwarranted claim that sentence-types are states of affairs. It is also worth mentioning that positing states of affairs on its own is a theoretical cost. Although there might be other good metaphysical reasons for positing these things, this is not a given. Artifactualism does not require us to posit states of affairs and does not give a problematic analysis of sentence-types. King’s 27 Although one can take the view that all objects are actually states of affairs, this is obviously a commitment laden position. 38 idea that propositions are very much like sentence-types is a great insight that Artifactualism builds off of, but taking sentence-types and propositions to be states of affairs is not necessary. Soames (2014b) gives the most detailed argument as to how his account of propositions satisfies Necessary Truth-Conditions. According to Soames, propositions are act-types of predication. To predicate being-F of o is basically just to mentally represent o as being F. For Soames, the proposition that snow is white is the act-type of predicating being-white of snow. The relevant question for his view is why we should think such an act-type has truth-conditions and is a proposition given the fact we do not pre-theoretically take any act-types to be true or false. We do not take any of their instances to have truth-conditions either. But Soames attempts to show that act-types of representing inherit their representational natures from their instances. Just as particular acts of representing snow as being white are representational, the act-type of representing snow as being white is representational in an “extended sense” (Soames 2014c: p.234). Soames goes on to argue that this is enough for the act-types to have truth-conditions. In this way, Artifactualism is similar to Soames’ view. The functional-type that is the proposition that snow is white represents snow as being white in an extended-sense and is true iff snow is white because of facts about its membership criteria. Similarly, the act-type of representing snow as being white is representational because all instances of this type of act must be an act of representation. Whether Soames can plausibly show that the relevant act-types satisfy Necessary Truth-Conditions is beyond the scope of what we are up to in this chapter, but, chapter 2 will give us very strong reasons for concluding that these entities satisfy Necessary Truth- Conditions. 28 But we can already see that Artifactualism clearly has an easier time establishing this simply because, for Artifactualism, the relevant types that are propositions have instances that have truth-conditions. Truth-apt sentence tokens have truth-conditions. Soames’ alleged propositions are such that there instances never have truth-conditions. Particular acts never have truth-conditions. What reason could we have for thinking that act-types count as having truth- conditions if their instances never have truth-conditions? It is not as if we have independent pre- theoretical reasons for thinking that act-types have truth-conditions. Soames’ view also has some obvious theoretical commitments that make it less attractive. The account is commitment laden when it comes to the nature of the mind. It requires Major Cognitive Contact to be true, whereas our theoretical needs only require Minor Cognitive Contact. It requires that extremely primitive agents must perform the special mental act of 28 See Caplan et al (2013) for some problems that Soames’ account faces in this regard. 39 predication. It is not enough for primitive agents to be in mental states that represent things as being ways. Soames must assume that these primitive agents are able to perform a special mental act. 7. Concluding Remarks Ultimately, King, Soames, and Speaks are right that we need metaphysically robust propositions in our ontology. But they get it wrong when it comes to the way that propositions need to be. Our theoretical needs only require an account of propositions that satisfies Necessary Truth- Conditions and Minor Cognitive Contact. Unlike its competitors, Artifactualism is a minimal theory of propositions that is well positioned to meet these theoretical needs without taking on unnecessary commitments. i 40 Chapter 2: Bridging the Gap between Representation and Truth Abstract An account of truth-aptitude must answer two questions. First, what way does an entity need to be in order to be truth-apt, i.e. have any truth-conditions at all. Second, how are the truth- conditions of a truth-apt entity determined by its other properties? Although these questions are worth investigating in their own right, this chapter argues for a theory of truth-aptitude that has important consequences for the nature of propositions. A careful investigation into the nature of truth-aptitude reveals that a truth-apt entity must have the function, i.e. serve the purpose, of representing. I argue that this result rules out nearly every metaphysically reductive theory of propositions that has been proposed so far. Artifactualism is the only view of propositions that is consistent with the demanding theory of truth-aptitude. We can all agree that there are true sentences and that there are false sentences. We all grant that some things such as rocks, birds, tables, and so on are neither true nor false. The things that are either true or false, like true sentences and false sentences, are truth-apt. Rocks, birds, and the like are not truth-apt. To be truth-apt is to have some truth-conditions rather than having no truth- conditions at all. 29 It would be quite surprising if we could not give an account of the underlying features that distinguish the truth-apt things from the not truth-apt things. Ideally, what we want is an account that gives us illuminating necessary and jointly sufficient conditions for when an entity has any truth-conditions and one that explains how the more basic properties of a truth-apt entity determine its truth-conditions. A persistent concern in giving such an account will be to explain why maps are not truth-apt despite the fact that they share many properties in common with truth- apt sentences. This chapter attempts to achieve three main results. First, it introduces and argues for a novel theory of truth-aptitude. Second, it argues that the correct theory of truth-aptitude places such a demanding constraint on giving a reductive metaphysics of propositions that every reductive 29 We can think of an entity e having truth-conditions as e having the property of being-an-x-s.t.-x-is-true- iff-S where ‘S’ is a schematic letter for a sentence. Note that having truth-conditions does not entail having a truth-value. 41 account of propositions proposed thus far is ruled out. Third, it sketches a novel and promising account of propositions that is not ruled out by the demanding theory of truth-aptitude. 1. Introduction There has been surprisingly little work done on investigating the nature of truth-aptitude. Recent philosophical interest in the nature of truth-aptitude has mostly focused on examining the relationship between minimalism about truth and minimalism about truth-aptitude. Minimalism about truth is roughly the view that understanding what it is for something to be true simply requires knowing that, for an arbitrary sentence ⎡S⎤, ⎡S⎤ is true iff S. 30 In contrast, minimalism about truth-aptitude is roughly the view that all it takes for an entity to be truth-apt is for that entity to have a certain syntactic form. There is good reason to think that minimalism about truth does not entail minimalism about truth-aptitude. 31 There is also good reason to think that minimalism about truth-aptitude is false. 32 It should be clear that only extreme theoretical considerations would motivate one to take on a minimalist view of truth-aptitude. 33 As Jackson et al (1994) argues, the syntax of a sentence alone is never enough to establish its truth-conditions. At the bare minimum, facts about patterns of use concerning the sentence need to be appealed to in order to explain why a sentence has any truth-conditions at all. Wright (1992) endorses a sophisticated version of truth-aptitude 30 See O’Leary-Hawthorne and Oppy (1997) for an overview of different ways to be a minimalist about truth. See Soames (2003) for an argument that the definition needs to be in terms of propositional-truth. 31 Burgess (2010), O’Leary-Hawthorne et al (1997), and O’Leary-Hawthorne and Price (1996), argue that a minimalist theory of truth does not entail a minimalist theory of truth-aptitude. If correct, this is a fortunate result for non-cognitivists about ethical discourse who hold that well-formed indicative claims about ethics are not truth-apt – the non-cognitivist need not reject a minimalist theory of truth like that in Horwich (1999). In general, this would be a fortunate result for anyone that held a non-cognitivist view in any domain of discourse such as aesthetics, mathematics, and so on. Holton (2000) also takes up these issues but has a more nuanced view about the relationship between minimalism about truth and truth-aptitude. 32 Jackson et al (1994) gives convincing arguments to show that minimalism about truth-aptitude is implausible. 33 Expressivist accounts of ethical discourse sometimes endorse a minimalist theory of truth-aptitude as a way to address the Frege-Geach problem. Stoljar (1993) and Price (1994) take minimalism about truth- aptitude to be a viable option for the expressivist. Basically, their idea is that, even though ethical statements do not represent the world as being any way and instead express various attitudes of the speaker, ethical statements still meet the minimal requirements to be truth-apt. They have the right syntactic form to be substituted into instances of the truth-schema: ⎡S⎤ is true iff S. And if such ethical statements have truth- conditions, then the Frege-Geach problem is far more tractable for the expressivist. Note that the hybrid- expressivism of Ridge (2006) tries to get by without doing this. See Schroeder (2008) for an overview of the Frege-Geach problem. 42 minimalism that is supposed to avoid this problem. 34 But rather than taking minimalist views head on, I am going to show how we can do better. We can give a theory of truth-aptitude that is plausible, explanatory, and not posited for ad hoc theoretical purposes. Outside of highly minimalistic theories of truth-aptitude, very little has been proposed. Jackson et al. (1994: p. 294) proposes that a necessary condition for an entity e being truth-apt is that e can be used to “give the content of a belief”. This might be a start but it is not much of an explanation. Furthermore, it seems to make the substantive and potentially problematic assumption that for any proposition p, it is possible that an agent A believe p. 35 Soames (2015) and King (2007) operate on the assumption that if an entity e has the truth-conditions of being- true-iff-o-is-F 36 , then e represents o as being F. 37 Sometimes Soames and King even appear to assume that e representing o as being F is sufficient for e having these truth-conditions. 38 Although I think that the few non-minimalist proposals given so far are on the right track, we would greatly benefit from an investigation exclusively on the relationship between representation and truth. The most immediate benefit concerns the nature of propositions. There has been a resurgence of interest in whether we need propositions in our ontology, and if so, whether we can give an account of what sort of things these entities are. 39 All involved agree that if propositions exist, then they have truth-conditions and they have their truth-conditions essentially. With a theory of truth-aptitude, we can greatly narrow down the plausible candidates for entities that could be propositions. In fact, it turns out that when we arrive at the correct theory of truth-aptitude, we see that all the reductive theories of propositions so far put forward, including King (2014b), Speaks (2014b), Hanks (2011), and Soames (2015), cannot be correct. Section 2 explains how we will go about investigating the nature of truth-aptitude. Section 3 argues that a necessary condition for an entity being truth-apt is that the entity must represent things as being some way. Section 4 goes on to present the Naïve Representational Theory of truth-aptitude, which is the view that representing things as being some way is sufficient for being 34 Boghossian (1990) also describes a sophisticated version of truth-aptitude minimalism meant to avoid this problem. 35 Considerations regarding both Kaplan’s paradox as well as the Russell-Myhill Paradox might start to get one worried about such an assumption. See Bueno et al (2013) for an overview of these paradoxes. 36 The locution of ‘having the truth-conditions of’ is being used here to pick out the special sense of ‘truth- conditions’ where ‘Snow is white’ has the truth-conditions of being-true-iff-snow-white and ‘Snow is white and arithmetic is incomplete’ does not. Spelling out what this special sense of having truth-conditions amounts to without appealing to propositions has been notoriously problematic. 37 I am using ‘F’ in this chapter as a schematic letter. 38 C.f. Soames (2015: p. 21) and King (2007: pp. 57-62). 39 C.f. Soames (2015), King (2014b), Speaks (2014b), and Hanks (2011). 43 truth-apt and that a truth-apt entity’s truth-conditions are determined by what things it represents as being what ways. The Naïve Representational Theory is promising but faces two serious problems that are respectively covered in sections 5 and 6. Section 7 presents a modification to the Naïve Representational Theory, the Functional Theory of truth-aptitude that solves it problems. Roughly, the Functional Theory is that only entities that have a certain representing function are truth-apt and that an entity’s truth-conditions are determined by its representing function. Section 8 argues that if the Functional Theory is correct, then it rules out nearly 40 all reductive theories of propositions proposed so far. The section ends by giving a full of account of Artifactualism and showing how it is consistent with the Functional Theory of truth-aptitude. 2. The General Demarcation Problem The general demarcation problem is the problem of giving an account of why some entities such as certain indicative sentences are truth-apt whereas other entities are not truth-apt. A theory of truth-aptitude solves the general demarcation problem by giving us necessary and jointly sufficient criteria for when a given entity is truth-apt. But how should we go about inquiring after these criteria? Although it is easy to lose track of, representation, accuracy, and truth are not technical notions merely posited to give theoretical explanations. The notion of a truth-condition is a technical notion but truth is not. 41 We know that the table in front of us is neither true nor false long before we do any semantic theorizing. We know that the cloud in the sky is neither accurate nor inaccurate. We know that Jupiter does not represent any thing as being any way. Since many students of the philosophy of language are simply told rather than persuaded that there is a relationship between representation and truth, they end up concluding that whatever is meant by ‘represent’ in linguistic theorizing is something different than our everyday notion of representing. The problem is exacerbated because representing is supposed to do all types of surprising work in the philosophy of mind. But we can show that there is a close relationship between representation and truth by looking at flat-footed cases where we intuitively judge that some types of things are representational, some types of things are accurate or inaccurate, and some types of things are true or false. 40 The caveat is explained in section 8. 41 See Cappelan and Hawthorne (2009) for an in-depth treatment of the relationship between truth- conditions and the property of being-true-or-false. 44 One potential stumbling block to this strategy is that, under any conception of propositions where propositions are the designata of certain embedded that-clauses, 42 propositions are assumed to be truth-apt. If we start with a view of propositions where they are, say, abstract entities from the third realm but are nonetheless truth-apt, this might make it quite hard to give an account of truth-aptitude that accommodates our flat-footed judgments about what sorts of worldly things are representational, accuracy-apt, and truth-apt. What could ethereal truth-apt propositions have in common with worldly truth-apt entities, assuming that there are any worldly truth-apt entities? Since I take our flat-footed judgments about worldly entities being representational, accuracy-apt, and truth-apt to be more certain than any theory of propositions offered thus far, propositions being truth-apt will not be part of the initial data we use to develop a theory of truth-aptitude. However, once we have a theory, we will use this to help us discover what propositions must be if they are indeed truth-apt. So in short, our pre-theoretic judgments about representation, accuracy, and truth will be our starting point. Sooner or later we are likely to face difficulties that our pre-theoretic judgments will not solve, but we can get well on our way towards a satisfactory theory of truth-aptitude by trying to accommodate them. 3. The Link between Representation, Accuracy, and Truth This section argues that there is a link between representation, accuracy, and truth. Specifically, we learn that an entity e representing things as being some way is a necessary condition for e having both accuracy and truth-conditions. I will first argue that there is a link between representation and accuracy. Although there are representational things that lack accuracy- conditions, some ways of being representational do give rise to accuracy-conditions. Next, I will show that there is a link between representation and truth. Again, there are indeed things like maps, two-dimensional graphs, and three-dimensional models that are all representational in nature, have accuracy-conditions, but are still neither true nor false. But by comparing and contrasting the nature of these things with truth-apt entities, it will become clear that representing things as being some way is a necessary condition for being truth-apt. What we first need to show is that there is a connection between representation and accuracy. We use ‘accurate’ in a variety of way. We talk about accurate dart throwers and accurate dart 42 Section 8 covers conceptions of propositions where propositions are not assumed to be truth-apt. 45 throws. We talk about accurate safety inspectors and accurate robot welders. But accuracy in these regards appears to have nothing to do with representation and truth. On the other hand, an accurate written report, say one about amoeba, does have something to do with representation. The report is accurate to the degree that it successfully represents amoeba as being as they actually are. This should lead us to think that there is a link between representation and accuracy only when we are using ‘accurate’ to mean that something is accurate at representing. With that in mind, from here on out I will only use ‘accurate’ in the sense of accurate-representing. So in the sense of ‘accurate’ we are considering, there is a trivial link between representation and being accuracy-apt. But it still is not true that everything that is representational has accuracy-conditions in the relevant sense. Consider cases where we represent one object with another object for some purpose. 43 We might use a floor lamp to represent our banker in order to practice our pitch for a loan. We might use a visualized green dot to represent the concept of world peace for the purposes of meditation. In short, we can use an entity e to represent some object o before we go on to represent o as being any particular way with e. When an entity e is a representation merely because we are representing o with e, e is neither accuracy-apt nor truth- apt. However, when we represent things as being some way, accuracy comes into the picture. To show that representing things as being some way is necessary for having accuracy- conditions, consider a map. One black dot on the map represents Los Angeles. Another black dot represents Portland, Oregon. Neither dot itself is accurate or inaccurate. But the map is accurate or inaccurate and its degree of accuracy depends on the spatial relationships between the things on the map. Given the orientation of the dots, the map represents Los Angeles as being south of Portland. The accuracy of the map depends on whether Los Angeles is south of Portland because the map represents Los Angeles as being this way. If Los Angeles is as the map represents it as being, this will contribute to the map’s accuracy. 44 So in general, which things e represents as being which way determine e’s accuracy-conditions. If e does not represent any things as being any way, then e is not accuracy-apt. Now we need to show how representation and accuracy relate to truth. To do this, we will first show that sentences also represent things as being ways. Imagine a written report about U.S. geography. It contains, among other English sentences, a token of ‘Los Angeles is south of 43 This type of distinction is noted by Goodman (1968) and also is often posited to give explanations about the nature of mental states, such as in Burge (2009). 44 See Kulvicki (2015), Bronner (2015), Rescorla (2009), and Casati and Varzi (1999) for arguments over the various nuances concerning what and how maps represent. 46 Portland’. The multi-sentence report is neither true nor false, but it is accurate or inaccurate. In fact, the degree to which the report is accurate, like the map, depends on whether Los Angles is south of Portland. Why? It must be because the report, just like the map, represents Los Angeles as being south of Portland. Unlike the map, the report manages to do this by having a proper part, a sentence token, that represents Los Angeles as being this way. But both the report and the map represent Los Angles as being south of Portland and this leads us to conclude that a sentence token on the written report represents Los Angeles as being south of Portland. So written reports that consist of multiple complete indicative sentences have accuracy-conditions that are determined by which things the sentences on the report represent as which ways. And this gets us the important result that individual well-formed complete indicative sentence-tokens represent things as being ways. We are now in a position to see an important relationship between representation and truth. Notice that whether a sentence token is true perfectly coincides with whether the things it represents as being some way are in fact that way. A token of ‘Los Angeles is south of Portland’ represents Los Angeles as being south of Portland. The sentence token is entirely accurate iff Los Angeles is the way that the token represents Los Angeles as being – namely as being south of Portland. Importantly, the sentence token is true in exactly the same circumstances. The token is true whenever it represents Los Angeles as being as Los Angeles actually is. Since it represents Los Angeles as being south of Portland, this just amounts to the token being true whenever Los Angeles is in fact south of Portland. This gives us very good reason to posit that e’s truth- conditions are determined by which things it represents as being which way and that e must represent some things as being some way in order to be truth-apt. So seeing the differences and similarities between a dot on a map, a map, a multi-sentence written report, and a sentence that is part of the report helped us see that there is a close link between representation, accuracy, and truth. Representing an object with another object does not give rise to accuracy or truth-conditions. But representing some things as being some way is a necessary condition for both being accuracy-apt as well as being truth-apt. 4. The Naïve Representational Theory of Truth-Aptitude Our findings in the previous section naturally lead to the Naïve Representational Theory of truth- aptitude. The Naïve Representational Theory of truth-aptitude consists of two claims: 47 NRC) For any entity e, e is truth-apt iff for n >0, e represents n things as being some way. NRD) For any truth-apt entity e, (e has the truth-conditions of being-true-iff-o-is-F) iff (e represents o as being F). 45 NRC gives the criteria for what it takes to be truth-apt. NRD tells us how the truth-conditions of a truth-apt entity are determined. The Naïve Representational Theory goes beyond our findings in the last section by supposing that representing things as being some way is also sufficient for being truth-apt. But the theory is still prima facie plausible. And as we have seen, it is a theory that Soames and King appear to operate on in several instances, so it is worth investigating. The Naïve Representational Theory, however, does face serious problems. Sections 5 and 6 will respectively cover what I take to be two insurmountable problems for it. The most pressing concern is that entities with accuracy-conditions, including maps, end up being truth-apt according to the Naïve Representational Theory. However, the remainder of this section will cover the problems the theory faces that I believe to be surmountable. These surmountable problems concern the nature of truth-apt complex sentences, sentences with empty-terms, potentially paradoxical sentences, and sentences in different grammatical moods. Examining both the surmountable and insurmountable problems will provide us with a way to ultimately arrive at a superior theory of truth-aptitude. The first surmountable problem the Naïve Representational Theory faces concerns complex sentences that do not appear to be about things in the everyday sense of ‘thing’. For sake of simplicity I will talk about sentence-types, but what I say here equally applies to sentence tokens. Obviously sentences like ‘If snow is white then grass is green’, ‘Everything is red’, and ‘The tallest man is fat’ are truth-apt. According to the Naïve Representational Theory, these sentences must represent things as being some way. But this is not a serious problem. We can either take ‘If snow is white then grass is green’ to represent snow and grass as having a complex property or take it to represent propositions as having a complex property. We can take ‘Everything is red’ to either represent the property of being-red as having the higher- order property of being-had-by-everything or take it to represent a propositional function as being 45 This definition lacks time indices. If one is a B-theorist about tense and a non-temporalist about propositions, one can easily modify the definition to include time indices. See Brogaard (2012) for more on issues about tense and propositions. 48 some way. If we accept the Russellian analysis of the definite description, 46 we can similarly analyze ‘The present king of France is sleeping’ as representing the property of being-a-present- king-of-France as having a higher order property or as representing a propositional function as having a property. 47 Chapter 4 covers these options in more detail. All of these suggestions require positing complex ways of being. I have been calling such ways-of-being ‘properties’, but the Naïve Representational Theory does not prohibit one from distinguishing between the “real” properties and what we might call ‘concepts’. 48 So positing complex ways of being is a cost for the theory, but it is a manageable cost. 49 If you want to be cheap here, I wish you luck in trying to get by in semantic theorizing with, say, your five favorite fundamental properties. The next potential problem for the Naïve Representational Theory concerns so-called “empty” terms. The first question we face is whether a sentence like ‘Pegasus is a mythical creature’ is truth-apt. If the sentence is not truth-apt then there is no problem if ‘Pegasus’ fails to refer. The sentence would not represent anything as being any way, and would not be truth-apt. However, if such sentences are truth-apt, we have two options. The first option is to posit that there is some thing that is ‘Pegasus’ and so ‘Pegasus’ does indeed refer and the sentence does indeed represent some thing as being some way. 50 The second option is to accept some version of free-logic. 51 Again commitments have to be made here, but there is a range of plausible options. 46 Note that the theory need not accept the Russellian analysis. 47 Note that none of these suggestions entail that the logical form of these sentences are of subject-predicate form. This might come at a surprise since it is natural to use interpreted logical forms to indicate that a sentence represents certain things as being a certain way. 47 But (interpreted) logical form does more than indicate what things are represented as being what way. It also gives us information about what a priori inferences can be made. That is why the logical form of ‘Everything is yellow’ is ∀xFx and the logical form of ‘The Sun is yellow’ is F(o). From the truth of the first sentence, but not the second, we can infer that there is nothing that is not yellow. Claiming that both sentences represent things as being ways does not force us to accept that these sentences have the same logical form. King (2007: pp. 27-29) has a good overview of how interpreted logical forms are used in linguistic theorizing. Also see Larson and Ludlow (1993). 48 Bealer (1982) is one place where such a distinction is made. 49 Audi (2013) is one instance of the view that there are no complex properties such as the disjunctive property of being-a-god-or-not-being-a-god. Proponents of such views need to give us an account of what we refer to when we refer to complex ways of being. For example, we can say ‘o is red or blue’ and then say ‘o* is also this way’. Although we could try to offer a highly revisionary account that fails to posit a referent for ‘this way’, the most common and reasonable alternative to positing complex properties is to posit concepts and claim that concepts are ways of being but are not identical to properties. 50 See Salmon (1998) as well as Ackerman (forthcoming) for examples of such views. 51 See Bacon (2013) for an overview of the various versions of free-logic. One could accept a positive free- logic and accept that the sentence represents Pegasus as being some way. However, one would also have to deny the inference to the conclusion that there is something such that the sentence represents it as being 49 The third surmountable problem concerns potentially paradoxical sentences. Let ‘S1’ name the sentence ‘S1 is not true’. The following claim τ1 seems true: τ1) S1 represents S1 as being not true. So given the Naïve Representational Theory, we would get the following claim τ2: τ2) S1 is true iff S1 is not true. There are two known options here. The first option is to retain a classical logic but be something of a semantic pluralist. That is to say, one can keep classical logic but posit that there are many properties that we could pick out with the predicate ‘is true’. The most well-know version of this move is the Tarski Hierarchy Theory, 52 but it is not the only way to go this route. 53 If we go this route, we will also have to say that there are many different relations we can pick out with ‘representing o as being F’ and we will have to modify our definition of the Naïve Representational Theory to accommodate this. The second option is to reject classical logic. What features of classical logic we reject will have different consequences for the Naïve Representational Theory. 54 One outcome would be that we can neither assert nor deny the left or right side of τ2’s bi-conditional even though we can assert τ2. Another outcome would be that τ2 is true even though the left and right side of its bi-conditional are neither determinately true nor determinately false. With all of these options, we can keep the natural assumptions that τ1 and τ2 are true. Note that these options get us that S1 has truth-conditions but, so to speak, lacks a truth-value. Now it would be a bit misleading to say that the Naïve Representational Theory has good options when it comes to potentially paradoxical sentences. The options are not great, but this is simply because the known options for solving the semantic paradoxes all come with significantly revisionary consequences. A final surmountable problem for the Naïve Representational Theory concerns sentences that are not in the indicative mood. Take a sentence like ‘Is Aristotle wiser than Plato’. Some semantic analyses of this sentence analyze it as representing Aristotle as being wiser than Plato, 55 so the sentence would appear to have truth-conditions according to the Naïve Representational Theory. Those that analyze sentential-interrogatives (as opposed to wh-interrogatives) in terms of a some way. And so one would necessarily have to present the Naïve Representational Theory with schematic letters so that the statement of the theory consisted of many theorems. To accept a negative free- logic, we would have to deviate further from the Naïve Representational Theory. We could accept a negative free-logic if we claimed that an entity e is truth-apt iff e has the function of representing some things as being some way. ‘Pegasus is a mythical creature’ would turn out to be truth-apt since the function of this sentence is to represent some thing as being some way even though the sentence simply fails, functionally speaking, to represent any thing as being any way. This deviation brings us very close to the preferred theory of truth-aptitude I am going to argue for. 52 See Halbach (1997) for more on the Tarski-Hierarchy Theory. 53 Bacon (MS) is a version of the semantic pluralist route that does not posit a hierarchy. 54 See Field (2008) for an account of various non-classical solutions. 55 See Vanderveken (2009). 50 proposition having a certain illocutionary force are already committed to sentential-questions having truth-conditions and must already explain away this result regardless of whether one accepts the Naïve Representational Theory. But there is also the option of analyzing sentential- questions in a way where they do not represent things as being ways. 56 In short, the Naïve Representational Theory is a promising view of truth-aptitude. The potential problems it faces concerning complex sentences, empty-terms, paradoxical sentences, and non-indicative sentences are surmountable. However, we will see in the following two sections that two serious problems force us to search for a better theory of truth-aptitude. 5. The Unity of the Sentence Disjunctions as well as sentences that contain sentential-negation give rise to problems for the Naïve Representational Theory. The problems arise because of NRD. Roughly, this is because NRD says that any representing done by an entity contributes to its accuracy-conditions. A proponent of the Naïve Representational Theory needs to explain how a complex sentence’s representational properties determine its accuracy-conditions and truth-conditions. Consider a sentence of the form ⎡F(α) v ¬F(α)⎤, such as ‘The Sun is a god or the Sun is not a god’. This sentence represents the Sun as being a god but it also represents the Sun as not being a god. Like a multi-sentence written report, the disjunctive sentence has parts that represent things as being ways. One disjunct represents the Sun as being god. The other disjunct represents the Sun as not being a god. The disjunctive sentence appears to represent these things in virtue of its parts. The Naïve Representational Theory wrongly entails that this disjunction is necessarily inaccurate. To make the problem more vivid, imagine a written report that consists of the two-sentences ‘The Sun is a god’ and ‘The Sun is not a god’. The report is necessarily inaccurate. Since the sentence ‘The Sun is a god or the Sun is not a god’ represents the same things as the written 56 Here is a sketch of how this might go. Think of representing o as being F as a set-up for performing illocutionary-acts like asserting, suggesting, and so on. Basically, we get our audience to think of o as being F as a set-up for asserting that o is this way, or suggesting that o is this way, etc. Now use ‘questioning’ to pick out something that is a lot like representing but a set-up for different illocutionary-acts. We might think that ‘Is Aristotle wiser than Plato’ questions whether Aristotle has the property of being wiser than Plato rather than that it represents Aristotle as being this way. To question whether an object has a property would be something like to get one’s audience to imagine this object and this property in order for you to go on and perform an illocutionary speech-act such asking whether Aristotle is this way, rhetorically asking whether Aristotle is this way, and so on. 51 report, this disjunction is also necessarily inaccurate according to the Naïve Representational Theory. But this is absurd. We can know a priori that this disjunction is true, 57 but the Naïve Representational Theory instead entails that the disjunction is necessarily inaccurate. So the Naïve Representational Theory must be false. A similar problem arises for a sentence like ‘It is not the case that the Sun is a god’. To put a phrase to this problem, the Naïve Representational Theory cannot account for the unity of the sentence. To get around this problem one might try to deny, for example, that a disjunctive sentence represents what its disjuncts represent. Obviously a whole does not always have all the properties of its parts. But, in this case, I believe it is implausible to claim that the disjunction does not represent what its disjuncts represent. First, notice that the sentence inherits at least some representational properties from its parts. The sentence represents the Sun. It counts as doing this because one of its parts, a token of the phrase ‘the Sun’, represents the Sun. Furthermore, when an agent utters ‘The Sun is a god or the Sun is not a god’ and thereby represents that either the Sun is a god or is not a god, the agent does this by representing the Sun as being a god and representing the Sun as not being god. The same should hold for a written disjunction. It is only by representing the Sun as being a god and representing the Sun as not being a god, in virtue of its parts, that the whole disjunction represents what it represents. It is not enough for the sentence to have parts that do this. Like an agent, the whole disjunction represents the Sun as being contradictory ways in order to represent the Sun as being at least one of these ways. 58 It is extremely important to see that the force/content distinction has nothing to do with this problem. 59 A sentence representing the Sun as being a god is not the same as a sentence having the illocutionary force of asserting that the Sun is a god. Asserting and representing are not one and the same thing. A storyteller can represent the Sun as being a god without asserting that this is the case. If we want to know whether Hesiod really was asserting that time is a god we will be annoyed if someone points out the he represented time as being a god – we know that. Embedded indicative sentences do not lack their representational properties. The token of ‘Aristotle is wise’ in a token of ‘If Aristotle is wise then Plato is wise’ does not have its assertoric force, but it still represents Aristotle as being wise. This is in fact how we are able to determine 57 Assuming classical logic is correct. 58 We need not assume that the disjunct represents the Sun as having a complex property. We could assume it represents two propositions about the Sun as being such that at least one of them is true. 59 Reiland (2012) argues that, despite what Hanks claims in Hanks (2011), Hanks’ view of propositions faces a problem involving complex propositions and the force/content distinction. 52 what the whole sentence represents. We look at what the parts represent. Obviously the accuracy and truth-conditions of the parts of a complex sentences with truth-functional operators determine the accuracy and truth-conditions of the whole. So, if the Naïve Representational Theory is correct, then the representational properties of the whole disjunction will determine the truth-conditions of the whole disjunction. The problem with the Naïve Representational Theory is that, given the whole disjunction’s representational properties, the theory wrongly entails that certain sentences are necessarily inaccurate. Barring some implausible moves about what the problematic sentences represent, the Naïve Representational Theory ought to be rejected. 6. The Special Demarcation Problem The Naïve Representational Theory gives rise to the special demarcation problem. The special demarcation problem is the problem of explaining why some entities are merely accurate or inaccurate where as other entities are also truth-apt. NRS is the reason the Naïve Representational Theory faces the special demarcation problem. Roughly, this is because NRS entails that any accuracy-apt entity is truth-apt. The real problem for the Naïve Representational Theory is that no slight modification to it can explain why things like maps are merely accuracy-apt and not truth- apt. The first thing to establish is that there are indeed entities that are accuracy-apt but not truth- apt. As noted earlier, maps, two-dimensional graphs, and three-dimensional models are all representational in nature but are neither true nor false. They are also accuracy-apt but not truth- apt. One might be inclined to ignore this data and claim that treating things like maps as being truth-apt will not create problems for our theoretical purposes. But we stand to make real progress in the philosophy of language and the philosophy of mind if we can figure out the relationship between representation and truth, so it is worth our effort to get things right. The Naïve Representational Theory gets the wrong results here. Since maps represent things as being ways they count as being truth-apt according to the theory even though they are not. The same goes for agents performing representational acts as well as representational acts themselves. When John represents the Sun as not being a god, it would appear that both John and his act count as representing the Sun as being some way. The agent is accurate in doing this. The agent’s 53 representing act is an accurate one. But there are no true agents. There are no true actions. 60 Agents and actions are not truth-apt. One way to try to amend the theory is to claim that what it is for an agent or action to represent is different from what it is for an artifact like a map to represent and that only things that represent in the way that artifacts represent can be truth-apt. This suggestion seems plausible and we will return to it later. But we still need to explain why maps are merely accuracy-apt. It is important to see that we cannot explain why maps are not truth-apt merely by appealing to the fact that maps exhibit a spatial mode of representing whereas truth-apt sentences do not. A multi-sentence written report is not truth-apt even though it exhibits the same mode of representing as a single truth-apt sentence. Consider all the different things that a map represents as being different ways. Assume it is possible to write out all the things the map represents. 61 We could make a written report that is representationally identical to the map. We could also make a giant sentence consisting of many conjunctions that is representationally identical to both the map and the written report. Yet only the giant sentence is truth-apt. Why? We are going to have to more radically diverge from the Naïve Representational Theory if we are to solve the special demarcation problem. We need a theory of truth-aptitude that can accommodate the important fact that some things are accuracy-apt but not truth-apt. 7. The Functional Theory of Truth-Aptitude This section introduces and argues for a novel theory of truth-aptitude that builds off of the shortcomings of the Naïve Representational Theory. What we want is a theory that can account for the unity of the sentence and solve the special demarcation problem. I will first present and explain the Functional Theory of truth-aptitude. Next, I will deal with some complications that arise for the theory. This section ends by going over the Functional Theory’s virtues. The Functional Theory consists of the following two claims: 60 When we assert that S, what we asserted is truth-apt but our action of asserting is not truth-apt. At least according to our pre-theoretical judgments, there is no particular act that we preform such that the act is either true of false. 61 In many cases, for some objects and some way, it might be indeterminate whether a map represents those objects as being that way. Furthermore, a map might end up representing infinitely many things as being ways. However, these complications do not matter for the point at hand since a multi-sentence written report is also not truth-apt. 54 FC) For any entity e, e is truth-apt iff there is some x such that e has the function of representing x as being-F and e does not have any other representing-as functions. FD) For any truth-apt entity e, (e has the truth-conditions of being-true-iff-o-is-F) iff (e has the function of representing o as being F). 62 So the basic idea behind the Functional Theory is that only entities that have a unique function of representing things as being some way are truth-apt. To be clear, the idea is not that truth-apt entities have a single function. An agent might create a truth-apt sentence token not only for the purpose of representing but also for the purpose of practicing her handwriting. FC roughly just amounts to the claim that a truth-apt entity has a single representing-function. A truth-apt entity can have parts that each have their own representing functions. But the Functional Theory claims that the truth-apt entity itself only has a single representing-as function. It will turn out that the Functional Theory solves the special demarcation problem by positing that maps have many representing-as functions whereas truth- apt entities only a have a single representing-as function. In order to explain what the Functional Theory exactly amounts to, I need to first explain what it is to have the function of representing. After that, I need to explain what having a unique representing-as function amounts to. To illustrate what it is to have the function of representing, it will be helpful to contrast agents performing the act of representing as opposed to what artifacts do when they count as representing. An agent representing o as being F is one and the same as an agent depicting o as being F, which in turn, is one and the same as portraying o as being F. Although we say things like ‘John depicted the Sun as being a god’ more often than ‘John represented the Sun as being a god’ these locutions are about the same event. Similarly, the verbs ‘representing’, ‘depicting’ and ‘portraying’, all pick out the same action. Sometimes speakers use the phrase ‘representing that’ and ‘asserting that’ interchangeably, but this can only be done in contexts where it is agreed that the purpose of representing is to make assertions. Representing o as being F might be an essential part of asserting that o is F, but to assert this is to do more than to represent this. When artifacts represent, an artifact is doing something because it is the function of the artifact to represent. This is no different from the fact that appliances do things too. Electric 62 See footnote 16 if lack of time-indices is a concern. 55 kettles boil water just as humans boil water. Clocks tell the time just as humans tell (others) the time. Are these artifacts performing an action? No, they are performing a function. We use action-words to pick out a function of artifacts when what it is for the artifact to perform that function shares similarities with what it is for the agent to perform the action. Representing is an action agents perform but representing is also a function that artifacts, like maps, perform. We will talk of boiling water, telling the time, and representing as doings of artifacts and agents alike, but we need not settle whether there is a univocal sense of ‘representing’ that applies to both what agents and artifacts do. Having artifacts that have the function of representing is very useful. Imagine a group of hunters hunting a fearsome and notorious boar that we will call ‘Grulos’. The lead hunter represents all the possible directions that Grulos could have gone by drawing arrow-shaped marks on the ground. When he is done, each arrow will respectively have the function of representing Grulos as having gone some direction. This will facilitate the other hunters in taking a vote as to which way Grulos went. Even if the lead hunter were to walk away and, unbeknownst to the rest of the hunters, be mauled and killed by the great Grulos himself, the arrows would, ceteris paribus, 63 continue to represent despite the fact that our poor lead hunter would not be using them anymore. Creating things (e.g. inscriptions, air-vibrations, gestures) that have the function of representing things as being some way makes things more efficient. We do not need to wait for the lead hunter to one by one diachronically represent Grulos as having gone each direction while we vote. We do not need to employ many tribe members to each represent Grulos as having gone some direction. We put the inscriptions in the dirt to work. Now that we have a sufficient understanding of the difference between the action of representing and artifacts having the function of representing, we can introduce some helpful distinctions. An entity e is a representation if e has the function of representing a particular object o or if e has the function of representing o as being some particular way being-F. The former have representing-with functions. The latter have representing-as functions. The dot on the map has the representing-with function of representing Los Angeles. The map has the representing-as function of representing Los Angeles as being south of Portland. It might also have the representing-as function of representing Texas as being larger in land area than Rhode Island. 63 As long as the tribe as a whole intended to use them in the same way and successfully did so when they did use the marks, the marks would still have the function of representing. 56 The Functional Theory claims that an entity is truth-apt iff it has a unique representing-as function. We now know what it is for an entity to have a representing-as function. But what is it for an entity to have a unique representing-as function? Consider a sentence token that has the function of representing o as being F. This token has a unique representing function iff it does not have any other functions that involve representing things as being some way. So, roughly speaking, if the token also had the function of representing o* as being G, then the token would not have a unique representing-as function. 64 The Functional theory also claims that an entity’s truth-conditions are determined by its unique representing-as function. An entity that has the unique representing-as function of representing o as being F is true iff o is F. There is an important complication that the Functional Theory gives rise to that we ought to address before we show how it solves the problems the Naïve Representational Theory faced. The options open to the Functional Theory for what complex sentences represent should be clear. As with the Naïve Representational Theory, the Functional Theory must posit the existence of complex ways of being in order to account for what complex sentences represent. However, what is not clear is how all these complex sentences could have a unique representing-as function. Consider a token of ‘Los Angeles is south of Portland’. It has the function of representing Los Angeles as being-south-of-Portland. But it also has the function of representing Portland as being- an-x-s.t.-Los-Angeles-is-south-of-x. This would appear to entail that it has at least two distinct representing-as functions since these functions involve representing different things as being different ways. But if this truth-apt token does not have a single unique representing-as function, then the Functional Theory must be false. 65 Fortunately, the truth-apt token does have a unique representing-as function. A token of ‘Los Angeles is south of Portland’ represents Los Angeles as being south of Portland. To represent this is to do one and the same as representing Portland as being such that Los Angeles is south of it. 66 These are identical doings. Claiming that these doings are identical is no more radical than claiming that to act as if chocolate ice cream tastes better than vanilla ice cream is to act as if vanilla ice cream is such that chocolate ice cream tastes better than it. So, there is one representing action that we can pick out in at least two different ways: 64 There is actually an important caveat to this claim that we will address shortly. 65 It will do no good to claim that it has the single function of representing Los Angeles and Portland as standing in the being-south of relation. ‘Portland is south of Los Angeles’ also does this. 66 The ‘just is’ statements put to work in Rayo (2013) would be a natural way to understand what I am claiming here if one does not want to admit so-called “doings” into ones ontology in order to claim that these things are identical. 57 (1) ‘The act of representing Los Angeles as being south of Portland.’ (2) ‘The act of representing Portland as being such that Los Angeles is south of it.’ 67 Similarly, a token of ‘Los Angeles is south of Portland’ has a unique representing function that we can pick out with different locutions: (1) ‘The function of representing Los Angeles as being south of Portland.’ (2) ‘The function of representing Portland as being such that Los Angeles is south of it.’ (3) ‘The function of representing Los Angeles as being an x and Portland as being a y such that x is south of y.’ It should be clear how this applies to all truth-apt complex sentences. For example, a token of ‘Grass is green and snow is white’ has the function of representing grass as being some way, but we can also pick out this unique representing-as function with ‘the function of representing snow as being-an-x-s.t.-grass-is-green-and-x-is-white’. 68 We are now in a position to see how the Functional Theory accounts for the unity of the sentence and solves the special demarcation problem. The Functional Theory accounts for the unity of the sentence because the theory claims that a truth-apt entity’s truth-conditions are determined by its representing-as function and not what that entity merely represents. ‘The Sun is a god or the Sun is not a god’ represents the Sun as being contradictory ways because its parts have the function of doing this. One disjunct has the function of representing the Sun as being a god and the other has the function of representing it as not being a god. But the disjunction itself 67 I have not answered whether we pick out an equivalent act with ‘representing Portland as being north of Los Angeles’. I believe an equivalent act is not picked out, but given everything I have said here, it is an open option to claim that the act is equivalent. Unfortunately, providing underlying criteria to establish whether this locution does not or does not pick out the same act as ‘representing Los Angeles as being south of Portland’ is beyond the scope of this dissertation. 68 There is a tempting alternative here. Instead of claiming that these sentences have a unique representing- as function, why not instead claim that truth-apt entities have a unique representing-as function up to logical equivalence? The idea would be that even though ‘Los Angeles is south of Portland’ has at least two representing functions-as involving representing Los Angeles and Portland as being different ways, there is a sense in which these representings are logically equivalent. The problem with this move is that a written report that has two tokens of ‘Snow is white’ written on it is not truth-apt even though its two representing- as functions that it inherits from its proper parts are certainly logically equivalent. There might be some way of modifying this move to deal with such a case, but more investigation is need here. 58 does not inherit these functions. Instead, it has the function of either representing the Sun as being a complex way or representing the proposition that the Sun is a god as being a complex way. To illustrate what is going on here, consider a clock that has gears as parts. The function of a gear g1 might be to make another gear g2 spin, but it is not one of the clock’s functions to spin g2. Spinning g2 is something the clock merely does in order to serve its function of telling the time. 69 Similarly, the disjunction counts as representing the Sun as being contradictory ways because of the functions of its parts, but the disjunction itself does not have the function of representing the Sun as being contradictory ways. And this is why the disjunction is not necessarily inaccurate. As for the special demarcation problem, this problem arose because there are entities that are not truth-apt but represent things as being ways. Agents represent things as being ways when they perform public acts of representation, but agents themselves are not truth-apt. John is not true when he is communicating and representing the Sun as being yellow. The Functional Theory easily solves this problem since agents do not have the function of representing things as being ways. On the other hand, truth-apt artifacts like sentences tokens do have the function of representing things as being some way. The more difficult issue that the special demarcation problem raises is how we account for the fact that maps are accuracy-apt but not truth-apt. Maps have the function of representing things as being ways just like truth-apt sentence tokens. The Functional Theory gets the right result here because maps do not have a unique representing-as function. Maps have many representing-as functions. A map might have the function of representing Los Angeles as being south of Portland and also have the function of representing New York as being east of Portland. On the other hand, a token of ‘Los Angeles is south of Portland and New York is east of Portland’ has the unique representing-as function of either representing Los Angeles or the proposition that Los Angeles is south of Portland as being some complex way. This is why a token of this conjunction is truth-apt whereas the map is not. Why think the map has many representing-as functions? Imagine we have a sign that reads ‘The Sun is a god or the Sun is not a god’. We then scratch the sign with a knife so that only the 69 This brings out that the difference between having the function of representing and merely representing is more complicated than the difference between what artifacts and agents count as doing. The clock example shows that there is a way for an artifact to count as φ’ing by having parts that have the function of φ’ing such that the parts φ’ing contributes to the clock performing one of its functions. 59 left disjunct is legible and the sign reads ‘The Sun is a god’. The sign no longer serves its original representing-as function. 70 Now, imagine that we have a map of North America. Among other things, it represents Canada as being North of the U.S. Now we cut out the top part of the map that contains Canada. It is not as if the map stopped serving its representing-as functions altogether. It used to have the function of representing Canada as being ways, but now it does not. Still, it has many functions of representing many other things as being many other ways. Cutting out Canada just made it such that the map served fewer representing-as functions. On other hand, scratching the sign “disabled” the signs original unique representing-as function. There are more data to support the Functional Theory. A speech that consists of many truth- apt sentences is neither true nor false. This is because the speech has many representing-as functions whereas each sentence only has one. Furthermore, imagine a sign that reads ‘I am Barrack Obama’. Now imagine you and Obama are holding the sign while standing near a freeway. Is the sign true? The sign is not truth-apt because it serves multiple representing-as functions. The sign is only partially accurate because it represents Obama as being how he actually is but it represents you as being as you are not. In fact, we can start to see how the Functional Theory might help us with currently puzzling semantic phenomena where we do not want to say that a given indicative sentence is true or false even though it is partially accurate. 71 For further support, imagine that a speaker is giving a speech that is being broadcast to both Earth and Twin Earth. The speaker intends to be speaking both English and Twinglish. The speaker says ‘Tigers are endangered’. There are no tigers on Twin Earth but only qualitatively identical twigers that Twinglish speakers refer to with the Twinglish word ‘tiger’. Given her intentions and her audience, the speaker’s sentence token has two representing-as functions. It has the function of representing tigers as being endangered but it also has the function of representing twigers as being endangered. The Functional Theory gets the right result that this token is not truth-apt. If tigers are endangered but twigers are not, we do not want to claim that the token is either true or false. Cases like these show the clear superiority of the Functional Theory over minimalist theories of truth-aptitude. The sentence on the sign and the sentence in the speech both 70 Perhaps the original sign went out of existence because its original representing-as function was essential to it. 71 For example, consider a token of the sentence ‘My computer, which I bought in 1992, is broken’. If my computer is broken but I did not buy it 1992, is this token sentence true or false? The Functional Theory gives a plausible answer. The token is not truth-apt because it has multiple representing-as functions. The sentence is accuracy-apt though. It is partially accurate. 60 pass all the relevant syntactic tests but are not truth-apt. Minimalism gets this wrong. The Naïve Representational Theory gets this wrong. The Functional Theory gets it right. To summarize, the Functional Theory distinguishes itself from the Naïve Representational Theory by claiming that all truth-apt entities have unique representing-as functions that determine their truth-conditions. This gives rise to some complications. However it also allows the Functional Theory to overcome the serious problems that the Naïve Representational Theory faces. The Functional Theory can account for the unity of the sentence and overcome the special demarcation problem. 8. Consequences for the Nature of Propositions If our findings are right about the nature of truth-aptitude, then all truth-apt entities have unique representing-as functions. This has important implications for giving a reductive metaphysics of propositions. In this section I will first go over a couple of ways of thinking about the roles of propositions in our theorizing. Next, I will show how our findings on truth-aptitude rule out just about every metaphysically reductive theory of propositions, including those recently put forward by King, Speaks, Hanks, and Soames. Finally, I will give a full account Artifactualism and show how it is consistent with the Functional Theory of truth-aptitude. As we saw in chapter 1, there is an important division in how propositions are thought about. On the one hand, there is the conception of propositions where they are entities that are the things asserted, the things believed, and are bearers of truth and falsity. We will say that this is the metaphysically robust conception of propositions. On the other hand, some creative philosophers have sought to show that we do not need metaphysically robust propositions. We just need appropriate theoretical tools for modeling linguistic and mental phenomena. Some of these theoretical tools are things that the theorist endows with truth-conditions and other intentional properties and we call these things ‘propositions’. 72 On some conceptions, the theorist not only endows a tool with truth-conditions, but intentionally creates a type of theoretical tool that has its 72 Sometimes the theoretical tools that get called ‘propositions’ are not endowed with truth-conditions. This happens when the theorist uses some entity to represent the conditions under which a sentence is true and calls these conditions a ‘proposition’. 61 truth-conditions essentially. 73 Our findings on the nature of truth-aptitude only have consequences for the metaphysically robust conception of propositions. Those sympathetic to a metaphysically robust conception of propositions agree that propositions must have some important properties beyond being the things asserted, the things believed, and being bearers of truth and falsity. They agree that propositions must have their truth-conditions essentially and that propositions were not intentionally endowed with their essential truth-conditions by the decisions of agents. It would be simpler to suppose that all those sympathetic to a metaphysically robust conception of propositions hold that propositions have their essential truth-conditions independently of the contingent intentions, actions, beliefs, and conventions of agents. However, King (2014b) claims to give us an account where propositions indirectly receive their essential truth-conditions via the contingent actions of agents. So for dialectical purposes, I am only assuming that propositions have their truth-conditions essentially and are not intentionally made this way by agents. If we accept a metaphysically robust conception of propositions, our findings on truth- aptitude entail that propositions essentially have representing-as functions and that agents did not intentionally endow propositions with their essential functions. This places an incredibly demanding constraint on what propositions can be. We need to find some entity that essentially has a representing-as function and that was not intentionally made this way by agents. Accordingly, propositions cannot be sets of possible worlds, 74 functions from worlds to truth- values, ordered pairs of objects and properties, or any mathematical structure. No mathematical structure essentially and without the intentional help of agents has the unique representing-as function of representing snow as being white. So no mathematical structure can be the proposition that snow is white. Even some non-reductive theories are threatened by the Functional Theory. Frege (1997c) claims that propositions are abstract entities from the third realm. But if these entities are not the sort of things that can have functions, then this account of propositions must also be false. The entities alleged to be propositions in recent theories also cannot be propositions. According to Speaks (2014), propositions are properties. The proposition that snow is white is the property of being-such-that-snow-is-white. According to King (2014b), propositions are facts, i.e. states of affairs. The proposition that snow is white is, roughly, the existential state of affairs of 73 See Soames (2010: pp. 80-98) for the details of how such an account would go. 74 Stalnaker (1976) presents such a view. 62 there being a syntactic structure that encodes whiteness being ascribed to snow. 75 For Hanks (2011) and Soames (2015), propositions are types of actions. For Hanks and Soames the proposition that snow is white is the act-type of predicating whiteness of snow. These entities do not have any function. Consequently they are not truth-apt and cannot be what propositions are under a metaphysically robust conception of propositions. These entities do not just fail to have essential truth-conditions. They lack truth-conditions altogether. The property of being-such-that-snow-is-white does not have the function of representing snow as being white. The same goes for an existential state of affairs that consists of a syntactic relation ascribing whiteness to snow. This thing has no function. 76 Similarly, the act-type of predicating whiteness of snow does not have the function of representing snow as being white. Seeing why act-types lack representing-functions is a bit more involved though. Agents might perform the act of predicating being-white of snow in order to serve some further purpose of theirs, but in the relevant sense of ‘function’, neither the act-instances nor the act-type have the function of representing snow as being white. An act-instance of predicating whiteness of snow is not like a sentence token that has the function of representing snow as being white. If it were like the token, then the act-instance would be truth-apt, but that is absurd. Particular acts are not true or false. 77 Like the act-instances, the act-type also lacks a function of representing. Perhaps the act-type “inherits” certain representational properties from its instances. But since the act-instances lack representing-as functions, we have no reason to think the act-type has a representing-as function. 75 As we see in King (2014b: p. 50), the things King calls ‘facts’ include an object possessing a property, n objects standing in an n place relation, and so on. So for King, facts are what are traditionally called ‘states of affairs’ and propositions are existential states of affairs. 76 King (2014b) presents a clever account of how the intentions and actions of agents indirectly lead to the creation of propositions that essentially have their truth-conditions. However, King at best gives us an entity that essentially represents. Our demanding account of truth-aptitude makes his view implausible because it is implausible that any state of affairs could have the function of representing things as being some way. Neither directly nor indirectly could agents create a state of affairs that essentially has a function. Now King claims that sentence-types are states of affairs, so he might object that because sentence-types have functions, he can conclude that some states of affairs have functions. But he would still need to show that the existential states of affairs he claims are propositions have functions. King can only get the result that some states of affairs have functions by relying on the pre-theoretic assumption that sentence-types have functions and then giving a revisionary account of what sentence-types are. We do not have pre-theoretic judgments that propositions have functions. So, King cannot give us independent reasons for why we should think that the existential state of affairs that he claims are propositions have functions just like the states of affairs he claims are sentence-types. To paint a picture, the existential states of affairs that are propositions in King’s theory are by-products of what agents do with the states of affairs that are sentence-types. These by-products have no function. 77 Soames (2015) acknowledges this fact although he does claim that act-types are truth-apt. See footnote 32 for more on acts being truth-apt. 63 So in short, none of the entities King, Speaks, Hanks, and Soames allege to be propositions have any representing-as functions. They are therefore not truth-apt. Consequently they cannot be what propositions are under a metaphysically robust conception of propositions. At this point, one might begin to suspect that our findings on truth-aptitude show that no entity could possibly have all the properties that propositions need to have under a metaphysically robust conception of propositions. There is the move of positing sui generis entities that essentially have unique representing-as functions. But we ought to do better than this if we can. But as we saw in chapter 1, there are entities that essentially have unique representing-as functions and are this way without the help of agents. Here again are the reasons why this is so. First, suppose there are functional-types which include things such as tool-types. 78 The arrowhead is a tool-type whose instances all share the common function of being for piercing flesh. If there is such a tool-type, it would be reasonable to suppose that this type essentially counts as having the function of piercing flesh. No particular rock essentially has function of piercing flesh, but the type/kind that is the arrowhead is this way. Roughly, the type is this way because of its membership criteria. All instances of the arrowhead must have the function of piercing flesh in order to be an instance of this type. Having this function explains why a given rock is an instance of this type. This is why the type itself counts as having the function of piercing flesh and we can sensibly say that the arrowhead is a device for piercing flesh. 79 For the same reason, the triangle, a type of shape, counts as being three-sided. The tiger, a type of animal, counts as being a mammal. Roughly speaking, the type “inherits” certain properties form its instances given the type’s membership criteria. A type like the arrowhead has its membership criteria essentially, so the arrowhead essentially counts as having the function of piercing flesh. Sentence tokens are artifacts or tools just like arrowheads. As we have seen, truth-apt sentence tokens have unique representing-as functions. Everyone already agrees that two tokens of ‘Snow is white’ both belong to a common type that is the English sentence ‘Snow is white’. 78 Ali Khalidi (2015), Thomasson (2003), Searle (1995), and Boyd (1991) argue for various views of kinds that allow for kinds of artifacts. Hawthorne and Lepore (2011: 480-485) present some worries for these inclusive approaches. Ali Khalidi (2015) provides a good overview of the debate and also a nuanced view about which agent-dependent kinds there are. Also see Massimi (2014) for an interesting Kantian alternative that also allows for the possibility of kinds of artifacts. 79 See Carlson and Pelletier (1995) for a collection of articles on the semantics of generics that share the common view that we ought to treat locutions like ‘the clock’, when used this way, as picking out a kind. 64 Sentence-types are plausibly taken to be functional-types since their instances must be particular communicational tools. 80 Now suppose that there is another functional-type that tokens of ‘Snow is white’, ‘La nieve es blanca’, ‘Schnee ist weiß’, and so on all belong to. This functional-type is such that all its instances must have the unique representing-as function of representing snow as being white. Given what we are supposing about membership criteria, it follows that this functional-type essentially counts as having the unique representing-as function of representing snow as being white. Again, roughly speaking, the type “inherits” its representational properties from its instances just as the Tiger “inherits” the property of being a mammal from its instances. But the type essentially counts as having a unique representing-as function because the type essentially has its membership criteria. Given that this functional-type essentially counts as representing snow as being white and given our findings on truth-aptitude, it follows that this functional type essentially counts as being such that it is true iff snow is white. Therefore, this functional type has its truth-conditions essentially and without the help of agents. This gives us very good reason to suppose that this functional-type is the proposition that snow is white. In general, it gives us good reason to suppose that propositions are functional-types whose instances all share a unique representing-as function. As we know, Artifactualism just is the view that propositions are these functional-types. Importantly, Artifactualsim is a view where propositions are not the primary representers. Rather, given their membership criteria, propositions count as representing and having truth-conditions by “inheriting” certain properties from worldly truth-apt things. 9. Concluding Remarks Our investigation into the nature of truth-aptitude led to a demanding account where only entities that have unique representing-as functions are truth-apt. We saw early on that there is a link between representation, accuracy, and truth by considering the relationship between a dot on a map, a map, a written report, and a token sentence on a written report. But without distinguishing the notion of having the function of representing from merely representing, we could not account 80 The fact is that we attribute all sorts of properties to linguistic-types in our semantic theorizing without making good on how a type or kind could have these properties. Kaplan (1990) and Wetzel (2009) are two of the few instances where these problems are taken up head on. But the lack of agreement in literature that has since developed in response to Kaplan (1990) is a testament to the problematic state we are in. 65 for the unity of the sentence or solve the special demarcation problem. The Functional Theory of truth-aptitude solved these problems by claiming that only entities with unique representing-as functions have truth-conditions. This conclusion about the nature of truth-aptitude ii lead to a very demanding constraint on what propositions must be. The constraint ruled out every reductive theory of propositions put forward so far. Only Artifactualism, the view that propositions are functional-types, appears positioned to meet the demands of the correct theory of truth-aptitude. 66 Chapter 3: Living without Representational Mental States Abstract That there are representational mental states is a widely held assumption in the philosophy of mind. This chapter argues that we ought to reject this assumption. The argument has two parts. The first part is to provide a plausible account of denying that there are representational mental states. I offer an account that both allows us to accommodate important truisms of folk-psychology and is not overly restrictive when it comes to which theories in the philosophy of mind it is compatible with. The second part is to show that once we rid ourselves of representational mental states we can make progress on problems we face in metaphysics, the philosophy of language, and the philosophy of mind. Its, perhaps, most important consequences, outside our inquiry into the nature of propositions, concerns the current debate on the nature of perception between relationalists and representationalists. 1. Introduction In chapter 1, we saw that we had two options for giving an account of propositional attitudes once we accept Artifactualism. One upshot of this chapter is that we ought to accept a view of propositional attitudes that does not require propositions to be mental entities. So, we can hold on to the view that propositions are kinds of artifacts rather than more general functional kinds whose instances include biological states. Our primary concern in this chapter, however, is to further investigate the nature of representation and to discover whether there are any good reasons for either positing or denying the existence of representational mental states. The argument has consequences for accounting for the nature of propositional attitudes, fitting representation into the natural order, what propositions are, and the nature of perception. It is worth saying again that our primary concern here is the nature of representation and whether there are representational mental states but it will quickly become clear how this question bears on Artifactualism. 67 What are we denying when we deny that there are representational mental states? 81 We are denying that mental states are about things, represent things as being ways, have accuracy- conditions, and have truth-conditions. Given that there are, for example, true beliefs that are about, say Abraham Lincoln, how can we possibly deny the existence of representational mental states without giving up on the most basic truisms of folk-psychology? Three simple assumptions will allow us to deny that there are representational mental states without barring us from accommodating the truisms of folk-psychology and without severely restricting our theoretical options for accounting for the nature of mental states in general. The first assumption is that representing is distinct from indicating or any other closely related notion. The second assumption is that the relevant relata of belief reports, traditionally thought to be propositions, necessarily are representational, have accuracy-conditions, and have truth- conditions. The third assumption is that our attitude words like ‘belief’ are ambiguous/polysemous 82 between the attitude of believing and what is believed. With these three assumptions we give an account where, although we refer to representational entities to help pick out mental states, the states themselves are not representational. The states and the representational entities we use to pick them out do share important common properties, but being representational is not one of them. The upshot of giving this account is that we will be able to make progress on problems in metaphysics, the philosophy of language, and the philosophy of mind. Section 2 covers the various reasons for positing that there are representational mental states. Section 3 enumerates and motivates the assumptions we will make about the nature of representation. Section 4 provides a plausible way of denying that there are representational mental states. Section 5 shows that this way of denying that there are representational mental states allows us to still account for our pre-theoretical judgments in folk-psychology. Section 6 shows that we can do this without severely limiting our theoretical options when it comes to accounting for the nature of mental states in general. Section 7 argues that our view allows us to give a naturalistic account of how there come to be representational things. Section 8 shows that our view has important implications concerning 81 What we are denying does not rule out the possibility that, contrary to how things are, there could be creatures that have representational mental states. 82 It does not matter to our purposes whether we claim that there are homophonous words pronounced ‘belief’ or claim that there is one word ‘belief’ with multiple related but distinct senses. The former would entail that the attitude words are ambiguous. The latter would entail that the attitude words are polysemous. 68 Artifactualism. We learn that we have reason to prefer a view of propositions where propositions are not mental entities. Finally, section 9 shows that our way of denying that there are representational mental states has important consequences for the current debate on perception going on between relationalists and representationalists. 2. Background and Motivation There are two broad groups of reasons for positing the existence of representational mental states. The first group of reasons concerns our pre-theoretical judgments about the nature of propositional attitudes as well as perception. The second group of reasons concern theoretical motivations that arise from trying to give a general account of various aspects of mental states. We will go over these reasons in turn. Some of our pre-theoretical judgments about mental states seem to give us extremely strong prima facie evidence that there are representational mental states. We hold that people have accurate/inaccurate/true/false beliefs and judgments. We hold that an agent’s faculty of perception can be more or less accurate. If we assume, as I think we should, that being representational is necessary for being accurate/inaccurate/true/false, 83 then it is almost irresistible to conclude that there are representational mental states that underlie belief, judgment, and perception. Believing, judging, and perceiving are, or at least are partially constituted by, mental states. A related motivation for positing representational mental states concerns occurrent judgments and belief formation. We are prone to say that A formed his belief about God and the origin of the Universe when A came to represent God as the creator of the Universe in his thoughts. Talk like this makes it natural to conclude that A judging as well as occurrently believing, that o is F is, in some sense, closely related to A representing o as being F in A’s thoughts. And how could A represent o as being some way in his thoughts without A being in a mental state that represents o as being some way? Finally, we pre-theoretically judge that to believe that o is F is to have an attitude towards o and the property of being F. It is very natural to conclude that having an attitude towards o is one 83 See Soames (2011) as well as Cappelan and Hawthorne (2009) for a more in depth treatment on the relationship between being-true-or-false and having truth-conditions. I will treat these properties as being one and the same for sake of simplicity. 69 and the same as having an attitude about o. What kind of entities can be such that they have the property of being-about-o? Entities that represent o or represent o as being some way are about o. We now turn to theoretical motivations for positing representational mental states. The two broad categories of reasons here concern trying to give a naturalistic account of mental states and trying to explain the peculiarities of perception. For sake of simplicity, let’s assume we can uniquely pick out what beliefs, desires, perceptions, and so on are by constructing a giant Ramsey-sentence that consists of all our pre-theoretical judgments we make when we engage in folk-psychology and then replacing the terms for psychological words with variables bound to quantifiers. 84 . There will be a lot in common between the properties ascribed to the things picked out by, for example, ‘believing that the sky is blue’ and ‘imagining that the sky is blue’. In general, the Ramsey-sentence will ascribe a lot of common properties to the things picked out by phrases of the form ⎡φ’ing that o is F⎤. And so it is reasonable to suppose that the things picked out by ⎡φ’ing that o is F⎤ will also share more basic underlying properties. For example, it could turn out that each thing picked out by ⎡φ’ing that o is F⎤ is some functional state and that the function of it somehow pertains to the object o and the property of being F. So far though, nothing commits us to the existence of representational mental states. Now, it is, or at least was, a very popular view that what all the things picked out by ⎡φ’ing that o is F⎤ share in common is that they partially consist of mental symbols that represent o as being F. The Computational Theory of Mind (CTM), most strongly argued for by Jerry Fodor, 85 is that inevitably our Ramsey-sentence is going to pick out states where the state picked out by ⎡φ’ing that o is F⎤ is a sentence in the language of thought standing is some computational relation in the agent’s brain. CTM is a more specific version of the Representational Theory of Mind (RTM). RTM is the claim ⎡φ’ing that o is F⎤ must partially consist in a mental state that represents o as being F. 86 A second theoretical motivation arises from trying to explain the peculiarities of perception. A promising way to explain what goes on when we hallucinate is to claim that our perceptual 84 That we can do this is basically assumed in Lewis (1972). For our purposes, it does not matter if we add any truths that we learn through the science of mind into the Ramsey-sentence. 85 See Fodor (1975) and well as Fodor (1990). 86 There is another theoretical motivation for positing representational mental states that has to do with CTM. The basic idea is the claim, made by Fodor (1975), is that there can be no computation without representation. So if the mind is sufficiently like a computer, then there must be representations that the mind uses in computing. It is beyond the scope of this article to argue that this is false, but see Piccinini (2008) for a good argument to this end. 70 states are much like our belief and judgments states. Just as we can inaccurately believe and judge, we can inaccurately perceive. The way this is spelled out almost always involves representational mental states. 87 The idea is that perceptual states “have content” just as belief and judgment states have content. The story goes, for an entity e to have content is to be representational in a way such that e has accuracy-conditions. So, the common conclusion is that perceptual states have accuracy-conditions just like belief-states. 88 3. Features of Representation There is a way of understanding the notion of representing that makes the conclusion of my argument seem wildly implausible. It is not uncommon for those working in the philosophy of mind to use the terms ‘content’, ‘representation’, and ‘about’ as technical terms, roughly defined in the following way. Assuming that all states we pick out with ⎡φ’ing that o is F⎤ must share common underlying features, content is simply taken to be whatever these shared underlying features are. To represent o as being F is taken as merely being a state whose content is picked out with ‘that o is F’. Anything that represents o as being F in the above sense is taken to be about o. If we define ‘content’, ‘representing’, and ‘about’ in these ways, then I do not deny that there are “representational” mental states. The problem with using these words as technical terms is that, as I will argue, they lead to confusion that hinders progress in various areas. ‘Content’ defined in this way is not one and the same as what it is to be content of sentences, diagrams, and other representational artifacts. ‘Representing’ defined this way is not the same as our pre- theoretical notion of representing. The same goes for ‘about’. Although it is not often talked about, 89 we have pre-theoretical judgments about representation, accuracy, and truth. Long before we theorize about linguistic or mental 87 See, for example, Evans (1982), Searle (1982), Dretske (1995), Chalmers (1996), and Byrne (2001). 88 Burge (2010) includes novel arguments that aim to show that the nature of perception entails the existence of representational mental states. The arguments basically claim that the success of certain explanations in mainstream psychology concerning perception require that perceptual states have accuracy conditions. Ganson, Bronner, and Kerr (2014) argue against Burge and claim that the success of these explanations only require states to have “functional sensitivity and responsiveness to stimuli”. Although we can examine the details of this debate here, it will turn out that our way of denying that there are representational mental states still allows someone like Burge to claim that perceptual states must share certain important features with belief states and these features are like having accuracy-conditions. 89 Goodman (1968) takes up just this project of explaining the pre-theoretical data concerning representation. 71 phenomena we know that the table in front of us neither true nor false. The cloud in the sky is neither accurate nor inaccurate. Jupiter does not represent anything as being any way. Jupiter is not about anything. Content on the other hand is a technical notion, but when it comes at least to complete well-formed indicative sentences, if we know their content, then we know the sentence’s truth-conditions, accuracy-conditions, and what the sentence represents as being what way. Turning to our assumptions, representing is both a public act agents perform and a function artifacts can perform. An agent can represent the Sun as being a god while telling a story to a group of people. An inscription in the dirt can have the function of representing the Sun as being a god. Complex agents like ourselves undeniably perform mental acts where we represent things to ourselves. We talk to ourselves all day. We visualize things to help in our reasoning. Our most important assumption is that to represent o as being F is not to indicate that o is F. To indicate that o is F is to provide evidence that o is F. Representing the Sun as being a god does not, in and of itself, provide evidence that the Sun is a god, although one could indicate the Sun is this way by representing the Sun as being this way. 90 What is the difference between a drawing of Che Guevara that represents him as wearing a hat and a photograph of him wearing a hat? The function of the drawing is, inter alia, to put you in a state of mind where you are thinking of Che as wearing a hat. 91 The photograph itself has no such function although someone can use the photograph to make you think of this. The photograph does indicate, given its nature, that Che was wearing a hat. The drawing can also have the function of indicating that Che was at one point wearing a hat, but the drawing need not have this function. In general even though there can be relationships between indicating that o is F and representing that o is F, they are distinct notions. To indicate that o is F is to provide evidence that o is F. To represent o as being F is, roughly, to attempt to gets ones audience to think of o as being F for some end or other, such as telling a fictional story or making an assertion. As for accuracy, there is a broad and a narrow notion of accuracy. An action is broadly accurate to the degree it accomplishes its intended outcome. An artifact is broadly accurate to the degree is accomplishes its function. The notion of narrow accuracy is just being broadly accurate 90 Furthermore, providing evidence that o is F is not one and the same as being a causal result of o being F. Bent rays of light indicate, i.e. provide evidence that the Universe is expanding, but they would still do this even if they were caused by something else. 91 Note that this is not one and the same as serving of the purpose of getting you to judge that Che is wearing a hat. 72 in representing things as being as they are. A map of the United States has, among other functions, the function of representing Los Angeles as being south of New York. The maps (narrow) accuracy-conditions are just the conditions under which the map represents the United States’ geography as being exactly as it is. Finally, we will assume that being representational is a necessary condition for having accuracy and truth-conditions. A map that is made more accurate by o being F, represents o as being F. An entity e that is true iff o is F is such that e at least represents o as being F. We are warranted in assuming this given our pre-theoretical judgments about representation, accuracy, and truth. A written geography report that includes the sentence ‘Los Angeles is south of New York’ will be more accurate if Los Angeles is south of New York. Why? Well, because the sentence on the report, like the map, represents Los Angeles as being south of New York. And notice that the sentence is true iff it represents Los Angeles as being as it actually is. 4. Denying that there are Representational Mental States Two additional assumptions are needed to plausibly deny that there are representational mental states. Our first assumption is that the relevant relatum of propositional attitude reports are representational. The relevant relatum is traditionally assumed to be proposition, and I think this is correct, but we need not assume that here. The entity just needs to represent some things as being some way, have accuracy-conditions, and have-truth-conditions. For ⎡A φ’s that o is F⎤, it needs to represent o as being F. The range of options open to us here span from Frege’s sui generis abstract propositions 92 all the way to symbol/sentence types from a logically unambiguous language, such as the interpreted logical forms of Larson and Ludlow (1993). 93 It is important to note that in order to assume the relevant relatum is representational, we must also assume that the logical form of propositional attitude reports are such that we can sensibly quantify over the that-clause. 94 Our second assumption is that our attitude words like ‘belief’, ‘desire’, and so on are either ambiguous or polysemous. In one sense of ‘belief’, two people can have the same belief because they believe one and the same thing. When we say that Kevin’s belief about God creating the 92 C.f. Frege (1918) 93 The relatum could even be a natural language sentence-type as Davidson (1968) claims. See Bach (1997) for other considerations and difficulties about what the relevant relatum is. 94 King (2002) gives a lot of good reasons for assuming this. 73 Universe is unshakeable, whereas Jane’s is less strong, we are characterizing the attitudes they have as being certain ways. Jane and Kevin can have the same attitude towards God, but their particular attitudes have different properties. I do not deny that necessarily, an agent A has a belief that p iff A has the attitude of believing that p. Still, what is believed is not the type of thing that is for example, unshakeable. What I believe can be what you deny (i.e. can be identical to what you deny), but my attitude cannot be identical to what you deny. This ambiguity exists for all the cognitive attitude words including ‘desire’, ‘fear’, and so on. 95 You and I can have one and the same desire that there be world peace, although the strength of our desires, i.e. how strongly we each desire this, can differ. We can both have the same fear that a meteor will hit the Earth, but your attitude of fearing this might be justified whereas my attitude of fearing this is unjustified. Our way of denying that there are representational mental states is simple once we have our assumptions on the table. The basic idea is that we refer to representational entities, specifically propositions, to pick out propositional attitudes like believing, but the underlying mental states we pick out are not themselves representational. The reason we use propositions to pick out these mental states is that there are similarities between the representational propositions and the underlying mental states we refer to. However, representing things as being ways is not something that the propositions and the state we pick out with the proposition share in common. Only the proposition is representational. We use the fact that propositions are about things to help categorize and give helpful names to the attitudes of agents. What better way to keep track of an attitude towards the things that that o is F is about than to refer to that o is F in picking out the attitude? As Stalnaker points out, this general way of conceiving of propositions makes them very similar to numbers. He says : The analogy between propositions and numbers suggests that there may be further alternative strategies for explaining how a person can be related to a proposition… What is it about such physical properties as having a certain height or weight that makes it correct to represent them as relations between the thing to which the property is ascribed and a number? The reason we can understand such 95 Reasons why we do not say desires are fears are true or false whereas we do say beliefs and judgments are true or false. Basically we can say what you fear is true and what you wish will become true. Desire involves a different type of relation. 74 properties––physical quantities––in this way is that they belong to families of properties which have a structure in common with the real numbers… That, I think, is all there is to the fact that weights and other physical quantities, are, or can be understood as, relations between physical objects and numbers. (Stalnaker 1987: pg. 9). Loar (1981: pg. 142) puts forth a very similar idea. Now when it comes to propositions and mental states, the question is what properties of mental states make them such that they have a structure in common with propositions? One can jump to the extreme conclusion that a mental state picked out with that o is F represents exactly what that o is F represents and therefore has the same truth-conditions. But there are obviously other options and we will cover them is section 5. For now, we are only making the very weak claim that a state we pick out with that o is F, for example when we say ‘A hopes that o is F’ is such that the mental state is, or at least partially consists in, having an attitude that is towards the things this propositions is about. The mental state is identical to or partially consists in an attitude that is towards o and is towards the property of being F. We are making this weak claim and furthermore denying that what they have in common is representing o as being F. Now one reason for doubting our account is that it might seem hard to square with the fact that the proposition that o is F is the object of A’s attitude when A believes that o is F. If we assume that we must stand in a cognitive relationship to the things we have attitudes towards, i.e. must be able to think of them, and assume that to be the object of an attitude is to be what an attitude is towards, it will follow that we must stand in a cognitive relationship to propositions if we have propositional attitudes. For example, we say our attitude is towards a particular spider when we fear the particular spider that is running across our floor. We also say that our attitude is towards a proposition when we fear that a meteor will hit the Earth. Does this entail that the relationship we stand in to the particular spider when we fear it must be one and the same relationship we stand in to the proposition that p when we fear that p? Standing in a causal relation to that spider is important part of what it is to fear it. Accordingly, propositions would also have to be something that we are causally related. Given our assumption that the relevant relata are representational, they would have to be something like causally efficacious representational mental states. 75 So, we need to make sense of what it is for a proposition to be an object of a propositional attitude if we are going to plausibly deny that there are representational mental states. Yes, propositions are the objects of attitudes like belief, but unqualified talk about which object an attitude is ‘towards’ can be misleading. If Aristotle generally wants a dog, in one sense his attitude is towards nothing in particular. In another sense of ‘towards’ we can say his attitude is towards whatever our analysis tells us the object of the wanting-relation is in this case. We are going to be forced to posit a relatively exotic entity as the object of the attitude. ‘Aristotle wants a dog’ of course represents Aristotle as having an attitude, but an attitude towards what? Van Geehoven and McNally (2005) treat complements of intensional verbs as denoting complex properties. Richard (2001, p. 111) claims that the complement of such verbs denote ‘possible worlds intensions of noun phrases’. Whatever the object is, it is not going to be something a child needs to stand in a cognitive relationship to in order for the child to want a dog in general. 96 Once a child runs into things that are certain way, e.g. that have the property of being-a-dog, it does not take much for the child to have a general attitude towards things that are this way. In the object-of-attitude sense of ‘towards’, there are cases where agents need not stand in a cognitive relationship to what their attitudes are towards and A can have an attitude where the proposition that o is F is the object of the attitude even though A cannot cognize it. There is another sense of ‘towards’, where having an attitude towards o does require the ability to think about o. When A fears a particular spider, this requires A to be able to think about this spider. Once we separate these sense of ‘towards’ we see that we are not committed to the existence of representational mental states even though representational entities are the objects of propositional attitudes. We merely use representational entities to pick out mental states. The mental states share some properties with the propositions we use to pick them out, but being representational is not one of those properties. 5. Accommodating the Truisms of Folk-Psychology The task now is to show that we can still accommodate the pre-theoretical judgments we make when we engage in folk-psychology. Specifically, there are three truisms we need to accommodate. First, beliefs and judgments are representational, about things, accurate/inaccurate, and true/false. Second, we often describe belief and judgment formation in terms an agent 96 I take Anscombe(1981) as inspiration for this way of thinking about the objects of attitudes. 76 coming to represent things as being some way in her thoughts. Third, propositional attitudes are attitudes towards things. First, beliefs and judgments, understood as what is believed and what is judged, are representational, have accuracy-conditions, and have truth-conditions. This just amounts to the fact the entities we use to pick out the mental states/attitudes are representational and so on. A has a true belief when what she believes is true. A’s judgments is accurate when what she judges is accurate. The states she is in when she believes and judges do not have truth or accuracy conditions. The states do not represent things as being ways. The same goes for thinking true/false thought. Note the difference between talking about the mental act that is picked out by ‘having the thought that p’ and the object of the act, i.e. what is thought, which is indeed true or false. It is clear upon inspection that to say that A had the thought that God created the Universe is to only say that A performed some mental act; it is not to say that A posses or contains what he thought. 97 In short, A can have the true/false thought that o is F without being in a mental state that represents o as being F. We can even allow that believing and judging can be accurate if we remember the distinction between broad and narrow accuracy. We can think that these propositional attitudes are somewhat akin to intentional actions, in that (some of) these attitudes have an intended outcome built in. 98 For example, we can be broadly accurate believers and judgers, because part of believing either does or ought to consist in only believing that o is F when o is indeed this way. 99 We are just denying that correctly believing that o is F requires being in a mental states that is itself narrowly accurate, i.e. accurately represents things as being as they are. Turning to the second truism, it often seems natural to say that when A judges or forms a belief that o is F, A represents o as being F in A’s thoughts. Talk like this makes it natural to conclude that A having an attitude of believing that o is F is at least in some sense closely related to A representing o as being F in A’s thoughts. Now, even if we grant that judging that o is F requires performing the private mental act of representing o as being F in our thoughts, this does not entail that there are representational mental states. The state underlying the act of representing, need not, for example, have accuracy conditions. But I think we should also deny 97 Schroeder (2008) makes a similar point for the notion of having a reason to φ. 98 See Chan (2013) for a range of views on these issues as well as Schwitzgebel (2013). 99 Such norms, are likely why we talk about true beliefs and judgments and eschew talk of true hopes and wonders. 77 that representing o as being F is a necessary part of judging that o is F. As a matter of fact we complex agents talk to ourselves and visualize things to ourselves, and in this way we make judgments by representing things to ourselves. But simple agents need not represent anything to make basic judgments. There is no pre-theoretical reason for thinking that an animal cannot judge that their prey went in a certain direction without representing their prey as being a certain way. It is just the fact that our most striking cases of judgment often involve representing things to ourselves. But you can still judge in a split second that bridge is unsafe to walk on without representing this to yourself. We now turn to the third truism that our propositional attitudes are attitudes towards objects and properties. When A believes that the Sun is a god, A has an attitude towards the Sun. We have already shown that A’s belief, as in what is believed, is indeed about the Sun. And the ambiguity in attitude words makes it easy to think that the attitude itself and the mental states that (partially) constitute it must also be about the Sun. But this obviously does not follow. Still we have the question of what it is for an attitude to be towards one object rather than another. The very rough answer is that the nature of the attitude partially consists in standing in important relations to whatever the attitude is towards. The next section will cover our options for strengthening this claim. 6. Options for Accounting for the Nature of Mental States We will now show that our way of denying that there are representational states is not overly restrictive when it comes to giving a theory as to what mental states generally are. There are various ways of categorizing the views one can take on this question. But I think the most insightful way to do this is via the question of what are the structural relations between mental states and the propositions we use to pick them out. We can roughly order the alternatives by how many important relationships hold between propositions and the states we use to pick them out. The strongest view is arguably Fodors’. According to Fodor’s view the mental states represent the same things as the propositions, have similar syntactic structure to propositions (according to his view of propositions), and stand in computational relationships that mirror the inferential relationships of propositions. The weakest view is arguably eliminativism as endorsed in Churchland (1981). Because this view claims that there is nothing we succeed in picking out with our attitude reports, it trivially is such that there are the least amount of things in common between what we pick out with the propositions and the proposition itself. What I intend to show 78 is that our way of denying that there are representational mental states only rules out the very strongest view. Even though we cannot allow that the states we pick out by referring to that o is F represent o as being F, we can allow for something almost as strong. Recall our assumption that representing o as being F is not one and the same indicating that o is F. It is also not one and the same as being a causal result of the states of affairs of o being F. This means it is not one and the same as “reliably indicating” that o is F. 100 So one could accept our view and also take on a view like either Fodor (1990), Field (1978), Stalnaker (1987), Loar (1981), or Dretske (1988) which all claim the mental state we pick out with that o is F stands in an important indicational/causal relationship with the state of affairs of o being F. Alternatively, one could take on a view like that in Millikan (1984) where the mental states we pick out with that o is F all have the function of indicating that o is F. We can also allow that the states picked out stand in computational relations that mirror the inferential relations of propositions. We just need to stand fast to the distinction between representing and indicating. Because of this flexibility, one might worry that our way of denying that there are representational mental states cannot be all that important if it does not rule out these views. Aren’t we just being picky about the meaning of ‘representing’? The remaining sections will make clear why it is so important to distinguish between indicating and representing. A weaker view we could take on is that the mental states picked out do not necessarily stand in an indication relation with obtaining states of affairs. Rather, a mental state picked out with that o is F might be a state whose functional role simply has to do with o and with being F. Various functionalists endorse this view. The idea is that once we have constructed our giant Ramsey-sentence, it will fall out of it that all the states we pick out with that o is F share the common property of having a function that has to do with o and the property of being F. 101 We can again allow, as Loar (1981), Block (1986), and Harman (1987) all claim, that there is some 100 Dretske (1999) defines the notion of reliably indicating and puts it to work in accounting for the nature of mental states. 101 Harman (1987) claims that the Ramsey-sentence itself will tell us what the common function is in terms of o and being F. Lycan (1981) also gives a view along these lines. Other versions of functionalism, such as Loar (1981) and Block (1986) hold that we will need to do further investigations to discern how exactly o and being F relate to the underlying functional state, with the upshot being that the state itself will have to stand in indicational/causal relationships with o and being F. So only the functionalism of Harman (1987) ends up being an actually weaker view. 79 underlying relationship between the mental states that mirrors the inferential relations between propositions used to pick them out. Dispositionalist views of mental states, such as the views in Baker (1995) and Switzgebel (2013), are also compatible with our account. According to these views the important relationship that an agent A stands in to o and being F when an agent is in a state picked out with that o is F is that either A has certain dispositions towards o and being F or certain conditional statements that have to do with o and being F are true of A. The remaining weaker views are obviously compatible. There is the view that all the states picked out with that o is F just share the property standing in some important relationship with o and being F, whatever that relation is. There is view that all the states picked out with that o is F just share some common underlying property, whatever it is. There is the instrumentalist view akin to Dennett (1981) that our attitude talk is merely a useful way of picking out physical patterns. There is the Churchland-style view that our attitude talk fails to pick out anything. Since we care about accommodating the truisms of folk-psychology, the Dennett-style view and Churchland-style view are not compatible with the spirit of our proposal. I think we should at least accept that all the states picked out with that o is F are such that they partially consist in standing in an important relationship to o and the property of being F. 102 That is to say, there is still the real problem of specifying what relationship an agent A, that has a propositional attitude picked out by that o is F, stands in to o and the property of being F. The project of giving this account should not be called the project of giving a “semantics” of mental states unless one holds the strongest view about the structural relationship between propositions and mental states. It just is the general question of what relationship an agent needs to stand in to o and the property of being F in order to be in the mental state we pick out with ⎡φ’ing that o is F⎤ when φ is an attitude 102 In many cases it seems that this relationship must be causal. My fear of the spider on the floor certainly partially consists in me standing in a causal relationship to that spider. Me believing that the Eiffel Tower is in Paris likely consists in me standing in a long causal relationship with the Eiffel Tower and Paris. But the same story will not work for our beliefs about Santa and numbers. Schwitzgebel (2013) convincingly argues that the relationship sometimes requiring a causal connection can even be accommodated by dispositional views. 80 verb. So far, no unproblematic account of this relationship has been proposed. And perhaps the relationship is so unnatural and disjunctive that we are unlikely to stumble upon an a priori analysis that gets it right. In any case, these are topics for another time. 7. Solving the Problem of Intentionality Denying that there are representational mental states makes the task of giving a metaphysical account of the nature of representation much easier. Although the so-called “problem of intentionality” is no longer a popular issue, it still remains a real problem. As Stalnaker aptly puts it: The problem of intentionality is a problem about the nature of representation. Some things in the world- for example, pictures, names, maps, utterances, certain mental states- represent, or stand for, or about other things- for example, people, towns, states of affairs…For various familiar reasons, intentional and representational relations seem unlike the relations that hold between things and events in the natural world…The challenge presented to the philosopher who wants to regard human beings and mental phenomena as part of the natural order is to explain intentional relations in naturalistic terms. (Stalnaker 1987, pg. 6) What I am suggesting here is that once we deny that mental states are among the representational things in the world, the problem of giving a naturalistic account representation becomes much easier. On the other hand, if mental states are representational, then we cannot appeal to them in our definition of what is, in general, to be representational and the problem of fitting representational things into the natural order appears nearly impossible. The following is a sketch of a novel account of how representation fits in with the natural order. Nothing is representational before there are agents with beliefs and desires and so on. Even when there come to be agents that have beliefs, desires, and so on, we are still a long way from representational things coming in to existence. Representational entities finally come into existence when agents begin to communicate with each other and themselves. Before any object has the property of being representational, agents first perform public acts of representation. What is it to perform the public act of representing o as being F? Roughly, to perform the public act of representing o as being F is to do some other action with the intention of getting 81 ones audience to imagine or think of o as being F and succeed in bringing this intention about as long as the audience is paying attention. Since we have shown that thinking of o as being F does not require being in a representational mental state, we are not forced to push the existence of representing farther back than these public representational acts. Now in order to intend to make her audience think of o as being F, doesn’t the representer need to have beliefs about her audiences beliefs and thoughts? And haven’t I claimed that we pick out such beliefs and thoughts by referring to representational entities? For example, I have claimed that we pick out the belief that o is F by referring to some entity that represents o as being F. It would seem to follow that the representer must be able to think thoughts about representational things in order to perform public acts of representing. But I am claiming that public acts of representing come into existence before any representational objects come into existence. The way this is possible is that primitive humans did not think of the mental states of their peers in terms of their peers standing in a relation to a representational entity. We modern humans often pick out these states in terms of them standing in relations to representational entities, but early humans did not do this, and we in fact do not always do this. I can think of your belief as the reason or thing that is making you act strangely. For example, a mom might fear that a stranger across the street will abduct her child. The child can start to have an attitude towards her mom’s propositional attitude by thinking of her mom’s propositional attitude as the thing that is making her mom act strangely. Pre-linguistic humans must have started thinking about the attitudes of others in a similar way. In a similar spirit, Butterfill and Apperly (2013, p. 607) claims that agents with limited conceptual sophistication can track the belief states of others without being able to cognize ‘propositional attitudes as such.’ In general, once we see that agents can cognize the propositional attitudes of others without cognizing propositions, we can start to sketch the origins of sophisticated communication that involves intentional and public representing. Why did the lead hunter of our tribe raise his hand? Perhaps he wants us to be quiet. We can cognize this attitude as the thing that made the hunter raise his hand. And now why did he draw that mark in the dirt? The hunter might draw a mark that looks like a boar to serve the function of indicating where we last saw the boar we are hunting. Later that night, we make a similar drawing by the fire to indicate that we are representing the boar as taking a spear to the eye. We are not indicating that this has happened or even that we want this to, or are trying to make this, happen. It just feels good to represent this for 82 ourselves and for the other hunters. We make a tool that has the function of getting our audience to think of the boar as taking a spear to the eye, and thinking this thought puts us all at ease. Bit by bit we form more attitudes towards the attitudes of others until we are able to use gestures, sounds, and symbols to efficiently and systematically indicate what we are trying to represent. We represent for amusement. We represent to aid in our survival. We make tools that represent for us. We start having more and more complex attitudes towards these representational tools. And finally, somewhere down the line, we find that certain kinds of representational tools are very useful for categorizing and keeping track of the attitudes and linguistic behavior of others. 103 This is how representation fits into the natural order. To be sure, there are many more details to fill in, but the task is no longer a daunting metaphysical problem. The task was only a daunting metaphysical problem when we assumed that mental states were representational. When we assumed this, it stopped us from explaining what it is to be generally representational in terms of mental states. 104 103 See Godfrey-Smith (2006) for a similar idea that our folk-psychologizing about mental pheneomena uses features of public symbols as a model for the nature of mental states. 104 Now one might object that we have only pushed back the problem since we still have the mystery of how we can have attitudes towards things that do not exist. In other words, we have shown how representation fits in the natural order, but our attitudes have puzzling features that cry out for explanation. Another way to put this objection is that the problem of intentionality is not so much how representation fits in the natural order but the problem of how we explain the fact that we appear to stand in relations to non-existent things. But I would like to suggest it was a mistake from the beginning to think that giving an account of representation and an account of the mental could on its own deal with these mysteries. After all, even if it turns out to be is false that our universe is expanding, bent rays of light indicate, as in provide evidence, that our Universe is expanding. The precession of Mercury’s orbit indicated that Vulcan existed, even though Vulcan does not exist. Without any minds around, we get the problem that things appear to stand in the indication relation to false things and non-existent things. This is a real metaphysical problem, and giving an account of how representation and the mental fits into the natural order will not solve it. 83 8. Consequences for the Current Debate on the Nature of Propositions Our way of denying that there are representational mental states makes it easier to give an account of what propositions are. As we have seen, an agent can have propositional attitudes without being able to think about propositions and without being causally related to propositions. So when it comes to giving an account of what propositions are, there is no reason to think that a proposition must be some mental thing. There is no need to think they existed before humans started making representational tools for communicating. Propositions just need to be representational. For King’s view and Artifactualism, propositions only come into existence when agents create representational artifacts. In light of this, King has conceded that there are no propositional attitudes before agents start using representational tools to represent things as being ways. But as we have seen, our way of denying that there are representational mental states is such that we need not concede this. We can hold that agents had propositional attitudes long before representational artifacts were created and still hold that propositions only came into existence when humans started making representational artifacts. So, if we are right, then there is no advantage to having a view of propositions where they are mental entities. There is no advantage to having a view where representation is a necessary part of belief and desire. In fact, there is a disadvantage to holding Soames’ and Hanks’ views. The disadvantage is that it makes the problem of intentionality a real metaphysical problem. If there are private acts of representing that occur long before agents communicate, then we need an account of how representing fits in with the natural world. We have given an account of what it is to perform the public acts of representing and to be an artifact with the function of representing in terms of propositional attitudes. Soames and Hanks cannot do this because private acts of representing as well as propositions are, according to their account, prior in explanation to propositional attitudes. Because of this, their views have the burden of explaining how these special private act of representing things as being ways fits in with the natural order. In short, denying that there are representational mental states not only opens up our options for accounting for what propositions are, it gives us reason to prefer theories where propositions are representational but not mental entities. 84 It is worth pointing out that the views of propositional attitudes put forward in sections 4-6 are compatible with not denying that there are representational mental states. We could hold that there are representational mental states even though we do not need them to account for propositional attitudes. We could also hold on to our view that propositions are kinds of artifacts and claim that mental states do in fact represent things as being ways. Instances of propositions would still be restricted to artifacts according to this view, but we pick out these mental states because the states represent what the proposition has the function of representing. In other words we can keep the view that instances of propositions only come into existence when agents start creating representational artifacts but also hold that the mental states of agents represented things as being ways long before public acts of representing were preformed and before any representational artifacts were created. We could use some of the moves we made in Section 7 to again explain how primitive agents can cognize that propositional attitudes of others without being able to think of propositions. However, our reason for preferring a view where we separate the notion of indicating and representing and where we deny that there are representational mental states is that the view does not make fitting representation into the natural order a daunting metaphysical problem. Any view where believing, desiring, and so on require the notion of representing in one way or another, is a view that cannot use our account in Section 7 to avoid the problem of intentionality. So, given our current inquiry into whether there are representational mental states, Artifactualism really allows for three options. The first option is that there are no representational mental states and propositions are kinds of artifacts. The second option is that there are representational mental states and propositions are kinds of artifacts. The third option is that there are mental states that have the function of representing and propositions are kinds whose instances include artifacts and mental states. The first option is the preferred one because it helps us make progress on the problem of intentionality, has the fewest number of theoretical commitments concerning the nature of the mind, and as we will now see, allows us to make progress on debates about perception. 9. Consequences for the Relationalism/Representationalism Debate In the broadest terms, representationalists maintain that our perceptual states share something important in common with belief and judgment states. Representationalists hold that just as we can believe and judge inaccurately, we can be in inaccurate perceptual states. This is because 85 believing, judging, and perceiving all (partially) consist in representing the world as being some way according to the representationalist. The upshot of this is supposed to explain a range of phenomena including the phenomenology of illusions and hallucinations. Austere Relationists deny that our perceiving can be inaccurate in this way. They deny that perceiving consists in representing the world as being some way. Some contemporary arguments for austere relationalism include Travis (2004), Martin (2002), Campbell (2002), Genone (forthcoming), Fish (2009), and Brewer (2007). Schellenburg (2011) has done arguably the most work recently to clarify the nature of this debate and argue on behalf of the representationalists. 105 Her way of ultimately describing representationalism is such that it is committed to perceptual states having-accuracy conditions, which is obviously incompatible with our view. 106 Schellenburg points out some good reasons to think that perceptual states have “content” and are therefore representational. One reason is that when we perceive things, our environment seems a certain way to us. The environment can be as it seems to us or it can fail to be as it seems to us. So, our “seemings” can be accurate or inaccurate. Another reason is to account for the phenomenology of illusions and hallucinations. What it is like for A to hallucinate something can arguably be exactly what it is like for A to perceive something. The same goes for illusions. A final reason is to account for cognitive penetration. It appears that sometimes what it is like for A to have an experience is affected by the concepts we posses and employ. For example, understanding Mandarin appears to affect the phenomenology of listening to the sounds produced by someone speaking Mandarin. Having the concept of a duck arguably can change how the classic rabbit-duck drawing seems to us. 105 See also Schellenburg (2014), Logue (2014), and Genone (forthcoming) for good overviews on the nature of the debate. 106 Now everyone should agree that there is a sense of ‘accuracy’ such that our faculty of perception can be more or less accurate. The more objects and properties we are able to perceive in our immediate environment the more accurate our faculties of perception are. But the representationalist claims that all perceptions can be accurate or inaccurate and this requires that perceptual states be representational and have accuracy conditions. 86 What we will now show is that our way of denying that there are representational mental states allows an austere relationalist to explain all of these phenomena without positing perceptual states that have accuracy conditions. First, we show how an austere relationalist can make sense of talk of ‘content’ when it comes to perceptual states. Next, we explain how the austere relationalist can accommodate the fact that when we perceive, the world seems a certain way to us. Finally, we give an account of the relationship between perception and how things seem that allows for the phenomenology of perceptual states to be multiply realizable, for hallucination and illusions to seem exactly like veridical perceptions, and for two perceptual states to differ in how they seem even if the exact same objects and property instances are being perceived via the same sense modalities. So, the first task for us is to make sense of what it is to be the “content” of a perception. For the propositional attitudes we were considering, we saw that when someone talks about the content of, for example a belief, she really means the propositions we refer to ultimately pick out the belief. For our everyday reports about perception, the relevant relatum of the perceiving relation is either an object or a property instance. For example when we say ‘John is perceiving the Sun’ the Sun is the content of his perception. And we know that this means John must stand in an important relation to the Sun to be in this mental state. Unlike the propositional attitudes, it seems pretty obvious upon a priori reflection that John must be immediately and presently causally related to the Sun when he is perceiving the Sun. Of course this is not sufficient to perceive the Sun, but is just necessary. This simple observation about necessary conditions to perceive o underlies the spirit of relationalism. In some sense, our immediate environment is what matters for what we are perceiving. 107 The important question, as Schellenburg points out, is how the austere relationalist is going to make sense of our ‘seems’-talk. It is undeniable that when an agent perceives, things seems a certain way to her. We say things like ‘It seems to A that the top line is longer than the bottom line’. We say that the way things seemed when I hallucinated was just like how it seemed when I was not hallucinating. So far, most relationalists have gone to great lengths to give revisionary accounts of what is going on with our ‘seems’ talk. Some have claimed that all ‘seems’ talk is 107 We also sometimes talk of perceiving things as being ways and perceiving that certain things are the case. We perceive one line as being longer than another line. We perceive that a robbery is in progress. A relationalist can hold that all these claims use a univocal sense of ‘perceiving’ as long as they hold that something like states of affairs in our immediate environment can causally affect our mental states or redescribe the cases in terms of property-instance causation. An austere relationalist can also claim that these ways of using ‘percieve’ pick out something like low-level judgment. Either option is open to us. 87 really about low-level judgments. 108 Others have simply denied the pre-theoretical data that the phenomenology of how things seem is multiply realizable. 109 But consider the Muller-Lyer illusion. Two lines that are actually equal in length can be such that the top line seems longer than the bottom line even if we are judging, and even know, that they are equal in length. Although we sometimes use ‘seems to A’ to mean ‘A judges that’, such a case gives strong prima facie evidence to the claim that there is a sense of ‘seems’ that is independent of judgment and closely related to perception. The austere relationist ought to grant that when A perceives o’s Fness, A is guaranteed to be in state that we pick out with ‘It seeming to A that o is F’. So, it is correct that perceiving guarantees that we are in a state that we can pick out with a proposition. And as we have already claimed, this state is such that it partially consists in standing in an important relationship to o and being F. We can also make sense of the fact that the world can either be as it seems to us or fail to be as it seems to us. If it seems to us that o is F then the world is as it seems if o is indeed F. So, if we want, we can say “seemings” are accurate or inaccurate, and this is just like the claim that beliefs, as in what is believed, can be accurate or inaccurate. 110 We have already shown that we do not need representational mental states for this to be the case. 111 This is not the first time it has been suggested that some views of perceptual content are such that an austere relationalist can accept that experiences have contents. Logue (2014) is a recent example that maps out the various ways an austere relationalist can think of content. What is novel about our approach is that our way of denying that there are representational mental states allows us to give a principled reason for taking a view of “content” such that a mental state having “content” does not entail that the state has accuracy-conditions. It would be completely ad hoc to hold the fact beliefs having “content” entail they have accuracy-conditions, but perceptual states having content do not entail this. Furthermore, everyone should grant that the “content” of a state, i.e. what we refer to help pick out the state, tells us something about the underlying nature 108 See, for example, Genone (forthcoming). 109 For instances, see Campbell (2002). 110 Bryne (2009) and Siegal (2010), among others, invent technical words to deal with the fact that English does not have an a verb φ that picks out the seeming-state an agent is in through the locution ⎡A φ’s that o is F⎤ . 111 Logue (2014) make a similar point that there are ways of thinking of content such that austere relationalism is consistent with perceptual states having content. However, unlike Logue, we are committing ourselves to the negative claim that perception is not a matter, fundamentally, or otherwise, of representing the environment as being some way. In other words, we are taking a stance on what the relationship between the propositions and the state could not be. We are also making this claim about all mental states in general. If we only made this claim for the relationship between content and the metaphysical nature of perceptual states, the account could rightfully taken to be ad hoc. 88 of the state. Our approach is also novel in that it allows us to have a principled reason for claiming that the relevant relatum of perception-report shares a relatively minimal amount of features with the state picked out. Even if we pick out the state with a representational entity, the state itself need not be representational. Our next step is to show how an austere relationalist can allow that the phenomenology of perceiving is multiply realizable. The austere relationalist can and should claim that what an agent perceives only partially, and asymmetrically, determines how things seem to her. Two experiences can seem the exact same way, even if the agent is perceiving different qualitatively identical things on different occasions. For example, when it comes to relationship between how things seem and what is perceived, the particulars perceived arguably do not matter for how things seem. Only the properties matter. Perceiving two qualitatively identical cups on separate occasions, ceteris paribus, might seem exactly the same way. This is not to say that there is no difference between perceiving cup1 and cup2. It is just to say that how things seem is not symmetrically determined by what an agent perceives. The same goes for hallucination and illusion. A veridical perception can seem identical to a hallucination, because how things seem are only partially and asymmetrically determined by what is perceived. How can the austere relationalist claim that how things seem is only partially determined by what is perceived? She can do so by being very careful when talking about ‘perceptual states’. There is a difference between the very particular perceptual-state we are in when we perceive o’s Fness versus when we hallucinate that o is F. In one state, we stand in an immediate causal relation to o’s Fness. In the other state, we still stand in some relation to o and being F but it is not an immediate causal one. However, there is a more general underlying state that is common to perceiving o’s Fness and hallucinating that o is F. We can pick out the general state with ‘The state A is in when it seems to A that o is F’. And we should assume that what it is like for an agent to be in one particular perceptual state P1 will, ceteris paribus, be identical to what it is like for an agent to be another particular perceptual state P2 when P1 and P2 are more particular ways of being in a state that we pick out with the phrase ‘The state A is in when it seems to A that o is F’. So an austere relationalist can and should grant that what it is like for A to hallucinate that o is F can be exactly what it is like for A to perceive o’s Fness. There is a common underlying mental state to these happenings. The “content” of hallucinations and illusions are indeed propositions, but this way of using ‘content’ is misleading. We pick out the states one is in when hallucinating and undergoing illusions by relating an agent to a proposition. The things the 89 proposition is about are things the agent stands in an important relationship to when she is in these states but it is of course not the same particular relationship an agent is in when she perceives o’s Fness. Accounting for the phenomenology of illusions and cognitive penetration is less clear than outright hallucination. It is not so clear that what it is like for it to seem to A that the top line is longer than the bottom is what it would be like for A to perceive the top line’s being-longer-than- the-bottom-line if the top line had this property. It is also not clear whether, when cognitive penetration occurs, we are perceiving property instances that we could not previously perceive, 112 or whether our low-level implementing of concepts affect how things seem, or some other explanation. 113 No matter what, the austere relationalist has a way of accommodating the phenomenology of illusions and cognitive penetration. If as, for example, Schellenburg claims, our employing of low-level perceptual concepts affects what it is like to perceive the line illusion or to look at the cartoon rabbit-duck, this is no problem for the austere relationalist. Schellenburg thinks it is a problem because she thinks the relationalist cannot appeal to content at all. But the relationalist can allow that employing low-level concepts puts A in a state that we pick out with ‘It seeming to A that o is F’ even if there is no property instance of being-F for A to be immediately causally related to. 114 In this way, it is possible for two agents to both be in a general perceptual state where they are perceiving the same things, but what it is like to be in these states can differ because of other happenings in the mind. So if our way of denying that there are representational mental states is correct, the austere relationalist can explain everything the representationalist seeks to explain. In fact, given how the relationalist ought to account for the data, unless one has reason to be a representationalist that 112 Siegel (2012) and Genone (forthcoming) both offer this explanation. 113 Levine (2011) examines these types of views when applied to the phenomenology of occurent cognitive states like occurent belief. 114 So it can be true that agents A and B perceive the same things, but perhaps due to employing different low-level perceptual concepts, it seems to A that there is a cartoon duck in front of her while it seems to B that there is a cartoon rabbit in front of him. Schellenburg attempts to come up with a technical way of specifying in one fell swoop the difference in “content” between the “perceptual state” of A and the “perceptual state” of B when this is happening. It surely is possible to come up with some technical machinery for doing this, but doing so does not tell us anything new about the nature of perception if we accept our way denying that there are representational mental states. In other words, there might be fruitful new ways of picking out the very particular state A is in that differs from B, but it does not give us any new information from our folk ways of describing the similarities and differences between A and B. Both A and B are perceiving the same dots of ink on the page and so on. But it seems to A that a cartoon duck is in front of her while it seems to B that a cartoon rabbit is in front of him. 90 believes that the objects of awareness are representational, mental, mind-dependant entities, there is no reason to hold onto the view that perceptual states have accuracy-conditions. Of course, that does not mean all is solved when it comes to perception. There is still the real debate as to how we are to characterize and explain the phenomenology of illusions and cognitive penetration. But regardless of the explanations given, our way of denying that there are representational mental states allows the austere relationalist to give any of the explanations on the table. Overall, Schellenburg’s recent work defending representationalism is very important because it forces austere relationalists to give a non-revisionary account of our ‘seems’-talk and to give an account that accommodates the truism that phenomenology is multiply realizable. Our way of denying that there are representational mental states rules out representationalism on the one hand, but on the other hand, allows the austere relationalist explain all the phenomena the representationalist seeks to explain. 10. Concluding Remarks We have taken on the wide spread assumption that there are representational mental states. We found a way to deny that these things exist while still accommodating the truism of folk- psychology. Our account leaves us with many options when it comes to giving a theoretical account of the nature of mental states in general. Most importantly, denying that there are representational mental states allowed us to make progress in metaphysics, the philosophy of language, and the philosophy of mind. We can explain how representation fits in with the natural world. We can move forward towards settling the question of what propositions are. We can simplify and make progress on current debates about the nature of perception. 91 Chapter 4: What Complex Sentences and Propositions Represent Abstract Chapter 2’s investigation into the nature of truth-aptitude got us the result that anything that has truth-conditions represents things as being some way. This result forces us to account for the way complex truth-apt sentence represent things as being. What do quantified sentences represent? How about sentence with sentential operators? The same questions arise for the propositions expressed by these complex sentences. This chapter argues for the preferred way of treating complex sentences and propositions. Chapter 2’s investigation into the nature of truth-aptitude got us the result that anything that has truth-conditions represents things as being some way. This result forces us to account for the way complex truth-apt sentence represent things as being. What do quantified sentences represent? How about sentence with sentential operators? The same questions arise for the propositions expressed by these complex sentences. This chapter argues for the preferred way of treating complex sentences and propositions. Section 1 defends the result that each type of complex truth-apt sentence and proposition represents some things as being some way. Although this assumption has already been defended, it creates difficulties in accounting for what complex sentences and propositions represent. I will argue that we cannot drop this assumption – not even for a similar but weaker claim. Section 2 argues that, for each complex sentence/proposition, we have intuitions about what the setenence/proposition is about. It turns out that examining these intuitions and attempting to give a theory of propositions that vindicates these intuitions helps rule out competing accounts of the way complex sentences/propositions represent things as being. Section 3 explains, for each type of complex proposition/sentence, what way it represents things as being. In this section, I sketch how agents can come to understand what these sentences/propositions represent in terms of understanding more basic properties. Section 4 shows how my account of what complex sentences/propositions represent, along with the Functional Theory of truth-aptitude and Artifactualism, can shed light on some semantic phenomena that we currently have a poor understanding of. 92 1. All Propositions Represent Things as Being Ways According to Artifactualism and the Functional Theory of Truth-Aptitude, the proposition that snow is white represents snow as being white. Well what does the proposition that that snow is white or grass is green represent? The natural thing to say is that this proposition represents that snow is white or grass is green. I believe that this answer is correct, however the answer does not tell us what things the proposition represents as being what way. In other words, the answer tells us that the proposition stands in the representing-that relation to something but does not tell us whether the proposition stands in the representing-as relation to anything and if so what things it stands in the representing-as relation to. I will not rehash any previous argument meant to show that propositions cannot just have the property of being-representational and must rather represent some things as being some way. Instead, I intend to forestall the following reaction. By ‘the representing-that relation’ I mean the relation that that p stands in to e when e represent that p. This relation has to be different than the relation e stands in to o when e represents a normal object o. To see why this is so, consider various fearing relations. When you fear that a meteor will hit the Earth you stand in a relation to that a meteor will hit the Earth that is different from the relation you stand in to a particular spider when you fear the spider. King (2002) gives a good overview of this phenomenon. In what follows I will respectively call the relations picked out by the relevant verb-phrase when the object of the verb is a that-clause: ‘fearing-that’, ‘desiring-that’, ‘representing-that’ and so on. As one works though my analysis of what way each type of proposition/sentence represents things as being, it will become clear that the assumption that each truth-apt proposition/sentence represents things as being some way leads to surprising results. One might reasonably try to accept the spirit of my previous arguments that propositions need to do more than be representational and adopt the view that all propositions, in addition to being representational, stand in the representing-that relation some entity although not all propositions stand in the representing-as relation. Now it is extremely important to note that the following is false: if e represents that p then e represents a proposition. It is true that any time e represents that p the accuracy of e will depend on whether it is the case that p. But something can represent, as in stand-for, a proposition without representing that proposition as being any way just as something can represent, or stand- 93 for, snow without representing snow as being any way. If a mark, for example, represents a proposition but does not represent the proposition as being any way, then mark neither has truth- conditions nor accuracy-conditions. Being an instance of a proposition and representing a proposition are not one and the same thing. On the other hand, whenever e represents that p, it appears that e does stand in a relationship to the proposition picked out by that p. It does not stand in the represents/standing-for relation. It stands in the represents-that relation. This is similar to the difference between fearing and fearing- that. Fearing that a meteor will hit the Earth appears to be a propositional relational. But to stand in this relation to the proposition that a meteor will hit the Earth is of course not one and the same as fearing a proposition. The proposition is not going to be the end of life as we know it. I will now argue that it is untenable to take a view where complex sentences truth-apt sentences and complex proposition do not represent things as being ways but merely stand in the representing-that relation to something. First, it appears that whenever we know that an entity e stands in the representing-that relation we are in a position to know that e stands in the representing-as relation and vice versa. Take the proposition that snow is white. It represents snow as being white. It also represents that snow is white. In at least a simple case like this, the proposition standing in the representing-that relation and standing in the representing-as relation are basically equivalent. You cannot have one relation without the other. Anyone that knows that the proposition represents snow as being white is in a position to know that the proposition represents that snow is white and vice versa. Why would things be different for more complicated propositions? Take, for instance, the proposition that snow is white or grass is green. It represents that snow is white or grass is green. Wouldn’t it be surprising to learn that that although it represents this it does not represent any thing as being any way? Perhaps more compellingly, any time we specify that a certain entity e represents some things as being some way, we are in a position to say what e stands in the representing-that relation to. If e represents the propositions p and q as being such that at least one of them is true we know that e represents that at least one out of p and q is true. If e represents Los Angeles as being such that it is south of Portland we know that e represents that Los Angeles is south of Portland. These observations are not conclusive, but they give us reason to believe that whenever an entity stands in the representing-that relation it also stands in the representing-as relation and vice versa. 94 The second reason to keep the assumption that all propositions stand in the representing-as relation is that representing-that appears to be a propositional relation. When e is not a proposition, we can learn about e when we learn what e stands in the represents-that relation to. If we learn that a mark represents that snow is white we have learned that the mark represents snow as being white. This gives us reason to think that we use the propositional relation represents-that to efficiently specify what the accuracy of a given entity depends on. If e represents that p, then the accuracy of e (partially) depends on whether p is true. If e represents that p, and does not stand in the representing-that relation to anything else, e is completely accurate if and only if p is true. As will become clear, in complicated cases such as where e represents that snow is white or grass is green, specifying what things e represents as being what way is long-winded and potentially confusing. If we have already gone through the process of understanding what the proposition that snow is white or grass is green represents as being what way, and therefore understand when this proposition is true, there is no need to repeat this process again. All we need to say is that e represents that p and we now know that e’s accuracy (partially) depends on whether p is true. For any other entity that represents the same things as being the same way as that p, we simply say the entity represents that p. However, when we are trying to understand what way things need to be in order for the proposition that p to be true we learn nothing when we learn that the proposition that p represents that p. We learn that whether the proposition is accurate depends on whether the proposition is true and of course we already knew that. At the end of the day, we need to know what things the proposition p represents as being what way. Once we know this, we know that p is true whenever those things are as that p represents them as being. And the accuracy of anything that represents that p will in turn depend on whether those things that that p represents as being some way are as that p represents them as being. The third reason concerns giving a theory of truth-aptitude. In as much as the Functional Theory of truth-aptitude is promising, we have reason to believe that all things that are capable of being true or false, which includes all propositions, represent things as being some way. More importantly, how else could we give a general account of truth and accuracy? If representing-that does not, in some sense, reduce to representing-as, our account will have to take as primitive a claim like ⎡if e represents that p, then e is true if and only if p⎤. However, as written this claim uses p as a meta-linguistic variable that can takes sentences as values. The claim basically gives a compressed version of a list of claims rather than giving an actual account of when entities are 95 true in general. We should prefer a view that gives an actual account rather than summarizes a list of truths in a compact way. Worse yet, the only ways to modify this claim so that p is a variable that ranges over objects rather than a sentences leaves us with trivial and uninformative claims. We end up with a claim like ‘For all propositions p and entities e, if e stands in the represents-that relation to p, then e is true if and only if p is true’. Since the clause after ‘if and only if’ needs to be of the form F(p), we end up having to use some property, such as being-true, being-the-case, and so on, which we are trying to analyze in the first place. Therefore, we need a theory that analyzes truth and accuracy in terms of representing-as since it allows us to give an actual account of when entities are true and accurate in general. In short, even if it entails some surprising results, we have good reason to hold fast to the view that all truth-apt entities, including complex propositions, represent things as being some way. 2. Intuitions about Aboutness Provide Constraints So far, for each atomic proposition p that we have considered, I have offered some object o and some way F for which p represents o as F without giving any argument as to why we should believe that p represents that object as being that way. For example, I have assumed that if that snow is white represents any thing as being any way then it represents snow as being white. But one might object that I have made an arbitrary decision. Why not think that this proposition instead represents the universe as being-such-that-snow-is-white. In fact, why not think that this proposition represents the number two as being-such-that-snow-is-white? Anything that represents the number two as having this strange (alleged) property will represent things as they are iff the number two has this property, and the number two will have this property iff snow is white. So if we assume that the proposition that snow is white is the functional-type whose instances all share the unique representing-as function of representing the number two as being such that snow is white then the proposition ends up having the correct truth-conditions; what more could we ask for? We can and should ask for a lot more. If we do not, as we just saw, there will be too many entities that can play the role of the proposition that snow is white. More importantly, we have intuitions about what propositions are about. At first this is likely to sound strange since we rarely 96 say ‘Tell me what that proposition is about’. However, it is completely natural to ask ‘What is Goldbach’s conjecture about?’, ‘What is the Phantom Time Hypothesis about?’, and so on. And if anything that we talk about in everyday life is a proposition, hypotheses and conjectures are propositions. Goldbach’s conjecture is the proposition that every even number can be expressed as the sum of two prime numbers. What is this hypothesis about? It is about being an even number, being a prime number, and so on. Imagine that John hypothesizes that the earth is 10 seconds old. What is John’s hypothesis about? John’s hypothesis is about the earth as well as about the property of being 10 seconds old. From this we get that the proposition that the Earth is 10 seconds old is about the Earth and the property of being 10 seconds old since John’s hypothesis is identical to the proposition that the Earth is 10 seconds old. These points concerning what propositions are about should seem obvious and I believe them to be incontrovertible. In the simple case of the proposition that snow is white, we rule out any entity that represents the number two as being such that snow is white since the proposition that snow is white is not about the number two and is not even about the property of being such that snow is white. The proposition is about snow and about the property of being white. Even though this seems like a minor point now, trying to vindicate our intuitions about what a given proposition is about will end up doing an immense amount of work when it comes to more complicated propositions. For example, consider the proposition that everything has mass. There is a tradition of analyzing such a proposition as partially consisting of a propositional function. 115 One might claim that this proposition represents a propositional function as being a certain way. But is this proposition about a function? No. If John’s hypothesis is that everything has mass, John’s hypothesis is not about a propositional function. Why would he bother testing a theory about a propositional function? Although one might despair such a result, we will make no progress in figuring out what each proposition is unless we try to honor the few intuitions we do have about them. 115 For example, see Russell (1903). 97 3. Complex Types of Sentences/Propositions This section explains what way each type of proposition represents things as being. The claims here apply equally to the corresponding truth-apt complex sentences that express these propositions. 3.1 Multi-Place Relations The proposition that Los Angeles is south of Portland represents Los Angeles as being south of Portland. To do this is one and the same as representing Portland as being such that Los Angeles is south of it. Therefore, this proposition also represents Portland as being such that Los Angeles is south of it. Each instance of this proposition is a single way representation that represents Los Angeles as being south of Portland and thereby represents Portland as being such that Los Angeles is south of it. In fact we can succinctly get this idea across by saying that the proposition represents Los Angeles as being an x and Portland as being y such that x is south of y. In general the proposition that R(o 1 , o 2 ,.., o n ) represents o 1 as being an x such that R(x, o 2 ,.., o n ). Representing this is one and the same as, for any m such that 0 ≤ m ≤ n, representing o m as being an x such that R(o 1 ,.., o m-1, x , o m+1, .., o n ). So, for each m such that 0 ≤ m ≤ n, the proposition that R(o 1 , o 2 ,.., o n ) represents o m as being an x such that R(o 1 ,.., o m-1, x, o m+1, .., o n ). Each instance of this proposition is a single way representation that represents o m as being an x such that R(o 1 ,.., o m- 1, x , o m+1, .., o n ). One might object to the claim that to represent Los Angeles as being south of Portland is to do one and the same thing as to represent Portland as being such that Los Angeles is south of it. First, to clarify, I am not claiming that whenever we represent Los Angles as being south of Portland we also end up doing something else– namely, representing Portland as being such that Los Angeles is south of it. This would mesh poorly with my claim that propositions are single way representations and furthermore mesh poorly with my account of truth. I am also not claiming that the property of being-south-of-Portland is identical to the property of being-an-x- such-that-Los-Angeles-is-south-of-x. Rather I claiming that representing Los Angeles as being south of Portland is one and the same doing as representing Portland as being such that Los Angeles is south of it. We use different phrases to pick out one and the same doing. Claiming that these doings are identical is no more radical than claiming that to act as if chocolate ice cream tastes better than vanilla ice 98 cream is to act is if vanilla ice cream is such that chocolate ice cream tastes better than it. These are not different ways to act. These are not different doings. Similarly to drive to Los Angeles from Portland is to drive to Portland from Los Angeles. Thankfully, doing this involves only doing one trip. If one is still uneasy about accepting that these doings are identical it is worth going over some alternative analyses of relational propositions that will not work. For example, the proposition that Los Angeles is south of Portland cannot represent the ordered pair <Los Angeles, Portland> as standing in the being south of relation. First of all, the proposition is not about an ordered pair. The proposition is about Los Angeles, it is about Portland, and it is about the being south of relation. Furthermore, there is no reason to think it represents this ordered pair as being some way rather than to think it represents <Portland, Los Angeles> as being some way. Sure the words that refer to each city appear in a certain order in English sentences that are instances of this proposition, but so what? Similarly, it is not clear what it is to represent <Los Angeles, Portland> as standing in the being south of relation. Does it represent Los Angeles as being south of Portland, Portland as being south of Los Angeles, or neither? From the difficulties we encounter with the alternative analysis it should be clear why it is natural to think that the proposition that Los Angeles is south of Portland “simultaneously” represents Los Angeles as being some way and represents Portland as being some way. The proposition is about these two cities so it better represent them as being some way. Furthermore, it will do us no good to say that the proposition represents the cities as standing in the being south of relation. The proposition that Portland is south of Los Angeles also represents the cities as being in this relation. So the proposition that Los Angeles is south of Portland represents Los Angeles as being south of Portland and this is one and the same as representing Portland as being such that Los Angeles is south of it. 3.2 Sentential Operators Discerning the nature of propositions expressed by sentences with sentential operators turns out to be somewhat problematic. Does the proposition that if Socrates ate olives then Plato olives represent the universe as being some way, two propositions as being some way, or does it represent Socrates and Plato as being some way? Without relying on our intuitions concerning what each proposition is about we will get nowhere on this question. To test our intuitions, call the hypothesis that if Socrates ate olives then Plato olives the ‘Philosopher King Hypothesis’. What is the Philosopher King Hypothesis about? Well, the hypothesis is about Socrates, it is 99 about Plato, it is about olives, and it is about the relation of one thing having eaten another. So it would seem that the proposition should represent these things as being some way. I claim that this hypothesis represents Socrates as being such that if he ate olives then Plato olives. Doing this is one and the same as representing olives as being such that if Socrates ate them, then Plato ate them, which is one and the same as representing Plato as being such that if Socrates ate olives then he ate olives. Each instance of the Philosopher King Hypothesis is a single-way representation that represents Socrates, Plato, and so on in these ways. A reasonable worry one might have at this point concerns the alleged ways I claim this proposition represents Socrates, Plato, and so on as being. What way is something when that thing is such that if it ate olives then, Plato ate olives? How do we come to understand what it is to have this property? If this is a property at all, it is a strange property to say the least. And my theory requires the existence of even stranger properties when sentences contain multiple sentential operators. I will attempt to explain how we come to understand these properties at the end of this section, but for now I want to show that any plausible analysis of propositions expressed by sentences with sentential operators requires positing seemingly strange properties. If we had instead claimed that the Philosopher King Hypothesis represents the universe as being such that if Socrates ate olives then Plato ate olives, we would be stuck explaining how we come to have an understanding of the property of being-such-that-if-Socrates-ate-olives-then- Plato-ate-olives. This would arguably be a worse predicament since we would have to explain how we can understand what it is to have this property without relying on a prior understanding of when the proposition that if Socrates ate olives then Plato ate olives is true. In any case, analyzing the proposition this way leads us to posit equally strange properties. And this analysis fares much worse when it comes to what the proposition is about. One might try to claim that every proposition is about the universe so it does not contradict our intuitions to claim that the Philosopher King Hypothesis is about the universe. But is the proposition that if 0 =1 then 1 =2 about the universe? I do not think so. This claim would be true in any alternate universe. One might instead claim that the proposition is about “reality”. But when we say ‘e represents reality as being some way’ we are really using ‘reality’ as a dummy variable. What is this object you call ‘reality’? So, claiming that these propositions represent the universe or reality as some way faces equally bad problems concerning strange properties and worse problems concerning what these propositions are about. 100 If we had instead claimed that the Philosopher King Hypothesis represents the proposition that Socrates ate olives as being such that if it is true then the proposition that Plato ate olives is true we would again be stuck explaining how we come to understand a strange property. How do we come to learn when an object has the property of being-such-that-if-it-is-true-then-the- proposition-that-Plato-ate-olives-is-true. Furthermore, this analysis contradicts our intuitions concerning what the Philosopher King Hypothesis is about. The hypothesis is not about propositions and it is not about the property of being true. It should also be clear why any variation of this analysis will not fare any better. The hypothesis is not about the property of being the case, it is not about the property of representing things as they are, it is not about a function that maps propositions to propositions. So we have good reason to assume that the Philosopher King Hypothesis represents Socrates as having a strange property. The proposition that not snow is white represents snow as being such that not it is white. The proposition that either snow is white or grass is green represents snow as being such that either it is white or grass is green. The proposition that necessarily snow is white represents snow as being such that necessarily, it is white. As for how we come to understand these properties, I intend to do the following. First I will give some cases where it is obvious that we can and do come understand somewhat complex and strange properties in terms of simpler properties. Next I will show that, independent of my theory of propositions, we are forced to choose among three surprising options when it comes to analyzing how we are able understand when objects have certain properties. I will then show that once we accept any of these three options we can easily explain how we are able to understand when an object has any of the properties that my theory posits as existing. First, our ability to recognize that certain objects do not have a property and our ability to recognize that both one object has one property and another object has another property is most likely not the type of ability that we can give a conceptual analysis of in terms of simpler abilities. We could claim that whenever one is able to recognize that both o 1 has the property of being yellow and o 2 has the property of being red, we are able to do so because we are able to recognize that o 1 has the property of being such that it is yellow and o 2 is red. But whether we claim this or not, there is nothing mysterious or suspect in claiming that we have the ability to recognize that both one object has one property and another object has another property. The same goes for recognizing that an object does not have a certain property. 101 I believe that, as a matter of fact, we often come to understand what it is to have somewhat strange and complex properties in terms of understanding what it is for objects to have simpler properties. For example, consider the property of being yellow and red. We know that an object o has this property if and only if o has the property of being yellow and o has the property of being red. Although there is nothing stopping nature from creating a creature that recognizes the more “complex” property innately, it is doubtful that humans are hardwired to recognize this property. But we are hardwired to recognize yellowness and redness and we can use these abilities to recognize new properties. Similarly, we know that an object o has the property of being-such- that-it-is-not-the-case-that-it-is-yellow if and only if o does not have the property of being yellow. We also know that o 1 has the property of being-such-that-it-is-yellow-and-o 2 -is-red if and only if o 1 has the property of being yellow and o 2 has the property of being red. Now it would be nice if we could explain our ability to recognize when objects have properties in terms of our ability to recognize things not having properties and our ability to recognize when both one thing and another have certain properties. However, simple cases of disjunction show that such an explanation will not work. When it comes to our ability of recognizing that either o 1 is yellow or o 2 is red we face three somewhat unpalatable options. The first option is that we just leave it unanalyzed how one is able to recognize that either o 1 is yellow or o 2 is red. The second option is that we analyze this ability in terms of our ability to recognize when o 1 has the property of being-such-that-either-it-is-yellow-or-o 2 -is-red, and in turn, leave this ability unanalyzed. The third option is that we analyze our ability to recognize that either o 1 is yellow or o 2 is red in terms of another primitive ability such as recognizing that it is not the case that both o 1 does not have the property of being yellow and o 2 does not have the property of being red. So we are forced to take one of these three options concerning our ability to recognize when two objects and two properties are such that either one object has one of the properties or the other object has the other property. Similar unpalatable options exist for material conditionals. I will now show that whatever option we choose, we can explain how one can understand and recognize the properties that my theory posits as existing. When it comes to properties such as being an x such that either x is white or grass is green, something has this property if and only if either it has the property of being white or grass has the property of being green. We can either take this ability as primitive, take our ability to recognize when something has the property of being-such-that-either-it-is-white-or-grass-is-green as 102 primitive, or take some equally strange ability as primitive. Similar options exist for being an x such that ((F(x) & G(b)), being an x such that ¬(F(x)), being an x such that ((F(x)→ G(b)), and so on. When it comes to propositions expressed by sentences with more and more sentential operators it quickly becomes implausible to take our abilities to recognize the properties involved in these propositions as primitive. Consider the proposition that ¬(F(a)→(G(b) v ¬F(a))). Somehow, we are able to “work through” a proposition like this and figure out when it is true. Therefore, if my theory is correct, it better also be true that we can figure out when an a has the property of being an x such that ¬(F(x)→(G(b) v ¬F(x))) . Our ability to do this will ultimately be explained in terms of the abilities concerning property recognition that we have already gone over. But part of the explanation also involves recognizing when propositions themselves have certain properties. First, we need to examine the property of representing things as they are. An entity e represents things as they are if only if e represents some things as being some ways and, for each object o that e represents as being some way F, o has the property of being F. The property of representing things as they are is closely related to the property of being true, but as we have seen, they are not one and the same. A map can represent things as they are but the map is neither true nor false. Next there is the property of representing things as they are not. An entity e represents things as they are not if only if e represents some things as being some ways and, for some object o that e represents as being some way F, o does not have the property of being F. Although I have just given definitions of the two properties I take them to be properties that we are already able to recognize even though we do not often explicitly talk about them. Now according to me the proposition that ¬(F(a)→(G(b) v ¬F(a))) represents a as being an x such that ¬(F(x)→(G(b) v ¬F(x))). When does a have this property? Consider the following propositions: Prop 1: that ¬(F(a)→(G(b) v ¬F(a))) Prop 2: that (F(a)→(G(b) v ¬F(a))) Prop 3: that F(a) Prop 4: that G(b) v ¬F(a) 103 Prop 5: that G(b) Prop 6: that ¬F(a) a has the property being an x such that ¬(F(x)→(G(b) v ¬F(x))) iff Prop 1 has the property of representing things as they are. Prop 1 has the property of representing things as they are iff Prop 2 has the property of representing things as they are not. Prop 2 has the property of representing things as they are not iff both prop 3 represents things as they are and prop 4 represents things as they are not. Prop 3 represents things as they are and Prop 4 represents things as they are not iff a has the property of being F, prop 5 represents things as they are not, and prop 6 represents things as they are not. A has the property of being F, prop 5 represents things as they are not, and prop 6 represents things as they are not iff a has the property of being F, b does not have the property of being G, and Prop 2 represents things as they are iff a has the property of being F and b has the property of being G. Therefore, a has the property of being an x such that ¬(F(x)→(G(b) v ¬F(x))) iff a has property of being F and b has the property of being G. No matter how many sentential operators are involved we can always analyze our ability to understand complex properties in terms of our ability to understand simple properties in this way. Modal operators only bring slight complications. First, we have the ability to recognize when it is possible that an object has a property of being F. We can leave this ability unanalyzed or analyze it in terms of the ability of recognizing when an object has the property of being such that possibly, it is F. The same options exist for recognizing when an object necessarily has a property, probably has a property, and so on. For propositions expressed by sentences with many sentential operators, including modal operators, we use the property of representing things as they possibly are and the property of representing things as they necessarily are. An entity e represents things as they possibly are iff e represents some things as being some ways and, for each object o that e represents as being some way F, o possibly has the property of being F. An entity e represents things as they necessarily are if only if e represents some things as being some ways and, for each object o that e represents as being some way F, o necessarily has the property of being F. The proposition that ◊(F(a) v !(G(b))) represents a as being an x such that that ◊(F(x) v !(G(b)). a has this property iff that (F(a) v !(G(b)) represents things as they possibly are iff possibly, that F(a) representing things as they are or that G(b) represents things as they necessarily are iff possibly, a has the property of being F or b necessarily has the property of being G. 104 In summary, a proposition such as if Socrates ate olives then Plato ate olives is not about propositions. It is about Socrates, Plato, and so on. Propositions expressed by sentences with sentential operators represent objects as having some seemingly strange and sometimes complex properties. Regardless of my theory of propositions, we have to assume that we are able to recognize some of these seemingly strange properties. For the remaining properties, we are able to understand what it is for an object to have these properties in terms of when other objects have simpler properties. We can use our understanding of when propositions have properties to understand the properties involved in propositions expressed by sentences with multiple sentential operators. 3.3 Quantifiers The proposition that something is red represents the property of being-red as being-had-by- something. Each instance of this proposition is a single-way representation that represents the property of being-red as being-had-by-something. The proposition that everything is red represents the property of being-red as being-had-by-everything. The proposition that most things are red represents the property of being-red as being-had-by-most-things. This analysis of propositions expressed by sentences with quantifiers raises several questions. First, we face the now familiar problem of strange properties. What way is something when it has the property of being-had-by-something? Well, whenever Oliver is red, being-red is had by Oliver. I claim that to represent Oliver as being red is do to one and the same as representing being-red as being had by Oliver. The ‘being had by’ construction is admittedly obtuse but it is needed since English appears to lack means of performing second order quantification. For example it would be nice if we could instead say ‘that Oliver is red represents the property of being red as being such that Oliver it’. If we switch to a mix of natural and artificial languages we can make the same point by saying that Oliver is red represents being-red as being a φ such that φ(Oliver) which is to do one and the same as representing Oliver as being an x and being-red as being a φ such that φ(x). In fact, up to this point, I have been purposely avoiding what way a proposition like that Los Angeles is south of Portland represents the being-south-of-relation as. But as we know from Frege 116 , this proposition is just as much about the being-south-of-relation as it is about Los Angeles and Portland. This proposition represents the being-south-of-relation as being a φ such 116 C.f. Frege (1997a: pp. 54-55) 105 that φ(Los Angeles, Portland) which is one and the same as representing Los Angeles as being an x, Portland as being a y, and the being-south-of-relation as being a φ such that φ(x,y). Without helping ourselves to formal apparatuses we would have to say that this proposition ‘represents the being-south-of-relation as being-such-that-Los-Angeles-it-Portland.’ Returning to quantifiers, to represent being-red as being-had-by-something is one and the same as representing being-red as being-a-φ-such-that-∃xφ(x). How do we come to understand what it is to have the property of being-had-by-something? We know that a property φ has the property of being-had-by-something if and only if φ is had by at least one object. When φ is being-red we see that this just amounts to the benign claim that we know being-red is had by something iff being-red is had by at least one object iff something is red. As with sentential operators, we face three options when it comes to which abilities we take as primitive. First, we could treat as primitive our ability to recognize when a property is had by at least one object and use this to explain how we understand when a property has the property of being-had-by- something. Or we could treat our ability to recognize when a property φ has the property of being-had-by-something as primitive. Or we could treat some other similar ability as primitive, such as the ability to recognize when it is not the case that φ has the property of being-had-by- nothing. No matter which option we choose we can then go on to explain how we understand the properties involved in propositions expressed with sentences that have multiple quantifiers. The proposition someone shaves everyone represents the property of being-one-who-shaves-everyone as being-had-by-something. 117 But what is this property of being one who shaves everyone? Well we know that some object o has this property iff the property of being-shaved-by-o is had by everyone iff the property of being-shaved-by-o has the property of being-had-by-everyone. To show how this procedure generally works, consider the proposition that ∀x∃y∀z(R(x, y, z)). Let ‘o 1 ’, ‘o 2 ’, and ‘o 3 ’ denote arbitrary objects. Let φ 1 be the property of being an x such that ∃y∀z(R(x, y, z)) . Let φ 2 be the property of a being y such that ∀z(R(o 1 , y, z)). 117 Note that it represents a property as having the property being-had-by-something. After all, the proposition that Russell shaves everyone represents the property of being-one-who-shaves-everyone as being had by something, namely, Russell. However, that Russell shaves everyone does not represent the property of being-one-who-shaves-everyone as having the property being-had-by-something. It represents this property as being had by Russell. 106 Let φ 3 be the property of being a z such that R(o 1 , o 2 , z). The proposition that ∀x∃y∀z(R(x, y, z)) represents φ 1 as being-had-by-everything. When does an object have φ 1 ? An object o 1 has φ 1 iff ∃y∀z(F(o 1 , y, z)) iff φ 2 has the property of being-had-by- something. An object o 2 has φ 2 iff ∀z(F(o 1 , o 2 , z)) iff φ 3 has the property of being-had-by- everything. An object o 3 has φ 3 iff R(o 1 ,o 2, o 3 ). Now consider a proposition with restricted quantification such as that ∀x:F(x)[G(x)]. Let φ 1 be the property of being an x such that G(x). Let φ 2 be the property of a being x such that F(x). Let φ 3 be the property of being had by anything that has the property φ2. Let φ 4 be the property of being an x such that F(x)→G(x). That ∀x:F(x)[G(x)] represents φ 2 as having φ 3 . A property ψ has φ 3 iff φ 4 has the property of being-had-by-everything. Since the section on sentential operators covered how we come to understand what it is to have φ 4, no more analysis is necessary. It is important to emphasize that although we can come understand what it is for being-G to have the higher-order property of being-had-by-anything-that-is-F(x) in terms of the conditional F(x)→G(x) this does not entail that restricted quantification is actually of the form ∀x(F(x)→G(x)). Finally, there are some remaining issues concerning what propositions expressed by sentences with quantifiers are about. Consider the proposition that everything has mass. Call this proposition the ‘Student of Physics Hypothesis’. What is the Student of Physics Hypothesis about? Well it certainly is about the property of having mass. Therefore, this proposition must represent the property of having mass as being some way. So far so good. However, according to my analysis the proposition represents the property of having mass as being-had-by-everything. Therefore, the Student of Physics Hypothesis must be about the property of being-had-by- everything. I take this to be a surprising but acceptable result. After all, consider the Student of Sophistry Hypothesis that nothing has mass. Both of these hypotheses are about the property of having mass but it is not true that what both these hypotheses are about is identical. One is about a property being had by everything and the other is about a property being had by nothing. 107 3.4 Remaining Types At this point it would be natural to consider propositions expressed by sentences that contain definite descriptions. However, confusion over whether certain features of assertions made using definite description is due to semantic versus pragmatic factors makes it unsuitable to analyze what these propositions are here. In the following section I will show how my theories of propositions, truth, and accuracy help us make progress in understanding poorly understood semantic phenomena, including features of definite descriptions. 4. Application to Poorly Understood Semantic Phenomena In this section, we use our findings on what way complex propositions represent things as being in conjunction with Artifactualism and the Functional Theory of truth-aptitude to shed light on some poorly understood semantic phenomena. 4.1 Multi-Clausal Sentences A multi-clausal sentence such as ‘John’s computer, which he bought in 2012, is broken’ is somewhat of a mystery in contemporary semantic theorizing. If John’s computer is broken but he bought the computer in 2010, is this sentence true or false? 118 When asked this question, neither answering true nor false seems satisfactory. With our theory of propositions, accuracy, and truth we can solve this mystery. The multi-clausal sentence is not an instance of a single conjunctive proposition. Rather the sentence is an instance of two propositions. The sentence itself does not have a unique representing-as function. It has multiple representing-as functions. However, some of its proper parts have a unique representing-as function. Therefore, the sentence is capable of being completely true as well as completely accurate but it is not capable of being true simpliciter. This explains our conflicted feelings when only one of the propositions is false and we are asked whether the sentence is true. The sentence is neither true nor false but rather partially true. When both propositions are true then we do want to say that whoever utters this sentence says something true. But according to the Functional Theory of truth-aptitude the sentence would not be true, but rather completely true. I think this is to be explained by pragmatics. When a speaker represents John’s computer as being one way and then also represent John’s computer as being 118 See both Bach (1999) and Potts (2005) for some discussion of this problem. 108 another way, if we all agree that John likely represented things as they are on both occasions, we take him to have also represented his computer as being-such-that-it-is-broken-and-it-was- bought-in-2012. So John might end up asserting the conjunction given certain pragmatic factors and therefore assert something that is true and not just completely true. Now one might wonder why I would not instead claim that the sentence itself is an instance of a single conjunctive proposition and instead use pragmatics to explain our unwillingness to ascribe falsity to John’s assertion when one of the propositions is false. The idea would be that if we suspect that only one of the propositions is false we treat John as asserting two propositions rather than one. However, I take this to be a worse explanation of the phenomena at hand because the sentence intuitively represents John’s computer as being one way with the subordinate clause and then represents the computer as being another way with the main clause. We put commas in such a sentence to indicate this. Of course these considerations are not conclusive, but I suspect that considering some other similar phenomena will make it clear that we ought to believe that multi-clausal sentences are instances of multiple propositions. In any case, regardless of whether the sentence itself is an instance of two propositions, our account of accuracy can explain our resistance to assert that what John said is false when his computer is broken but was not bought in 2012. 4.2 Descriptions We often say that someone’s description is accurate or inaccurate rather than saying that the description is true or false. For example, if John describes the suspect he saw by saying ‘The suspect was short, dark-haired, pigeon-footed, and spoke with a Sicilian accent’ the police will want to figure out how accurate John’s description is. If it turns out that the suspect actually spoke with a Northern Italian accent we would not say that John’s gave a false description but rather say that his description was partially inaccurate. Furthermore, although it is a weaker intuition, we probably would neither say that John’s description was partially true nor say that it was partially false. This data concerning descriptions is truly confounding to contemporary semantics. Here we have one sentence that seems to describe the world. It is a description for heaven’s sake. How could it neither be true nor false. The description represents its subject as being one way, then another, and then another way rather than representing its subject as being one complicated way. However, unlike multi-clausal sentences, no proper part of the description is an instance of any proposition. In the sentence ‘John’s computer, which he bought in 2012, is broken’ the main clause is an instance of one 109 proposition and the subordinate clause contains an instance of another proposition. Since it has proper parts that are instances of propositions it has conditions for when is completely/partially true. The sentence ‘The suspect was short, dark-haired, pigeon-footed, and spoke with a Sicilian accent’ represents the suspect as having been short, represents the suspect as having been dark- haired, and so on but no part of the sentence is an instance of the proposition that the suspect was short, an instance of the proposition that the suspect was dark-haired, and so on. As with multi-clausal sentences, there may be some instances where someone who utters this sentence ends up asserting, via pragmatics, a conjunctive proposition or ends up asserting that the suspect had a complex property. In such cases we are willing to say that what John asserted is true. As with multi-clausal sentences, one could instead claim that the sentences is an instance of a conjunctive proposition and use pragmatics to explain why we hesitate in attributing truth, partial truth, and falsity to what someone asserts when he normally utters this sentence. Again, I think the original explanation is correct since the commas in the sentence are indicators that the subject is being represented one way and then being represented another way. Contrast this with descriptions that we are willing to call true or false such as ‘The suspect was pigeon-footed’ and ‘The suspect’s coat was light green’. Each of these sentences are instances of a single proposition. Notice that the adjectives in the second sentence are not separated by commas since being-light-green is a single way the coat is represented as being. The coat is not represented as being light and then represented as being green. On the other hand ‘The suspect’s coat was light-green, polyester, and smelled of cigars’ contains commas since it represents the coat as being several ways. 4.3 Lists, Paragraphs, Speeches, and Discourses What we have learned about multi-clausal sentences and descriptions can be easily be applied to more complex linguistic structures such as lists, paragraphs, speeches, and discourses. We now know that any theory that tries to tell us that these things are true or false must be an incorrect theory. 119 We also now have the tools to explain a previously puzzling feature of entities that contain multiple sentences. If John gives a speech, writes a paragraph, or so on it might turn out that everything John said is true. But isn’t everything John said identical to his speech, his paragraph, and so on? And we would not say that John’s speech is true or that his paragraph is true. Well, as we know, the proposition that everything John said is true represents the property 119 Ignoring very short lists, paragraphs, speeches, and discourses that are instances of a single proposition. 110 of being-said-by-John as being such that anything that has it also has the property of being-true. It does not represent John’s speech, paragraph, or so on as having the property of being-true. 4.4 Plurals Applying our theory to questions about plurals also leads to some interesting results. Let ‘Al’, ‘Bob’, and ‘Christie’ be names of some rocks. Now consider the sentence ‘Al, Bob, and Christie are granite rocks’. If it turns out that only Al and Bob are granite rocks would we say that this sentence is false? Unlike the cases of multi-clausal sentences and descriptions, we are not as hesitant to say that the sentence is false. Similarly if we refer to Al, Bob, and Christie by saying ‘They are granite rocks’ we seem to have said something false. We have represented them as being some way and they are in fact not that way. We represented Al, Bob, and Christie as being granite rocks. We also know that the proposition that Al, Bob, and Christie are granite rocks cannot represent a set or even a group as being some way. The proposition is about Al, Bob, and Christie. It certainly is not about a set containing these rocks since that set does not have the property of being a granite rock let alone being granite rocks. It also appears that the proposition is not about any group that these rocks might be a part of. The group does not have the property of being granite rocks. The group has the property of having members that are granite rocks. These observations concerning aboutness lead us to suspect that the proposition that Al, Bob, and Christie are granite rocks represents these rocks as being granite rocks. Now I would imagine that the majority of philosophers recoil at the idea that there is a property of being-granite-rocks. What does this property apply to? Plural things? The metaphysical questions that underlie this worry are certainly substantive. However, we have no problem considering representations that represent some things as being one way and we know that representing things as being one way cannot be identical to representing a group or set as being some way. In fact, a simple proposition like that snow is white has to represent multiple things as being one way. It represents snow as being an x and the property of being white as being a φ such that φ(x), which is to do one and the same as represent snow as being white, which is in turn, to represent being whiteness as being had by snow. So there should be no mystery how we represent Al as being-such-that-he,-Bob,-and-Christie-are-granite-rocks. But one might still try to find a way out of admitting that there is such a property as being- granite-rocks. First, we know that representing these rocks as being granite rocks cannot be one 111 and the same as representing that Al is a granite rock, representing that Bob is a granite rock, and representing that Christie is a granite rock since any representation that represented these three things could only have accuracy-conditions rather than truth-conditions. Furthermore, there is reason to doubt that representing these rocks as being granite rocks is one and the same as representing Al as being-such-that-he-is-a-granite-rock-and-Bob-is-a-granite-rock-and-Christie- is-a-granite-rock. Here’s why. Consider the proposition that these rocks surround a shrub. Assume that these rocks do surround a shrub. Representing these rocks as surrounding a shrub cannot be one and the same as representing Al as being such-that-he-surrounds-a-shrub-and-Bob- surrounds-a-shrub-and-Christie-surrounds-a-shrub. No single rock, in this case, surrounds a shrub. The rocks surround the shrub. So, in summation, our account of propositions, accuracy, and truth rules out several ways one might go about understanding sentences with plurals. It forces us to posit what one might call ‘plural properties’, i.e. a way of being that only multiple things can be. However we are already completely comfortable asserting that a representation can represent multiple things as being one way and we need to posit these “plural properties” to explain the nature of all propositions since all propositions represent things as being some way. 4.5 Tense and Aspect So far I have ignored the question of whether propositions can change in truth-value. 120 Since the time of Frege it has been a widely held assumption that propositions do not change in truth-value. Rather, propositions represent some objects as being some way at some time t and the way those objects were, are, or will be at t never changes. We now have strong reasons to doubt this claim. Consider the sentence ‘Detroit is a populous city’. According to the Fregean picture this sentence alone cannot be an instance of a proposition where as the sentence ‘On 8/9/2013 a.d. at 11:39 PST, Detroit is a populous city’ is an instance of a proposition. But if this is correct, then every proposition is about a time since every proposition must represents a time as being some way. 121 This is absurd. Yes, other stronger considerations could force us to accept this result, but to accept this result is to undeniably bite the bullet. Furthermore, there appears to be real differences between the properties of representing things as they were, representing things as they are, and things as they will be. A proposition is 120 See Richard (1981) and Brogaard (2012) for detailed discussion on this issue. 121 Strictly speaking every proposition would be about a time or the property of being-a-time. 112 true only if it represents things as they are. I do not think the word ‘are’ we are using to pick out the property representing-things-as-they-are is a tenseless verb. If it were, we would say ‘A proposition is true only if it represents things as they be.’ The proposition that Detroit is a populous city is true only if it represents things as they are. If it represents things as they were but not as they are, then it is simply not true. It was true, but it no longer is. The same goes for representing things as they will be. The propositions that there are 10 billion people living on the Earth might very well represent things as they will be but it does not now represent things as they are. On the other hand the proposition that there were once less than a billion people living on Earth represents things as they are and, for all we know, the proposition that there will be 10 billion people living on the Earth represents things as they are. These considerations suggest that truth involves representing things they are and not “how they be a t”. Of course there a metaphysical considerations that might lead one to doubt that there really are distinct properties of representing things as they were, are, and will be respectively, but we have at least shown that to deny that these are distinct properties is to go against clear intuitions that we have. 4.6 Non-Referring Demonstratives Although non-referring demonstratives are perhaps better understood than the phenomena we have been considering, our theory of propositions, accuracy, and truth gives an unequivocal answer as to whether a sentence or utterance made with a non-referring demonstrative expresses a proposition and is capable of being true or false. If someone fails to refer with a demonstrative then they are not able to represent anything as being anyway. If the sentence/utterance does not represent anything as being any way then it is not an instance of, and therefore does not express, any proposition. Furthermore, the sentence/utterance can neither be true or false given our definition of truth since it does not represent anything as being anyway. 4.7 Definite Descriptions Our theory of propositions, accuracy, and truth also helps us make progress when it comes to definite descriptions. First, we have a way to check the predictions made by a given theory of definite descriptions concerning when a sentence containing a definite description expresses a proposition and when the sentence is either true or false. Furthermore, we can use our intuitions about what propositions/hypotheses are about to help evaluate which theories are better than others. 113 If we treat a sentence such as ‘The king of France is bald’ as representing the property of being-king-of-France as being-such-that-one-and-only-thing-has-it-and-anything-that-has-it-also- has-the-property-of-being-bald then this sentence will be an instance of, and therefore express a proposition, regardless of whether there is a king of France. When there is no king of France this sentence as well as the proposition that the king of France is bald will be false. On the other hand, if we treat the sentence as representing whatever ‘the king of France’ denotes as being some way then the sentence will express a proposition and either be true or false only when there is a king of France. Finally, if we treat the sentence as representing the property of being-king-of- France as being one way and also representing being-king-of-France another way, the sentence will not be an instance of a proposition and only have accuracy-conditions. Keeping with our previous analyses of other types of propositions, we should think that the proposition that the king of France is bald represents the property of being-king-of-France as being-a-φ-such-that-the-φ-is-bald, i.e. being-a-φ-such-that-ιx:φ(x)[bald(x)]. Of course we need to explain how we come to understand the property of being-a-φ-such-that-the-φ-is-bald. But even without giving this explanation, we can see that the proposition represents being-king-of-France as being some way and therefore appears to not be truth-valueless. In fact the only way the sentence could be truth-valueless when there is no king of France is to claim that the phrase ‘being-a-φ-such-that-the-φ-is-bald’ does not pick out a property/way something could be when there is no king of France. Although it is possible that such a phrase could fail to pick out a property, we now have a completely different question to focus on when trying to answer whether the sentence is false or truth-valueless. Furthermore, it appears to be a question that we can make progress on. After all, even if there is no king of France there is a unique president of the United States and the sentence ‘The president of the United States is bald’ is certainly not truth-valueless. Doesn’t this sentence represent the property of being-president-of-the-United-States as being-a-φ-such-that-the-φ-is- bald? If so, then the property of being-a-φ-such-that-the-φ-is-bald exists whether or not there is a king of France. Therefore the sentence ‘The king of France is bald’ appears to be false even when there is no king of France. As for what propositions expressed by sentences with definite descriptions are about, consider what we will call the ‘Anti-Euclid Hypothesis.’ The Anti-Euclid Hypothesis is that the largest prime number is odd. If someone asks you what the Anti-Euclid Hypothesis is about you will hesitate to say it is about the largest prime number since you know that there is no such 114 number. On the other hand you will not say that the hypothesis is about nothing. This gives us good reason to think that the hypothesis is about the property of being-the-largest-prime-number and not about whatever is denoted by phrase ‘the largest prime number’. Now there certainly are cases where we use a sentence like ‘The president of the United States is wise’ to assert something about a person rather than the property of being-president-of- the-United States. I suggest that this phenomenon is best explained via pragmatics. In many or perhaps even most cases where we use a definite description we succeed in asserting a proposition that is about the thing uniquely picked out by the description or at least uniquely picked out by the description and some other pragmatic features. Giving an account of what needs to be the case in order to succeed in making an assertion about some particular object rather than just making an assertion about the property used in the definite description is no small task. And nothing I have said in this section conclusively shows that the semantics of definite descriptions are a certain way. However, I hope to have shown that our theory of propositions, accuracy, truth at least gives us some needed leverage in answering the long-standing and recalcitrant questions about the semantics of definite descriptions. 4.8 Phenomena Often Categorized as “Convential Implicature” Our inability to figure out which linguistic phenomena are due to semantics rather pragmatics has been a hindrance in making progress in semantics. The problem has only been worsened by an inability to precisely specify what role semantic content is to play in our theorizing. But when we understand the relationship of a sentence/utterance expressing proposition in terms of the sentence/utterance being an instance of that proposition, the role of content is much clearer. Discerning the semantic content of an indicative sentence/utterance becomes a task of figuring out what the sentence/utterance represents as being what way. If the sentence/utterance does not represent anything as being anyway, it has no semantic content. This is not to say the sentence/utterance is meaningless since we have not, and should not, assume that to be meaningful is to be an instance of a proposition. Furthermore, if a part of sentence does not affect what the sentence represents as being what way, then that part of the sentence is not “part” of the semantic content. This observation is important since there continues to be a lot of debate about what “enters into” the semantic content without any clear conception of what counts as “entering into” semantic content. It should be clear that utterance modifiers like ‘therefore’, ‘in any case’, and ‘roughly’ do not affect what 115 proposition a sentence is an instance of because they do not affect what the sentence represents as being what way. Utterance modifiers are therefore not “part” of the semantic content of sentence. With these facts in mind about what it is to express a certain content, we have a lot more traction on problems such as whether ‘Shaq is huge but agile’ has a different content than ‘Shaq is tall and agile’. The problems are still too difficult to begin to solve here, but our theory of propositions provides a way forward. 5. Concluding Remarks Assuming that all truth-apt complex sentences and complex propositions represent things as some way leads to some surprising but also ends up doing important work in helping us better understand certain semantic phenomena we are currently puzzled by. In short, the surprising results we get by assuming that all truth-apt complex sentences and complex propositions represent things as being way are worth it. Without assuming this we appear doomed in giving an account of truth-aptitude. And if we do assume this, we can make progress on in semantics theorizing. 116 Chapter 5 : Artifactualism and the Puzzles of Belief Abstract In this chapter we see how Artifactualism can give an elegant solution to long-standing puzzles of belief such as Frege’s puzzle. The solution is Fregean in one sense because it posits a proposition distinct from that Venus is identical to Venus that the ancients came to know. The basic idea is that some artifacts are specialized representers. Specifically, some artifacts have the function of representing Venus as being identical to Venus, but other artifacts have the more specialized function of representing Venus as being identical to Venus by referring to Venus with ‘Hesperus’ in one instance and referring to Venus with ‘Phosphorus’ in the other instance. These specialized token tools are instances of a specialized proposition that is representationally identical to the plain proposition that Venus = Venus. Importantly, this specialized proposition is what the ancients came to know when they learned that Hesperus is Phosphorus. I argue that although Soames’ and Hanks’ theory of propositions can give a similar solution the puzzles, Artifactualism gives the most plausible and straightforward account of how tokens of sentences like ‘The ancients learned that Hesperus is Phosphorus’ can be true. 1. Introduction The recent renewal of interest in the nature of propositions is in part motivated by the hope that doing so will provide us with superior ways of solving longstanding puzzles of belief. Puzzles like Frege’s Hesperus/Phosphorus cases, Kripke’s Pierre and Paderewski cases, and Perry’s essential indexical cases, to name a few, still loom large. 122 One promising view recently put forward is what we will call Actionism. Actionism is the view that propositions are act-types. As we saw in chapters 2 and 3, Peter Hanks and Scott Soames both offer us Actionist views where propositions are act-types of predicating/representing. 123 According to such views, the proposition that snow is white is the act-type of predicating whiteness of snow, which is basically the act-type of representing snow as being white. Agents perform token acts of representing snow as being white. The proposition that snow is white is the type of action that these token acts are instances of. 122 C.f. Frege (1997b), Kripke (2011), and Perry (1979) respectively. 123 See Hanks (2011) and Soames (2015). 117 Actionists can offer what we will call a Propositionalist Solution to the long-standing puzzles. Propositionalist Solutions claim that, for example, there is a proposition distinct from that Venus is identical to Venus that the ancients came to know when they learned that Hesperus is Phosphorus. We all know the classic propositional solution that Frege’s “Sense and Reference” offers and we all know its problems. Actionism is positioned to offer a Propositionalist Solution that avoids these problems. Following the demise of descriptivism, the dominant trend in the last forty years has been Non-Propositionalist Solutions. The general idea is that we can, for example solve, Frege’s puzzle without positing a distinct proposition that the ancients came to know when they learned that Hesperus is Phosphorus. Instead, the view is that the ancients merely came to stand in a different cognitive relationship to the proposition that Venus is identical to Venus. As we will see shortly, there are good reasons for preferring a Propositionalist Solution, all other things being equal. So, Actionism is positioned to give an attractive solution the puzzles. However, it also has bizarre consequences. If act-types are propositions, then act-types have truth-conditions. Actionism entails that the proposition that snow is white is the type of thing agents do. Actionism entails that certain types of things we do are true and certain types of things we do are false. These consequences are undoubtedly revisions to how we pre-theoretically think. More importantly, chapter 2 gave us an important reason for preferring Artifactualism over Actionism. The Functional Theory of Truth-Aptitude rules out Actionism. My aim here is to show that Artifactualism is a promising alternative to Actionism when it comes to solving the long-standing puzzles of belief. It can offer a similar Propositionalist solutions to the puzzles without entailing the bizarre results that propositions are things we do and that things we do can be true or false. As for how we will proceed, Section 2 motivates both Actionsim and Artifactualism by arguing that we ought prefer a view of propositions that can give a Propositionalist Solution to the puzzles. Section 3 covers Actionism in more detail. We see more motivation for the view, how it can solve the puzzles, and exactly what problems it faces. Section 4 shows how Artifactualism can solve the puzzles in a similar fashion to Actionism. Section 5 compares how Actionism and Artifactualism fare when it comes to explaining the nature of attitude reports in general. The upshot is that Artifactualism has a slight advantage in this regard. In general, this chapter does not give significant reasons for preferring Artifactualism over Actionism when it comes to solving the puzzles of belief. It simply shows that Artifactualism does as well if not slightly better in this regard. This result should ultimately be 118 considered in conjunction with the arguments in previous chapters meant to establish that Artifactualism is superior to competing views in other regards. 2. For Propositionalist Solutions Non-propositionalist solutions to the puzzles of belief have been something of the status quo in the last 30 years. 124 It is not hard to see why propositionalist solutions have fallen out of favor. If there is indeed another proposition the ancients came to know when they learned that Hesperus is Phosphorus, what is this entity and how does it differ from the proposition that Venus is identical to Venus? This new proposition must have the same truth-conditions as that Venus is identical to Venus, the same modal profile, be about the same things, and yet still be distinct. Without beginning a serious inquiry into the metaphysics of propositions, we can have no firm point of agreement concerning the properties propositions have and no principled way to argue that there is indeed a distinct proposition that the ancients came to know. The metaphysics behind the propositions in the most well known propositionalist solution, Frege (1997c), is obscure to say the least. More importantly, although Frege’s propositionalist solution gives us a distinct proposition that the ancients came to know, this proposition appears to have the wrong modal profile. At least on a naive reading of Frege, the proposition expressed by ‘Hesperus is Phosphorus’ could equally well be expressed by a sentence of the form ⎡The F is the G⎤ and so this proposition is a contingent, rather than a necessary, truth. Contemporary non-propositionalist solutions avoid embarking on an inquiry into the metaphysics of propositions and still seem to do pretty well accommodating our pre-theoretical judgments. For example, Salmon (1986) and Braun (1998) claim that there is not a new proposition the ancients came to know, but rather the ancients came to have a new way of grasping, or standing in an important cognitive relationship with, that Venus is identical to Venus. Salmon claims that our judgments concerning how the semantics of ‘A believes that Hesperus is Phosphorus’ differs from ‘A believes that Venus is Venus’ are to be explained away by pragmatic phenomena. Crimmins and Perry do think there are semantic differences between these belief reports, but share the view that there is not a new proposition the ancients came to know. 125 124 Bealer (1993), along with the current resurgence of interest in propositions, is an example that goes against this trend 125 See both Crimmins and Perry (1989) and Crimmins (1992) 119 Rather, the ancients simply came to stand in a new cognitive relationship to the proposition that Venus is identical Venus. 126 There are problems with the details of pretty much every non-propositionalist solution given to date. 127 But my aim here is not summarize and add to the list of problems for specific versions of non-propositionalist solutions. Rather, I think non-propositionalist solutions in general all share a serious defect that should not be taken lightly. It is easy to become satisfied with non- propositionalist solutions given the failure of past propositionalist solutions. What more could we want, as long as we give a plausible account where there is some new mental state that the ancients came to be in after they learned more about Venus? The problem with all non-propositionalist views is that they cannot straightforwardly accommodate the simple fact that there is something new the ancients came to know. It is not just that the ancients ended up being in a new belief state. They learned something new. There is one more thing that they knew after their discovery. They did not come to know a belief state. They did not come to know that Venus is identical Venus; they already knew that. The thing they came to know is that Hesperus is identical to Phosphorus. Without being highly revisionary in accommodating our pre-theoretic judgments, it is unavoidable that our pre-theoretic ways of talking and thinking about beliefs and knowledge commit us to there being some thing that is distinct from that Venus is identical to Venus that the ancients came to know. It is plausible that when the ancients learned this new thing, they came believe that Venus is identical to Venus in a more particular/special way. 128 But we cannot escape the fact that we pre-theoretically judge that they were not just in a new belief state, but also came to know one more thing. 126 Other views that might be considered non-propositionalist are veiews such as Richard (1990), Higginbotham (1991), and Larson and Ludlow (1993). These views share the common feature that they posit a different entity as the content of the complement clause in attitude reports. These views do not face the problems we are going to cover for the non-propositionalist views, because these views do, for example give us a new entity that the ancients came to know when they learned that Hesperus is Phosphorus. In this way the are more akin to Actionism and Artifactualism and might as well be considered propositionalist views. These views do however, have their own unique problems that we cannot go into here. See Sider (1995) and Nelson (2005) for problems with Richard’s view. See Clapp (2002) for problems with Higginbotham’s and Ludlow and Larson’s views. 127 See Schiffer (2006) and Richard (1997) for problems with the views that attempt explain away our semantic intuitions. See Stanley (2002), Reimer (1995), Schiffer (1996), and Bach (1993) for problems with non-propositionalist views that claim attitude reports are context sensitive. 128 We can even grant that learning this new thing is partially constituted by being in this new more specific state of believing that Venus is identical to Venus in a more particular way. But there is no way that being in this state is identical standing in the knowing relation to this belief state. 120 When an inquiry into the metaphysics of propositions looks daunting and the failures of past propositionalist solutions are fresh in our minds, it easy to look over the fact that non- propositionalist solutions fail to give us a new proposition that the ancients came to know. But this is not some obscure theoretical failure. We pre-theoretically judge that there is some new thing that is the ancients stand in the knowing relation to, and non-propositonalist theories fail to give us such an entity. Now, there will always be revisionary and tricky ways to get around problems like this. A proponent of non-propssitionalist solutions might find a tricky way to claim that the sentence ‘The ancients came to know one more thing when they learned that Hesperus is Phosphorus’ is true even though there is not one more thing that they came to know. Some revisionary account of the semantics of attitude reports would be the way to fudge the truth here. But, without being highly revisionary, propositionalist-solutions can accommodate our simple, but important, pre- theoretic judgment that the ancients came to know one more thing when they learned that Hesperus is Phosphorus. And for this reason, all other things being equal, a propositionalist- solution is to be preferred. This is not to say that our folk pre-theoretic ways of picking out mental states are the be all and end all when it comes to the tools we have for describing the mental states of agents. There are surely more specific way of believing that Venus is identical to Venus that our folk-practices fail to capture. There is also surely something right in Kripke’s claim that the puzzles present cases where “our normal apparatus for the ascription of belief is placed under the greatest strain and may even break down”. (Kripke 2011: p.158) The strain is real. The breakdown seems imminent when we try to give a theory of what is going on. Yet, on a day to day level, we sill operate as if the apparatus of making propositional attitude reports almost always works. Everything seems to be working fine when I tell you that the ancients came to know something new when they learned that Hesperus is Phosphorus. We just need a theory of propositions that can accommodate this, and both Actionism and Artifactualism are positioned to do just that. 3. Actionism As we have seen, Actionism is the view that propositions are act-types. Agents perform particular representational acts. While telling a story, an agent might represent the Sun as being god. While day dreaming, an agent might perform the mental act of representing the Eiffel Tower as being made of toothpicks. Representing the Sun as being a God and representing the Eiffel Tower as 121 being made of toothpicks are types of acts that agents do. Actionism claims that such representational act-types are in fact what propositions are. According to Actionism, the proposition that the Sun is a god just is the type of action of representing the Sun as being a god. 3.1 Details and Motivation Hanks (2011) and Soames (2015) are the two prominent Actionist views. Both claim that propositions are acts of “predication”, but upon inspection of their respective views, it is clear that to predicate a property of an object is to represent that object as having that property. So in what follows, we will stick to talking about acts of representing rather than acts of predicating to keep things simple. One difference between Soames and Hanks concerns whether the acts of representing must be mental acts in order to count as propositions. According to Hanks both public communicative acts of representing, like telling a story, and private mental acts of representing, like imagining, are instances of the act-types that are propositions. According to Soames, propositions are only mental act-types. 129 This difference between Hanks’ and Soames’ views, however, will not matter for our purposes. What reason do Actionists give for claiming that propositions are types of actions? The main reason given is that types of actions appear to have an important property that propositions must have. Without going into the bulk of the details, if propositions are the things asserted and believed, then they must have their truth-conditions independently of the intentions and conventions of agents. Sentences from natural and artificial languages mean what they do, and have the truth-conditions that they have, in virtue of the actions, intentions, and conventions of agents. But propositions are not this way. If we hold fixed what someone actually asserted when they assert that snow is white, we necessarily hold fixed the conditions under which what they actually asserted would be true or false. In a word, propositions have their truth-conditions inherently. Actionists are optimistic that types of actions are the kinds of things that could have truth- conditions inherently. The act of representing snow as being white is inherently a representational act. It is the kind of thing, by its very nature, where one represents snow as being white. The act does not have this property in virtue of the action, intentions, or conventions of agents. Rather, by its very nature, it is the type of act such that a performance of it requires a token representing of snow as being white. Actionists use the fact that the act-type of representing snow as being white 129 C.f. Soames (2015: 31) 122 is representational by its very nature to go on an argue that this act-type not only has truth- conditions, but has the truth-conditions it has, simply in virtue of its nature. 130 According to Actionists, the act-type of representing snow as being white is true iff snow is indeed white and the act-type has this property merely in virtue of its nature and not in virtue of the actions, intentions, and conventions of agents. So Actionists claim that it inherently represents snow as being white and is therefore inherently such that it is true iff snow is indeed white. 3.2 How it can Solve the Puzzles The basic idea behind the Actionist solutions is that there are propositions which are identical when it comes to what things they represent but are still distinct. For example, there can be two propositions that both represent Venus as being a planet but are still not one and the same proposition. Roughly, there are different ways of performing the action of representing o as being F, and when we build these different ways into the act-type we get propositions that each represent o as being F but are still distinct because each type of action is a different way of representing o as being F. In general, if two act-types A and B both are acts of representing o as being F, and B is a more particular way of performing A, then A and B are representationally identical but distinct propositions and B is an “enriched” version of A. For example, we first note that, according to Actionism, the “plain” proposition that Venus is a Planet is simply the act-type of representing Venus as being a planet. But one can perform this general act in many more particular ways. One can represent Venus as being a planet by referring to Venus with ‘Hesperus’. So there is the basic proposition that Venus is a planet which is the general act of representing Venus as being a planet, where no way of doing the act is built into the act-type itself. But there are also the “enriched” propositions that are representationally identical to that Venus is a planet. One such enriched proposition is the act-type of representing Venus as being a planet by referring to Venus with ‘Hesperus’. Both the general and more specific act- types are acts of representing Venus as being a planet, but one is a more determinate way of doing so. Actionists claim that the more determinate act-type is an enriched version of the proposition that Venus is a planet. The plain proposition and the enriched propositions are identical in what they represent but are still not one and the same proposition. To give another example, there is the plain proposition that Mark Twain is Sam Clemens, which according to Actionism, is just the act-type of representing Mark Twain as being identical 130 C.f. Soames (2015: pp. 24-26). 123 to Mark Twain. But there are also enriched versions of this proposition. There is the proposition that is the act-type of representing Mark Twain as being identical to Mark Twain by referring to Mark Twain with ‘Mark Twain’ in one instance and referring to him with ‘Same Clemens’ in the other instance. Once we see the basic idea behind the representationally identical but distinct enriched propositions that Actionism allows for, we can go about solving the long-standing puzzles of belief. Starting with Frege’s famous Hesperus/Phosphorus cases, there is of course the basic proposition that Venus is identical to Venus which is the act-type of representing Venus as being identical to Venus. Note that we could pick out this act-type with the locution ‘the act-type of representing Hesperus as being identical to Phosphorus’. You cannot represent Venus as being someway without representing Hesperus as being some way because they are one and the same thing. You cannot represent some thing as being some way without representing that thing as being that way. When the ancients discovered that Hesperus is Phosphorus they did not come to know the plain proposition that is the act-type of representing Venus as being identical to Venus. They already knew this. Rather they came to know an enriched proposition that is the act-type of representing Venus as being identical to Venus by referring to Venus with ‘Hesperus’ in once instance and referring to Venus with ‘Phosphorus’ in the other instance. It is important to note that solving Frege’s puzzles this way is a vast improvement over meta- linguistic solutions. Meta-linguistic solutions claim that what the ancients came to know is the proposition that the words ‘Hesperus’ and ‘Phosphorus’ co-refer. But this meta-linguistic proposition is a contingent one. What the ancients came to know is a necessary identity claim. Actionism gets this right because the proposition the ancients came to know is still about Venus and the relation of being identical. If Actionism is correct, the enriched act-type represents Venus as being identical to Venus and is therefore true iff Venus is indeed this way. The meta-linguistic proposition represents the wrong things as being the wrong way and therefore has the wrong properties. Even though the enriched proposition is an act-type that involves referring with particular words, it is still an act-type of representing Venus as being identical to Venus. As for Kripke’s Paderewski case, the same solution will not quite work. In Kripke’s case, Peter thinks that someone named ‘Paderewski’ is a statesmen and thinks that some other person he is familiar with, also name ‘Paderewski’, is a musician. As a matter fact, Peter is mistaken and 124 he is just familiar with one and the same person named ‘Paderewski’ who is both a statesmen and a musician. Peter is willing to assert in one breath, and seems to believe, that Paderewski has musical talent and that Paderewski does not have musical talent. The problem is that he uses one and the same word to refer to Paderewski twice even though he thinks that he is referring to two different individuals. Now, there is the basic act-type of representing Paderewski as being such that he has musical talent and as being such that he does not have musical talent. But there is also the act-type of doing this by referring to Paderewski with one and the same word on each occasion. This enriched proposition builds in a way of performing the act, but it is less specific than the ways built into the enriched propositions we needed for Frege’s puzzles. It just builds in that Paderewski is referred to by the same word on both occasions. Peter does not believe the enriched proposition, because if he recognized that he was thinking of one and the same thing as being contrary ways, he would not believe this. The enriched proposition is the one we all trivially and a priori know to be false. The plain proposition is the one we believe if we are the victim of a belief puzzle. Soames and Hanks solve the Paderewski-type puzzles in slightly different ways from the solution given above. 131 Soames claims that the enriched proposition that Peter does not believe is the act-type of representing Paderwski as having contradictory properties while recognizing the recurrence of one and the same referent being picked out twice. (Soames 2015: 144-146) For various reasons that we cannot fully go into here, I think Soames should avoid claiming that act- types of representing o as being F while doing φ count as enriched propositions. 132 Hanks on the other hand claims that acts of referring can differ not just by what name we use to refer but how we are thinking of the referent when we use that name. (Hanks 2011: pp. 45-48) So Hanks basically solves the Paderewski case in the way he solves Frege’s puzzle, by claiming that what Peter believes is an enriched proposition that involves two different ways of picking out a referent. Hanks’ view is that the enriched proposition that Peter believes is the act-type of representing Paderewski as being contradictory ways by referring to Paderwski one way and then 131 Soames’s view of what it is to entertain a proposition bars him from this solution since he claims that to entertain a proposition is to perform the representational act that is the proposition. Soames (2015: p.24). Since Pierre does unknowingly use one and the same name to pick out Paderewski in thinking this thought and is still disposed to accept the thought when he does so, Soames cannot claim that this is the trivially a priori false proposition that Pierre rejects. 132 We do not want a multitude of theoretically inept enriched proposition. Soames (2015: pp. 87-89) gives us an account of which acts count as enriched propositions that actually appears to rule out having enriched propositions that are distinguished by having recognition of recurrence built in. The recognition of occurrence is “representationally irrelevant” and plays no role in brining about what is represented. Recognizing the recurring constituents is not a more determinate way, not even partially a more determinate way, of represent Paderewski as being musical and unmusical. 125 referring to Paderewski another way. The first way of referring is referring to Paderewski with ‘Paderewski’ while thinking of the referent as a statesmen. The second way is referring while thinking of the referent as a musician. Again, for reasons I cannot go into here, I believe the first solution I sketched is the best way for actionist to go, but all three options are open to one that accepts Actionism. 133 We now turn to Perry’s puzzle of the essential indexical. In this case, someone is in a supermarket pushing around a shopping cart with a ripped bag of sugar. The messy shopper sees a trail of sugar on the floor but does not know that he himself is the creator of this trail. He even sees himself in the store mirror, leaving the trail of sugar, but does not realize that the person he is seeing is himself. We want a theory where an agent can fail to know that he himself is the messy shopper even though he knows that that thing he de re picks out by pointing at the store mirror is the messy shopper. To make this case easier to talk about, imagine President Obama is the messy shopper, getting back to normal daily life after the end of his second term. We want there to be a proposition that is distinct from, but representationally identical to, that President Obama is the messy shopper. One way to do this is to build in that Obama is picked out via self-reference. An Actionist can claim that the special proposition that Obama comes to know when he learns that he himself is the messy shopper is the act-type of representing Obama as being the messy-shopper by self-referring to Obama. Hanks (2011) does not actually address these puzzles, but Soames claims that the enriched proposition that Obama comes to know is the act-type of representing Obama as being the messy shopper by thinking of Obama in the first-person way. Strictly speaking, an Actionist need not posit a special first-person way of thinking of themselves. They just need a distinct enriched proposition that Obama comes to know. Puzzles like the essential indexical do raise a particular complication for propositionalist solutions. Only Obama can be in the particular mental state we pick out with ‘believing he himself is the messy shopper’ when ‘he himself’ refers to Obama. In other words, only Obama can be in the state of self-realizing that Obama is the messy shopper. This entails that only Obama 133 Like Soames’ actual solution, this solution gives rise to a multitude of theoretically inept propositions . Is there the enriched proposition of representing Venus as being a planet by referring to Venus with ‘Venus’ while thinking of the referent as being a cool vacation spot? This is why I suggest Actionists should take the route I first presented, which is, in fact, closer to Soames’ solution. Simply claim that Peter is not irrational because he does not believe an enriched proposition we all trivially a priori know is false. The enriched proposition is the act-type of representing Paderewski as being contradictory ways by using one and the same name to refer to Paderewski. 126 can stand in the believing relation to the enriched proposition that he comes to know. But this also entails that only Obama can stand in the ‘thinking that’ and ‘wondering whether’ relation to the enriched proposition we pick out with the complement clause ‘he himself is the messy shopper’ in this case. On the other, hand agents can of course think about the enriched proposition he comes to learn. We are doing that right now. Going into the details of how exactly Actionists have dealt with this complication will take us too far off course. 134 But we will see shortly how an Artifactualist should account for the fact that can we think about the enriched proposition that Obama comes to know even though we cannot “have the thought” that Obama has when thinks that he himself is the messy shopper. There are of course more puzzles such as those in Church (1988), Mates (1952), and Richard (1983). But we have seen enough of how Actionism can handle the puzzles that have long plagued us to rest assured that there are going to be similar solutions to these remaining puzzles. 3.3 Actionism’s Problem with Truth Although Actionism allows for an attractive solution to the longstanding puzzles of belief, it also has a striking and troubling feature. It entails that things we do are true or false. It entails that propositions are things we do. Just as we can do the twist like we did last summer, we can do the Phantom Time Hypothesis and Logicism until we are nauseous. Actionism entails that that snow is white is something philosophers of language have been doing an awful lot in the last 50 years. Hanks and Soames take different routes to addressing this apparent absurdity. Hanks (2011, p.41) claims that “token assertions [are]… the sort of things that can be true or false” and that a proposition “inherits its truth-value from its tokens”. This only sounds plausible when we think of ‘assertion’ as meaning what is asserted. When we use ‘assertion’ to mean the act of making an assertion, we clearly do not pre-theoretically judge that acts of making assertions are true or false. You are not performing a true action when you assert that 2 + 2 =4. What you are saying and asserting is true, but what you are doing is not the kind of thing that we pre-theoretically take to be true or false. Soames grants that Actionism does not appear to mesh well with our pre-theoretical judgments and has “highly counterintuitive consequences”. (Soames 2015: p.37) He attempts to give a sort of error theory as to why we should not take much stalk in our pre-theoretical judgments here, claiming that “every other articulated conception of what propositions are has 134 As mentioned before, Hanks (2011) does not actually address these puzzles so he does not consider this complication, but see Soames (2015:pp. 169-171). 127 counterintuitive consequences”. (ibid) Without going into the details of Soames’ argument that all articulated theories must go against our intuitions in some regard, 135 we simply need to see that Actionism requires giving a revisionary account of our pre-theoretical judgments in this regard. If we can give an alternative view of propositions that does not have these bizarre and revisionary consequences and can do all the work that Actionism does, then we will have reason for preferring that view. Furthermore, as we saw in chapter 2, the most promising view of truth-aptitude rules out Actionism since there is no sense in which action-types do have unique representing-as functions. According to the Functional Theory of truth-aptitude, they are therefore not truth-apt. Given our findings in chapter 1, this entails that action-types cannot be propositions and that Actionism is therefore false. 4. How Artifactualism Solves the Puzzles We saw that Actionism can solve the puzzles of belief by plausibly positing that there are some propositions that are representationally identical yet distinct. According to Actionism, if we have two act-types A and B, and A counts as an act of representing o as being F, and B is a more particular way of performing A, then A and B are representationally identical yet distinct propositions. Artifactualists, like myself, should similarly claim that if there are two representational tool-types A and B, and A is the type of tool has the function of representing o as being F, and B is the type of tool that has function of representing o as being F through some particular means, then A and B are representationally identical but distinct propositions. In other words, if we have two tool-types A and B, and they are both tool-types for representing o as being F, and B is a more specific version of A, then the proposition that is B is a “specialized” version of the proposition that is A. For example, the proposition that Venus is a planet is just the general tool-type of tools that have the function of representing Venus as being a planet. A token of ‘Venus is a planet’ is an instance of this tool-type. However, there are more specific versions of this representational tool. Just as there are clocks and then there are digital clocks, there are tools that have the function of 135 Soames (2015: pp.34-36) argues that any theory that takes propositions to be the meanings of at least some sentences will have counterintuitive results. The basic idea is that this assumption will entail counterintuitive results like that agents believe the meaning of the sentence ‘Snow is white’ when they believe that snow is white. Evaluating this argument is unfortunately beyond the scope of this chapter. 128 representing Venus as being a planet in a specific way. There is the tool-type that has the function of representing Venus as being a planet by referring to Venus with ‘Hesperus’. This more specialized tool-type is akin to the enriched propositions that Actionism posits to solve the puzzles of belief. An agent can believe that Venus is a planet and also rationally disbelieve the specialized proposition mentioned above if they do not know that Venus is Hesperus. Sometimes specific ways of achieving a result are built into the function of an artifact. A flair gun has the function of indicating the position of the shooter of the gun not by any old means. A broken flair gun can be hit against a rock to make loud sounds so that the lost person can tap out Morse code and indicate her position. But the non-functioning flair gun is not performing its indicational function in this case. In order for it to perform its indicational function, the gun must indicate the location of the shooter by launching a flare up in the air. As we will see in the next section, some particular representational artifacts have specific ways of achieving a result built into their function. There are particular sentence tokens that do not just have the function of representing Venus as being a planet. They have the function of doing this by referring to Venus with ‘Hesperus’. These token specialized representers are instances of the specialized proposition that is the tool-type of tools that have the function of representing Venus as being a planet by referring to Venus with ‘Hesperus’. Once we see that Artifactualism allows for specialized propositions that are very similar to the enriched propositions of Actionism, Artifactualism’s solutions to the puzzles are straightforward. In the Hesperus/Phosphorus case, the ancients came to know the specialized proposition that is the specialized tool-type of tools that have the function of representing Venus as being identical Venus by referring to Venus with ‘Hesperus’ in one instance and referring to ‘Phosphorus’ in the other instance. For the Paderewski case, Pierre does believe the proposition that Paderewski has musical talent and that Paderewski does not have musical talent. But there is a trivially false specialized proposition that he does not believe and a priori knows to be false. This proposition is tool-type of tools that have the function of representing Paderewski as being such that he has musical talent and such that he does not have musical talent by referring to Paderewski with one and the same name in both instances. This is why we can say let ‘a’ and ‘b’ refer to Paderewski and then truly claim that Peter of course does not disbelieve that a = a. But he also is the victim of puzzle and believes that a ≠ b. Finally, for the essential indexical puzzle, Obama comes to know another specialized proposition. To make it easier to describe the nature of this specialized tool-type, we need to 129 introduce some terminology. In a attitude report that represents an agent A as, for example, believing that o is F, we will say that A is the attitude-haver and we will call o the attitude-object. The specialized proposition that Obama comes to know is the tool-type of tools that have the function of representing Obama as being the messy shopper by referring to Obama such that Obama is the attitude-object in an attitude report and the way the attitude-object is referred to guarantees that it is identical to the attitude-haver. To illustrate this idea, consider the attitude- report ‘Obama believes that he himself is the messy shopper’. The attitude-object here is referred to in a way that guarantees that it is identical to the attitude-haver. The attitude-have is Obama, and the way of referring to the attitude-object guarantees that the attitude-object is also Obama. For ‘I learned that I am the messy shopper’ the nature of ‘I’ guarantees that the attitude-haver is identical to the attitude-object. The same goes for ‘Obama learned that he is the messy shopper’ as well as way of using ‘Obama learned that Obama is the messy shopper’ where is clear that using one and the same name is supposed to make a difference. Recall the complication we addressed earlier that only Obama can be in the mental states of self-realizing, self-suspecting, and self-judging that Obama is the messy shopper. This entails that only Obama can stand in believing-that, thinking-that, and wonder-whether relations to the specialized proposition. Yet, we can still think about this specialized proposition. The short answer to how this is possible is that the relation an agent A is represented as standing in to a proposition that p in ⎡A thinks that p⎤ is a different relation from the relation represented in ⎡A thinks about the proposition that p⎤. 136 But how can it be that only Obama can stand in the former relation to the specialized proposition mentioned above? This question only seems puzzling when we forget that our practice of relating agents to propositions is just one of the folk ways humans came up with for picking out mental states. It is just a fact that these mental states have these special properties. Our practices are adapted to accommodate these special properties. This is a case where our folk-practices are quite clever. If we try to represent another agent as standing in a propositional attitude relation to the specialized proposition Obama comes to know, we will instead represent that agent as standing in a relation to a different specialized proposition. For example, we cannot say ‘Biden believes that Obama himself is the messy shopper’ and relate Biden to the specialized proposition that Obama comes to know. Our practices of picking out this specialized mental state that only Obama can be in are cleverly designed so that we can only 136 See King (2002) for more on this linguistic phenomenon. 130 represent Obama as standing in a propositional attitude relation to the specialized proposition, and ultimately, only represent Obama as being in this special state that only he can be in. 137 But we can still think about this specialized proposition about Obama. We can report our beliefs about Obama believing this proposition by saying ‘I believe that Obama believes that he himself is the messy shopper’. And if we are not using complement clauses with anaphoric pronouns to pick out the specialized proposition, but rather using a name, or definite description, or so on, we can directly talk and think about the thing that Obama learns. 138 137 With some minor fiddling we can represent the impossible scenario of someone besides Obama believing the proposition that Obama comes to know. We just say let t refer to the thought Obama comes to know and then assert that I believe the proposition = t. No one should deny that it is possible to represent this absurdity. Our task is just to explain why our practices are the way they are and how our theory of propositions can explain exactly what is going on with these practices. 138 The messy shopper case provides strong support for the suggestion we made earlier about the nature of propositional attitudes. We use a proposition to help pick out a mental state when the proposition shares important features with the mental state. We refer to the proposition that Venus is a planet in order to pick out attitudes that are towards Venus, the property of being a planet, and so on because the proposition represents, and is therefore about, these things. The specialized proposition we refer to in order to talk about the new state the ancients were in when they made their astronomical discovery has referring to Venus with ‘Hesperus’ and ‘Phosphorus’ built into its function because the state we pick out has a similar property. It is a state having to do with dispositions towards thoughts and sentences that refer to Venus with these words. And now we see for the case of the essential indexical that we build in referring with an anaphoric pronoun to the function of the specialized proposition because of the unique properties that the mental state and the specialized proposition share. The other option we covered was to expand our account of propositions so that they are not tool-types but more general functional-types. If we do this we can claim that believing that o is F is partially constituted by being in a mental state M where M is an instance of the functional-type of things that have function of representing o as being F. If we accept this, then we can give a different solution for the messy shopper case. We could that the proposition Obama comes to know is one that represents him as being the messy shopper by self-referring to Obama. Obama would need to be in a mental state M that has the function of representing Obama as being the messy shopper by referring to Obama via self- reference. 131 It is worth pointing out that Obama does not need think that Obama-believes-that-he-himself- is-the-messy-shopper when he ends up in the special state of knowing that he himself is the messy shopper. He can say to himself, ‘My goodness I am the messy shopper’. Being in the state of believing that oneself is the messy shopper does not require representing oneself as being in this state by relating oneself to a specialized proposition. But when we want to represent Obama as being in the state by using a propositional attitude verb, we do need to represent Obama as standing in a relationship to a specialized proposition. Specifically, we represent Obama as standing in a relationship to the tool-type of tools that have the function of representing Obama as being the messy shopper by referring to Obama in attitude report in a way that guarantees that the attitude-haver is identical to the attitude-object. 5. The Nature of Attitude Reports in General This section argues that Artifactualism can give a superior account of the nature of attitude reports in general. The main problem with the solutions given by both Actionism and Artifactualism is that we appear to use one and the same complement clause in an attitude report to sometimes pick out a plain proposition and other times to pick out an enriched/specialized proposition. For example, sometimes we use ‘Hesperus is not Phosphors’ in ‘John believes that Hesperus is not Phosphorus’ to pick out a plain proposition that merely represents Venus as not being identical to Venus. We do this when we are characterizing John as being truly irrational. But other times we use the complement clause to instead pick out an enriched/specialized proposition that one can end up believing when one does not realize that one is talking or thinking about one and the same thing. The dilemma we appear to face is whether the fact that we sometimes pick out the enriched/specialized propositions is to be fully explained by the semantics of attitude reports or if part of the account will be due to pragmatics features. 5.1 The Range of Actionist Options Hanks gives an account of the semantics of attitude reports where they are context-sensitive. Soames gives an account where enriched/specialized propositions are never semantically expressed by complement clauses in attitude reports. It is only through pragmatic phenomena that we manage to perform a speech act where we represent an agent as standing in a relation to an enriched/specialized proposition. I will first go over Hanks’ and Soames’ accounts respectively, noting their strengths and weakness. I will then show that although either of these options are open Artifactualists, there is a novel and attractive option open to Artifactualists that avoids the 132 problems with the limited options open to Actionists. We will only be able to sketch the general details of the alternative option exclusively open to the Artifactualist. Hanks claims that the “context-sensitivity in the word ‘believes’” explains how we can use one and the same complement clause but still give different belief reports in different contexts. (Hanks 2011: p.39) Instead of claiming that what the complement clause denotes varies from context to context, Hanks claims that the complement clause always determines a range of propositions. For ‘A believes that o is F’ the range of propositions include the enriched versions of the plain proposition that o is F. According to Hanks, the context determines which enriched propositions in this range an agent must believe in order for the sentence to be true. One puzzling thing about this account is that Hanks says that ‘believes’ is context-sensitive but then uses ‘believes’ in giving the truth-conditions of ‘A believes that o is F’. This like giving the truth conditions for ‘I am F’ by saying that it is true iff I am F. Hanks wants to deny that the relation picked out with ‘believes’ varies from context to context but rather that which propositions count as being “suitably related” to the plain proposition that o is F vary from context to context. But imagine we pick out an enriched proposition by specifying the particular act-type that it is and call that act-type α. Surely Hanks will grant that ‘A believes the proposition that is identical α’ is not context-sensitive. Perhaps Hanks holds that ‘the believing-relation’ is not context-sensitive whereas ‘believes’ is context- sensitive when it takes a complement clause. If so, he might be able to sensibly say that ‘A believes that o is F’ is true when A stands in the believing-relation to a proposition (or the propositions) suitably related to the plain proposition that o is F. Besides it being unclear what exactly Hanks’ account of the context-sensitivity of attitude- reports amounts to, there are general problems with claiming that words like ‘believes’, which we do not pre-theoretically judge to be context-sensitive, are in fact context-sensitive. ‘Believes’ for example, seems to fail all the tests we would employ for testing for context sensitivity. 139 There are many philosophers that think we must posit other ways of being context-sensitive to explain other semantic phenomena, 140 but it is still a theoretical cost to have to do this. Soames account of the semantics of attitude reports is more straightforward. The proposition semantically expressed by ⎡A φ’s that o is F⎤ is a proposition that is true iff A stands in the φ’ing- 139 See Cappelen and Lepore (2005) for an overview on such tests. 140 A notable example is MacFarlane (2011). See Bach (2003) for considerations against such views. 133 relation to the plain un-enriched proposition that o is F. In other words, the proposition semantically expressed by the complement clause of an attitude report is always an un-enriched proposition. It is only through pragmatic means that a speaker succeeds in asserting a proposition that is true iff an agent A stands in, say, the believing-relation to an enriched proposition. (Soames 2015: p.94). The benefit of Soames’ view is that we do not have to posit exotic forms of context- sensitivity or give a complicated account of propositional attitude verbs. The cost is that we have to accept that the proposition expressed by a sentence token of ‘The ancients did not always know that Hesperus is Phosphorus’ is always false. An Artifactualist is free to take either Hanks’ route or Soames’ route. One could claim that propositional-attitude-verbs are context-sensitive like Hanks does. One could, in a similar spirit, alternatively claim that the complement clause somehow manages to be context-sensitive. Or one could claim that it is only through pragmatic devices that we succeed in representing agents as standing in propositional attitudes to specialized propositions, as Soames does. But Artifactualism is unique in that it allow for a third alternative. 5.2 An Option Exclusive to Artifactualism An Artifactualist can claim that a token of a sentence ⎡A believes that S⎤ is true iff the proposition p that S is an instance of is such that A stands in the believing relation to p. This option is uniquely open to Artifactualists because sentences are instances of propositions according to Artifactualism. So, take a token of ‘The ancients knew that Hesperus is Hesperus’. The token of ‘Hesperus is Hesperus’ will itself be an instance of a proposition. Which proposition p it is an instance of, i.e. which tool-type(s) it belongs to, is determined by the sentence token’s representational function. The sentence is true iff the ancients knew the proposition p where p is what the token of ‘Hesperus is Hesperus’ is an instance of. Some sentence tokens are specialized representers. The sentences Kripke (2011: 151) calls “word salads”, like ‘Pierre believes that Londres is pretty’ are actually perfect examples of specialized representers. The token of ‘Londres is pretty’ in a token of ‘Pierre believes that Londres is pretty’ will not just have the function of representing London as being pretty. It will have the function of representing London as being pretty by referring to London with ‘Londres’. The specialized representer token of ‘Londres is pretty’ has a particular way of representing London as being pretty built into its function. We can tell that this token has a specialized representing function because one who assents to ‘Pierre believes that Londres is pretty’ will not 134 also assent to ‘Pierre believes that London is pretty’. The token of ‘Londres is pretty’ is clearly a specialized sentence token because it is being used with a specialized representing function, to pick out a specialized proposition, and ultimately pick out a mental state Pierre is in that is more specific than the general state of believing that London is pretty. Besides word salad cases, the essential indexical cases also nicely illustrate the concept of a specialized sentence token. When we say ‘Obama believes that he himself is the messy shopper’ we could use ‘that he himself is the messy shopper’ to just pick out the plain proposition that Obama is the messy shopper. If we are doing this, then asserting that Obama believes that he himself is the messy shopper would be one and the same as asserting that Obama believes that Obama is the messy shopper. However, we can make a token of ‘that he himself is the messy shopper’ such that it has a specialized representing function when it is the complement clause of a token attitude report. We can use it so it has the function of representing Obama as being the messy shopper by referring to Obama in a way that guarantees that attitude-haver is identical to the attitude-object. We create a sentence token with this special representing function, for example, when we are prepared to not assent to ‘Obama believes that he himself is the messy shopper’ but prepared to assent to ‘Barrack believes that Obama is the messy shopper’. In such a case, we clearly intend for the anaphoric pronoun to be doing something special. Notice how we often emphasize the pronoun by emphatically saying ‘he’. Although more details ultimately need be given here, the basic idea is just that our intentions bring it about that we create a token of ‘he himself is the messy shopper’ that has the special representing function of representing Obama as being the messy shopper by referring to Obama in a way that guarantees that the attitude-have is identical to the attitude-object. 141 141 For the notoriously hard Paderewski case, we can clarify what Peter believes in the following way. Yes, we assent to ‘Peter believes that Paderewski is both musical and Paderewski is not musical.’ But let us agree to give Paderewski some new nicknames, say ‘Paddington’, ‘Pitter Padder’, and ‘The Ski’. Now we can communicate what we mean when we assent to ‘Peter believes that Paderewski is both musical and Paderewski is not musical’, by also assenting to ‘Peter believes that Paderewski is musical and Paddington is not musical’, to ‘Paddington is both musical and Pitter Padder is not musical’, to ‘the Ski is musical and Paderewski is not musical’, and so on. We show that the original complement clause does not have a specialized representing function because we can swap it out with other complement clauses and mean the same thing–namely that Peter believes the unspecialized proposition that Paderewski is musical and Paderewski is not musical. But when we communicate what Peter does not believe we might say ‘Peter does not believe that Paderewski is musical and Paderewski is not musical’. We will emphasize the word ‘Paderewski’ in the same way each time and repeat some gesture each time we say it to indicate that we are using the complement clause with a special function. Or we might say ‘Let the constants “a” and “b” refer to Paderewski. Peter of course believes that a is identical to a, but is confused, so he also believes that a is not identical to b’. Again we show that the names are being used in a way that gives the complement clause a special function by indicating how the complement clause can and cannot be changed if our meaning is to 135 Now the big question is how we thread the line between semantics and pragmatics in explaining what is going on here. So far, I have told you that the sentence token can be created with a special representing function. But we also can create a token of ‘Hesperus is Phosphorus’ where the token complement clause does not have a special representing function. 142 Could we give an account that explains the special representing function of the specialized sentence token purely in terms of its semantic properties? This is not possible unless we radically rethink how we do semantics. 143 But the Artifactualist can leave it open whether the sentence token ends up getting a specialized representing function purely in virtue of its semantic properties or whether we must appeal to pragmatic features of the context to explain why the token has a specialized representing function. What is potentially confusing about this claim is that we are sometimes prone to think of the distinction between a sentence’s meaning due to it semantics properties versus what a speaker means in uttering the sentence as exhaustive. The mistake is to think that sentence tokens themselves cannot have properties due to pragmatic features. Our talk of ‘sentence-meaning’ is often ambiguous between the representational function a sentence-token would have if its representational function were purely determined by its semantic properties and the representational function a sentence token actually has. The representational function a token actually has can differ from what a speaker succeeds in representing by using that sentence token. For example, take a written token of ‘It is my favorite song’ where the writer succeeds in referring to a song with ‘it’. It is contentious whether the semantic properties alone of this token are enough to express a proposition. 144 Yet, given features of the context, the token of ‘it’ itself refers to a song. The writer does not just refer to the song by using ‘it’. The token of ‘it’ refers to stay the same. We also say he ‘of course believes’ the first proposition to help get across that we mean the trivially a priori one. Now, Kripke is absolutely right here that there is an incredible strain on our practice of using that-clauses in attitude reports. But we still can get across what is going with just our propositional attitude reports and without resorting to some other mode of description. 142 One might worry whether a given complement clause token could really lack a specialized function. Consider a token complement clause of ‘Hesperus is identical to Hesperus’. Does it just have the vanilla function of representing Venus as being identical to Venus or does it also have the more specialized function of representing Venus as being identical to Venus by referring to Venus with ‘Hesperus’ in both instances. It is undeniable that what this complement clause “does” is represent Venus as being identical to Venus by referring to Venus with ‘Hesperus’ in both instances. However, I claim that even though all token complement clauses of ‘Hespersus is identical to Hesperus’ do this, not all of them have the specialized function where the way in which the representing is brought about is built into the function of the sentence token. Arguing for this however is delicate, and needs to be taken up on another suitable occasion. 143 See Ackerman (MS) for a view of linguistic meaning that can allow for this. 144 Bach (2001) argues that the token sentence does not semantically express a proposition whereas King (2013) argues that it does. 136 the song. Only the most extreme views deny this. 145 But what the writer represents in asserting this may not be what the sentence token represents if the writer is using the token sentence sarcastically to assert that he hates that song. So if we do not deny that the representational properties of a sentence token can be determined by more that just its semantic properties, then it is not problematic to hold that token complement clauses of ‘Hesperus is Phosphorus’ can have a specialized representing function. It is understandable why we do not want to say the sentence-type we utter when we say ‘The ancients came to know that Hesperus is Phosphorus’ semantically expresses a true proposition. But if we can, we should take on a theory where tokens of this sentence can be true because tokens of the complement clause have a special representing-function. Whether we explain this feature of the token complement clauses through semantic properties alone does not matter for the intuitive data we are trying to accommodate. The intuitive data is just that tokens of this attitude- report are sometimes true. This is just like the situation with tokens of ‘It is my favorite song’. To put it another way, the following is a pre-theoretic datum our theory ought to accommodate: some tokens of ‘It is my favorite song’ are true. It is not enough to give a theory where we explain how speakers/writers sometimes make true assertions with tokens of this sentence. Similarly, it is not enough to give a theory where speakers/writers sometimes make true assertions with tokens of ‘The ancients eventually came to know that Hesperus is Phosphorus’. Some tokens of this sentence are themselves true. Artifactualism is in the unique position of being able to give such an account without positing exotic forms of context-sensitivity. 6. Concluding Remarks We have seen that Actionism allows for attractive and plausible solutions to the long-standing puzzles of belief such as Frege’s puzzles. But Actionism also has troubling consequences when it comes to what sorts of entities end up having truth-conditions. Artifactualism turns out to be a superior alternative to Actionism because it allows us to solve the puzzles in a similar way to Actionism, does not commit us to revisionary consequences about what sorts of things are true or false, and is better positioned to account the nature of propositional attitude reports. 145 Strawson (1950) is an example of an extreme view that denies this. Bach (2008) also gives arguments meant to cast doubt that in most cases singular term tokens themselves refer. 137 Bibliography Ali Khalidi, M. 2015. ‘Three Kinds of Social Kinds’. Philosophy and Phenomenological Research 90, pp. 96-112. Anscombe, G.E.M. 1981. ‘The Intentionality of a Sensation: A Grammatical Feature’. In The Collected Papers of G.E.M. Anscombe: Volume Two Metaphysics and the Philosophy of Mind. Oxford: Basil Blackwell Publisher, pp. 3-20. Armour-Garb, B. and Woodbridge, J.A. 2012. ‘The Story about Propositions’, Noûs 46(4), pp. 635-674. Audi, P. 2013. ‘How to Rule out Disjunctive Properties’, Noûs 47(40), pp. 748-766. Ackerman, G. (forthcoming). ‘Securing Singular Thought about Merely Hypothetical Entities’, Philosophical Studies. Ackerman (MS). ‘Reference, Content, and Function’. Bach, K. 1993. ‘Sometimes a Great Notion: A Critical Notice of Mark Crimmins’ Talk about Beliefs’. Mind and Language, 8: 431-441. Bach, K. 1997. ‘Do Belief Reports Report Beliefs?’, Pacific Philosophical Quarterly 78, pp. 215- 241. Bach, K. 1999. ‘The Myth of Conventional Implicature’, Linguistics and Philosophy 22: 327–66. Bach, K. 2001. ‘You Don’t Say’, Synthese 128: 15-44. Bach, K. 2003. ‘Context ex Machina’. In Szabo, Z. (ed.), Semantics versus Pragmatics, New York: Oxford University Press. Bach, K. 2008. ‘On Referring and not Referring’. In Gundel, J. and Hedberg, N. (eds.), Reference: An Interdisciplinary Perspective. New York: Oxford University Press, pp. 13-60 Bacon, A. 2013. ‘Quantificational Logic and Empty Names’, Philosophers’ Imprint 13(24), pp. 1- 21. Bacon, A. (MS). ‘Radical Anti-Disquotationalism’. Baker, L. R. 1995. Explaining Attitudes. Cambridge: Cambridge University Press. Bealer, G. 1993. ‘A Solution to Frege's Puzzle’, Philosophical Perspectives 7, pp. 17-60. Bealer, G. 1998. ‘Propositions’, Mind 107, pp. 1-32. Block, N. 1986 ‘Advertisement for a Semantics for Psychology’, Midwest Studies in Philosophy 10, pp. 615–678. Boghossian, P. 1990. ‘The Status of Content’, Philosophical Review, 99, pp. 157-184. 138 Boyd, R. 1991. ‘Realism, Anti-Foundationalism, and the Enthusiasm for Natural Kinds’, Philosophical Studies 61, pp. 127-148. Braun, D. 1998. ‘Understanding Belief Reports’, Philosophical Review 107, pp. 555-595. Brewer, B. 2007. ‘Perception and Its Objects’, Philosophical Studies 132, pp. 87-97. Brogaard, B. 2012. Transient Truths: An Essay in the Metaphysics of Proposition. New York: Oxford University Press. Bronner, B. 2015. ‘Maps and Absent Symbols’, Australasian Journal of Philosophy 93(1), pp. 43-59. Bueno, O., Menzel, C., and Zalta, E. ‘Worlds and Propositions Set Free’, Erkenntnis 79, pp. 797- 820. Burge, T. 2009. ‘Five Theses on De Re States and Attitudes’. In Almog, J. and Leonardi, P. (eds.), The Philosophy of David Kaplan. New York: Oxford University Press, pp. 246-316. Burgess, A. 2010. ‘How to Reconcile Deflationism and Non-Factualism’, Nous 44(3), pp. 433- 450. Butterfill, S. and Apperly, I. 2013. ‘How to Construct a Minimal Theory of Mind’. Mind & Language 28, pp. 606-637. Byrne, A. 2001. ‘Intentionalism Defended’, The Philosophical Review 110, pp. 199-240. Byrne, A. 2009. ‘Experience and Content’, Philosophical Quarterly 59, pp. 429-451. Campbell, J. 2002. Reference and Consciousness. Oxford: Oxford University Press. Caplan, B., Tillman, C., McLean, B., and Murray, A. 2013. ‘On Act- and Language-Based Conceptions of Propositions: Not the Optimistic Type’, Canadian Journal of Philosophy 43, pp. 575-589. Cappelan, H. and Hawthorne, J. 2009. Relativism and Monadic Truth. New York: Oxford University Press. Cappelen, H. and Lepore, E., 2005. Insensitive Semantics, Oxford: Blackwell Publishing. Carlson, G. and Pelletier, F. (eds.) 1995. The Generic Book. Chicago: University of Chicago Press. Casati, R and Varzi, A. 1999. Parts and Places: The Structures of Spatial Representation. Cambridge: MIT Press. Chalmers, D. 1996. The Conscious Mind. Oxford: Oxford University Press. Chan, T. (ed.) 2013. The Aim of Belief. New York: Oxford University Press. Church, A. 1988. ‘A Remark Concerning Quine’s Paradox about Modality’. In Salmon, N. and Soames, S. (eds.), Propositions and Attitudes, Oxford: Oxford University Press. 139 Churchland, P. 1981. ‘Eliminative Materialism and the Propositional Attitudes’, Journal of Philosophy 78, pp. 67-90. Clapp, L. 2002. ‘Davidson's Program and Interpreted Logical Forms’, Linguistics and Philosophy 25, pp. 261-97. Crimmins, M. 1992. ‘Context in the Attitudes’, Linguistics and Philosophy 15, pp. 185-198. Crimmins, M. and Perry, J. 1989. ‘The Prince and the Phone Booth: Reporting Puzzling Beliefs’, Journal of Philosophy 86, pp. 685-711. Davidson, D. 1968. ‘On Saying that’. Synthese 19.1, pp. 130-146. Dennett, D. 1981. ‘True Believers: The Intentional Strategy and Why It Works’. In Heath, A.F. (ed.), Scientific Explanation: Papers Based on Herbert Spencer Lectures Given in the University of Oxford, Oxford: Clarendon Press. Dretske, F. 1988. Explaining Behavior: Reasons in a World of Causes. Cambridge: MIT Press. Dretske, F. 1995. Naturalizing the Mind. Cambridge, MIT Press. Dretske, F. 1999. Knowledge and the Flow of Information. Stanford: CSLI Publications. Evans, G. 1982. The Varieties of References, (ed.) McDowell J. Oxford: Oxford University Press. Field, H. 1978. ‘Mental Representation’, Erkenntnis 13, pp. 9-61. Field, H. 2008. Saving Truth from Paradox. New York: Oxford University Press. Fish, W. 2009. Perception, Hallucination, and Illusion. New York: Oxford University Press. Fine, K. 1994. ‘Essence and Modality: The Second Philosophical Perspective Lecture’, Philosophical Perspectives 8, pp. 1-16. Fodor, J. 1975. The Language of Thought. New York: Cromwell. Fodor, J. 1990. A Theory of Content. Cambridge: MIT Press. Foster, J.A. 1976. ‘Meaning and Truth Theory’. In Evans, G. and McDowell, J. (eds.), Truth and Meaning: Essays in Semantics, New York: Oxford University Press. Frege, G. 1980. The Foundations of Arithmetic. Evanston: Northwestern University Press. Frege, G. 1997a. ‘Begriffsschrift’. In Beaney, M. (ed.), The Frege Reader, Oxford: Blackwell, pp. 47-78. Frege, G. 1997b. ‘On Sinn and Bedeutung’. In Beaney, M. (ed.), The Frege Reader, Oxford: Blackwell, pp. 151-172. Frege, G. 1997c. ‘Thought’. In Beaney, M. (ed.), The Frege, Oxford: Blackwell, pp. 325-346. Ganson, T., Bronner, B., and Kerr, A. 2014. ‘Burge’s Defense of Perceptual Content’, Philosophy and Phenomenological Research 88(3), pp. 556-573. Genone, J. (forthcoming). ‘Appearance and Illusion’. Mind. 140 Godfrey-Smith, P. 2006. ‘Mental Representation, Naturalism, and Teleosemantics’. In Papineau, D. and MacDonald, G. (eds.), Teleosemantics, New York: Oxford University Press. Goodman, N. 1968. Languages of Art: An Approach to a Theory of Symbols. New York: The Bobbs-Merrill Company. Halbach, V. 1997. ‘Tarskian and Kripkean Truth’, Journal of Philosophical Logic 26, pp. 69-80. Harman, G. 1987. ‘(Nonsolipsistic) Conceptual Role Semantics’. In LePore, E. (ed.), New Directions in Semantics, London: Academic. Hanks, P. 2011. ‘Structured Propositions as Types’, Mind 120, pp. 11-52. Hawthorne, J. and Lepore, E. 2011. ‘On Words’, The Journal of Philosophy 108, p. 447-485. Hawthorne, J. and Manley, D. 2012. The Reference Book. New York: Oxford University Press. Higginbotham, J. 1991. ‘Belief and logical form’. Mind and Language, 6: 344-69. Higginbotham, J. 1992. ‘Truth and Understanding’, Philosophical Studies, 65(1), pp. 3-16. Holton, R. 2000. ‘Minimalism and Truth-Value Gaps’. Philosophical Studies, 97(2), pp. 137-168. Horwich, P. 1999. Truth. New York: Oxford University Press. Jackson, F., Oppy, G., and Smith, M. 1994. ‘Minimalism and Truth-Aptness’, Mind 103, pp. 287- 302. Kaplan, D. 1990. ‘Words’. Proceeding of the Aristotelian Society, Supplementary Volumes 64, pp. 93-119. King, J. 2002. ‘Designating Propositions’, The Philosophical Review 111, pp. 341-371. King, J. 2007. The Nature and Structure of Content. New York: Oxford University Press. King, J. 2013. ‘Speaker Intentions in Context’, Noûs 47, pp. 218–237. King, J. 2014a. ‘What Role do Propositions Play in our Theories?’ In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 5-8. King, J. 2014b. ‘Naturalized Propositions’. In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 47-70. Kripke, S. 2011. ‘A Puzzle about Belief’. Philosophical Troubles, Volume I. New York: Oxford University Press, pp. 125-161. Kulvicki, J. ‘Maps, Pictures, and Predication’, Ergo 2(7), pp. 149-174. Larson, R. and Ludlow, P. 1993. ‘Interpreted Logical Forms’. Synthese 95(3), pp. 305-355. Lewis, D. 1972. ‘Psychophysical and Theoretical Identifications’, Australasian Journal of Philosophy 50, pp. 249-258. Loar, B. 1981. Mind and Meaning. Cambridge: Cambridge University Press. 141 Logue, H. 2014. ‘Experiential Content and Naïve Realism: A Reconciliation.’ In Brogaard, B. (ed.), Does Perception have Content? New York: Oxford University Press. Lycan, W. 1981. ‘Form, Function, and Feel’, The Journal of Philosophy 78(1), pp. 24-50. MacFarlane, J. 2011. ‘Epistemic Modals are Assessment-Sensitive,’ in Egan, A. and Weatherson, B. (eds.), Epistemic Modality, Oxford: Oxford University Press, pp. 144–178. Martin, M.G.F. 2002. ‘Particular Thoughts and Singular Thoughts’. In O’Hear, A. (ed.), Logic, Though, and Language. Cambridge: Cambridge University Press. Massimi, M. 2014. ‘Natural Kinds and Naturalized Kantianism’, Noûs 48(3): 416-449. Mates, B. 1952. ‘Synonymity’. In Linsky, L. (ed.), Semantics and the Philosophy of Language. University of Illinois Press. Oxford. McGlone, M. 2012. ‘Propositional Structure and Truth Conditions’, Philosophical Studies 157(2), pp. 211–225. Merricks, T. 2015. Propositions. New York: Oxford University Press. Millikan, R.G. 1987. Language, Thought, and other Biological Categories. Cambridge: MIT Press. Millikan, R.G. 1990. ‘The Myth of the Essential Indexical’, Noûs, 24(5), pp. 723-734. Nelson, M. 2005. ‘The Problem of Puzzling Pairs’, Linguistics and Philosophy 28, pp. 319-350. O’Leary-Hawthorne, J., Jackson, F., and Oppy, G. 1997. ‘Minimalism and Truth’, Nous 31(2), pp. 170-196. O’Leary-Hawthorne, J. and Price, H. 1996. ‘How to Stand Up for Non-Cognitivists’, Australasian Journal of Philosophy 74(2), pp. 275-292. Perry, J. 1979. ‘The Problem of the Essential Indexical’, Noûs 13, pp. 3-21. Piccinini, G. 2008. ‘Computation without Representation’, Philosophical Studies, 137(2), pp. 205-241. Potts, C. 2005. The Logic of Conversational Implicatures. Oxford: Oxford University Press. Price, H. 1994. ‘Semantic Deflationism and the Frege Point’. In Tsohatzidis, S. (ed.), Foundations of Speech Act Theory: Philosophical and Linguistic Perspectives, London: Routledge, pp. 132-156. Rayo, A. 2013. The Construction of Logical Space. New York: Oxford University Press. Reiland, I. 2012. ‘Propositional Attitudes and Mental Acts’, Thought 1(3), pp. 239-245. Reimer, M. 1995. ‘A Defense of de re Belief Reports’, Mind and Language 10: pp. 446-463. Rescorla, M. ‘Predication and Cartographic Representation’, Synthese 169(1), pp. 175-200. Richard, M. 1981. ‘Temporalism and Eternalism’, Philosophical Studies, 39(1): pp. 1–13. 142 Richard, M. 1983. ‘Direct Reference and Ascriptions of Belief’, Journal of Philosophical Logic, 12, pp. 425-452. Richard, M. 1990. Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them, New York: Cambridge University Press. Richard, M. 1997. ‘Propositional attitudes’. In Hale, B. and Wright, C. (eds.), A Companion to Philosophy of Language, Oxford: Blackwell, pp. 197-226. Richard, M. 2001. ‘Seeking a Centaur, Adoring Adonis: Intensional Transitives and Empty Terms’, Midwest Studies in Philosophy 25, pp. 103-127. Ridge, M. 2006. ‘Ecumenical Expressivism: Finessing Frege’, Ethics 116, pp. 302-336. Russell, B. 1903. The Principles of Mathematics. London: Norton. Salmon, N. 1986. Frege's Puzzle. Cambridge (Massachusetts): MIT Press. Salmon, N. 1998. ‘Nonexistence’, Noûs 32, 277-319. Schellenburg, S. 2011. ‘Perceptual Content Defended’, Nous 45(4), pp. 714-750. Schellenburg, S. 2014. ‘The Relational and Representational Character of Perceptual Experience’. In Brogaard, B. (ed.), Does Perception have Content, New York: Oxford University Press. Schiffer, S. 1996. ‘The Hidden-Indexical Theory's Logical-Form Problem: A Rejoinder’, Analysis 56, pp. 92-97. Schiffer, S. 2003. The Things We Mean. New York: Oxford University Press. Schiffer, S. 2006. ‘A Problem for a Direct-Reference Theory of Belief Reports’, Noûs 40, pp. 361-368. Schroeder, M. 2008. ‘What is the Frege-Geach Problem’, Philosophy Compass 3.4, pp. 703-720. Schwitzgebel, E. 2013. ‘A Dispositional Approach to Attitudes: Thinking Outside the Belief Box’. In Nottelman, N. (ed.), New Essays on Belief, New York: Palgrave Macmillan. Searle, J. 1982. Intentionality: An Essay in the Philosophy of Mind. Cambridge: Cambridge University Press. Searle, J. 1995. The Construction of Social Reality. New York: Free Press. Sider, T. 1995. ‘Three Problems for Richard's Theory of Belief Ascriptions’, Canadian Journal of Philosophy 25, pp. 487-513. Siegal, S. 2010. ‘Do Visual Experiences have Content?’ In Nanay, B. (ed.), Perceiving the World, New York: Oxford University Press. Siegal, S. 2012. The Contents of Visual Experience. New York: Oxford University Press. Soames, S. 1992. ‘Truth, Meaning, and Understanding’, Philosophical Studies 65(1), pp.17-35. 143 Soames, S. 2002. Beyond Rigidity: The Unfinished Semantic Agenda of Naming and Necessity. New York: Oxford University Press. Soames, S. 2003. ‘Understanding Deflationism’, Philosophical Perspectives 17(1), pp. 369-383. Soames, S. 2010. What is Meaning. Princeton: Princeton University Press. Soames, S. 2011. ‘True at’, Analysis 71, 124-133. Soames, S. 2014a. ‘Why the Traditional Conception of Propositions Can’t be Correct’. In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 25-46. Soames, S. 2014b. ‘Cognitive Propositions’. In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 91-120. Soames, S. 2014c. ‘Clarifying and Improving the Cognitive Theory’. In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 226-240. Soames, S. 2015. Rethinking Language, Mind and Meaning. Princeton: Princeton University Press. Speaks, J. 2006. ‘Truth-Theories, Translation Manuals, and Theories of Meaning’, Linguistics and Philosophy 29(4), pp. 487-505. Speaks, J. 2014a. ‘What’s Wrong with Semantic Theories that Make No Use of Propositions?’. In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 9-24. Speaks, J. 2014b. ‘Propositions are Properties of Everything or Nothing’. I In King, J., Soames, S., and Speaks, J. (eds.), New Thinking about Propositions, New York: Oxford University Press, pp. 71-90. Stalnaker, R. 1976. ‘Propositions’, Issues in the Philosophy of Language, pp. 79-91. Stalnaker, R. 1987. Inquiry. Cambridge: MIT Press. Stanley, J. 2002. ‘Making it Articulated’. Mind and Language 17, pp. 149-168. Stoljar, D. 1993. ‘Emotivism and Truth-Conditions’. Philosophical Studies 70, pp. 81-101. Strawson, P. F. 1950. ‘On Referring’, Mind 59, pp. 320-44. Thomasson, A.L. 2003. ‘Realism and Human Kinds’. Philosophy and Phenomenological Research 67, pp. 580-609. Travis, C. 2004. ‘Silence of the Senses’, Mind 113, pp. 57-94. Van Geenhoven, V. and McNally, L. (2005). ‘On the Property Analysis of Opaque Complements’, Lingua 115, pp. 885-914. Vanderveken, D. 2009. Meaning and Speech Acts. Cambridge: Cambridge University Press. 144 Wetzel, L. 2009. Types and Tokens. Cambridge: MIT Press. Wright, C. 1992. Truth and Objectivity. Cambridge.: Harvard University Press.

Abstract (if available)

Abstract This dissertation is in an investigation into the nature of propositions and propositional attitudes. It is inspired by the growing optimism that returning to a serious investigation into the nature of propositions is a promising way to make progress on long standing problems in the philosophy of language and the philosophy of mind. The view of propositions I offer here, 'Artifactualism', is the view that propositions are kinds of artifacts. Propositions are kinds. Their instances are particular representational artifacts. The proposition that snow is white is the kind of artifact that has the function of representing snow as being white and has no other functions involving representing things as being ways. Instances of this kind include tokens of the English sentence ‘Snow is white’, tokens of the Spanish sentence ‘La nieve es blanca’, and many other particular linguistic artifacts that have the function, i.e. serve the purpose, of representing snow as being white. In general, propositions are representational kinds/types whose instances all share a common representing function.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Telling each other what to do: on imperative language

PDF

Instantial terms, donkey anaphora, and individual concepts

PDF

Meaningfulness, rules, and use-conditional semantics

PDF

Comparative iIlusions at the syntax-semantics interface

PDF

Conceptually permissive attitudes

PDF

Reasons, obligations, and the structure of good reasoning

PDF

Positivist realism

PDF

Unlimited ontology and its limits

PDF

Responsibility and the emotional structure of relationships

PDF

Constraining assertion: an account of context-sensitivity

PDF

Reasoning with uncertainty and epistemic modals

PDF

The constitution of action

PDF

The metaphysics of social groups: structure, location, and change

PDF

The neural representation of faces

PDF

Vagueness, ontology, and the social world

PDF

Mistaken defense and normative conventions

PDF

Positivity and generalized Teichmüller theory: dynamical and algebraic aspects

PDF

Subjectivity, commitments and degrees: on Mandarin hen

PDF

Endocrinology of social behavior and reward

PDF

Process-oriented rationality

Asset Metadata

Creator Ackerman, Gregory (author)

Core Title Representation, truth, and the metaphysics of propositions

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Philosophy

Publication Date 04/21/2016

Defense Date 03/11/2016

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,propositional attitudes,propositions,representation,Truth,truth-aptitude

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Soames, Scott (committee chair), Jeshion, Robin (committee member), Schein, Barry (committee member), Uzquiano Cruz, Gabriel (committee member)

Creator Email ackermag@gmail.com,gsackerm@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-238917

Unique identifier UC11279348

Identifier etd-AckermanGr-4340.pdf (filename),usctheses-c40-238917 (legacy record id)

Legacy Identifier etd-AckermanGr-4340.pdf

Dmrecord 238917

Document Type Dissertation

Format application/pdf (imt)

Rights Ackerman, Gregory

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA