Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Constraining assertion: an account of context-sensitivity
(USC Thesis Other)
Constraining assertion: an account of context-sensitivity
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
CONSTRAINING ASSER TION:
AN A CCOUNT OF CONTEXT -SENSITIVIT Y
b y
E duar do V illanuev a Chigne
A D isser tation P r esented to the
F A CUL T Y OF THE USC GRADU A TE SCHOOL
UNIVERSIT Y OF SOUTHERN CALIFORNIA
I n P ar tial F ulllment of the
R equir ements for the D egr ee
DOCT OR OF P HIL OSOP HY
(P HIL OSOP HY )
M ay Copyright E duar do V illanuev a Chigne
A ckno wledgements
I would like to expr ess my gratitude to the people whose wor k and/or help hav e inuenced the
contents of this disser tation. F irst and for emost, I would like to thank my advisor , Scott S oames,
who has been an amazing mentor during my time at USC. Scott ’ s ex ceptional philosophical
acuity and meticulously as w ell as his deep understanding of philosophy of language hav e been
wonder ful sour ces of kno wledge and inspiration since my rst day at USC. As any student of
his wor k will notice, his inuence in this disser tation is ev er ywher e. I was also lucky to hav e
M ar k Schr oeder in my committee. I hav e benetted gr eatly fr om his ex cellent and always timely
feedback along with his constant encouragement.
I was v er y for tunate to spend the S pring semester of at the D epar tment of Linguistic
and P hilosophy of MIT . I would like to thank Agustín Ray o for making it possible. D iscussions
with him about some sections of this disser tation as w ell as his o wn wor k helped me a gr eat deal
in clarifying my o wn thoughts. D uring my time at MIT , I also benetted fr om some seminars
in Linguistics, especially fr om Kai v on F intel ’ s seminar on intensional semantics. I would also
like to thank D ilip N inan for inviting me to spend the enormously philosophically stimulating
S ummer of at A r ché , the P hilosophical R esear ch Centr e for Logic, Language, M etaphysics,
ii
and E pistemology of the U niv ersity of S t Andr e ws. A substantial por tion of chapter was
written during that time.
I hav e also benetted fr om comments fr om K enny Easwaran, B arr y Schein, E kain G ar-
mendia, J oanna B locho wiak, and audiences at the F irst Annual M eeting of the Latin-American
Association for Analytic P hilosophy , the USC P hilosophy Colloquium, the M ind and
Language G r oup at the P onticia U niv ersidad Católica del P er ú, and the V W or kshop on Cog-
nitiv e P erspectiv es on M ind and Language at the I nstitute for P hilosophical R esear ch of the
U niv ersidad N acional A utónoma de M éxico . P or tions of this disser tation w er e written during
the tenur e of a USC G raduate School D isser tation Completion F ello wship , for which I am v er y
grateful.
I o w e a gr eat debt to my family . anks to my par ents, O swaldo and R osa E lvira, and my
sister , Claudia, for their enthusiastic suppor t and much-needed encouragement thr oughout all
these y ears. F inally , thanks to S andra, my wife, whose daily companionship and lo ving suppor t
hav e made this disser tation, and many mor e things, possible.
iii
T able of Contents
A ckno wledgements ii
A bstract vii
S emantic M inimalism . I ntr oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e Argument b y E limination . . . . . . . . . . . . . . . . . . . . . . . . . . e Argument fr om E xplanator y F or ce . . . . . . . . . . . . . . . . . . . . . T esting for Context-S ensitivity . . . . . . . . . . . . . . . . . . . . . . . . . .. T est : B locking I nter-Contextual D isquotational I ndir ect R epor ts . . .. T est : B locking Collectiv e D escriptions . . . . . . . . . . . . . . . .. T est a: P assing an I nter-Contextual D isquotational T est . . . . . . . .. T est b: Allo wing for R eal Context S hifting Arguments . . . . . . . . S emantic E v aluation under Conditions of I gnorance . . . . . . . . . . . . . . .. Asser ting M ultiple P r opositions . . . . . . . . . . . . . . . . . . . . .. N on-Asser ted S emantic Contents . . . . . . . . . . . . . . . . . . . .. Ambiguity And V agueness . . . . . . . . . . . . . . . . . . . . . . . . Context-S ensitivity vs. I ndexicality . . . . . . . . . . . . . . . . . . . . . . P r enominal P ossessiv es . r ee eses A bout S emantic Content . . . . . . . . . . . . . . . . . . . . . . P ossessiv e R elations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e Argument fr om M assiv e Ambiguity . . . . . . . . . . . . . . . . . . . . . e Argument fr om E xistential G eneralization . . . . . . . . . . . . . . . . . e Argument fr om Co v er t I ndexicality . . . . . . . . . . . . . . . . . . . . . T wo Conceptions of M eaning and Asser tion . . . . . . . . . . . . . . . . . . . S emantic G uidance B asics . . . . . . . . . . . . . . . . . . . . . . . . . . . .. What Ar e Asser tion Checklists? . . . . . . . . . . . . . . . . . . . . .. Contexts v ersus Communicativ e S ituations . . . . . . . . . . . . . . .. G lobal Constraints v ersus Local Con straints . . . . . . . . . . . . . . . A ccounting for P ossessiv e R elations . . . . . . . . . . . . . . . . . . . . . . .. I nher ent and E xtrinsic R elations . . . . . . . . . . . . . . . . . . . . iv
.. e U niqueness Constraint . . . . . . . . . . . . . . . . . . . . . . .. T ying U p S ome Loose E nds . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compound N ominals . I ntr oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compound N ominals: S ome F eatur es . . . . . . . . . . . . . . . . . . . . . .. D ir ectionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. E ndocentric, E x ocentric, and Copulativ e Co mpounds . . . . . . . . . e Argument fr om Ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . e Argument fr om E xistential G eneralization . . . . . . . . . . . . . . . . . e Argument F r om Co v er t I ndexicality . . . . . . . . . . . . . . . . . . . . . e S emantic G uidance A ppr oach . . . . . . . . . . . . . . . . . . . . . . . .. G lobal Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . .. Local Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . .. F iltering and O r dering . . . . . . . . . . . . . . . . . . . . . . . . .. U nspecic R elations . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I ndexicals . I ntr oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S emantic G uidance & S emantic D eterminism . . . . . . . . . . . . . . . . . . e S tandar d S emantic D eterministic P ictur e . . . . . . . . . . . . . . . . . .. T wo G oals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. I ndexicals & I ntensional O perators . . . . . . . . . . . . . . . . . . .. D ouble I ndexing . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Contexts as I ndices . . . . . . . . . . . . . . . . . . . . . . . . . . .. Logical T r uth & Logical Consequence for LD . . . . . . . . . . . . . .. Character: D eterminacy & Kno wability . . . . . . . . . . . . . . . . . e M ultiple O ccurr ences P r oblem . . . . . . . . . . . . . . . . . . . . . . .. D emonstrativ es & Dthat . . . . . . . . . . . . . . . . . . . . . . . .. r ee U nsatisfactor y S olutions to the M ultiple O ccurr ences P r oblem . . e S emantic G uidance P ictur e . . . . . . . . . . . . . . . . . . . . . . . . .. What ’ s O ur G oal? . . . . . . . . . . . . . . . . . . . . . . . . . . . .. e M eanings of I ndexicals . . . . . . . . . . . . . . . . . . . . . . .. R evisiting the M ultiple O ccurr ences P r oblem . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e S emantics/P ragmatics D istinction . I ntr oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S emantics vs. P ragmatics: r ee V ie ws . . . . . . . . . . . . . . . . . . . . . e P r oblem of U nder determination . . . . . . . . . . . . . . . . . . . . . . . I nterlude: T wo P otentially P r oblematic N otions . . . . . . . . . . . . . . . . v
.. ought . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. What is S aid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . r ee W ays of D ealing with U nder determination . . . . . . . . . . . . . . . .. S emanticism on U nder determination . . . . . . . . . . . . . . . . . .. P ragmaticism on U nder determination . . . . . . . . . . . . . . . . . .. Cooperativism on U nder determination . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B ibliography vi
A bstract
W e utter declarativ e sentences to asser t pr opositions. M any of the sentences w e utter for this
purpose ar e context-sensitiv e: the pr opositions they ar e literally used to asser t can change fr om
one context of utterance to another . F or example, if I tell y ou ‘I am talking to y ou ’, I asser t a
pr oposition that is differ ent fr om the pr oposition y ou would asser t if y ou utter ed it. H o w ev er ,
the r elation betw een these pr opositions and the linguistic meaning of the sentence is the same:
the linguistic meaning of ‘I am talking to y ou ’ fully determines, for ev er y context, what is
asser ted b y a literal utterance of it. Cases like this hav e led many philosophers and linguists to
believ e that ifS is an unambiguous, context-sensitiv e, declarativ e sentence andp is a pr oposition
asser ted (without conv ersational implicatur es) b y a literal utterance ofS in a contextc , thenp is
fully determined b y the linguistic meaning ofS inc . I call this belief ‘S emantic D eterminism ’.
is disser tation has two main goals. e rst is to challenge S emantic D eterminism and
the second is to pr opose an alternativ e theor y of the r elationship betw een the linguistic mean-
ings of context-sensitiv e sentences and the contents they ar e literally used to asser t. As for the
rst goal, I sho w that semantic theories committed to S emantic D eterminism hav e r eal dif-
culties accounting for or dinar y uses of sentences containing linguistic constr uctions such as
vii
pr enominal possessiv es, compound nominals, and indexicals. As for the second goal, I pr opose
an alternativ e vie w I call ‘S emantic G uidance ’. A ccor ding to S emantic G uidance, the linguistic
meaning of an unambiguous, context-sensitiv e, declarativ e sentenceS mer ely pr o vides speakers
with an asser tion checklist , i.e., a non-exhaustiv e set of conditions that ev er y literal use ofS must
satisfy in or der to r esult in an asser tion. Asser tion checklists play two fundamental r oles: they
help speakers to ensur e their literal uses ofS r esult in asser tions of the pr opositions they intend
to communicate, and (ii) they guide hear ers to war ds the asser ted contents intended b y speakers
without xing them. I sho w ho w this account not only can explain the data semantic theories
committed to S emantic D eterminism explain best, but also is able to handle a wider range of
data.
e disser tation pr oceeds as follo ws. I n the rst chapter , I examine a r ecent theor y called ‘S e-
mantic M inimalism ’, which holds that the meanings of unambiguous indexical-fr ee declarativ e
sentences ar e always full-edged pr opositions. I t fur ther maintains that since those pr opositions
can be grasped and r epor ted acr oss differ ent contexts, those sentences ar e context- in sensitiv e.
e fe w context-sensitiv e expr essions in natural language ar e the Kaplanian indexicals. I argue
against this vie w and sho w the differ ent ways in which indexical-fr ee sentences can be context-
sensitiv e. O nce this is seen, the S emantic D eterminism pr esupposed b y S emantic M inimalism
falls b y the wayside.
e next two chapters ar e dev oted to detailed analyz es of two linguistic constr uctions that
pose serious difficulties to any theor y committed to S emantic D eterminism: pr enominal pos-
sessiv es and compound nominals. P r enominal possessiv es ar e linguistic constr uctions such as
viii
‘J ohn ’ s book ’ and ‘J ohn ’ s mother ’ that contain the morpheme ’ s . Compound nominals ar e se-
quences of two or mor e nouns that function as a single noun, e.g., ‘ steel knife ’ and ‘ pie char t ’.
S e mantic D eterminists ar e baffled b y these constr uctions because competent speakers normally
use them to asser t a div ersity of r elations betw een their constituent par ts that ar e no wher e to
be found in the expr essions themselv es. F or example, although ‘ steel knife ’ and ‘kitchen knife ’
exhibit the same grammatical str uctur e, the r elations betw een knife and steel (e.g., knife made
of steel), and knife and kitchen (e.g., knife used in the kitchen) ar e completely differ ent.
M or eo v er , typically differ ent literal uses of a single constr uction can conv ey differ ent r ela-
tions depending on the context of use. F or instance, the possessiv e phrase ‘J ohn ’ s pictur e ’ can
be used to r efer to a pictur e por traying J ohn, a pictur e taken b y J ohn, a pictur e o wned b y J ohn,
etc. I argue that theorists hav e found these expr essions deeply puzzling, because they hav e been
assuming S emantic D eterminism. I pr opose that the linguistic meanings of sentences contain-
ing them just giv e us asser tion checklists that mer ely constrain the range of pr opositions that one
can asser t b y literal uses of those sentences. I sho w ho w , as soon as w e adopt this vie w , the
puzzle disappears.
e four th chapter extends the S emantic G uidance frame wor k to the home territor y of
S emantic D eterminism: indexicals. I challenge the traditional tr eatment of indexicals as w ell as
some inuential alternativ e accounts in contemporar y linguistics and philosophy b y pr esenting
some data that these theories fail to account. I argue that in or der to account for these data,
w e should tr eat indexical sentences as semantically under determined, i.e., as expr essions whose
linguistic meanings determine, for ev er y communicativ e situation, sets of pr opositions that one
ix
is allo w ed to asser t b y uttering them. is account leav es the job of identifying the intended
asser ted contents in those sets to conv ersational par ticipants, who would carr y out this task
guided b y conv ersational maxims and principles, their backgr ound pr esuppositions, and the
purposes of their par ticular communicativ e situations. I sho w ho w , b y adopting this vie w , the
pr oblematic data can be explained away .
e last chapter attempts to situate S emantic G uidance in the semantics-vs-pragmatic de-
bates. T o do so, I characteriz e thr ee positions one can take about the division of labor betw een
semantics and pragmatics in the explanation of the information a giv en linguistic expr ession
can be literally used to asser t: S emanticism, P ragmaticism, and Cooperativism. M y charac-
terization of them focuses on thr ee main issues: (i) their conception of the r elation betw een
semantic content and asser ted content, (ii) the r ole they assign to context in the determination
of asser ted content, and (iii) the bur den they impose on semantics and pragmatics in the ex-
planation of asser ted content. I n addition, I explain the challenge posed b y the phenomenon
of under determination to the traditional division of labor betw een semantics and pragmatics,
and why being able to handle it matters. F inally , I argue that the division of labor pr esupposed
b y S emantic G uidance (Cooperativism) has the r esour ces to handle under determination better
than the alternativ es consider ed.
x
Chapter S emantic M inimalism
. I ntr oduction
H o w per v asiv e is context-sensitivity in natural language? M ost contemporar y philosophers
tell us that it ’ s much mor e common than our philosophical ancestors thought. Cappelen and
Lepor e disagr ee. ¹ I n their book I nsensitiv e S emantics ( ), they claim that, with the ex ception
of a v er y small set of expr essions, natural language expr essions ar en ’ t context-sensitiv e. eir
theor y is called ‘S emantic M inimalism ’. S emantic M inimalism is the vie w that the semantic
content of a declarativ e sentenceS is the complete pr oposition thatS expr esses in ev er y context
in whichS can be asser tiv ely utter ed. is pr oposition is just minimally determined b y context: ²
once ev er y expr ession inS has been disambiguated and the r efer ents of the ob vious indexicals
¹ A t least they used to disagr ee back in .
² F or the purposes of this chapter , b y ‘ context ’ I will mean ‘ extra-linguistic context ’ unless I explicitly say oth-
er wise.
(the so-called ‘B asic S et ’ ³ ) that S contains ar e contextually x ed, the pr oposition that r esults
is contextually in sensitiv e. Cappelen and Lepor e hold that the context- in sensitivity of most
sentences in natural language is what enables us to grasp and r epor t what they say ev en when
w e ignor e their contexts of utterance.
e purpose of this chapter is to challenge S emantic M inimalism. I will argue that (i) the
main argumentativ e strategies Cappelen and Lepor e use in fav or of it (i.e., an argument b y elim-
ination and an argument fr om explanator y for ce ) fail, (ii) it ’ s based on the unjustied assumption
that ultimately all context-sensitivity in natural language is Kaplanian indexicality , and (iii) its
main methodological tools to determine context-sensitivity ar e aw ed. M y discussion of S e-
mantic M inimalism is divided in v e par ts (sections . – . ). I n section , I briey argue that
the argument b y elimination is unsound. I n sections . and . , I analyz e the argument fr om
explanator y for ce; in par ticular , I discuss Cappelen ’ s and Lepor e ’ s Context S ensitivity T ests.
I attempt to sho w that these tests ar en ’ t effectiv e tools for detecting context-sensitivity and,
thus, since Cappelen ’ s and Lepor e ’ s critique of some Contextualist accounts (e.g. contextual-
ism about ‘kno ws ’, v ague pr edicates, and gradable adjectiv es) is based on these tests, it fails.
S e ction . is dev oted to S emantic M inimalism ’ s methodological ignor ance , i.e., the pr ocedur e
accor ding to which in or der to get at the semantic content of a declarativ e sentenceS one should
ev aluate S under conditions of ignorance about the context in which an utterance of S took
place. Against this method, I raise thr ee pr oblems: (a) the pr oblem of sentences utterances of
³ e B asic S et include the follo wing expr essions: the personal pr onouns ‘I’, ‘ y ou ’, ‘he ’, ‘ she ’, ‘it ’ in their v arious
cases and number , the demonstrativ e pr onouns ‘ that ’ and ‘ this ’ in their v arious cases and number , the adv erbs ‘her e ’,
‘ ther e ’, ‘ no w ’, ‘ today ’, ‘ y ester day ’, ‘ tomorr o w ’, ‘ ago ’, ‘hence(for th)’, and the adjectiv es ‘ actual ’ and ‘ pr esent. ’ W or ds
and aspects of wor ds that indicate tense ar e also context sensitiv e.
which asser t multiple pr opositions in ev er y context, (b) the pr oblem of sentences utterances of
which don ’ t always asser t their semantic contents, and (c) the pr oblem of ambiguity and v ague-
ness. F inally , the chapter ends up with some speculativ e r emar ks about the differ ent kinds of
semantic context sensitivity that S emantic M inimalism fails to consider , but that a theor y of
context sensitivity ought to explor e.
. e Argument b y E limination
H o w do Cappelen and Lepor e (hencefor th ‘C&L ’) establish their radical r esults? e main
argument of I nsensitiv e S emantics ( ) might be r econstr ucted as follo ws:
(P
1
) er e ar e only thr ee vie ws one can hold about context-sensitivity: (i) Radical
Contextualism, (ii) M oderate Contextualism, and (iii) S emantic M inimalism.
(P
2
) M oderate Contextualism leads inex orably to Radical Contextualism.
(P
3
) Radical Contextualism is false.
(C
1
) M oderate Contextualism is false.
(C
2
) S emantic M inimalism must be tr ue.
I think that both (P
1
) and (P
2
) ar e false. S o, let ’ s star t with (P
1
) . Why do C&L think that it
is tr ue? e r easoning seems to be this: Let ’ s call the set of all and only the genuinely context-
sensitiv e expr essions in E nglish the ‘B asic S et ’. Let ’ s call the set of all and only the Kaplanian
indexicals ‘K ’. Assuming that all the members ofK ar e genuinely context-sensitiv e, ther e ar e
only thr ee positions one can take: (a) e B asic S et is the set of ev er y linguistic expr ession in
E nglish, (b)K is a pr oper subset of the B asic S et, and (c) e B asic S et isK . G iv en that (i)
Radical Contextualism defends (a), M oderate Contextualism defends (b), S emantic M inimal-
ism defends (c), and (ii) that (a), (b), and (c) exhaust all the possibilities av ailable in this r espect,
(P
1
) is tr ue.
e pr oblem with this line of r easoning is that the semantic theories in question don ’ t limit
themselv es to defending (a), (b), and (c) or their logical consequences. I n par ticular , S emantic
M inimalism (hencefor th ‘SM’) doesn ’ t. Among the central theses that SM holds ar e these:
(SM
1
) e semantic content of a declarativ e sentence S is a complete pr oposition , i.e.
a pr oposition whose tr uth-v alue and tr uth-conditions ar e speciable indepen-
dently of any context of use.
(SM
2
) O ne of the pr opositions conv ey ed b y a genuine asser tiv e utterance of S is the
semantic content ofS .
(SM
3
) Any coher ent account of linguistic communication must pr esuppose (SM
1
) and
(SM
2
). ⁴
U nless (SM
1
)–(SM
3
) ar e deriv able fr om (c) alone, w e ha v e good r eason to think that (P
1
) is
false. And it ’ s clear that these claims ar en ’ t deriv able fr om (c) alone. I n fact, one could coher-
ently deny each of these claims and, at the same time, defend (c). F or example, take (SM
1
).
A ccor ding to (SM
1
), the semantic content of any giv en declarativ e sentence S is a complete
pr oposition. N onetheless this doesn ’ t follo w fr om (c). F or one could coher ently hold (c) and
⁴ S ee Cappelen & Lepor e ( , –).
simultaneously claim that some sentences expr ess incomplete pr opositions , i.e., str uctur ed enti-
ties whose tr uth-v alues and tr uth-conditions can ’ t be specied independently of ev er y context
of use. ⁵ F or example, one could hold that ( ) semantically expr esses such an incomplete pr opo-
sition,
() T ipper is r eady .
is doesn ’ t entail—contrar y to what C&L believ e—the claim that ( ) is semantically context-
sensitiv e. A ccor ding to the vie w I am considering, the semantic content of ( ) is context-
insensitiv e, i.e. ( ) semantically expr esses the same incomplete pr oposition in ev er y context
of utterance. is incomplete pr oposition would only be pr agmatically completed in a giv en
context, i.e. ( ) may be used to asser t differ ent complete pr opositions depending on the con-
text in which it ’ s utter ed. C&L don ’ t accept this as a coher ent alternativ e. ey think that
only a metaphysically confused philosopher could think of semantic incompleteness as a viable
position. I disagr ee. I will addr ess this worr y later on.
I n addition, I think that(P
2
) is false. What r easons do C&L hav e to hold(P
2
) ? A ccor ding
to them, ther e ar e only two kinds of argument in fav or of M oderate Contextualism: (i) Con-
text S hifting Arguments, and (ii) I ncompleteness Arguments. H o w ev er , C&L claim that these
arguments ar e unstable: they inex orably lead to Radical Contextualism. I n other wor ds, con-
sistent M oderate Contextualists should become Radical Contextualists. I will focus on C&L ’ s
critique of I ncompleteness Arguments. ⁶
⁵ I n fact, this vie w is defended in B ach ( ).
⁶ I will discuss C&L ’ s critique of Context S hifting Arguments in section . .
S o, let ’ s r econsider sentence ( ). A ccor ding to C&L, the M oderate Contextualist r easons
as follo ws: she asks ‘ What is T ipper r eady for?’. S ince she feels that w e can ’ t offer a satisfactor y
answ er to this question without appealing to the context of utterance, she thinks that what ( )
says independently of any context can ’ t be a complete pr oposition. is intuition is r einfor ced
b y the intuition that w e seem unable to determine the tr uth-v alue and tr uth conditions of ( )
independently of any context of use. S ince the M oderate Contextualist kno ws that complete
pr opositions ar e the sor t of entity that can hav e both tr uth-v alues and tr uth-conditions, she
concludes that ( ) doesn ’ t expr ess such a pr oposition. S o, she conjectur es that ( ) expr esses a
sub-pr opositional content, i.e. an incomplete logical form or a semantic skeleton or a semantic
scaffolding or a semantic template or a pr opositional scheme, etc. C&L ’ s objection to this line
of r easoning is basically this. Consider sentence ( ),
() T ipper is r eady for a math exam.
S i nce ( ) pr o vides an answ er to the question ‘ What is T ipper r eady for?’, the M oderate Con-
textualist should conclude that ( ) is semantically complete. B ut, why doesn ’ t the M oderate
Contextualist keep on asking mor e questions about ( )? After all—C&L claim—ther e ar e in-
denitely many ways to be r eady for an exam. S o, C&L wonder: what exactly distinguishes
( ) fr om its suggested completion ( )? F or , why should w e suppose that ( ) is semantically
complete, while ( ) isn ’ t? What is the principled differ ence? eir answ er: ‘N one, as far as w e
can tell. ’ H ence, they conclude that if the M oderate Contextualist holds that ( ) is semanti-
cally incomplete, she should also consider ( ) to be semantically incomplete. S ince C&L think
that this objection generaliz es, they conclude that a consistent M oderate Contextualist should
embrace Radical Contextualism.
I think that C&L ’ s objection is aw ed. ey assume that the question ‘ What is T ipper
r eady for?’ is as r elev ant as any other question about the way T ipper is r eady when it comes to
determining the semantic contents of ( ) and ( ). ey claim that since ther e isn ’ t a principled
r e ason for pr eferring one o v er the others, the M oderate Contextualist ’ s choice is arbitrar y and,
thus, her position is unstable (i.e., it leads inex orably to Radical Contextualism). I n r eply , I
will claim that ther e is a principled r eason for pr eferring the M oderate Contextualist ’ s question
o v er the others and, hence, that C&L ’ s assumption is false. M y argument is r oughly this:
(P
1
) F or ev er y pr edicateP and termt , if the atomic sentence⌜P(t
1
...t
n
)⌝ expr esses
a complete pr oposition, then (i) each one of the n terms t r efers t o an object o ,
and (ii) P is ann -ar y pr edicate that r efers to ann -ar y pr oper ty .
(P
2
) I f the sentence ‘ T ipper is r eady ’ expr esses a complete pr oposition, then (i) ‘ T ipper ’
is a term that r efers to T ipper , and (ii) ‘is r eady ’ is a monadic pr edicate that r efers
to the monadic pr oper ty of being r eady .
(P
3
) I t isn ’ t the case that ‘is r eady ’ is a monadic pr edicate or that the pr oper ty of being
r eady is monadic.
(C) I t isn ’ t the case that the sentence ‘ T ipper is r eady ’ expr esses a complete pr oposition.
C&L would hav e to concede that (P) is tr ue, since this is what they assume in one of their
objections to M oderate Contextualism:
ink about what people who ar e r eady hav e in common. T o make this vivid, imagine
A ’ s being r eady to commit a bank r obber y , B’ s being r eady to eat dinner , and C’ s being
r e ady to take an exam. […] S uppose y ou ’ r e a contextualist about ‘r eady . ’ Y ou cer tainly
could agr ee with us that they all ar e r eady for their r espectiv e tasks . er e ’ s nothing in
contextualism as such that blocks this question fr om arising. I f so, the contextualist
too would hav e to say something about what exactly it is to stand in that r elation (the
r elation of r eadiness) to a task or pr oject . is is not an issue that magically ev aporates
at the moment y ou become a contextualist about ‘r eady . ’ As far as w e can tell, the
metaphysical and the semantic issues ar e or thogonal to one another . ( Cappelen &
Lepor e ( , ), my emphasis)
Although C&L draw the wr ong lesson fr om their discussion (i.e. that sentences of the form ‘A
is r eady ’ ar e semantically complete), it r ev eals that they don ’ t take the pr oper ty of being r eady
as a monadic pr oper ty or the pr edicate ‘is r eady ’ as a monadic pr edicate. O n the other hand,
since(P
2
) is simply an instance of(P
1
) , if(P
1
) is tr ue, then(P
2
) is also tr ue. Although I don ’ t
hav e a good argument in fav or of(P
1
) , I think it ’ s intuitiv ely tr ue. I n any ev ent, nothing C&L
say goes against (P
1
) , especially since they don ’ t seem to be concerned about empty names or
genuine pr edicates that fail to r efer to a pr oper ty (they might think that these ar e pr oblems for
the metaphysician).
I n sum, a r eason one could offer in fav or of the claim that ( ) is semantically incomplete
and ( ) isn ’ t is that the pr edicate ‘is r eady ’ in both sentences is dyadic and, thus, it ’ s satisable
b y pairs of objects—or better , b y pairs of individuals and ev ents. S ince in ( ) both argument
places ar e saturated and in ( ) only one of them is so, w e can conclude that the former is seman-
tically complete and the latter isn ’ t. er efor e, while the question ‘ What is T ipper r eady for?’
is r elev ant to the determination of whether the semantic content of ( ) is complete, questions
about the way T ipper is r eady ar en ’ t. N otice that the r easons giv en in fav or of the claim that ( )
is semantically incomplete don ’ t commit us to M etaphysical N ihilism (contrar y to what C&L
claim). e pr oblem in question isn ’ t whether being r eady is a pr oper ty or not. e pr oblem
is whether it ’ s monadic or not. S ince the pr edicate ‘is r eady ’ isn ’ t monadic, w e hav e good r ea-
sons to deny that the pr oper ty it r efers to is monadic. I f this is on the right track, then w e can
conclude that the argument b y elimination is unsound.
. e Argument fr om E xplanator y F or ce
C&L claim that SM is able to explain the data that ev er y semantic theor y should explain, but
fails to . I n par ticular , they claim that SM has the follo wing vir tues:
I. SM claims that only expr essions that pass the follo wing tests ar e semantically
context-sensitiv e: (i) the I nter-Contextual D isquotational I ndir ect R epor t T est,
(ii) the Collectiv e D escriptions T est, and (iii) the ICD/R CSA T est. C&L claim
that no semantic theor y should classify as context-sensitiv e any expr ession that
doesn ’ t pass these tests.
II. SM, and no other vie w , can account for ho w I nter-Contextual D isquotational I n-
dir ect R epor ts can be tr ue wher e the r epor ter and the r epor tee nd themselv es in
radically differ ent contexts. A ccor ding to SM, in such cases the r epor ted content
is the semantic content. eories that incr ease the magnitude of contextual inu-
ence on content can ’ t explain the ease with which w e make such I nter-Contextual
D isquotational I ndir ect R epor ts.
III. SM, and no other vie w , can account for ho w the same content can be expr essed,
claimed, asser ted, questioned, inv estigated, etc. in radically differ ent contexts.
A ccor ding to SM, it is the context-insensitiv e semantic content that enables au-
diences who nd themselv es in radically differ ent contexts to understand each
other , to agr ee or disagr ee, to question and debate with each other . ⁷
What ar e these tests? Why do C&L think they ar e so impor tant? I n the next section, I will
examine C&L ’ s tests.
. T esting for Context-S ensitivity
O ne of C&L ’ s central objections to both Radical Contextualism (R C) and M oderate Contex-
tualism (MC) is that these theories ar e empirically inadequate. ey claim that one of the most
common strategies used b y the Contextualist to convince us that a giv en linguistic expr essiona
is context-sensitiv e is to build stories in which a declarativ e sentenceS(a) is asser tiv ely utter ed
and ask us to tap into our intuitions about those utterances. P r esumably , sinceS(a) would be
tr uly utter ed in some stories but not in others, the Contextualist takes this to sho w that a is
context-sensitiv e.
Although C&L agr ee with the Contextualist that in or der to determine whether or not a
is context-sensitiv e one should appeal to competent speaker ’ s intuitions about differ ent uses of
a , they think that the Contextualist ’ s methodology is fundamentally aw ed. ey write,
⁷ S ee Cappelen & Lepor e ( , –).
e Radical Contextualist ’ s seduction wor ks only on someone whose focus is on an as-
tonishingly limited range of communicativ e acts. As soon as one tries to accommodate
a wider range of data, R C r uns into insurmountable empirical obstacles. ( Cappelen &
Lepor e , )
S i nce—accor ding to C&L—MC collapses into R C, it follo ws that MC also r uns into insur-
mountable empirical obstacles. I n shor t, C&L criticism might be put as follo ws:
(P
1
) A linguistic expr essiona is context-sensitiv e only if competent speakers hav e cer-
tain intuitions i about uses of cer tain sor ts of sentences containing a .
(P
2
) I f a semantic theor yT claims that a
′
is context-sensitiv e but competent speak-
ers don ’ t hav e intuitions i about sentences containing a
′
, then T is empirically
inadequate.
(P
3
) Both R C and MC claim that ther e is a set of context-sensitiv e expr essions about
which competent speakers don ’ t hav e intuitions i .
(C) Both R C and MC ar e empirically inadequate.
B u t, what ar e those intuitions i and ho w do w e kno w they ar e r elev ant to the determination
of context-sensitivity? C&L ’ s pr oposal is to design some tests such that (i) they succeed in
isolating that which a competent speaker can grasp if her kno wledge of the context of utterance
is minimal or inexistent, and (ii) they can help us to compar e the behavior of that piece of
information encoded b ya with that of the members of the B asic S et —which accor ding to SM
is identical to the set of the Kaplanian indexicals (giv e or take a fe w contextuals). S o, they design
thr ee tests for identifying context-sensitiv e expr essions and claim: ‘N o semantic theor y should
classify as context-sensitiv e any expr ession that doesn ’ t pass these tests. S emantic M inimalism
doesn ’ t. All other theories do . ’ I n what follo ws, I will discuss C&L ’ s Context-S ensitivity T ests .
.. T est : B locking I nter-Contextual D isquotational I ndir ect R epor ts
An inter-contextual disquotational indir ect r epor t is a sentence of the form ⌜A said that S⌝
utter ed in a context other than S ’ s original context of utterance. S uppose y ou want to kno w
whether or not a is context-sensitiv e. Let ‘S(a) ’ be a declarativ e sentence containing a . H er e
is ho w the test wor ks:
(DIR
1
) I f a speaker A asser tiv ely uttersS(a) in a contextc
1
, and w e can ’ t tr uly in di-
r ectly r epor t that utterance with⌜A said thatS(a)⌝ in a r elev antly differ ent
contextc
2
, then that is evidence that a is context-sensitiv e.
(DIR
2
) I f a speaker A asser tiv ely uttersS(a) in a contextc
1
, and w e can tr uly indi-
r ectly r epor t that utterance with⌜A said thatS(a)⌝ in a r elev antly differ ent
contextc
2
, then that is evidence that a isn ’ t context-sensitiv e.
H er e is an example. S uppose y ou want to kno w whether or not the rst person pr onoun ‘I’
is context-sensitiv e. Consider the sentence ‘I’ m writing this chapter ’ as utter ed b y me while
writing this chapter . I n my context of utterance, the sentence is clearly tr ue. B ut, if y ou r e-
por t my utterance b y saying ‘E duar do said that I’ m writing this chapter ’, y ou would be saying
something false. us, the occurr ence of ‘I’ in the sentence blocks the disquotational indir ect
r epor t in question. S ince no sentence containing ‘I’ can be disquotationally indir ectly r epor ted
(ex cept b y self-r epor ters), w e hav e evidence that ‘I’ is context-sensitiv e.
C&L claim that all and only asser tiv e utterances of sentences containing the members of
the B asic S et block disquotational indir ect r epor ts. S ince both R C and MC hold that the B asic
S et is a pr oper subset of the set of context-sensitiv e expr essions, C&L conclude that R C and
MC ar e empirically inadequate. B ut, why do they think of this test as a tool for identifying
semantic context-sensitivity? eir r easoning seems to be mor e or less this. A ccor ding to them,
a linguistic expr ession a is context-sensitiv e iff a can expr ess differ ent contents r elativ e to dif-
fer ent contexts of use. S o, if a speakerA asser tiv ely uttersS(a) in contextc
1
anda is genuinely
context-sensitiv e, then the occurr ence ofa inS(a) will get a semantic contentk r elativ e toc
1
.
S ince the semantic content of a declarativ e sentence is a function of the semantic contents of
its par ts and of the way they ar e syntactically connected, it follo ws that ifa is context-sensitiv e,
both S(a) and⌜A said that S(a)⌝ will be context-sensitiv e too . S o, if a speaker B asser tiv ely
utters⌜A said thatS(a)⌝ in a r elev antly differ ent contextc
2
, it shouldn ’ t be surprising that the
corr esponding occurr ence of a in the disquotational indir ect r epor t gets a differ ent semantic
contentk
′
r e lativ e toc
2
. ⁸
B ased on these r easons, C&L r epor t their o wn intuitions about some expr essions that ar en ’ t
members of the B asic S et, but that Contextualists claim to be context-sensitiv e. C&L allege that
they fail the I nter-Contextual D isquotational I ndir ect R epor t T est (‘DIR T est ’ for shor t), and
that this giv es us str ong evidence that they ar en ’ t context-sensitiv e. Among those expr essions,
C&L mention the follo wing: ⁹
⁸ C&L allo w for accidental cases in whicha gets the same semantic content in bothc
1
andc
2
, ev en thougha
is semantically context-sensitiv e, i.e., it ’ s a member of the B asic S et.
⁹ S ee Cappelen & Lepor e ( , –).
Q uantiers
Any utterance of ‘er e is at least one duck in N or way ’ can be r epor ted b y ‘S he said
that ther e is at least one duck in N or way ’ and any two such utterances can be r epor ted
b y ‘ey both said that ther e is at least one duck in N or way . ’
P o ssessiv es
Any utterance of ‘R udolf ’ s penguin is happy ’ can be r epor ted b y ‘S he said that R udolf ’ s
penguin is happy ’ and any two such utterances can be r epor ted b y ‘ey both said that
R udolf ’ s penguin is happy . ’
G radable A djectiv es
Any utterance of ‘A is tall ’ can be r epor ted b y ‘S he said that A is tall ’ and any two such
utterances can be r epor ted b y ‘ey both said that A is tall. ’
W eather R epor ts
Any utterance of ‘I t ’ s raining ’ can be r epor ted b y ‘S he said that it ’ s raining ’ and any two
such utterances can be r epor ted b y ‘ey both said that it ’ s raining. ’
T emporal R epor ts
Any utterance of ‘I t ’ s thr ee p .m. ’ can be r epor ted b y ‘S he said that it ’ s thr ee p .m. ’ and
any two such utterances can be r epor ted b y ‘ey both said that it ’ s thr ee p .m. ’
V ague P r edicates
Any utterance of ‘J ackie has blue ey es ’ can be r epor ted b y ‘S he said that J ackie has blue
ey es ’ and any two such utterances can be r epor ted b y ‘ey both said that J ackie has
blue ey es. ’
Kno wledge Ascr iptions
Any utterance of ‘A kno ws that he has a hand ’ can be r epor ted b y ‘S he said that A
kno ws that he has a hand ’ and any two such utterances can be r epor ted b y ‘ey both
said that A kno ws that he has a hand. ’
F r om these examples, C&L conclude:
e main point of these illustrations is that if y ou agr ee with the data, then y ou ar e
endorsing our objection against any contextualist about the expr essions just discussed:
I n ter-Contextual Disquotational I ndir ect R epor ts, though per fectly natur al and ubiquitous,
ar e inexplicable on the assumption t hat R C (or some v ersion of MC that extends to any
of these abo v e cases) is tr ue . e challenge is to explain away or challenge this data (or
sho w that the test is no good). (K eep in mind, though, that in or der to do that a single
counter example is not enough. W e pr esent these claims as generalizations about all
utterances of v arious sentences because that ’ s what w e believ e, but this generalization
is not essential to our objection. I f a Radical (or M oderate) Contextualist agr ees that
ther e ar e some tr ue r epor ts of these kinds, then she has a pr oblem. er e should be no
such r eadings accor ding to R C or MC.) ( Cappelen & Lepor e , –)
is is a peculiarly str ong conclusion. Why do C&L think that b y ackno wledging that ther e
ar e some tr ue r epor ts of the kind they pr esent one is ther eb y endorsing their objection against
the Contextualist? Why do they think that if the Contextualist manages to come up with a
counter example to the test ’ s pr edictions, the bur den of pr oof is still on the Contextualist to
explain the data and not simply giv e up the test? I suspect that ther e ar e basically two r easons
for this: (i) C&L take their test to be an ob vious a priori extension of a specic understanding
of the notion of context-sensitivity , and (ii) they assume that this understanding of the no-
tion of context-sensitivity is the only/most plausible one. H o w ev er , I think that these r easons
ar e incorr ect. B ut befor e I explain why I think so, let me star t b y offering some prima facie
counter examples to their claims:
Q uantiers
C : B ill says to his classmates: ‘E v er ybody is invited to my bir thday par ty on
S atur day ’.
C : J ackie does some v olunteer wor k at an institution that pr o vides shelter to
homeless people. S he asks J ohn to help her to think about a fun activity for those
people on S atur day . J ohn ex citedly says: ‘ T ake them to B ill ’ s bir thday par ty! B ill said
that ev er ybody is invited to his bir thday par ty on S atur day!’
P o ssessiv es
C : B ill, who wor ks for a car-r ental company , lends J ohn the ne w F errari the
company has acquir ed, which J ohn later smashes. J ohn calls J ackie for assistance and
says: ‘I’ v e had an accident with B ill ’ s car ’.
C : A fe w minutes later a police officer arriv es at the scene of the accident and
asks J ackie (who has also arriv ed at the scene): ‘ Ar e y ou the o wner of the car?’ J ackie
r e plies: ‘N o . B ill S mith is. ’ e police officer asks: ‘H o w do y ou kno w?’ J ackie says:
‘J ohn said that he had had an accident with B ill ’ s car . ’ ¹⁰
G radable A djectiv es
C : T alking about his v e-y ear-old niece, E mily , J ohn says to B ill: ‘E mily is
tall. ’
C : e chief of the local police depar tment tells B ill that they ar e tr ying to
r e cr uit y oung tall females. H e asks B ill whether he kno ws of a y oung tall female he
could talk to . B ill r eplies: ‘ T alk to E mily . J ohn said that E mily is tall. ’
W eather R epor ts
C : J ackie, who is tr ekking in P er u, is speaking to B ill o v er the phone. B ill,
who is bor ed at his windo wless USC office, asks her: ‘H o w is the w eather?’ J ackie says:
‘I t ’ s raining ’.
C : J ohn, who is at the office next to B ill ’ s, is deliberating whether it would be
a good idea to go on a picnic to G riffith P ar k. S o, after B ill hangs up on J ackie, he asks
B i ll: ‘I s it raining?’ B ill r eplies: ‘J ackie says it ’ s raining. ’
T emporal R epor ts
C : J ackie is no w in Kathmandu. B ill, who is in L A, calls her . H o w ev er , since
B i ll doesn ’ t kno w the time z one differ ence betw een L A and Kathmandu, he ’ s worried
¹⁰ N otice that although I hav e changed the tense of the v erb in the disquotational r epor t, this doesn ’ t disqualify
the example. F or C&L allo w for what they call ‘ disquotational
⋆
indir ect r epor ts ’: ‘ T o be ‘ disquotational
⋆
’ just
means y ou can adjust the semantic v alues of components of S that ar e generally r ecogniz ed as context-sensitiv e,
i.e., w e just test for the contr o v ersial components. ’ ( Cappelen & Lepor e , )
J ackie might be sleeping. J ackie picks up the phone and after saying hello, B ill asks her
whether it ’ s a good time for him to call her . S he appr o vingly r eplies: ‘I t ’ s thr ee p .m. ’
C : B ill ’ s r oommate, J ohn, has an impor tant job inter vie w at thr ee fteen p .m.
in do wnto wn L A. W orried that he might be r unning late, he asks B ill: ‘D o y ou kno w
what tim e is it?’ B ill, who is still talking to J ackie on the phone, r eplies: ‘J ackie says it ’ s
thr ee p .m!’
V ague P r edicates
C : M att is in S omalia doing some v olunteer wor k among poor semi-nomadic
people. O ne day , he o v erhears two of them, Akua and A tu, talking about N ur u, who
wor ks for a local farmer cleaning his barn. S peaking about N ur u ’ s economic situation,
Akua says: ‘N ur u is rich. ’
C : M att is on the phone speaking to his father—a w ell-kno wn W all S tr eet
ty coon. I n the middle of the conv ersation, M att ’ s father tells him that the curr ent
economic crisis has for ced him to desperately look for a rich business par tner . M att
immediately says: ‘ Akua said that N ur u is rich!’
As far as I can tell, the r epor ts in C sound natural and ar e intuitiv ely false. I f so, they
pass the DIR T est. S hould w e take this to be evidence in fav or of the claim that the expr essions
under consideration ar e context-sensitiv e? C&L don ’ t think so, for they don ’ t include these
expr essions in their B asic S et. S o, what justies C&L ’ s decision to ex clude these terms fr om the
set of context-sensitiv e expr essions? As I hav e mentioned befor e, I think their r easons cr ucially
depend on their understanding of the notion of context-sensitivity . H o w about C&L ’ s claim
about kno wledge ascriptions? I s it ob vious that sentences containing it fail the DIR T est? I don ’ t
think so; intuitions about kno wledge attributions ar e notoriously inconclusiv e. S o, I don ’ t think
that appealing to pr e-theor etical intuitions about them will tip the scales either way . I will say
mor e about this later on.
N o w , the M inimalist might r espond: ‘I t ’ s tr ue that the r epor ts in the illustrations sound
false. H o w ev er , it ’ s also tr ue that the speakers in C in fact said the wor ds r epor ted in
C . And this is all I need to make my point: no member of the B asic S et allo ws the
disquotational r epor t to come out tr ue. S o, these scenarios fail to pose a pr oblem to the DIR
T est in par ticular , and to SM in general. ’ is r esponse, I think, doesn ’ t addr ess the worr y these
cases pose to the test and to SM.
W e hav e to be car eful about the intuitions w e ar e tracking when w e say that the speakers in
C in fact said the wor ds r epor ted in C . As H awthorne ( ) points out,
“ many ‘ say that ’ r epor ts hav e a feel of ‘ mix ed quotation ’. ” P ut differ ently , competent speakers
fr equently use ‘ say that ’ r epor ts in conv ersation as a kind of hidden quotation mar ks. F or
example, consider the follo wing case:
W
FBI agents Alice and Bob hav e nally managed to inter cept suspect S mith ’ s cell phone.
ey ar e lucky , because S mith is in the middle of an impor tant conv ersation with an-
other high pr ole suspect. As soon as agent Alice puts on the headphones and eav es-
dr ops S mith ’ s rst wor ds, agent Bob asks him: “ What did he say?” Alice answ ers: “H e
said that Corleone would ”. P uzzled, Bob asks: “Corleone would what?” Alice r eplies:
“I don ’ t kno w!”
N otice that in this scenario, Alice successfully r epor ted S mith ’ s utterance; S mith in fact pr o-
nounced the wor ds ‘Corleone would ’. H o w ev er , this r epor t fails to captur e any complete pr opo-
sition (which, accor ding to C&L, is the semantic content of a sentence). Alice is able to pr oduce
the r epor t without having a clue about what was said at all. All Alice might r eally car e about
is getting S mith ’ s exact wor ds right. B ut, of course, C&L don ’ t want their test to appeal to
this use of ‘ say that ’ r epor ts. ey want the test to be an effectiv e tool for detecting seman-
tic context-sensitivity . S o, if w e scr een off these uses, then our intuitions about the r epor ts in
C tend to fav or the claim that they ar e false. And, as I hav e suggested, this should
pr ompt C&L to r evise either their test or their theor etical commitments. B ut, what has gone
wr ong with the DIR T est? I n the next section, I will tr y to answ er this question.
R epor ting I ndexicals
I n or der to motiv ate the DIR T est, C&L discuss the semantic behavior of pur e indexicals such
as ‘I’, ‘ no w ’ and ‘ tomorr o w ’ in disquotational indir ect r epor ts. ey point out that these expr es-
sions ar e differ ent fr om the Contextualist ’ s candidates (e.g. ‘ ev er ybody ’, ‘ tall ’, ‘blue ’, ‘kno ws ’,
etc.) in that sentences containing them can ’ t be tr uly indir ectly r epor ted (ex cept b y accident) in
any context c that differs fr om their original contexts of utterance with r espect to the r elev ant
context coor dinates, i.e. the coor dinates that determine the semantic content of the indexi-
cal expr essions in question. B ased on this obser v ation, C&L constr uct the DIR T est as a tool
for detecting this semantic pr oper ty of the Kaplanian pur e indexicals. eir basic assumption
seems to be that this pr oper ty should be shar ed b y all context-sensitiv e expr essions. is, I
believ e, is an unjustied assumption.
Kaplan ’ s pur e indexicals ar e special. When I utter ‘I’ I r efer to E duar do V illanuev a ev en
if I am absolutely cer tain that I am Christopher Columbus. M y intentions, beliefs, thoughts,
wishes, etc. about who I am make no d iffer ence. e r efer ence of ‘I’ is x ed independently
of my mental states. I f I utter ‘ today ’ on M onday , M ar ch , I r efer to M onday , M ar ch
, . I t doesn ’ t matter whether at that moment I am cer tain it ’ s F riday , O ctober , .
S imilarly for the other pur e indexicals; giv en the meaning and the context of utterance, the
r efer ence is—in P err y ’ s wor ds—automatic. ¹¹
us, if y ou want to r epor t what I said b y my utterance of a sentence containing a pur e
indexical, y ou can ’ t use the v er y same wor ds I utter ed unless y ou ar e in the same time, or in the
same position, or in the same world-state, or y ou ar e me. And the r eason y ou can ’ t use the v er y
same wor ds I hav e utter ed when r epor ting what I said is that pur e indexicals get their r efer ence
fr om the r elev ant objectiv e featur es of the context of utterance, no matter what speakers intend.
is isn ’ t the only kind of context-sensitivity that Kaplan considers. H e also dev elops a
semantic theor y for demonstrativ es. A ccor ding to Kaplan, demo nstrativ es r equir e, in or der to
determine their r efer ents, an associated demonstration. Kaplan holds that without a demonstra-
tion, demonstrativ es ar e incomplete. I n D emonstr ativ es ( b ), he describes demonstrations
as ‘ typically , though not inv ariably , a (visual) pr esentation of a local object discriminated b y
¹¹ S ee P err y ( a ).
a pointing ’ ( Kaplan b , ). H ence, since the range of things I can demonstrate is v er y
limited, if I hear y ou uttering a sentence containing a demonstrativ e, but either I don ’ t see the
object y ou ar e demonstrating or I want to r epor t y our utterance in a context in which the object
demonstrated isn ’ t av ailable, I can ’ t use y our v er y same wor ds.
H o w ev er , this isn ’ t the whole stor y . I t ’ s quite ob vious that some utterances of demonstra-
tiv es successfully r efer ev en though the speaker doesn ’ t demonstrate anything. W e typically
succeed in r eferring to objects using demonstrativ es without an associated demonstration when
it ’ s ob vious to ev er y/most conv ersational par ticipants what object w e ar e talking about. P ut
differ ently , w e succeed in r eferring to objects when our r eferring intentions ar e clearly grasped
b y our interlocutors in a giv en context. is obser v ation is among the r easons why Kaplan
changes his mind about the natur e of demonstrations in his After thoughts . H e says:
O n the theor y of tr ue demonstrativ es in D emonstr ativ es , a demonstration accompanies
ev er y demonstrativ e and determines its r efer ent. O n my curr ent vie w , the r efer ent of a
tr ue demonstrativ e is determined b y the utter er ’ s intention. ( Kaplan a , )
S o, what if ther e ar e other expr essions in E nglish semantically sensitiv e to speaker ’ s intentions?
W ell, ‘ no w ’ and ‘ w e ’ depend to a cer tain ex tent on the intentions of the speaker: a speaker might
use ‘ no w ’ to r efer to a v er y pr ecise moment in time (‘ N o w it ’ s :: AM, but no w it ’ s not ’) or
to r efer to a longer undened period of time (‘ N o w people ar e living longer than ev er befor e ’).
S i milarly for ‘ w e ’: a speaker might use ‘ w e ’ to r efer to a specic number of individuals (‘ W e ar e
a married couple ’) or to r efer to a gr oup of people whose specic number isn ’ t impor tant for the
purposes of the speaker ’ s utterance (‘ W e ar e gather ed her e to mourn the death of D r . X ’). S o,
what if ther e ar e other expr essions sensitiv e to speaker ’ s intentions that don ’ t appear in Kaplan ’ s
list of indexicals? Let ’ s nd out.
R epor ting N on-I ndexical Context-S ensitiv e E xpr essions
S uppose that the semantic content of a pr edicate F depends on context featur esd
1
,d
2
,...,d
n
such that for any two contextsc
1
andc
2
that differ with r espect tod
1
,d
2
,...,d
n
, the seman-
tic content of F r elativ e to c
1
and c
2
will also differ . P r esumably , F would count as a genuine
context-sensitiv e expr ession. N o w , suppose that the semantics forF is r oughly as follo ws. F irst,
d
1
,d
2
,...,d
n
ar e degr ees of some contextually determined standar d D such that ev er y utter-
ance F picks out pr oper ty P to some degr eed
i
. S econd, b y default, ev er y utterance of F picks
out P to degr ee d
j
and speakers kno w this on the basis of their semantic competence with F .
ir d, whether a giv en use of F picks out P to a higher degr ee than d
j
(default) depends on
the speakers ’ backgr ound pr esuppositions, their mutual expectations, and the purposes of the
conv ersation in which they utter F .
I f this w er e a plausible description of the semantic behavior of F , ther e ar e at least two
ways I can think of in which one could explain why tr ue disquotational indir ect r epor ts of
declarativ e sentences containing F ar e possible. O ne possibility is that, for the purposes of
ev er y day communication, if an or dinar y declarativ e sentence S(F) is asser tiv ely utter ed in c
1
and a speaker wants to r epor t it in a differ ent context c
2
without kno wing which d of D was
determined b y c
1
, the r epor ter could assume that F picks out P to degr ee d
j
(default). e
other possibility is this. S uppose the r epor ter doesn ’ t kno w the standar d the r epor tee is using in
c
1
. S till, it seems plausible that the r epor ter would be able to r epor t the utterance in question
b y simply intending to pick out whatev er contentF expr essed inc
1
, just as one is able to r epor t
a sentence containing a pr oper name of a person one isn ’ t acquainted with.
us, if ther e was any such expr ession in a natural language, competent speakers would
be able to successfully indir ectly r epor t sentences containing it ev en though the expr ession in
question would be semantically context-sensitiv e. I n other wor ds, such an expr ession would fail
C&L ’ s DIR T est despite being semantically context-sensitiv e. e key question, then, is this: is
ther e any E nglish expr ession that r oughly ts this pictur e? S ome scholars think ther e ar e. F or
example, some Contextualists think that ‘kno w ’ and v ague pr edicates in E nglish behav e mor e
or less in the ways described.
A ccor ding to some contextualist accounts, ¹² the semantic content of ‘kno w ’ is sensitiv e to
contextually salient standar ds. ese standar ds can be thought of as standar ds of degr ee of
some factor K , wher e K could be the str ength of the r elev ant evidence, the per tinent practical
inter ests, etc. S ince the r equir ed standar ds for the application of ‘kno w ’ can v ar y fr om context
to context, declarativ e sentences containing it can expr ess differ ent pr opositions in differ ent
contexts. S o, for example, since in or dinar y conv ersational contexts, the standar ds ofK r equir ed
for the corr ect application of ‘kno w ’ ar e lo w , in those contexts ‘kno w ’ will normally expr ess a
pr oper ty that corr esponds to those lo w standar ds (call it ‘kno w
low
’). O n the other hand, in
contexts in which the standar ds for the corr ect application of ‘kno w ’ ar e extraor dinarily high
(e.g., a cour t of law or a scientic confer ence), ‘kno w ’ will normally expr ess a pr oper ty that
corr esponds to those high standar ds (call it ‘kno w
high
’). S ince lo w-standar d contexts ar e way
¹² F or example, Le wis ( ), Cohen ( ), and D eR ose ( ).
mor e common than high-standar d contexts, competent E nglish speakers will normally assume
and expect other speakers to assume that ‘kno w ’ expr esses kno w
low
b y default.
O n this pictur e, if A asser tiv ely utters ⌜S( kno ws)⌝ in c
1
and I ignor e the standar ds for
kno wledge inc
1
, I can successfully disquotationally r epor t it inc
2
b y assigning ‘kno w ’ its default
extension, i.e., kno w
low
. S o ev en if the standar ds for the corr ect application of ‘kno w ’ in c
1
w er e extraor dinarily high, my r epor t inc
2
will come out tr ue, because the claim that one kno ws
p at lo w er standar ds is trivially entailed b y the claim that one kno ws p at higher standar ds.
Alternativ ely , my r epor t in c
2
will come out tr ue if in r epor ting the sentence in question I
simply intend whatev er interpr etation the r epor tee had in mind. Arguably , in such a case my
utterance of ‘kno w ’ would, as it w er e, inherit whatev er content it had in c
1
. I f either of these
possibilities obtains, then ‘kno w ’ will count as a genuine context-sensitiv e expr ession ev en if
the uses of it under consideration fail C&L ’ s DIR T est.
A similar account can be giv en of v ague pr edicates such as ‘blue ’ or ‘bald ’. ¹³ F or exam-
ple, S oames ( ) argues that v ague pr edicates ar e context-sensitiv e and par tially dened.
R oughly , the idea is this. T ake the pr edicate ‘blue ’. N ormally , E nglish nativ e speakers learn it
b y ostention, i.e. they learn it b y being sho wn uncontr o v ersial examples of blue things and un-
contr o v ersial examples of things that ar en ’ t blue. As a r esult, they acquir e beliefs about the range
of things to which ‘blue ’ applies and the ra nge of things to which it doesn ’ t. E v entually , when
their application of ‘blue ’ r egularly and systematically coincides with their linguistic commu-
nity ’ s application of it, they become competent with the pr edicate in question. ese systematic
¹³ F or instance, see S oames ( ), Raffman ( ), and F ara ( ).
coincidences can be understood as r ules that go v ern the application of ‘blue ’. H o w ev er , since
ther e ar e objects not co v er ed b y the r ules and, thus, ther e ar e no gr ounds for accepting either the
claim that ‘blue ’ applies to them or it doesn ’ t (i.e., ‘blue ’ is undened for these objects), speakers
ar e fr ee to apply the pr edicate ‘blue ’ (or its negation) to them accor ding to their conv ersational
purposes and backgr ound beliefs.
O n this account, then, ‘blue ’ is context-sensitiv e and, as in the case of ‘kno ws ’, the context
standar ds for being blue might go fr om lax to stringent. S o, ev en if a competent speaker o v er-
hears an utterance of a declarativ e sentence containing ‘blue ’ or applying ‘blue ’ to an object
and doesn ’ t kno w which standar ds ar e r elev ant in the context of utterance, she will kno w that
the speaker is using ‘blue ’ with an extension that includes its default extension—i.e. the set of
objects to which the pr edicate applies in vir tue of the community-wide r ules that go v ern its use.
B ut if so, then she doesn ’ t hav e to kno w the standar ds in play to successfully disquotationally
r epor t such sentences. F or if something is blue r elativ e to a stringent standar d, it sur ely is blue
r elativ e to the default one. er efor e, on this pictur e, ‘blue ’ is context-sensitiv e ev en though
some uses of it fail C&L ’ s DIR T est.
I f these obser v ations ar e not off track, then w e hav e good r easons to doubt the r eliability of
the DIR T est as a tool for detecting context-sensitivity . N otice that w e don ’ t hav e to endorse
the Contextualist accounts I hav e just sketched. e point is rather tha t whether or not they
ar e plausible can ’ t be decided based on a test that simply r ules out these possibilities. I f w e
ar e to challenge the Contextualist about ‘kno w ’ or v ague pr edicates w e hav e to do it based
on substantiv e arguments. I t ’ s simply methodologically bad to r eject these vie ws because they
don ’ t pass a test built on an unexamined understanding of the notion of context-sensitivity .
F or these r easons, I conclude that the DIR T est is an inadequate tool for detecting semantic
context-sensitivity . Let ’ s mo v e on to r evie w C&L ’ s second test.
.. T est : B locking Collectiv e D escr iptions
H er e is ho w the test wor ks. Let ‘v ’ be a v erb phrase. Let ‘u ’ be an asser tiv e utterance at c
1
of
the sentence⌜Av -s⌝ . Let ‘u
′
’ be an asser tiv e utterance atc
2
of the sentence⌜Bv -s⌝ . S uppose
that both u and u
′
ar e tr ue. A ccor ding to this test, if v is context-sensitiv e, then w e can ’ t
automatically infer that ther e is a contextc
3
in which the collectiv e description⌜A andB both
v⌝ comes out tr ue. O n the other hand, if it ’ s ob vious that ther e is such c
3
, that is evidence
in fav or of the claim that v isn ’ t context-sensitiv e. I n shor t, one might r econstr uct the test as
follo ws:
(CD
1a
) I f⌜Av -s⌝ is tr ue in a contextc
1
,⌜Bv -s⌝ is tr ue in a r elev antly differ ent con-
text c
2
, but w e can ’ t tr uly asser tiv ely utter the collectiv e description⌜A and
B bothv⌝ , then that is evidence that the v erb phrasev is context-sensitiv e.
(CD
2a
) I f⌜A v -s⌝ is tr ue in a context c
1
,⌜B v -s⌝ is tr ue in a differ ent context c
2
,
and w e can tr uly asser tiv ely utter the collectiv e description⌜A and B both
v⌝ , then that is evidence that the v erb phrasev isn ’ t context -sensitiv e.
e r eason C&L think of this test as a tool for identifying semantic context-sensitivity is this:
I f ‘v ’ is a context-sensitiv e term, then its semantic v alue can change fr om one utterance
to another . S o, ‘Av -s ’ and ‘Bv -s ’ might attribute differ ent pr oper ties to A andB . B ut
it doesn ’ t follo w that ‘v ’ can be used to describe what A andB shar e. M aybe b y chance
someone might be able to use ‘v ’ in some context to r efer to a pr oper ty they both shar e,
but that would be a coincidence. ( Cappelen & Lepor e , , ft. )
C&L make similar r emar ks about singular terms. I f the singular termn is context-sensitiv e and
asser tiv e utterances of⌜n is F⌝ and⌜n isG⌝ ar e tr ue in r elev antly differ ent contextsc
1
andc
2
r espectiv ely , then ther e is no guarantee that ther e will be a context c
3
such that the collectiv e
description⌜n is F and G⌝ is tr ue in c
3
. O n the other hand, if n isn ’ t context-sensitiv e and
asser tiv e utterances of⌜n is F⌝ and⌜n isG⌝ ar e tr ue in r elev antly differ ent contextsc
1
andc
2
r e spectiv ely , then ther e will be a contextc
3
such that the collectiv e description⌜n is F andG⌝
is tr ue inc
3
. I n shor t,
(CD
1b
) I f⌜n is F⌝ is tr ue in a context c
1
,⌜n is G⌝ is tr ue in a r elev antly differ ent
contextc
2
, but w e can ’ t tr uly asser tiv ely utter the collectiv e description⌜n is
F andG⌝ , then that is evidence that the singular termn is context-sensitiv e.
(CD
2b
) I f⌜n is F⌝ is tr ue in a contextc
1
,⌜n is G⌝ is tr ue in a differ ent contextc
2
,
and w e can tr uly asser tiv ely utter the collectiv e description⌜n is F and G⌝ ,
then that is evidence that the singular termn isn ’ t context-sensitiv e.
F or example, suppose y ou want to kno w whether or not ‘ y ester day ’ is context-sensitiv e. S up-
pose y ou kno w that the sentence ‘ Y ester day J ohn left ’ is tr ue in c
1
, and als o that the sentence
‘ Y ester day B ill left ’ is tr ue in a differ ent context c
2
. H o w ev er , suppose that y ou ignor e the
times of these contexts. S ince fr om this information y ou can ’ t automatically infer that ther e
is a context c
3
in which ‘ Y ester day J ohn and B ill left ’ is tr ue, y ou hav e evidence that the term
‘ y ester day ’ is context-sensitiv e. ¹⁴
As in the case of the DIR T est, C&L conjectur e that all and only asser tiv e utterances of
sentences containing the members of the B asic S et block collectiv e descriptions. S ince both R C
and MC hold that the B asic S et is a pr oper subset of the set of context-sensitiv e expr essions,
C&L conclude that R C and MC ar e empirically inadequate. Among the expr essions that the
Contextualist considers to be context-sensitiv e but fail the Collectiv e D escription T est (‘CD
T est ’ for shor t) C&L discuss ar e the follo wing: ¹⁵
Kno wledge Ascr iptions
I f ther e is a tr ue utterance of ‘A kno ws that he has a hand ’ in context of utterancec , and
another tr ue utterance of ‘B kno ws that he has a hand ’ in r elev antly differ ent context
c
′
, the follo wing collectiv e description is per fectly natural: ‘Both A and B kno w that
they hav e hands. ’
V ague P r edicates
I f ther e is a tr ue utterance of ‘J ackie has blue shoes ’ in context of utterance c , and
another tr ue utterance of ‘J ackie has blue sungla sses ’ in a r elev antly differ ent contextc
′
,
then the follo wing collectiv e description is per fectly natural: ‘J ackie has blue shoes and
sunglasses. ’
¹⁴ I nter estingly enough, despite C&L ’ s militant Kaplanianism about context-sensitivity , they tr eat ‘ y ester day ’ as
a singular term—not as a sentential operator , as Kaplan does. H er e, I think, C&L ar e right.
¹⁵ S ee Cappelen & Lepor e ( , –).
G radable A djectiv es
I f ther e is a tr ue utterance of ‘M ount E v er est is tall ’ in a context of utterance c , and
another tr ue utterance of e E mpir e S tate B uilding is tall in a r elev antly differ ent
contextc
′
, then the follo wing collectiv e description is per fectly natural: ‘M ount E v er est
and the E mpir e S tate B uilding ar e both tall. ’
P o ssessiv es
I f ther e is a tr ue utterance of ‘J ill ’ s daughter is happy ’ in a context of utterance c , and
another tr ue utterance of ‘J ill ’ s dog is happy ’ in a r elev antly differ ent contextc
′
, then the
follo wing collectiv e description is per fectly natural: ‘J ill ’ s dog and daughter ar e happy . ’
Q uantiers
I f ther e ar e tr ue utterances of ‘J ill bought at least two penguins ’ and ‘J ill bought at least
two ducks ’ in two r elev antly differ ent contexts, then the follo wing collectiv e description
is per fectly natural: ‘J ill bought at least two penguins and ducks. ’
And, nally , C&L end up their discussion of these cases b y posing the same challenge they
posed in connection with the DIR T est:
I f R C (or MC) is tr ue, none of these collectiv e sentences should be tr ue. I f y ou hav e
the intuition that the collectiv e sentences ar e tr ue in at least some of these cases, then
y ou need either to nd a way to explain away those intuitions or to nd a way to
accommodate them. O ur hypothesis: ese intuitions can ’ t be explained away and
they can ’ t be accommodated within the frame wor k of R C or MC. ( Cappelen & Lepor e
, )
N otice that this test (like the others) inv okes the notion of r elev antly differ ent contexts. H o w-
ev er , C&L don ’ t tell us exactly what this notion amounts to, and I suspect that the v agueness
of this notion might be playing an impor tant r ole in the r esults of their test. T ake for example
the term ‘kno w ’. As I hav e alr eady mentioned, accor ding to some Contextualists, the semantic
content of ‘kno w ’ is sensitiv e to contextually salient standar ds. S uppose that those standar ds
ar e standar ds of degr ee of some factorK . H ence, for any two uses of ‘kno w ’ (e.g. in kno wledge
ascriptions) either they will pick out the same degr ee of K or not. I t follo ws fr om this that
for any two tr ue kno wledge ascriptions in contexts in which the degr ee ofK is differ ent, ther e
will always be a context in which the c ollectiv e description is tr ue, namely , one with the same
standar d as the context with the lo w er standar d. I f the Contextualist is right, then ‘kno w ’ is
context-sensitiv e ev en though it fails C&L ’ s CD T est.
A similar stor y can be told about v ague terms such as ‘blue ’ or ‘ tall ’. S uppose that ‘J ackie has
blue shoes ’ is tr ue r elativ e to a context in which the standar ds for counting as blue ar e v er y lax.
S uppose that ‘J ackie has blue sunglasses ’ is tr ue r elativ e to a context in which the standar ds ar e
v er y stringent. I n that s cenario, ther e will always be a context in which the collectiv e description
‘J ackie has blue shoes and sunglasses ’ is tr ue, namely , a context with the same standar d as the
context with the lax standar d. er efor e, if ‘blue ’ wor ks as described, it ’ s context-sensitiv e, ev en
though it fails C&L ’ s test.
What about possessiv e constr uctions and quantiers? I must confess that I don ’ t see the
r elev ance of C&L ’ s discussion of examples described abo v e. F irst, they don ’ t t the v erb-phrase
v ersion or the singular-term v ersion of the CD T est. S econd, it seems that w e can easily con-
str uct cases inv olving possessiv es and quantiers that will fail to hav e a collectiv e description
—contrar y to what C&L think. F or example,
P o ssessiv es
C : T alking about the G r eat D ane J ill o wns, B ill says: ‘J ill ’ s dog is huge. ’
C : J ill has taken her neighbor ’ s Chihuahua to a D og P ar k. S eeing the dog J ill
is walking, B ill says to J ohn: ‘J ill ’ s dog is tiny . ’
O b viously , fr om the tr uth of these utterances in their r espectiv e contexts it doesn ’ t follo w that
the collectiv e description ‘J ill ’ s dog is huge and tiny ’ is also tr ue. O r consider the follo wing
scenarios,
Q uantiers
C : After sear ching unsuccessfully thr ough the supermar ket shelv es, B ill calls
J ohn to tell him that ther e ar e no water bottles. D isappointed, J ohn says to J ill: ‘B ill
says that ther e ar e no water bottles. ’
C : B ill is a w ell-kno wn nominalist about abstract entities. I n a discussion
betw een J ohn and J ill about this issue, J ohn says to J ill: ‘B ill says that ther e ar e no
abstract entities. ’
H o w e v er , the tr uth of these utterances in their r espectiv e contexts doesn ’ t guarantee the tr uth
of the collectiv e description ‘B ill says that ther e ar e no abstract entities and water bottles. ’ ¹⁶
S o, what is wr ong with the CD T est? C&L might be right that fr om the fact that v is
context-sensitiv e (in some way or other) plus the fact that the individual attributions ar e tr ue
¹⁶ H er e I am using a disquotational v ersion of the CD T est. is v ersion of the test is pr oposed in Cappelen &
Lepor e ( b ). is is ho w they formulate it: ‘ T est : Collectiv e I nter-Contextual Disquotational I ndir ect R epor ts :
T ake sev eral utterances, u
1
u
n
, o f S(e) in contexts C
1
C
n
, all of which ar e r elev antly differ ent. I f , in a C
′
,
w e can indir ectly r epor tu
1
u
n
with ‘ey said thatS(e) ’ (wher e ‘ they ’ r efers to the speakers ofu
1
u
n
), then
that ’ s evidence e is semantically stable, i.e. that its semantic v alue doesn ’ t v ar y betw een contexts of utterance. ’
( Cappelen & Lepor e b , ).
in the two contexts, it doesn ’ t follo w that the collectiv e attribution will be tr ue in some context.
H o w ev er , this may follo w fr om the par ticular way in which v is context-sensitiv e—especially
ifv is ‘kno w ’ or a v ague pr edicate. As in the case of the DIR T est, C&L build the CD T est on
the assumption that context-sensitivity is basically Kaplanian indexicality . And as I argued in
my discussion of the DIR T est, this assumption is unjustied.
.. T est a: P assing an I nter-Contextual D isquotational T est
is test states that a is context-sensitiv e only if ther e is a tr ue utterance of an instance of any
of the follo wing schemata for I nter-Contextual D isquotation:
(ICD
1
) er e is (or can be) at least one false utterance of⌜S(a)⌝ ev en thoughS(a) .
(ICD
2
) er e is (or can be) at least one tr ue utterance of⌜S(a)⌝ ev en though it ’ s not
the case thatS(a) .
e idea behind this test is this: since a isn ’ t context-sensitiv e unless its semantic content can
shift fr om context to context, to test whether a is context-sensitiv e or not, the theorist should
puta in a declarativ e sentenceS(a) and then utter it while per forming the test. I fa is genuinely
context-sensitiv e, the tr uth-v alue ofS(a) will v ar y fr om one context of utterance to a r elev antly
differ ent one. And this shift in tr uth-v alue is pr ecisely what the ICD T est is designed to make
explicit.
F or instance, suppose that w e want to kno w whether ‘ she ’ is context-sensitiv e. A ccor ding to
the ICD T est, one should star t b y selecting a sentence containing ‘ she ’ (and no other context-
sensitiv e expr ession), say the follo wing:
() S he is F r ench.
en, one should use ( ) to generate an instance of either (ICD
1
) or (ICD
2
),
() er e is (or can be) at least one false utterance of ‘S he is F r ench ’ ev en though
she is F r ench.
S i nce an asser tiv e utterance of ( ) in a context in which the speaker is pointing at a F r ench
woman would be tr ue, w e hav e evidence that ‘ she ’ is context-sensitiv e.
H o w ev er , suppose that w e constr uct an instance of either (ICD
1
) or (ICD
2
) fora . S uppose
w e ar en ’ t sur e whether the instance in question is tr ue. H o w could w e convince ourselv es and
others that the instance is in fact tr ue? C&L think that the best way to do it ’ s b y pr o viding a
R eal Context S hifting Argument for a .
.. T est b: Allo wing for R eal Context S hifting Arguments
H er e is C&L ’ s pr oposal for eliciting intuitions about the r esults of the ICD T est: an expr ession
a passes the ICD T est just in case it ’ s possible to constr uct a R eal Context S hifting Argument
(‘R CSA ’ for shor t) inv olving a . N o w , what is an R CSA? I n or der to answ er this question, w e
hav e to intr oduce the follo wing notions:
Context S hifting Argument (CSA)
A kind of thought experiment in which one imagines an alleged context-sensitiv e ex-
pr e ssiona being corr ectly used in a range of r elev antly differ ent contexts. e r esulting
data consists of the theorist ’ s r epor ts of her , and the audience ’ s o wn intuitions about
the semantic content of a at those imagined contexts.
S tor ytelling Context: e context in which a CSA is told.
T arget Context: A context about which a CSA is told.
S o, an R CSA is a description of a single context in which the alleged context-sensitiv e expr ession
a is both used in the S tor ytelling Context and also mentioned in describing its uses in T arget
Contexts. P ut simply , an R CSA is a narrativ ely enriched ICD. H er e is an example of a use of
an R CSA to sho w that ‘ no w ’ is context-sensitiv e:
N
Right no w , S tephen is not w earing a hat. Y ester day he was w earing a hat. And when
M ar y then utter ed ‘S tephen is w earing a hat no w ’ what she said then was tr ue, ev en
though S tephen is not w earing a hat no w . ¹⁷
S ince ‘ no w ’ is used both in the S tor ytelling Context and in the T arget Context, and the pr oposi-
tions expr essed b y the sentences that inv olv e ‘ no w ’ change tr uth v alue r elativ e to these contexts,
C&L conclude that N pr o vides str ong evidence in fav or of the thesis that ‘ no w ’ is context-
sensitiv e. A ccor dingly , w e can generate the follo wing instance of (ICD
2
) for ‘ no w ’:
() er e is at least one tr ue utterance of ‘S tephen is w earing a hat no w ’ ev en though
it isn ’ t the case that S tephen is w earing a hat no w .
C&L hold that it ’ s only thr ough this kind of stories (i.e., R CSAs) that one should be convinced
that an expr essiona passes the ICD T est, and that only such expr essions ar e context-sensitiv e.
¹⁷ A dapted fr om Cappelen & Lepor e ( , ).
C&L claim that neither the M oderate nor the Radical Contextualist r ely on R CSAs to
argue for their vie ws. I nstead, they typically r ely on I mpo v erished Context S hifting A rguments
(ICSA), i.e., arguments in which the alleged context-sensitiv e expr ession doesn ’ t get used (in the
S tor ytelling Context), but is instead only mentioned in describing its uses in T arget Contexts.
F or example, take a commonly used contextualist strategy for getting evidence in fav or of the
thesis that gradable adjectiv es such as ‘ tall ’ ar e context-sensitiv e. S he will ask y ou to consider
two T arget Contexts: (i) a context in which the topic of conv ersation is the heights of NBA
play ers, and (ii) a context in which the topic of conv ersation is the heights of U.S. pr esidents.
en she will ask y ou (fr om within the S tor ytelling Context) to consider two distinct utterances
of ( ) in these two T arget Contexts:
() B ill Clinton is tall.
S i nce Clinton isn ’ t tall for an NBA play er , our intuitions lean to war ds concluding that the
utterance of ( ) in the rst T arget Context is false. B ut, since Clinton is taller than most U.S.
pr esidents our intuitions seem to suppor t the claim that the utterance of ( ) in the second T arget
Context is tr ue. e Contextualist will take this as evidence in fav or of the context-sensitivity of
‘ tall ’. A ccor ding to C&L, since ( ) is nev er used in the S tor ytelling Context, this is an example
of an ICSA; sentence ( ) is only mentioned in describing its uses in the two T arget Contexts.
H ence—C&L conclude—this case doesn ’ t pr o vide evidence in fav or of the claim that ‘ tall ’ is
context-sensitiv e.
S imilarly for other terms such as ‘kno ws ’ and ‘r ed ’. C&L argue that the stories pr esented
in defense of the context-sensitivity of these terms nev er take the form of an R CSA, but that
of an ICSA. As an illustration of ho w har d it is to constr uct R CSAs for these terms, C&L giv e
the follo wing:
K R
Right no w , I’ m doing philosophy and thinking about R uper t. R uper t, ho w ev er , is not
doing philosophy . I nstead, he ’ s home making tea. R uper t doesn ’ t kno w he is y ears
old. F or R uper t to kno w he is y ears old, he has to r ule out the possibility that he is
a brain in a v at. R uper t, ho w ev er , is unawar e of (or not thinking about) this possibility .
And so he ’ s ignoring a possibility that must be r uled out in or der for any one to kno w
anything at all. S till, when R uper t utters in the comfor t of his home, ‘I kno w I am
y ears old ’ what he says is tr ue, because he ’ s ignoring this possibility , ev en though
this possibility has got to be consider ed in or der for R uper t to kno w anything at all.
( Cappelen & Lepor e , )
R R
I n or der to be r ed, an apple has to hav e r ed skin. at ’ s a necessar y condition for being
a r ed apple. I t is irr elev ant, for instance, whether an apple is r ed on its inside. H er e ’ s an
apple, call it R uper t; R uper t is r ed. O n the inside, R uper t is white. N onetheless, ther e
ar e utterances of ‘R uper t is r ed ’ that ar e false, not because R uper t ’ s color changes, but
because the speaker car es about what ’ s inside R uper t rather than whether it is r ed or
not. is affects the tr uth v alue of the utterance ev en though the color of the inside of
the apple is completely irr elev ant to whether R uper t is r ed. ( Cappelen & Lepor e ,
)
A ccor ding to C&L, if ‘kno ws ’ and ‘r ed ’ w er e genuinely context-sensitiv e, these R CSAs would
sound per fectly ne. B ut they don ’ t, ev en though they parallel exactly N . S o, they take this
as prima facie evidence that ‘kno w ’ and ‘r ed ’ ar en ’ t context-sensitiv e.
I believ e that this line of argument is v er y unsatisfactor y . F or take K R . I s
it ob vious that it ’ s defectiv e? I am not sur e. I n fact, it wouldn ’ t surprise me at all if a gr oup
of sincer e, r eectiv e, competent E nglish speakers nd it unpr oblematic or , at least, ar e unsur e
about it (ar en ’ t cases of this kind pr ecisely the ones on which one would normally expect the
intuitions of philosophers to differ?). S o, it seems that C&L ’ s appeal to K R is
operating on the unjustied assumption that if a is context-sensitiv e, then it should be ob vious
that it ’ s so to competent speakers in vir tue of their being competent with the language. Clearly ,
this is a consequence of their main implicit thesis that context-sensitivity is ultimately Kaplanian
indexicality . I will say mor e about this in the sections to come. ¹⁸
I n the mean time, let ’ s addr ess the follo wing question: Why do C&L think of this test as
an adequate tool for identifying context-sensitivity? B ecause it ’ s build ar ound a central featur e
of all context-sensitiv e expr essions, namely , their ability to be literally used to asser t differ ent
contents fr om one context of utterance to another . As they put it:
I t is a constitutiv e mar k of a context-sensitiv e expr ession e that it can be used with
differ ent extensions (semantic v alues) in differ ent contexts of utterance. […] I t follo ws
fr om this constitutiv e fact alone that for any context-sensitiv e expr essione our use ofe
in this context […] with whatev er extension it takes on in this context need not be the
same as whatev er extension it takes on in another context. ( Cappelen & Lepor e ,
)
S u ppose C&L ar e right. Let ’ s assume that the ICD/R CSA T est is a genuine test for context-
sensitivity . D oes it pr o vide evidence against (M oderate) C ontextualism and in fav or of S emantic
M inimalism? I don ’ t think so . F or consider the follo wing R CSA for ‘ tall ’:
T
Although S tev e is just y ears old, he is alr eady ′
′′
. H e is tall . S o, when a G uinness
W orld R ecor ds r epr esentativ e came to to wn, S tev e ’ s friends thought it would be a good
idea for S tev e to apply for the tallest--y ear-old-person r ecor d. H o w ev er , as soon as
they told him their idea, he r eplied: “ Ar e y ou crazy? I n or der to hav e a chance, I’ d hav e
¹⁸ S ee in par ticular sections . and . of this chapter .
to be at least ′
′′
tall. ose guys ar e huge! I ’ m not tall . ” H e was right; although S tev e
is tall, compar ed to the other applicants, he is not.
As far as I can tell, the abo v e stor y is a legitimate R CSA. And if I am right, w e can constr uct
the corr esponding ICD:
() er e ar e (or can be) false utterances of ‘S tev e is tall ’ ev en though S tev e is tall.
O r consider the follo wing R CSA for a possessiv e constr uction:
H ’
R osemar y is sho wing her old photo album to her friend Alicia. ey ar e looking at
pictur es of their childhood and having a conv ersation about the good old times. S ud-
denly , they come acr oss H ar old ’ s pictur e . H ar old is almost unr ecognizable: he looks
much y ounger , slimmer , and w ears a funny mustache. U nsur e of whom he was, Alice
asks: “I s that H ar old ’ s pictur e ?” R osemar y answ ers: “O h no . at ’ s not H ar old ’ s pictur e .
at ’ s my pictur e. H ar old must hav e his o wn copy of it ”. S he was right. at wasn ’ t
H ar old ’ s pictur e . H e had his o wn copy .
Again, I don ’ t see any pr oblem with the R CSA abo v e or with its ICD:
() er e ar e (or can be) false utterances of ‘at ’ s H ar old ’ s pictur e ’ [pointing ata ]
ev en though that ’ s H ar old ’ s pictur e [pointing at a ].
H o w e v er , neither gradable adjectiv es like ‘ tall ’ nor possessiv e constr uctions like ‘H ar old ’ s pic-
tur e ’ ar e members of C&L ’ s B asic S et. F ur ther , I suspect that w e could easily constr uct R CSAs
for many other terms that seem to be context-sensitiv e but that C&L ex clude fr om the B a-
sic S et, i.e., v ague terms, compound nominals (e.g., ‘ candy stor e ’), w eather r epor ts (e.g., ‘I t ’ s
raining ’), time r epor ts (e.g., ‘I t ’ s thr ee o ’clock ’), and quantiers among others.
B ut if this is so, why do K R and R R sound odd to many competent
E nglish speakers? M y guess is that they sound odd to them, because they ar e theor etically
loaded. S keptical scenarios or scenarios in which the topic of conv ersation is the necessar y
conditions for the application of the pr edicate ‘r ed ’ ar e less common than some philosophers
seem to think. is seems to be par ticularly clear in the case of K R . C&L ’ s
discussion of the R CSA T est intr oduces as the only r equir ement for considering an expr ession
a to be semantically context-sensitiv e to be able to constr uct an R CSA for it. B ut, they don ’ t
say what it takes for a giv en R CSA to be good enough.
I f they r equir e the R CSA to be simply acceptable, then K R fails to pr o vide
evidence in fav or or against M inimalism about ‘kno w ’, for it ’ s acceptable to some competent
speakers, e.g., Contextualists about ‘kno w ’. I f C&L r equir e the R CSA to be clearly and ob vi-
ously non-defectiv e, then K R seems to pr o vide evidence in fav or of M inimalism
about ‘kno w ’. B ut, why should anybody accept such a str ong standar d for testing context-
sensitivity? As I said befor e, their assumption is that if a is context-sensitiv e, it should be
ob vious that it is. I n any ev ent, since K R isn ’ t clearly and ob viously defectiv e ei-
ther (as it should be if C&L ’ s assumption in question w er e tr ue) appealing to it doesn ’ t help at
all to C&L ’ s defense of SM.
P erhaps R R is less contr o v ersial. H o w ev er , unless C&L giv e an argument as to why
R R , or any R CSA for that matter , should be taken mor e seriously than the arguments
the Contextualist about v ague pr edicates such as ‘r ed ’ offers, it ’ s unclear what to make of it. I n
fact, the contextualist about v ague pr edicates might ev en agr ee with C&L that R R
sounds odd, but claim at the same time that it fails to test the par ticular way in which ‘r ed ’ is
context-sensitiv e. F ur ther , she might suggest that the ar ticiality of R R is what causes
our tendency to r eject it. I f this is corr ect, then w e should be able to build mor e compelling
cases. P erhaps, this will do:
R R II
R uper t is a r edhead. When he visited a small tribe in the Amaz on rainfor est, people
ther e w er e amaz ed to see a r edhead for the rst time in their liv es. I n fact, one girl was
so confused that she asked him: ‘I s the color of y our hair r ed?’ R uper t answ er ed: ‘ Y es;
my hair is r ed ’ and what he said was tr ue. N onetheless, strictly speaking, R uper t ’ s hair
isn ’ t r ed, but orange.
I f these obser v ations ar e on the right track, then many expr essions C&L ex clude fr om the
B asic S et pass the ICD/R CSA T est. H ence, if the ICD/R CSA T est is a corr ect test for context-
sensitivity , then it follo ws that C&L ’ s B asic S et is just a pr oper subset of the set of context-
sensitiv e expr essions ther e ar e in E nglish. Alternativ ely , if the B asic S et exhausts all the genuine
context-sensitiv e expr essions in E nglish, then the ICD/R CSA T est isn ’ t an effectiv e tool for
detecting context-sensitivity . S ince accor ding to C&L ’ s S emantic M inimalism (i) ther e ar e no
context-sensitiv e expr essions that ar en ’ t members of the B asic S et, ¹⁹ and (ii) the ICD/R CSA T est
is an effectiv e tool for detecting context-sensitivity , C&L ’ s S emantic M inimalism is inconsistent.
A t this point one might wonder: what motiv ates C&L ’ s condence that the ICD/R CSA
T est will gather evidence in fav or of S emantic M inimalism and against Contextualism? Why
ar e they optimistic that the Contextualist will be unable to build R CSAs for expr essions that
ar en ’ t members of the B asic S et? I think that the r eason for this is basically the same I hav e
discussed in r elation to the pr evious tests: C&L assume that context-sensitivity is ultimately
Kaplanian indexicality .
O n Kaplan ’ s pictur e, ²⁰ indexicals ar e expr ession-types, not expr ession-tokens. e meaning
of an indexical has two basic aspects: content and char acter . I n his formal system, Kaplan r ep-
r esents the content of an indexicali as a function fr om cir cumstances of ev aluation (world-time
pairs) to extensions, and the character of i as a function fr om contexts (quadr uples of speaker ,
time, location, and possible world) to contents. is implies that if ther e ar en occurr ences o f
a giv en indexical term i in a sentence or a discourse, then ev er y occurr ence of i has the same
character . H ence, it follo ws fr om this that if the sentence or discourse is ev aluated r elativ e to
one context c , then all occurr ences of i will get the same content. N o w , if y ou ar e a S emantic
M inimalist and r educe semantic context-sensitivity to indexicality a là Kaplan, it follo ws that
if ther e ar e n occurr ences of a context-sensitiv e expr ession a in an R CSA, they will all get the
exact same content.
¹⁹ E x cept for , perhaps, a fe w contextuals.
²⁰ F or details, see chapter .
H o w ev er , most context-sensitiv e expr essions in E nglish (and in other natural languages)
don ’ t wor k like this, as the examples I hav e giv en attempt to sho w . I n fact, not ev en demon-
strativ es always wor k as described. F or consider the follo wing sentence:
() at is heavier than that.
S u ppose that a speaker asser tiv ely utters ( ) in a r oom in which ther e ar e only two salient things,
say , a feather and an anvil. I f w e literally (mis)apply Kaplan ’ s account to ( ), then since both
occurr ences of ‘ that ’ hav e the same character , both will g et the same content. B ut this implies
that ( ) is necessarily false; nothing is heavier than itself . O b viously this is a misapplication of
Kaplan ’ s logic of demonstrativ es to or dinar y E nglish; a misapplication that seems to be at the
hear t of the ICD/R CSA T est and ev en of SM itself .
Kaplan ’ s account of demonstrativ es doesn ’ t apply to E nglish demonstrativ es, but to their
r egimentation using thedthat operator . O n his account, occurr ences of E nglish demonstrativ es
ar e formally r epr esented b y means of thedthat operator plus a description someho w abstracted
fr om the context in the pr ocess of r egimentation. S ince differ ent occurr ences of dthat terms
can be supplemented with differ ent descriptions in Kaplan ’ s formal r epr esentations, differ ent
dthat terms corr esponding to differ ent occurr ences of the same E nglish demonstrativ e can be
assigned differ ent contents in the same context. S o, the alleged pr oblem that ( ) poses doesn ’ t
ev en appear in Kaplan ’ s formal system; for it will r epr esented r oughly as⌜dthat[ the x : Fx]
is heavier than dthat[ the y : Gy]⌝ . Kaplan ’ s main goal in D emonstr ativ es is to giv e a system
of logic for indexicals, not a semantics for context-sensitiv e expr essions in or dinar y E nglish.
ese enterprises ar e differ ent: while the former foc uses on languages whose grammars ar e
mathematically stipulated with the purpose of studying logical tr uth and logical consequence,
the latter focuses on languages whose grammars ar e disco v er ed, rather than stipulated, with the
purpose of explaining empirical facts about meaning.
I n sum, C&L claim that no semantic theor y should classify as context-sensitiv e any expr es-
sion that doesn ’ t pass the thr ee tests w e hav e discussed, on pain of being empirically inadequate.
ey think that one of SM’ s main vir tues is that it claims that only expr essions that pass these
tests ar e semantically context-sensitiv e. H o w ev er , I hav e argued that C&L ’ s tests ar e tests for de-
tecting Kaplanian indexicality , not semantic context-sensitivity . S ince C&L fail to justify SM’ s
key assumption, i.e., that all semantic context-sensitivity is ultimately Kaplanian indexicality ,
SM loses one of its main selling points.
What about the other claims defended b y SM? I sn ’ t it tr ue that competent speakers can
understand a declarativ e sentenceS without having a clue aboutS ’ s context of utterance? And
isn ’ t this possible because two of SM’ s main theses ar e right, namely , (i) that the semantic content
ofS is the minimal non-context-sensitiv e complete pr oposition thatS expr esses in ev er y context
in whichS is asser tiv ely utter ed, and (ii) that competent speakers ar e able to grasp that minimal
semantic content? I n the next section, I will examine these claims.
. S emantic E v aluation under Conditions of I gnorance
O ne of the central ideas behind SM is this: semantic content is sharply distinct fr om speech act
content. C&L argue that the sour ce of both M oderate and Radical Contextualism is the failur e
to make this distinction. is failur e motiv ates what C&L call the ‘M istaken Assumption ’:
(MA) A theor y of semantic content is adequate just in case it accounts for all o r most
of the intuitions speakers hav e about speech act content, i.e., intuitions about
what speakers say , asser t, claim, and state b y uttering sentences. ( Cappelen &
Lepor e , )
C&L argue that acceptance of (MA) implies that Radical Contextualism is tr ue. eir idea
might be r econstr ucted as follo ws:
(P
1
) F or ev er y declarativ e sentence S and context of utterance c , what speakers say ,
asser t, claim, and state b y utteringS v aries accor ding toc .
(P
2
) Competent speakers hav e r eliable intuitions about what speakers say , asser t, claim,
and state b y utteringS in a giv en contextc
′
.
(P
3
) MA
(P
4
) I f competent speakers hav e the intuition that pr oposition p is what speakers say ,
asser t, claim, and state b y uttering S in c
′
, then an adequate theor y of semantic
content will assign p as the semantic content ofS inc
′
.
(C) e semantic content that an adequate theor y of semantic content will assign to
S will v ar y accor ding toc .
H ence, acceptance of (MA) plus some r easonable assumptions about speech act content would
commit us to the claim that semantic context-sensitivity is ubiquitous. S o, C&L deny (MA)
(and, thus, also (P
4
) ). M or eo v er , they claim that ther e is no close and immediate connection
betw een semantic content and speech act content : although competent speakers hav e immediate
access to speech act content, they don ’ t hav e the same kind of access to semantic content.
B ut then, ho w do w e get at the semantic content of a sentence S ? C&L ’ s answ er is this: b y
interpr etingS under conditions of ignor ance . I n fact, this answ er is implicit on the v er y idea that
motiv ates SM:
e idea motiv ating S emantic M inimalism is simple and ob vious: e semantic content
of a sentence S is the content that all utterances of S shar e. I t is the content that all
utterances ofS expr ess no matter ho w differ ent their contexts of utterance ar e. I t is also
the content that can be gr asped and r epor ted b y someone who is ignor ant about the r elev ant
char acteristics of the context in which an utter ance of S took place. ( Cappelen & Lepor e
( , ), my emphasis)
S o, C&L spell out the general strategy for getting at the semantic content ofS as follo ws:
(i) ink about ho w to interpr et utterances ofS under conditions of ignorance, i.e.
when y ou kno w little or nothing about the context of utterance. at will for ce
y ou to r emo v e ‘ contextual inuence ’.
(ii) ink about collectiv e r epor ts: what utterances of S acr oss many contexts of ut-
terance shar e; this is r eected in what w e call collectiv e r epor ts, i.e. r epor ts of the
form ‘ey all said that p ’, when ‘ they ’ r efers to speakers v ariously situated who
hav e all utter edS . ( Cappelen & Lepor e a , )
S teps (i) and (ii) ar e supposed to wor k, because C&L hold the follo wing theses (or a close
v ersion of them):
(a) U tterances of S in differ ent contexts often succeed in conv eying mor e than one
complete pr oposition (the speech act content).
(b) e semantic content of S is the complete pr oposition that ev er y utterance of S
conv eys in any giv en context.
(c) Competent speakers who understand S hav e immediate access to some pr oposi-
tions conv ey ed b y a giv en utterance ofS .
(d) O ne of the pr opositions any competent speaker who understands S hav e access
to is the semantic content ofS .
H o w ev er , ther e ar e at least thr ee two major pr oblems C&L hav e to face in or der for their strategy
to wor k: (a) the pr oblem of sentences utterances of which asser t multiple pr opositions in ev er y
context, (b) the pr oblem of sentences utterances of which don ’ t always asser t their semantic
contents, and (c) the pr oblem of ambiguity and v agueness. Let me elaborate a little bit on these
pr oblems.
.. Asser ting M ultiple P r opositions
C&L want to defend these thr ee claims:
(a) Competent speakers who understand S hav e immediate and r eliable intuitions
aboutS ’ s speech act content.
(b) Competent speakers who understand S don ’ t hav e immediate and r eliable intu-
itions aboutS ’ s semantic content.
(c) S ’ s semantic content is a pr oper par t ofS ’ s speech act content.
A ccor ding to C&L, thinking about interpr etations of S under conditions of ignorance would
help us identify the semantic content of S fr om the set of pr opositions that a giv en utterance
of S would conv ey in a giv en context. H o w ev er , suppose that after car efully follo wing C&L ’ s
pr ocedur e, w e end up with mor e than one pr oposition, i.e., after thinking about S out of any
context w e disco v er that utterances ofS asser t mor e than one pr oposition acr oss contexts. What
should w e say about the semantic content of S ? S hould w e say that S semantically expr esses
mor e than one pr oposition? S hould w e say that S expr esses the conjunction of those pr oposi-
tions? I f w e adopt the former suggestion, then w e hav e to conclude that SM’ s methodological
ignor ance is aw ed and that one of SM’ s central thesis—i.e., the thesis that the semantic con-
tent ofS is the complete pr oposition that ev er y utterance ofS conv eys in any giv en context—is
false. O n the other hand, the latter suggestion will not wor k in many cases. F or consider the
follo wing sentences:
() S no w is white.
() I t ’ s tr ue that sno w is white.
Arguably , the semantic contents expr essed b y ( ) and ( ) differ , ev en though an asser tiv e
utterance of either of them always asser ts both pr opositions. O r compar e ( ) with ( ),
() A ctually , sno w is white.
e pr opositions that ( ) and ( ) semantically expr ess hav e differ ent modal and epistemic
pr oles, ²¹ ev en though an asser tiv e utterance of one of them always asser ts the pr opositions
that both sentences expr ess. I f these obser v ations ar e corr ect, then w e hav e sufficient r easons to
r e ject SM. O r take F r ege ’ s famous example: ²²
²¹ S ee S oames ( ). F or example, S oames argues that while the pr oposition expr essed b y a sentence of the
form ‘S iffS ’ is necessar y a priori, the pr oposition expr essed b y a sentence of the form ‘S iff actuallyS ’ is contingent
a priori.
²² S ee F r ege ( ).
() N apoleon, who r ecogniz ed the danger to his right ank, himself led his guar ds
against the enemy position.
F r ege ’ s o wn analysis of ( ) is that it expr esses two pr opositions ( thoughts ), i.e. the pr opositions
expr essed b y the follo wing sentences:
() a. N apoleon r ecogniz ed the danger to his right ank.
b . N apoleon himself led his guar ds against the enemy position.
I f F r ege is right, then w e can conclude that SM is false. H o w ev er , ther e ar e two ways C&L might
r espond: (i) they could say that ( ) s emantically expr esses the conjunction of the pr opositions
expr essed b y ( a ) and ( b ), or (ii) they could r estrict SM to simple declarativ e sentences.
e pr oblem with the rst option (which is the one F r ege fav ors) is that it entails that if the
subor dinate clause is false, the whole sentence would be false. H o w ev er , it isn ’ t clear that this is
so . F or consider the follo wing sentence:
() N apoleon, who was the last F r ench monar ch, himself led his guar ds against the
enemy position.
e F r egean analysis of ( ) in question would take the pr oposition expr essed b y it as the
conjunction of the pr opositions expr essed b y ( a ) and ( b ):
() a. N apoleon was the last F r ench monar ch.
b . N apoleon himself led his guar ds against the enemy position.
As it ’ s w ell-kno wn, the pr oposition expr essed b y ( a ) is false. H o w ev er , many competent
speakers would take the tr uth-v alue of ( ) to be tr ue if the pr oposition expr essed b y ( b ) is
tr ue. I n fact, when ev aluating the tr uth-v alue of ( ) many competent speakers will not car e
about the tr uth-v alue of ( a ). S imilarly for the negation of ( ):
() I t is not the case that N apoleon, who was the last F r ench monar ch, himself led
his guar ds against the enemy position.
M any (if not most) competent speakers would take ( ) as negating the main clause ( b ), not
the subor dinate one ( a ). H ence, they would take the tr uth-v alue of ( ) as depending on the
tr uth-v alue of ( b ) alone. O r consider the follo wing:
() I f N apoleon, who was the last F r ench monar ch, himself led his guar ds against
the enemy position, then N apoleon was the last F r ench monar ch.
I f the F r egean analysis w er e right, ( ) would be necessarily tr ue. H o w ev er , it ’ s doubtful that it
is so . F or compar e it with this:
() I f N apoleon, who was the last F r ench monar ch, himself led his guar ds against
the enemy position, then N apoleon himself led his guar ds against the enemy
position.
Arguably , many competent speakers would take this sentence as expr essing a necessar y tr uth,
ev en though the subor dinate cl ause is false. W e can complicate the pictur e ev en mor e if w e
intr oduce pr opositional attitude contexts. F or example,
() Ralph believ es that J ohn L. A ustin, who was an inuential Cambridge philoso-
pher , coined the term ‘illocutionar y act ’.
I ntuitiv ely , if Ralph in fact believ es that J ohn L. A ustin coined the term ‘illocutionar y act ’, then
( ) is tr ue, ev en though A ustin was not a Cambridge philos opher . B ut if this is so, then things
ar e mor e complicated than C&L seem to ackno wledge.
S ur ely , nothing I am saying her e is conclusiv e. H o w ev er , it ’ s unclear to me that SM—as
C&L pr esent it—is able to giv e us a satisfactor y account of these cases, especially because C&L
want to appeal to untutor ed intuitions of speakers who ignor e the r elev ant contexts of utter-
ance to determine the semantic contents of declarativ e sentences. S ince, pr esumably , intuitions
about the semantic contents of the sentences consider ed in this section can signicantly v ar y
fr om speaker to speaker , it seems r easonable to doubt (to say the least) SM’ s theses. And ev en if
SM had a way to av oid the pr oblem discussed in this section, it would hav e to face the follo wing
pr oblem: some simple declarativ e sentences ar e used to asser t a set of pr opositions (speech act
content) none of which can be taken to be their semantic contents. I f these cases ar e possible,
then C&L hav e a pr oblem. Let ’ s r evie w this issue.
.. N on-Asser ted S emantic Contents
Asser ting is a speech act. I t ’ s a kind of o v er t action that speakers per form in conv ersational
contexts b y uttering declarativ e sentences. What is asser ted, on the other hand, is typically
a pr oposition. Although one might agr ee that the default option for what is asser ted is the
pr oposition semantically expr essed b y the sentence utter ed, the mechanisms and norms that
go v ern our conv ersational practices sometimes make the asser ted content div erge fr om the de-
fault option. T o illustrate this point, consider the follo wing example of the so-called ‘ deferr ed
r efer ence, ’ ²³
() e ham sandwich is at T able .
I n a communicativ e situation in which ( ) is utter ed b y a waiter A to a co-wor ker , A can
succeed in asser ting the pr oposition expr essed b y ( ):
() e ham sandwich or der er is at T able .
B ut clearly , the semantic contents of ( ) and ( ) ar e differ ent. C&L would call the pr oposi-
tion expr essed b y ( ) ‘ the speech act content ’ of ( ) r elativ e to the context in question. N o w ,
the question is this: is the semantic content of ( ) among the pr opositions asser ted b y waiter
A in the context described? I f so, what r easons suppor t this claim?
er e is r eason to believ e that the semantic content of a sentence S isn ’ t always among
the pr opositions asser ted b y an utterance ofS . is happens, for example, when the semantic
content of S is p , and ther e is a conv ersational implicatur e that the speaker can ’ t be pr operly
understood as intending to asser t p , as the utterance of ( ) in question seems to sho w . I r ony
and metaphor ar e the most ob vious cases of this phenomenon.
²³ S ee N unberg ( ).
B ut this phenomenon doesn ’ t mer ely occur in non-literal utterances of sentences. I t can
ev en happen when speakers ar e speaking literally . T o illustrate this point, consider the follo wing
sentence: ²⁴
() I hav e often wonder ed whether the I ranian diplomat w e met last month was a
diplomat at all.
I n a communicativ e situation in which ( ) is asser tiv ely utter ed b y a sincer e, r eectiv e, com-
petent speaker , she will normally asser t that she has often wonder ed, of an individuali she and
her interlocutor(s) met the pr evious month, whetheri was a diplomat. I n such a case, she might
be using the denite description ‘ the I ranian diplomat ’ to get the hear er to r emember a cer tain
person who, at the time, they thought was an I ranian diplomat. H ence, absent a commitment
to the pr oposition semantically expr essed b y ( ), it seems r easonable to think that the speaker
didn ’ t asser t the semantic content of ( ). I f this is so, then one of SM’ s main theses is false;
it isn ’ t the case that one of the pr opositions asser ted b y any successful asser tiv e utterance of a
declarativ e sentenceS in a contextc isS ’ s semantic content.
.. Ambiguity And V agueness
Another serious pr oblem that C&L ’ s theor y has to face is this: many sentences ar e such that
if w e w er e pr esented with their utterances and w e didn ’ t kno w anything about the context, w e
²⁴ is is a slightly modied example borr o w ed fr om S oames ( , –). S ee also footnote on pp .
–.
would be unable to identify one pr oposition asser ted b y them. S uch cases include sentences
that contain some sor t of ambiguity or v agueness. Consider the follo wing sentence: ²⁵
() P eter ’ s bat is gr ey .
S e ntence ( ) poses thr ee pr oblems to C&L ’ s strategy for getting at the semantic content of
a sentence: (a) ‘bat ’ is ambiguous and so depending on its disambiguated meaning ( ) will
expr ess differ ent pr opositions, (b) ‘ gr ey ’ is v ague and so depending on its pr ecisication ()
will expr ess differ ent pr opositions, and (c) the possessiv e constr uction ‘P eter ’ s bat ’ is under-
determined, i.e. it will pick out, depending on the context of utterance, differ ent r elations
betw een P eter and a bat. T aken out of context, competent speakers will be unable to identify
one pr oposition expr essed b y ( ).
e same point holds for sentences that contain quantiers such as ‘ many ’ and ‘ fe w ’. F or
example, consider the follo wing: ²⁶
() M any men date many women.
() F e w of J ohn ’ s fe w friends ar e still ar ound.
S u ppose that w e utter ( ) in a context in which the topic of conv ersation ar e the dating habits
of American men. A normal r eectiv e competent conv ersational par ticipant might expect the
rst occurr ence of ‘ many ’ to pick out millions of men, and the second to pick out, perhaps,
six or sev en women. S imilarly for ( ): it seems per fectly possible to asser tiv ely utter ( ) in
²⁵ I borr o w this sentence fr om B ach ( ).
²⁶ E xamples borr o w ed fr om P ar tee ( ).
a context in which each occurr ence of ‘ fe w ’ picks out a differ ent number of individuals. I f
this is so, then taken out of context, it ’ s unclear whether ( ) and ( ) expr esses each a unique
complete pr oposition.
O ne way one might tr y to r espond to this worr y is b y r estricting one ’ s semantic theor y to
a disambiguated, pr ecisied v ersion of E nglish. I n fact, this might be what C&L had in mind
when they described in detail the steps one should follo w to determine the semantic content of
a giv en sentenceS :
(a) S pecify the meaning (or semantic v alue) of ev er y expr ession in S (doing so in
accor dance with y our fav orite semantic theor y).
(b) S pecify all the r elev ant compositional meaning r ules for E nglish (doing so in ac-
cor dance with y our fav orite semantic theor y).
(c) D isambiguate ev er y ambiguous/polysemous expr ession inS .
(d) P r ecisify ev er y v ague expr ession inS .
(e) F ix the semantic v alue of ev er y context-sensitiv e expr ession in S . ( Cappelen &
Lepor e , –)
H o w ev er , this alternativ e isn ’ t av ailable to C&L. F or the disambiguation and pr ecisication of
the examples giv en abo v e r equir e a pr evious kno wledge of the context of utterance. M or eo v er ,
the pr ocedur e was supposed to tell us what the semantic content is. B ut the disambiguation
and pr ecisication via (a)–(e) seems to r equir e that this has alr eady been settled. ese r easons,
I take it, undermine the v er y idea that motiv ates SM, namely , that the semantic content of
S is the minimal complete pr oposition ‘ that can be grasped and r epor ted b y someone who is
ignorant about the r elev ant characteristics of the context in which an utterance ofS took place ’
( Cappelen & Lepor e , ).
. Context-S ensitivity vs. I ndexicality
A central thesis defended in I nsensitiv e S emantics is that semantic context-sensitivity is ob vious:
I f an expr essione has its semantic v alue x ed in a context of utterance, that had better
be ob vious to all of us. Context-sensitivity can ’ t be some obscur e phenomenon that
y ou need to r ead scholarly books and ar ticles about in or der to r ecogniz e and mas-
ter . Context-sensitivity is a sur face phenomenon. E v er y speaker kno ws it when he ’ s
confr onted with it; and he kno ws that ev er y other competent speaker of his language
kno ws it as w ell, and all speakers kno w ho w to exploit context-sensitivity in the heat of
a conv ersation. N one of this should come as a surprise. I f y ou ask a speaker of some
par ticular expr ession whether or not it ’ s context-sensitiv e, she should be able to tell y ou
right away that it is, and ho w it is, context-sensitiv e. ( Cappelen & Lepor e , )
What ar e those ob vious context-sensitiv e expr essions? A ccor ding to C&L, they ar e the Kapla-
nian indexicals. I n fact, right at the beginning of the book, they say: ‘O ne central goal of this
book is to defend the uncontaminated intuitions that underlie Kaplan ’ s methodology fr om a
wide range of popular objections ’ ( Cappelen & Lepor e ( , ), my emphasis).
P r esumably , then, C&L think that Contextualists (i) ar en ’ t competent speakers of E nglish,
or (ii) ar e insincer e, or (iii) hav e lost their uncontaminated intuitions. H ypotheses (i) and (ii) ar e
clearly implausible. S o, perhaps they hold (iii). H o w ev er , I doubt that (iii) holds water . F irst,
it ’ s doubtful that competent speakers with or without some lev el of philosophical sophistication
hav e uncontaminated intuitions about context-sensitivity and ar e ‘ able to tell y ou right away ’
whether a giv en expr ession is context-sensitiv e or not. S econd, ev en if they had those uncon-
taminated intuitions it ’ s unclear whether those intuitions would infallibly and r eliably identify
ev er y semantically context-sensitiv e expr ession ther e is. C&L don ’ t offer any argument as to
why w e should adopt any of these assumptions. S o, in the absence of such an argument, they
r emain unjustied. P ut differ ently , C&L hold the follo wing set of claims:
. I n general, speakers don ’ t hav e r eliable or uncontaminated semantic intuitions.
. H o w ev er , speakers hav e r eliable and uncontaminated intuitions about context-
sensitivity .
. Context-sensitivity is a semantic phenomenon.
H o w ev er , C&L don ’ t giv e any sound argument as to why context-sensitivity intuitions hav e a
special status among other semantic intuitions that competent speakers hav e. B ut C&L o w e us
an explanation. F or ev en if they w er e right that context-sensitivity is ob vious to ev er y competent
speaker , they should not assume that the r eason this is ob vious is also ob vious. E xplaining this
(alleged) fact is par ticularly impor tant for their pr oject because they model ev er y semantically
context-sensitiv e expr ession ther e is in E nglish after the (ob vious) Kaplanian indexicals. M or e
specically , their theor y operates on the follo wing assumption:
(A) A linguistic expr essiona is semantically context-sensitiv e iffa has the semantic
pr oper ties that the Kaplanian indexicals hav e.
is assumption explains why C&L fail to consider a hypothesis that is mor e plausible than
(i)-(iii) about why many philosophers and linguists consider expr essions such as ‘r ed ’, ‘ tall ’,
‘rich ’, or ‘brick factor y ’ as context-sensitiv e: er e is mor e to semantic context-sensitivity than
mer e Kaplanian indexicality . I f this hypothesis is corr ect, then ev en if C&L ’ s arguments against
Contextualism w er e sound, the most they would sho w is that the expr essions the M oderate
Contextualist characteriz es as semantically context-sensitiv e ar en ’ t Kaplan indexical. B ut, unless
C&L offer good r easons in fav or of the claim that context-sensitivity is ultimately Kaplanian
indexicality , the M oderate Contextualist might ev en agr ee with C&L ’ s r esults without ther eb y
abandoning her semantic vie ws.
I n this chapter I hav e argued that w e should r eject (A). I hav e done so b y sho wing that
the two main argumentativ e strategies C&L emplo y in defense of SM (i.e., the argument b y
elimination and the argument fr om explanator y for ce ) ar e aw ed. I n addition, I hav e argued
that the central methodological assumption behind SM, namely , that competent speakers hav e
immediate and r eliable intuitions about semantic context-sensitivity , is false. F inally , I hav e
suggested that the elements in play that determine whether a giv en expr ession a is context-
sensitiv e or not may v ar y fr om expr ession to expr ession and that, for this r eason alone, any
test that attempts to tr eat them all equally is doomed for failur e. B ut independently of the
r easons offer ed her e against SM and, in par ticular , against assumption (A), it should be rather
clear why they ar e highly implausible: ther e is mor e than one legitimate notion of context
and—one may naturally infer—mor e than one kind of context-sensitivity . I f this is corr ect,
then a semantic theor y should be v er y car eful about distinguishing the sor ts of intuitions that
competent speakers hav e about the informational content a linguistic expr ession a conv eys in
c , on pain of making false generalizations and wr ong pr edictions about i ts semantic content.
And this is pr ecisely what C&L fail to do . S o, w e had better check mor e car efully the linguistic
constr uctions they quickly classify as semantically context-insensitiv e, i.e., quantiers, gradable
adjectiv es, v ague pr edicates, possessiv e constr uctions, and kno wledge attributions. B efor e I end
this last section of the chapter , let me say a fe w wor ds about what I hav e in mind.
er e ar e at least thr ee commonly used notions of context. F irst, ther e is the notion of con-
text as spatio-temporal setting in which a giv en utterance or conv ersation takes place. is kind
of context has indenitely many featur es none of which ar e linguistic or depend on the r ules
and conv entions of the language used b y the speaker or conv ersational par ticipants. Let ’ s call
this kind of context ‘e S patio-T emporal S etting ’. S econd, ther e is the notion of context as
communicativ e situation. is kind of context includes only the featur es of the spatio-temporal
setting that ar e r elev ant to the purposes of the conv ersation as w ell as the pr esumed common
backgr ound kno wledge, shar ed inter ests, mutual expectations, etc. of the conv ersational par-
ticipants. Let ’ s call this kind of context ‘e Communicativ e S ituation ’. ir d, ther e is the
formal notion of context as set of elements other than meaning (or character) r equir ed to fully
determine the semantic content of an indexical sentence. A ccor ding to this notion, contexts
ar e n -tuples of parameters, whole unique r ole is to pr o vide the contents that the characters of
indexicals will pick out. Let ’ s call this kind of context ‘e I ndex ’. ese differ ent kinds of
context hav e differ ent pr oper ties. T o mention a fe w:
. While the index has all of its elements essentially—i.e., if y ou change one of them,
y ou change the index—intuitiv ely , spatio-temporal settings hav e contingent ele-
ments and featur es.
. While the communicativ e situation is dynamic and normally ev olv es in the course
of a conv ersation as a r esult of the conv ersation, the spatio-temporal setting isn ’ t
r elev antly alter ed b y a normal conv ersation.
. While the index doesn ’ t captur e salience r elations, they ar e essential to commu-
nicativ e situations.
S o, our hypothesis is this: depending on which kind of context y ou take into consideration, y ou
will get differ ent r esults with r egar ds to what expr essions count as context-sen sitiv e. Although
not all of them might be equally semantically r elev ant, which ones ar e isn ’ t something that
can be decided pr e-theor etically . I t ’ s something that should be decided b y a semantic theor y
accor ding to what it aims at explaining. S ince semantics is a highly theor etical enterprise, w e
should not expect speakers to hav e crisp intuitions about ev er y kind of context-sensitivity ther e
is in the language they use.
Chapter P r enominal P ossessiv es
. r ee eses A bout S emantic Content
O ne of the central notions in semantics is that of semantic content. A ccor ding to the or tho-
do x pictur e in semantics, if a sentence S doesn ’ t contain any occurr ence of a context-sensitiv e
term, then its meaning just is its semantic content. H o w ev er , if S contains some occurr ence
of a context-sensitiv e expr ession, then its meaning will be a r ule that fully determines, for ev-
er y context, S ’ s semantic content. B ut, what is exactly semantic content? e follo wing ar e
commonly held theses about it:
S tr ong P r opositionalism
e semantic content of a declarativ e sentenceS (r elativ e toc ) is always pr opositional.
S tr ong Asser tiv e P r opositionalism
I f p is the semantic content of S (r elativ e to c ), then ev er y literal, implicatur e-fr ee
utterance ofS inc asser ts p .
S tr ong Linguistic D ependence
I f a declarativ e sentenceS is literally used to asser t (without conv ersational implicatur es)
a pr oposition p inc , then ev er y constituent of p is the semantic content (r elativ e toc )
of some constituent ofS .
e purpose of this chapter is to sho w that semantic theories committed to these theses hav e
r e al difficulties accounting for or dinar y uses of sentences containing pr enominal possessiv es. I n
addition, I would like to pr opose an alternativ e pictur e of the r elationship betw een linguistic
meaning and asser ted content that aims at o v er coming those difficulties.
. P ossessiv e R elations
er e ar e two main kinds of possessiv e constr uctions acr oss differ ent languages: v erbal posses-
siv es (‘J ohn has a book ’) and nominal possessiv es. E nglish nominal possessiv es hav e differ ent
forms:
. P r enominal P ossessiv e: J ohn ’ s book
. P ostnominal P ossessiv e: e book of J ohn
. D ouble G enitiv e: A book of J ohn ’ s
. P r edicativ e P ossessiv e: at book is J ohn ’ s
Let ’ s focus on pr enominal possessiv es. T wo characteristics ar e immediately appar ent: rst, al-
though pr enominal possessiv es such as ‘J ohn ’ s book ’, ‘J ohn ’ s mother ’ and ‘J ohn ’ s mountain ’ ex-
hibit the same grammatical str uctur e, the r elations they ar e used to conv ey betw een the nominal
possessor (the possessor) and the nominally possessed (the possessed) can v ar y signicantly fr om
context to context. I n fact, only a subset of these constr uctions conv eys a r elation of posses-
sion. F or consider the phrase ‘J ohn ’ s ex-girlfriend ’—men don ’ t o wn their girlfriends, let alone
their ex-girlfriends. O r consider ‘D iana ’ s tragic death ’, ‘Kripke ’ s W ittgenstein ’, ‘M ichelangelo ’ s
P ietà ’, ‘ po v er ty ’ s impact ’, ‘ today ’ s ne wspaper ’, and ‘ the city ’ s destr uction ’. I ntuitiv ely , none of
them conv eys a r elation of possession either .
S econd, the majority of pr enominal possessiv es in E nglish ar e such that a single possessiv e
constr uction can be legitimately used to conv ey differ ent r elations in differ ent contexts. T o
illustrate this point, consider the follo wing scenarios:
H J
Alice is talking to Bob about J ohn. S he is telling him ho w har dwor king J ohn is. Bob ,
who doesn ’ t kno w J ohn v er y w ell, asks Alice why she thinks so . S he tells Bob about the
hours and hours J ohn spent at the librar y writing his latest book. Alice nally adds:
() J ohn ’ s book is the pr oduct of ten y ears of r esear ch.
L J
Alice is talking to Bob about J ohn. S he is telling him ho w lucky J ohn is. Bob , who
doesn ’ t kno w J ohn v er y w ell, asks Alice why she thinks so . S he tells Bob about the
time when J ohn got so dr unk that he bought a raffle ticket instead of the bus ticket he
intended to buy to get home. F or tunately for him, he found a friend who took him
home and also won a book in the raffle. Alice nally adds:
( ) J ohn ’ s book is the pr oduct of ten y ears of r esear ch.
H er e w e hav e one sentence utter ed at two differ ent scenarios. I n the rst scenario, Alice asser ts
a singular pr oposition about a book J ohn author ed. I n the second scenario, Alice asser ts a
singular pr oposition about a book J ohn won in a raffle when he was dr unk. H o w should w e
account for the r elation betw een the linguistic meaning of ( ) and the pr opositions it is used to
asser t in these scenarios? I f y our theor y is committed to the conjunction of the theses described
in the pr evious section, ther e ar e thr ee w ell-kno wn strategies y ou can adopt:
(i) H old that sentences containing pr enominal possessiv es such as ( ) ar e inher ently
ambiguous.
(ii) H old that sentences containing pr enominal possessiv es such as ( ) semantically
encode complete pr opositions that nonetheless leav e the possessiv e r elation un-
specied.
(iii) H old that sentences containing pr enominal possessiv es such as ( ) ar e indexical.
I n the next section, I will examine these alternativ es.
. e Argument fr om M assiv e Ambiguity
A ccor ding to this alternativ e, declarativ e sentences containing pr enominal possessiv es seman-
tically encode each a multiplicity of specic pr opositions. H ence, when a speaker s asser tiv ely
utters S , s puts in play a set of pr opositions fr om which, pr esumably , the hear er—guided b y
pragmatic principles—will select the most r elev ant one in his context. us, on this vie w , the
possessiv e phrase ‘J ohn ’ s book ’ is ambiguous betw een ‘book written b y J ohn ’, ‘book bought b y
J ohn ’, ‘book r ead b y J ohn ’, ‘book about J ohn ’, ‘book checked out fr om the librar y b y J ohn ’,
‘book photographed b y J ohn ’, ‘book lo v ed b y J ohn ’, etc.
N otice that this appr oach obser v es the theses described in section . . I f pr enominal pos-
sessiv es expr ess multiple specic meanings , then sentences containing them expr ess multiple
specic pr opositions (S tr ong P r opositionalism), the constituents of which will be contributed
b y the contents encoded b y the constituents of the sentences utter ed (S tr ong Linguistic D epen-
dence). H o w ev er , of those pr opositions, only one of them will be asser ted b y an utterance of
any such sentence S at a giv en context based on the inter ests and expectations of the conv er-
sation in which it ’ s utter ed. at pr oposition will be the semantic content of S r elativ e to the
context of use (S tr ong Asser tiv e P r opositionalism).
I t should be clear that the Argument fr om M assiv e Ambiguity doesn ’ t adequately explain
the data. O n this account, the contextual v ariability of the pr opositions asser ted b y literal
utterances of a pr enominal possessiv e in differ ent occasions is explained at the lexical lev el. I t
claims that pr enominal possessiv es ar e massiv ely ambiguous betw een all the possessiv e r elations
they can be used to asser t. B ut, is this a genuine case of ambiguity? I don ’ t think so . N otice
that not ev er y piece of information a term t can be literally used to conv ey is a meaning of t .
I n or der for some piece of inform ationi to count as a meaning oft ,i has to be associated with
t as a r esult of a pr ocess of conv entionalization in the language.
Although this pr ocess of conv entionalization might v ar y fr om language to language, and
perhaps ev en fr om term to term, it ’ s clear thatt can ’ t hav e a meaning anybody or har dly anybody
has used t to conv ey . H o w ev er , this would be a consequence of taking pr enominal possessiv es
to be lexically ambiguous. F or giv en the appr opriate cir cumstances, backgr ound beliefs, and
pr esuppositions, a pr enominal possessiv e can be naturally used to asser t an unusual or ev en
no v el r elation betw een a possessor and a possessed. H ence, to say that pr enominal possessiv es
ar e lexically ambiguous is to say that the open-ended list of unusual and no v el r elations they
can be used to asser t ar e par t of their r espectiv e sets of meanings prior to any pr ocess of lin-
guistic conv entionalization or ev en use b y speakers of the language. U ndoubtedly , this sole
consequence makes the Argument fr om M assiv e Ambiguity untenable.
B ut the pr oblem doesn ’ t stop her e. O ne inter esting featur e of E nglish pr enominal posses-
siv es is that they can be iterated multiple times:
() J ohn ’ s father ’ s friend ’ s boat ’ s name is ‘D ay dr eams ’.
() A colleague ’ s wife ’ s sister ’ s son is in Los Angeles.
I f pr enominal possessiv e constr uctions w er e lexically ambiguous, then iterated pr enominal pos-
sessiv es such as the ones in ( ) and ( ) would also be lexically ambiguous. I n fact, the number
of ambiguities would multiply so rapidly , than soon they would become useless in conv ersa-
tion. B ut, giv en that competent E nglish speakers normally use and easily understand multiply
iterated pr enominal possessiv e constr uctions, w e hav e evidence that they ar en ’ t massiv ely am-
biguous.
. e Argument fr om E xistential G eneralization
e idea behind this pr oposal is that since the identity of the r elation that holds betw een the
possessor and the possessed could be any of a considerably big set of r elations, the corr ect
analysis of the meaning of a pr enominal possessiv e should be mer ely existential:
() J J ohn ’ s bookK = lx[ Book(x)^9R[R( J ohn,x)]]
O n this pictur e, sentences containing pr enominal possessiv es expr ess complete pr opositions
that quantify o v er possible r elations that might hold betw een the possessor and the possessed.
us, instead of explaining the v ariability of r elations that can be asser ted b y a literal use of a
pr enominal possessiv e b y appealing to massiv e ambiguity , this account deals with it b y assigning
just one semantic content to each pr enominal possessiv e that is to be instantiated b y differ ent
r elations in differ ent contexts.
e Argument fr om E xistential G eneralization clearly obser v es S tr ong P r opositionalism:
the semantic content of a sentence containing a pr enominal possessiv e is always a pr oposition;
mor e pr ecisely , it ’ s an existentially quantied pr oposition. I t ’ s also compatible with S tr ong As-
ser tiv e P r opositionalism and S tr ong Linguistic D ependence. O n the one hand, a commitment
to the former entails a commitment to a multiple-pr oposition vie w of asser tion, i.e., b y uttering
a sentence containing a pr enominal possessiv e a speakers is asser ting an existentially generaliz ed
pr oposition as w ell as one with the specic r elation r elev ant to the context of use. O n the other
hand, a commitment to the latter commitss , perhaps, to posit a co v er t existential quantier in
the morphosyntactic str uctur e of pr enominal possessiv es.
e pr oblem with this pr oposal is that it attempts to sav e S tr ong P r opositionalism, S tr ong
Asser tiv e P r opositionalism, and S tr ong Linguistic D ependence at too high a cost. F or on this
pictur e, whenev ers points at an arbitrar y book and says ‘at ’ s not J ohn ’ s book ’,s is asser ting,
among other things, something false, i.e., that it isn ’ t the case that the book in question bears
a r elation to J ohn. B ut, clearly , that semantic content isn ’ t explaining anything, since for any
book in the univ erse, J ohn is r elated to it. P ut differ ently , since the existentially quantied
expr ession giv en b y ( ) is trivially tr ue, its negation is blatantly false. G iv en that uttering such
transpar ent trivialities or falsities so ob viously violates the conv ersational maxims and the norms
of asser tion, it ’ s unclear in what sense the semantic content assigned b y this vie w is asser ted.
A v ariant of this pr oposal might be to take the domain of the existential quantier in the
pr oposed analysis of the possessiv e to be contextually r estricted. O n this v ersion, ‘J ohn ’ s book ’
means, appr o ximately , that ther e is someR in the context of utter ance such that bookb bearsR to
J ohn. S till, this r evision doesn ’ t help . F or as long as ther e is at least one book—any one—in the
context of utterance it will bear someR to J ohn. H ence, in any context in which ther e is mor e
than one book (e.g., a librar y), an utterance of ‘J ohn ’ s book ’ will asser t falsities no one—ex cept,
perhaps, the defender of this vie w—will hear . S o, although on this pictur e the semantic theses
in question ar e sav ed, they ar e sav ed at the cost of semantic idleness.
H er e is a mo v e a friend of the semantic theses under discussion might be tempted to make.
H e could keep S tr ong P r opositionalism and S tr ong Linguistic D ependence, but d r op S tr ong
Asser tiv e P r opositionalism. I n other wor ds, he could claim that, although the semantic content
of ‘J ohn ’ s book ’ is giv en b y ( ), asser tiv e utterances of sentences containing ‘J ohn ’ s book ’ don ’ t
always asser t its semantic content. S ince this would also be tr ue of the negation of sentences
containing the pr enominal possessiv e in question, our pr evious objection would fail.
er e is a major difficulty with this mo v e. I f the semantic content of a sentence S isn ’ t
always asser ted, then—absent a theor y about the r ole S ’ s semantic content plays when it isn ’ t
asser ted—ther e is no appar ent r eason to keep it. I n fact, asser ting such an extr emely w eak
semantic content would be utterly uncooperativ e in most conv ersational contexts: to be told
that ther e is a book someho w r elated to J ohn is, in most conv ersational contexts, a useless piece
of information. S ince—on this alternativ e—the semantic content of a sentenceS containing a
pr enominal possessiv e isn ’ t always asser ted, and ther e is no appar ent r ole that the existentially
quantied semantic content plays in an account of the meaning of S , w e can do without it.
I t ’ s note wor thy that although this semantic pr oposal is unsatisfactor y , its main motiv ation is
on the right track. I t ’ s tr ue that ev er y time a speakers asser tiv ely utters a sentence containing a
pr enominal possessiv e, ther e will be a r elationR thats is asser ting to hold betw een the possessor
and the possessed (or denying that ther e is such a r elation in the case of a sentence with a wide
scope negation). S ince, as w e hav e seen, the identity of R can v ar y signicantly fr om one
context to another , what this pr oposal is tr ying to get at is the maximal piece of information
that is common to what is asser ted b y utterances of such sentences in all standar d cases. S o,
although the defender of this vie w is right in his conception of semantic content as that which
is common to what is asser ted, his commitment to S tr ong P r opositionalism, S tr ong Asser tiv e
P r opositionalism, and S tr ong Linguistic D ependence made him unable to do justice to that
insight.
. e Argument fr om Co v er t I ndexicality
A common featur e of the pr evious alternativ es is that they take sentences containing pr enominal
possessiv es to be semantically context- in sensitiv e. I f my objections to these accounts ar e corr ect,
a natural alternativ e for the defender of the thesis under discussion is to claim that sentences
containing pr enominal possessiv es ar e indexical. I n this section, I will explor e this alternativ e.
A ccor ding to Kaplan ( b ), a linguistic expr essiona is context-sensitiv e iff the meaning of
a fully determines differ ent contents acr oss contexts. I n Kaplanian terminology , a is context-
sensitiv e iff a has a context-sensitiv e character . e paradigmatic cases of semantic context-
sensitiv e expr essions ar e indexicals. F or example, since the character of the rst person singular
pr onoun ‘I’ determines differ ent contents fr om one context to another , w e say that it ’ s context-
sensitiv e.
H o w does the Kaplanian semantics for indexicals wor k? I n a nutshell, it wor ks as follo ws. ¹
T ake the sentence,
¹ I will discuss Kaplan ’ s system in detail in chapter .
() H e ’ s the author of D emonstr ativ es .
F irst, w e notice that whenev er a speaker s utters it while gesturing to war ds Kaplan, s asser ts
the pr oposition that Kaplan is the author of D emonstr ativ es , and whenev er s utters it while
gesturing to war ds Kripke,s asser ts the pr oposition that Kripke is the author of D emonstr ativ es .
S econd, borr o wing fr om the intensional systems dev eloped in the s and early s, w e
think of linguistic meaning as a function that mapsn -tuples of entities (or contexts) to contents.
W e put Kaplan in one of those n -tuples and Kripke in another , and call the function fr om
contexts to contents ‘ character ’. S ince w e also notice that gestur es play an impor tant r ole in
letting s ’ s audience identify the r efer ents of ‘he ’, w e incorporate them into of the character of
‘he ’ r oughly as follo ws:
() I n any context c , an occurr ence of the term ‘he ’ r efers to the demonstrated
individual inc .
Can w e use the same strategy to account for sentences containing pr enominal possessiv es? e
indexicalist about pr enominal possessiv es thinks w e can. H er e is the idea. R ecall the scenarios
intr oduced in section . : H J and L J . A pplying the Kaplanian
strategy , the indexicalist thinks of each scenario as a package of parameters—call them ‘c
1
’ and
‘c
2
’ r espectiv ely—containing each a parameter for the r elev ant r elation betw een J ohn and a
book, and models the meaning of ( ) as a function fr om contexts to pr opositions—i.e., as a
character . us, c
1
will contain the r elation x author ed y , while c
2
will contain the r elation x
won y in a r affle while dr unk . After a fe w applications of the semantic r ules of the language,
w e will arriv e at p for c
1
and q for c
2
, wher e p is a singular pr oposition about a book J ohn
author ed, andq a a singular pr oposition about a book J ohn won in a raffle when he was dr unk.
N o w , the challenge for the indexicalist is to identify a constituent in ‘J ohn ’ s book ’ the char-
acter of which fully determines, for ev er y context, a unique possessiv e r elation. F or the indexi-
calist ’ s purposes, this might just be the possessiv e morpheme ’ s . S o, the idea would be that the
possessiv e morpheme ’ s contributes a r elational v ariable at LF , whose content is pr o vided b y
context.
Although this looks plausible, ther e ar e a fe w worries that the indexicalist would hav e to
addr ess. F irst, adding co v er t constituents wher ev er y ou need them just to deal with pr oblematic
cases doesn ’ t explain anything. What the indexicalist needs for such a mo v e to be r eally explana-
tor y is to nd independent evidence that those constituents ar e wher e y our theor y pr edicts they
ar e. S o, ho w could the indexicalist establish the fr ee occurr ence of anR at a pr enominal posses-
siv e ’ s LF? O ne w ell-kno wn way of doing so is the argument fr om binding defended b y S tanley
( , ), who appeals to the fact that occurr ences of v ariables can be bound b y v ariable-
binding operators. ² I n a nutshell, the idea is that w e can demonstrate the existence of a fr ee
occurr ence of a co v er t v ariable in the str uctur e ofS b y sho wing thatS can hav e bound r eadings.
e main pr oblem with this alternativ e is that pr enominal possessiv es don ’ t seem to be
bindable. F or compar e the follo wing sentences:
() F or any r elation R , if x is a book that bears R to J ohn, then Alice believ es that
x bears R to him.
² I will examine the argument fr om binding in chapter .
() F or any r elationR , ifx is J ohn ’ s book, then Alice believ es thatx bearsR to him.
I f pr enominal possessiv es had a fr ee occurr ence of a v ariable at LF , then ( ) and ( ) should
expr ess the same pr oposition. B ut they don ’ t. I ntuitiv ely , ( ) says that if a book bears any r ela-
tion to J ohn, Alice believ es that book bears the same r elation to her . B y contrast, ( ) intuitiv ely
says that if a book is J ohn ’ s, Alice believ es that book bears ev er y r elation to him. S o, while ( )
would be tr ue in a world w in which Alice suffers fr om some sor t of delusional identication
with J ohn, ( ) would be tr ue in a world w
′
in which she suffers fr om a similar condition but
with r espect to a book. D oes this sho w that ther e isn ’ t any co v er t R in ‘J ohn ’ s book ’? N o, it
doesn ’ t. F or the indexicalist might claim that the quantier in ( ) and ( ) isn ’ t of the right kind.
B ut if so, then bur den of pr oof is on the indexicalist to come up with a pr oper quantier that
does the job . Alternativ ely , the indexicalist might r eject the argument fr om binding. I n that
case, he should be able to pr o vide ne w evidence in fav or of his vie w .
S econd, whatev er linguistic meaning is, it must be that which an speaker s would hav e to
kno w , o v er and abo v e s ’ s world kno wledge, in or der to pr operly use linguistic expr essions to
make asser tions. is is one of the pr oper ties Kaplan ’ s notion of character was designed to
captur e. F or example, accor ding to Kaplan, the character of ‘I’ is this:
(Ch
I
) I n any contextc , an occurr ence of the term ‘I’ designates its utter er inc .
O n this pictur e, a competent user s of ‘I’ is one whose kno wledge of (Ch
I
) allo ws him to
identify the content of an occurr ence of ‘I’ in a giv en context. All s needs to do is identify
the r elev ant context, and apply (Ch
I
) to it. S ince contexts ar e n -tuples of objectiv e, easily
identiable featur es of spatiotemporal r egions, s will normally be able immediately determine
who an occurr ence of ‘I’ picks out.
ings ar e much mor e complicated for the indexicalist about pr enominal possessiv es. S ince,
on this pictur e, character (linguistic meaning) is a r ule that fully determines, for ev er y context,
a unique content, kno wing the character of ‘ J ohn ’ s book ’ should enables to automatically select
the r elev ant r elation fr om ev er y other r elation that obtains betw een J ohn and a book. What
would such a r ule look like? S ince ‘J ohn ’ s book ’ can be used acr oss possible contexts to pick
out differ ent books r elated to J ohn in almost ev er y possible way one can think of , its character
must be capable of grabbing all those r elations fr om those contexts while capturing two things:
(i) what speakers kno w when they kno w the linguistic meaning of ‘J ohn ’ s book ’, and (ii) the
conv entions of the language r egar ding this expr ession.
Constr ucting a character satisfying these demands is extr emely difficult. I n or der for it to
captur e all the possible r elations uses of ‘J ohn ’ s book ’ can conv ey , it must be highly abstract
and general. B ut, at the same time, in or der for it to captur e what speakers kno w when they
kno w the linguistic meaning of ‘J ohn ’ s book ’, and the conv entions of the language r egar ding
this expr ession, it must be highly specic and simple enough to be graspable b y agents like us
(including y oung childr en) with limited computational po w er . Although, it might be possible
to achiev e this, I think it isn ’ t necessar y . I n what follo ws, I will dev elop a differ ent pictur e of the
r elationship betw een linguistic meaning and the contents asser ted b y literal uses of pr enominal
possessiv es that offers a smoother account of the data.
. T wo Conceptions of M eaning and Asser tion
What pictur e of the r elationship betw een linguistic meaning and asser ted content of a context-
sensitiv e expr ession does the defender of S tr ong P r opositionalism, S tr ong Asser tiv e P r oposi-
tionalism, and S tr ong Linguistic D ependence pr esuppose? H e pr esupposes a pictur e accor ding
to which linguistic meaning fully deter mines asser ted content in ev er y context. Let ’ s call this
pictur e ‘S emantic D eterminism ’:
S D :
I f S is an unambiguous, context-sensitiv e, declarativ e sentence and p is a pr oposition
asser ted (without conv ersational implicatur es) b y a literal utterance ofS in a contextc ,
then p is fully deter mined b y the linguistic meaning ofS inc .
I believ e tha t S emantic D eterminism is an unnecessar y tenet of our semantic theorizing. e
pictur e I want to defend in what follo ws is one accor ding to which linguistic meaning mer ely
imposes a non-exhaustiv e set of constr aints that my literal uses of context-sensitiv e sentences
must satisfy if I want to successfully asser t a pr oposition b y uttering them. ³ O n this pictur e,
linguistic meaning doesn ’ t x asser ted contents, but simply pr o vides some guidance to war ds
them. S o, let ’ s call this v ie w of the r elation betw een linguistic meaning and asser ted content
‘S emantic G uidance ’:
³ is pictur e of linguistic meaning as set of constr aints on asser tion has been defended r ecently b y Scott S oames
in a number of places. S ee S oames ( , , , ).
S G :
I f S is an unambiguous, context-sensitiv e, declarativ e sentence and p is a pr oposition
asser ted (without conv ersational implicatur es) b y a literal utterance ofS , thenp satises
the non-exhaustiv e set of constr aints imposed b y the linguistic meaning ofS .
A ccor ding to S emantic G uidance, the linguistic meaning ofS just giv es us the building blocks
ar ound which p is to be constr ucted. e r est of the component par ts needed to assemble p
will come fr om differ ent sour ces. What s our ces? Any sour ce conv ersational par ticipants hav e
av ailable to them in their par ticular communicativ e situation. N ormally , they will include our
audience ’ s and our o wn general kno wledge of the world, our shar ed pr esuppositions, the specic
purposes of our conv ersations, and our mutual expectations.
S o, ho w does this vie w help us explain the pr opositions asser ted b y sentences such as ( )
as utter ed in H J and L J ? I n a nutshell, the account is this. e
linguistic meaning of ( ) simply giv es us a set of constraints on the pr oposition asser ted in
each scenario . Y ou can think of linguistic meaning as pr o viding y ou with an abstract asser tion
checklist . F or instance, the checklist for ( ) would be r oughly this:
A C r is sentence is being used to pr edicate the pr oper ty of being the pr oduct of
ten y ears of r esear ch of the individual the sentence is used to talk about.
r e individual this sentence is being used to talk about is a book.
r e individual this sentence is being used to talk about is signicantly r elated
to J ohn.
I f my use of ( ) is such that it complies with each and ev er y item in this checklist, I will succeed
in asser ting a singular pr oposition. Which pr oposition would I asser t? e checklist doesn ’ t
say . All that it giv es me is a set of constr aints that my literal uses of ( ) must satisfy if I want to
successfully asser t any pr oposition.
Asser tion checklists play two r oles. F irst, they ar e per for mance assessment tools : they help
speakers to ensur e their asser tions ar e in good shape. S econd, they ar e interpr etation guiding
tools : they help hear ers to narr o w do wn the set of possible pr opositions that a speaker can liter-
ally asser t b y uttering a sentence at a giv en moment. us, rather than being fully determined,
the pr opositions asser ted b y Alice ’ s utterance of ( ) in H J and L J
must be simply allo w ed b y the linguistic meaning of ( ).
F or instance, A C together with the information av ailable in H -
J will determine a set of r elev antly acceptable r elations that might include:
• R
1
= x author edy
• R
2
= x checked outy fr om the librar y
• R
3
= x studiedy
F r om these r elations, Bob will hav e to r ule outR
2
andR
3
to get at the asser ted content. I f the
information pr o vided b y Alice during the conv ersation is rich enough, he will be able to r ule
out R
2
and R
3
without any effor t. I f it isn ’ t, he might need to ask Alice for fur ther dir ections.
F or instance, he might ask ‘ Ar e y ou talking about the book J ohn checked out fr om the librar y
or the book he author ed?’ After hearing Alice ’ s answ er , he will arriv e at the pr oposition asser ted.
W e can giv e a similar stor y for L J . W e will use the exact same asser tion checklist,
but no w the information av ailable in this communicativ e situation ⁴ will determine a differ ent
set of r elev antly acceptable r elations that might include the follo wing:
• R
1
= x wony in a raffle while dr unk
• R
2
= x got dr unk celebrating the publication ofy
• R
3
= x author edy
I f ev er ything goes w ell in the conv ersation, Bob will be able to r ule out R
2
and R
3
and get at
the pr oposition asser ted. I f not, he might need to ask Alice for fur ther dir ections befor e getting
at it.
. S emantic G uidance B asics
An inquisitiv e r eader might think that things look too easy on the S emantic G uidance vie w and
demand a mor e detailed pictur e of it. M y goal in this section is to satisfy that demand. S o, let ’ s
star t with some key notions and distinctions that the S emantic G uidance frame wor k emplo ys.
⁴ I take communicativ e situations to be differ ent fr om contexts (as conceiv ed in indexical logic). B riey , the
differ ence is this. Contexts ar e sequences of parameters whose sole purpose is to pr o vide r efer ents to occurr ences
of indexical terms. B y contrast, communicativ e situations ar e bodies of information that include the backgr ound
pr esuppositions and purposes of specic conv ersations. As such, they ar e r esour ces av ailable to conv ersational
par ticipants in par ticular cir cumstances. F or mor e details on this distinction, see section .. .
.. What Ar e Asser tion Checklists?
S o far , I hav e appealed to asser tion checklists to illustrate the main insight of S emantic G uid-
ance. H o w ev er , y ou might be wondering what exactly an asser tion checklist is. H er e I would
like to say mor e about it.
I n a nutshell, the idea is this. An asser tion checklist forS is a r epr esentation of the context-
inv ariant information conv entionally associated with S that a speaker would hav e to kno w in
or der to make and understand literal asser tions of S . As such, it doesn ’ t include information
about the world or about the specic featur es of the conv ersational situations in which the
asser tions take place. ⁵ Asser tion checklists ar e heuristic devices that play two main r oles: (i)
they enable speakers to v erify whether the pr opositions they intend to communicate ar e, in
fact, allo w ed b y the linguistic meanings of their sentences, and (ii) they guide hear ers to war ds
the asser ted contents intended b y competent speakers when literally uttering those sentences.
H o w ar e they generated? e rst step is to distinguish betw een semantic content and as-
ser ted content . e second step is to characteriz e a constraint on asser ted contents. Let ’ s begin
with the former . Asser ting is a speech act that speakers typically per form b y making literal
utterances of declarativ e sentences. e asser ted contents of those utterances ar e pr opositions.
Although the default option for asser ted contents ar e the pr opositions that r esult fr om the
combination of the contents contributed b y the wor ds and phrases occurring in the sentences
utter ed—i.e., their semantic contents—sometimes the asser ted contents ofS hav e constituents
⁵ is idea is closely r elated to S oames ’ s notion of linguistic meaning as least common denominator . S ee S oames
( , , ).
not contributed b y any of the wor ds or phrases occurring inS . F amiliar examples of this phe-
nomenon include sentences such as ‘I t ’ s raining ’, ‘I’ m r eady ’, and ‘J ohn has had enough ’. ⁶ S o,
a plausible characterization of asser ted content might go as follo ws:
Asser ted Content
p is the asser ted content of a sentence S in a context c only if b y uttering S (i) p
is made av ailable to ev er y rational, competent conv ersational par ticipant in c , (ii) p is
inferable b y ev er y rational, competent conv ersational par ticipant inc , and (iii) it would
be appr opriate to tr eat the utter er ofS as committed to the tr uth of p .
F or the friend of S emantic G uidance, the distinction betw een semantic content and asser ted
content is cr ucial. R ecall our pr edicament. I n or der to understand an utterance of the expr es-
sion ‘J ohn ’ s book ’ w e hav e to be able to determine (i) the meaning of ‘J ohn ’, (ii) the meaning
of ‘book ’, and (iii) the r elation R that J ohn bears to a par ticular book. S ince the identity of R
isn ’ t par t of the information encoded b y the expr ession itself , it ’ s r easonable to suppose that it
must come fr om the communicativ e situation.
Both the friend of S emantic D eterminism and the friend of S emantic G uidance agr ee on
all of this. H o w ev er , while the friend of S emantic D eterminism infers fr om it that it ’ s the r ole
of linguistic meaning to someho w be able to catch that R fr om the communicativ e situation
and put it in the semantic content of ‘J ohn ’ s book ’, the friend of S emantic G uidance sees the
identity ofR as par t of its asser ted content, not as par t of its semantic content. S ince, accor ding
⁶ S ee, for example, P err y ( b ) and B ach ( ).
to S emantic G uidance, linguist ic meaning simply constrains asser ted contents, it ’ s not the job
of linguistic meaning to giv e us the semantically absent R .
e second step in the generation of an asser tion checklist is to characteriz e a constraint on
asser ted content. W e say thatC is a constraint on asser ted content iffC is a condition that ev er y
acceptable asser ted content must satisfy . O n the pictur e put for war d b y S emantic G uidance,
each meaning ful constituent ofS will add a constraint on the possible contentsS can be used
to asser t. W e can think of a constraint as a pair⟨t,k⟩ wher e t is a meaning ful constituent of
S and k is a condition that must be satised b y ev er y candidate asser ted content of t . e set
of constraints determined b y the linguistic meanings of all the meaning ful constituents of S
together with the constraints imposed b y S ’ s syntactic str uctur e is S ’ s asser tion checklist. I f S
contains an occurr ence of a pr enominal possessiv e, its asser tion checklist will not specify the
identity of R . I will elaborate on the natur e of constraints in section .. .
W ith these elements in place, the challenge for the S emantic G uidance theorist is to for-
mulate adequate asser tion checklists for the linguistic constr uctions she wants to examine. B ut,
what counts as an adequate asser tion checklist? An asser tion checklist A for a sentence S is
adequate iff (i)A contains each and ev er y constraint contributed b y the linguistic meaning of
each and ev er y meaning ful constituent of S , and (ii) ev er y pr oposition asser table b y a literal
utterance of S satises ev er y constraint in A . N otice that, on this conception of adequacy ,
what ’ s not pr ohibited b yA is permitted b y it. is means that asser ted contents that satisfy a
giv enA might include information that goes bey ond what minimally satises A ’ s constraints.
F or instance, if the communicativ e situation is such that it ’ s clear to all conv ersational par tic-
ipants that ‘J ohn ’ s book ’ denotes a contextually salient gr een book, the pr enominal possessiv e
might be used in a sentence to asser t a pr oposition that pr edicates the pr oper ty of being gr een
of a book r elev antly r elated to J ohn. H o w much information can be added to asser ted contents
will depend on the backgr ound pr esuppositions and purposes of the r elev ant communicativ e
situation.
.. Contexts v ersus Communicativ e S ituations
An impor tant distinction for S emantic G uidance is that betw een the notions of context and
communicativ e situation. M y aim in this section is to explain some signicant differ ences
betw een them.
e notion of context standar dly used in semantics can be traced back to M ontague ( ,
) and Kaplan ( b ). M ontague modeled Carnapian intensions as functions fr om pos-
sible worlds to extensions, and extended this analysis to interpr et what he called ‘ pragmatic
languages ’, i.e., formal languages that include indexical and temporal expr essions.
P ragmatic languages posed a ne w challenge to M ontague ’ s intensional frame wor k. F or the
interpr etation of indexical and temporal expr essions r equir ed the interpr etation function to
map something mor e ne-grained than possible worlds to extensions. H is solution was the
intr oduction of sequences of entities called ‘ parameters ’ whose r ole was to supply extensions
to those expr essions. F or example, the interpr etation of ‘I’ r equir ed an agent-parameter , ‘ no w ’
r equir ed a time-parameter , ‘her e ’ a location-parameter , etc. M ontague called these sequences
of parameters ‘indices ’ or ‘ points of r efer ence ’.
Kaplan made the most of this strategy and gav e it a ne w twist. H e r ealiz ed that r elativizing
intensions to just one index doesn ’ t captur e the right tr uth-conditions of sentences that hav e
indexicals occurring in clauses under the scope of intensional operators. F or example, consider
the follo wing sentence:
() I n London, in the y ear , those who ar e her e and no w won ’ t be aliv e.
A system with just one index pr edicts that ( ) is tr ue iff those who ar e in London in ar en ’ t aliv e at that place at that time. Clearly , this is incorr ect. F or ( ) is tr ue r elativ e to my
spatiotemporal coor dinates right no w iff those who ar e in Los Angeles in ar en ’ t aliv e
in London in . S o, the only way for the intensional frame wor k to captur e these tr uth-
conditions was to assign the place of the utterance to the occurr ence of ‘her e ’ and the time of
the utterance to the occurr ence of ‘ no w ’, irr espectiv ely of the fact that they occur in ( ) under
the scope of a locativ e and temporal operators. Kaplan achiev ed this b y incorporating Kamp ’ s
( ) double-indexing technique: one index would be in charge of the extensions of indexicals,
and the other would be in charge of expr essions whose extensions can v ar y whenev er they occur
under the scope of an intensional operator . Kaplan called the rst index ‘ the context of use ’,
and the second ‘ the cir cumstances of ev aluation ’. ⁷
F ollo wing this traditional usage, let ’ s r eser v e the term ‘ context ’ for n -tuples of parameters.
A key featur e of contexts is that their sole purpose is to pr o vide r efer ents to occurr ences of
context-sensitiv e expr essions. Consider , for instance, the agent-parameter . I ts r ole is exhausted
as soon as it pr o vides a r efer ent to an occurr ence of the wor d ‘I’. T o achiev e its purpose the
⁷ I will explain these issues in detail in chapter .
agent-parameter doesn ’ t hav e to captur e the fact that the speaker is speaking, uttering the wor d
‘I’, or using the language in any way . e interpr etation function doesn ’ t car e about what the
speaker is doing or what her mental states ar e like: the agent-parameter doesn ’ t ev en hav e to
captur e the fact that the speaker is rational, or that she is capable of understanding the wor d ‘I’.
is pr oper ty of contexts is cr ucial for Kaplan ’ s pr oject. H e thought of contexts as these
n -tuples of parameters because he wanted to build a logic of indexicals. H e wanted to trace
logical consequence r elations betw een indexical sentences and dene ‘logical tr uth ’ for indexical
languages. is can only be achiev ed in the intensional frame wor k he was wor king on, if ev er y
sentence of such a language giv es us a determinate content that can be ev aluated for tr uth or
falsity . S o, he used M ontague ’ s and Kamp ’ s insights and built in the semantic content of an
indexical sentence r elativ e to one of thosen -tuples.
U nderstanding contexts in this way is per fectly legitimate if y ou want to dev elop a logic.
H o w e v er , if y our pr oject is to captur e the linguistic meanings of natural language expr essions
and their r elation to the contents those expr essions can be used to asser t, y ou will need a richer
notion of context. Y ou will need to appeal to what I call ‘ communicativ e situations ’. Commu-
nicativ e situations ar e a lot mor e accommodating than contexts ar e. I n addition to the infor-
mation pr o vided b y contexts, they include ev er y extra-linguistic piece of information publicly
av ailable that is r elev ant to the interpr etation of the asser ted content of a giv en expr ession. F or
instance, they include the shar ed pr esuppositions, purposes of par ticular linguistic ex changes,
and mutual expectations of conv ersational par ticipants. I n shor t, a communicativ e situation is
a body of information shar ed b y a gr oup of conv ersational par ticipants that allo ws them to go
fr om semantic contents to asser ted contents. ⁸
F or the friend of S emantic G uidance, the distinction betw een context and communicativ e
situation is cr ucial. R ecall our pr edicament. W e needed an account of ho w to get at the missing
R in ‘J ohn ’ s book ’. F or the friend of S emantic D eterminism, context comes in handy . H e would
putR in the context and analyz e the linguistic meaning of ‘J ohn ’ s book ’ as a function that maps
R into its semantic content. B y contrast, the friend of S emantic G uidance would put R in
the communicativ e situation, out of r each of linguistic meaning, and leav e to conv ersational
par ticipants the task of inferring it using the r esour ces av ailable ther e.
I s ther e any adv antage to adopting the option fav or ed b y S emantic G uidance? I think so .
H er e ’ s one. A t a v er y intuitiv e lev el, if a theor y of meaning assigns m to an expr ession a ,
thenm must be what competent users ofa kno w when they kno w the linguistic meaning ofa .
M or eo v er , one would expectm to corr espond to what is set b y the conv entions of the language. ⁹
As I mentioned at the end of section . , these featur es ar e v er y har d to captur e if y ou adopt
S e mantic D eterminism.
⁸ Although the notion of communicativ e situation is r elated to S talnaker ’ s notion of context set S talnaker ( ),
and to B ach ’ s notion of br oad context B ach ( ), it differs fr om them. S ince communicativ e situations include
only information av ailable to ev er y conv ersational par ticipant, it differs fr om S talnaker ’ s notion as he allo ws indi-
viduals to hav e access to context sets not shar ed with other conv ersational par ticipants. I t also differs fr om B ach ’ s
notion of br oad context in that B ach seems to ex clude fr om br oad contexts what he calls ‘ narr o w contexts ’ (or what
I’ m simply calling ‘ contexts ’ her e). B y contrast, communicativ e situations encompass both narr o w and br oad
contexts.
⁹ F or example, her e is ho w Kaplan describes char acter : ‘e character of an expr ession is set b y linguistic
conv entions and, in turn, determines the content of the expr ession in ev er y context. B ecause character is what is
set b y linguistic conv entions, it ’ s natural to think of it as meaning in the sense of what is kno wn b y the competent
language user . ’ Kaplan ( b , )
B y contrast, the friend of S emantic G uidance has a less daunting semantic task. H er task is
to formulate adequate sets of constraints or asser tion checklists for the linguistic constr uctions
she is studying. F or example, no adequate asser tion checklist for ‘J ohn ’ s book ’ would r ule out
an R
′
that is intuitiv ely acceptable b y the conv entions of the language. B ut most impor tantly ,
an adequate asser tion checklist wouldn ’ t hav e to identify an R . O n the S emantic G uidance
pictur e, identifying the specic R asser ted in a communicativ e situation is a decision pr oblem
the solution of which is not its job to pr o vide.
.. G lobal Constraints v ersus Local Constraints
A ccor ding to S emantic G uidance, the linguistic meanings of all the meaning ful constituents
of S impose constraints on S ’ s asser ted contents. H o w ev er , they ar en ’ t the only sour ces of
constraints. S yntax and communicativ e situations also impose constraints on asser ted contents.
e purpose of this section is to distinguish betw een them.
S emantic G uidance agr ees with S emantic D eterminism on the follo wing thesis: the linguis-
tic meaning of a fully determines its semantic content. H o w ev er , it disagr ees with S emantic
D eterminism in that, accor ding to S emantic G uidance, the object, if any , fully determined b y
a ’ s linguistic meaning doesn ’ t hav e to coincide with any ofa ’ s asser ted contents—ev en whena
is used with its standar d meaning. Rather , the linguistic meaning ofa mer ely places constr aints
on its possible asser ted contents.
Communicativ e situations also place constraints on asser ted contents. S ome of their fea-
tur es such as shar ed pr esuppositions, shar ed purposes of the conv ersation, as w ell as conv er-
sational maxims and principles, play a vital r ole in helping conv ersational par ticipants get at
intended asser ted contents. F or r econsider sentence ( ):
( ) J ohn ’ s book is the pr oduct of ten y ears of r esear ch.
R elativ e to H J , the asser ted content is a singular pr oposition about a book
J ohn author ed, wher eas r elativ e to L J it ’ s a singular pr oposition about a book J ohn
won in a raffle when he was dr unk.
S yntax also plays an essential r ole in narr o wing do wn the set of candidate asser ted contents.
D espite appearances, the range of r elations that hold betw een the possessor and the possessed
is not completely unlimited. I t ’ s subject to grammatical constraints. T o appr eciate this point,
compar e the follo wing constr uctions: ¹⁰
() a. the table ’ s leg
b . the leg ’ s table
Although both constr uctions can be used to conv ey the par t-whole r elation, they can ’ t be used
inter changeably . F or I could asser tiv ely utter:
() M y knee hit the table ’ s leg and wine spilled into the plates,
to talk about a leg that is par t of a table but not:
() *M y knee hit the leg ’ s table and wine spilled into the plates,
¹⁰ S ee B ar ker ( ).
to talk about that v er y same leg. I f ther e wasn ’ t any grammatical constraint on the r elation
a pr enominal possessiv e can be used to asser t, no such asymmetr y would take place. ¹¹ ese
considerations suggest that ther e is mor e than one sour ce of constraints. S o, let ’ s tr y to sor t
them out.
F irst, r ecall our characterization of a constraint on asser ted content: C is a constraint on
asser ted content iffC is a condition that ev er y acceptable asser ted content must satisfy . Let ’ s
call a set of acceptable asser ted contents an ‘ acceptability range ’. us, on this pictur e, ev er y
constraint on asser ted content determines an acceptability range.
S econd, judging fr om the kinds of acceptability range they determine, constraints can be
categoriz ed into two main gr oups: global constraints and local constraints. G lobal constraints
ar e constraints that determine constant acceptability ranges acr oss possible communicativ e sit-
uations. at is to say , global constraints impose conditions that must be satised b y ev er y
acceptable asser ted content of a giv en a irr espectiv ely of wher e, when, ho w or for what pur-
posea is used. B y contrast, local constraints ar e constraints that determine acceptability ranges
that v ar y acr oss possible communicativ e situations. I n other wor ds, local constraints impose
conditions that must be satised b y ev er y acceptable asser ted content of a only at specic com-
municativ e situations.
O n the S emantic G uidance frame wor k, linguistic meaning and syntax impose global con-
straints, while shar ed pr esuppositions, conv ersational maxims, and conv ersational purposes im-
¹¹ I will discuss ho w these grammatical constraints operate in section .. .
pose local constraints. ¹² F or instance, the fact that ‘ the table ’ s leg ’ and ‘ the leg ’ s table ’ can ’ t be
used inter changeably to conv ey the v er y same par t-whole r elation is due to a global constraint
on the or der and dir ectionality that the constituents of a possessiv e constr uction must satisfy
together with the constraints imposed b y the linguistic meanings of the nouns ‘ table ’ and ‘leg ’.
I will say mor e about this in the next section.
B y contrast, communicativ e situations together with conv ersational maxims impose local
constraints on asser ted contents. eir r ole is to narr o w do wn acceptability ranges determined
b y global constraints accor ding to the purposes of par ticular conv ersations. er e ar e two main
kinds of mechanism b y which this is achiev ed: ltering and or dering . F iltering is the pr ocess
b y which members of global acceptability ranges that don ’ t ser v e the purposes of par ticular
communicativ e situations get ex cluded fr om local acceptability ranges. F or instance, although
the global acceptability range for ‘J ohn ’ s book ’ includes the x -won-y -in-a-r affle-while-dr unk
r e lation, it will be lter ed out of the local acceptability range corr esponding to H -
J because that r elation is completely irr elev ant to the purposes of the linguistic ex-
change that takes place ther e. I n highly cooperativ e communicational situations, ltering will
be enough for conv ersational par ticipants to get at intended asser ted contents.
O r dering is the pr ocess b y which members of local acceptability ranges get ranked accor ding
to ho w w ell they ser v e the purposes of the communicativ e situation. O r dering is par ticularly
useful in less cooperativ e communicativ e situations, wher e ltering is insufficient to single out
¹² is distinction is closely r elated to the distinction Ray o ( for thcoming ) makes betw een G lobalism and Local-
ism. H o w ev er , Ray o intr oduces it to distinguish betw een differ ent appr oaches to linguistic competence, r emaining
neutral about linguistic meaning. B y contrast, I use it to explain linguistic meaning, r emaining neutral abo ut lin-
guistic competence.
the intended asser ted content. F or example, imagine a communicativ e situation in which Alice
is talking to Bob about a book author ed b y J ohn. S uppose that in the middle of the conv ersation
she also tells Bob that she is r eading a book about J ohn. I f she w er e to utter ( ), it might not
be clear for Bob which book she is talking about. Assuming for simplicity that ev er y other
candidate r elation has been lter ed out, Alice ’ s utterance of ‘J ohn ’ s book ’ might determine two
competing or der ed local acceptability ranges:
O O R
1
= x author edy R
1
= x is abouty
R
2
= x is abouty R
2
= x author edy
I n a situation like this, Bob would need fur ther dir ections fr om Alice to get at the intendedR .
F or instance, he could wait for Alice ’ s next utterances or ask her dir ectly about it. ¹³
e distinction betw een global and local constraints on asser ted content is cr ucial for S e-
mantic G uidance. Local constraints ar e purpose-driv en, wher eas global constraints ar en ’ t.
us, although most of the heavy lifting in getting at intended asser ted contents is done b y
local constraints, global constraints such as linguistic meaning determine the range of contents
on which local constraints operate.
¹³ An alternativ e way of looking at this is the follo wing. I nstead of distinguishing betw een ltering and or dering,
w e could simply think of the communicativ e situation as determining an or dering, which might be v er y long and
strict, with the “lter ed ” ones all way do wn at the bottom. Although this alternativ e is formally equiv alent to
the one I’ m pr oposing, it would be unnecessar y in fairly cooperativ e communicativ e situations, i.e., situations
in which it ’ s r easonably clear to conv ersational par ticipants which r elation is being conv ey ed. I n situations like
these, ltering would be vir tually automatic without the need to rank all the implausible and irr elev ant r elations
someone could think of .
. A ccounting for P ossessiv e R elations
N o w that w e hav e the main ingr edients of the frame wor k in place, let ’ s look mor e closely at the
pr oper ties that pr enominal possessiv e r elations exhibit.
.. I nher ent and E xtr insic R elations
T radition in semantics has it that one-place pr edicates denote monadic pr oper ties. H o w ev er ,
some linguists—e.g. P ar tee ( ) and B ar ker ( , )—hav e argued that some one-place
pr e dicates denote two-place r elations. Compar e, for in stance, the pr edicates ‘is a woman ’ and
‘is a mother ’. Although Alice could satisfy the former all b y herself , satisfying the latter r equir es
her to be r elated to another individual in a v er y par ticular way . is r equir ement is semantic:
what y ou learn when y ou learn the linguistic meaning of ‘ mother ’ is that it denotes an individual
r e lated in a par ticular way to another . F or these authors, this suggests that while the extension
of the pr edicate ‘is a woman ’ is a set of entities, the extension of ‘is a mother ’ is a set of pairs:
() a. J womanK = lx[ woman(x)]
b . J motherK = lxly[ mother(x,y)]
N ouns that gur e in one-place pr edicates that denote r elations ar e called ‘r elational nouns ’.
e most common r elational nouns acr oss languages include those denoting family r elations
(‘ child ’, ‘ sister ’, ‘ cousin ’) and body par ts (‘head ’, ‘hand ’, ‘leg ’). ¹⁴
¹⁴ S ee B ar ker ( ).
e distinction betw een monadic and r elational nouns is impor tant for determining the
identity of the r elation holding betw een the possessor and the possessed in pr enominal posses-
siv e constr uctions. When a pr enominal possessiv e contains a r elational head noun, the posses-
siv e r elation will be naturally taken to be the one denoted b y the r elational noun. B y contrast,
when its head is a monadic noun, the identity of the possessiv e r elation will heavily depend on
the communicativ e situation. T o appr eciate this, compar e ‘J ohn ’ s mother ’ with ‘J ohn ’ s moun-
tain ’. While the default interpr etation the former is that it r efers to the woman who gav e bir th to
J ohn, the latter leav es us wondering. F ollo wing the terminology used in the literatur e, let ’ s call
pr enominal possessiv es like the former ‘ extrinsic possessiv es ’ and those like the latter ‘inher ent
possessiv es ’.
e extrinsic/inher ent distinction betw een pr enominal possessiv es can also help us to ex-
plain the intuitiv e differ ence betw een ‘ the table ’ s leg ’ and ‘ the leg ’ s table ’. T wo global constraints
ar e in play her e. e rst is that pr enominal possessiv es in E nglish ar e right-headed, ¹⁵ and the
second is that, of the two nominals that constitute it, only ‘leg ’ is r elational. S ince the head
of ‘ the table ’ s leg ’ is a r elational noun, it ’ s an inher ent possessiv e. us, b y default, conv er-
sational par ticipants will take the possessiv e r elation to be the par t-whole r elation wher e ‘leg ’
denotes the par t and ‘ table ’ the whole. N o w , compar e this with ‘ the leg ’ s table ’. e r eason
this constr uction sounds odd when hear d out of the blue is that it ’ s an extrinsic possessiv e (its
head is a monadic noun) whose modier (‘leg ’) is r elational. S ince normally the default pos-
¹⁵ R oughly , the head of a phrase is that constituent of it which dominates the whole phrase, i.e., it determines
its syntactic type and/or its basic semantic content. e other constituents of the phrase simply modify the head.
F or a thor ough discussion of the natur e and r ole of heads, see Corbett ( ).
sessiv e r elation is contributed b y the head and not b y the modier , w e can ’ t assume anything
about its identity without kno wing something about the communicativ e situation in which the
pr enominal possessiv e in question is utter ed.
F or the friend of S emantic D eterminism, this appears to be good ne ws. H e would claim
that at least in the case of inher ent possessiv es, linguistic meaning fully determines the posses-
siv e r elation. H e might tr y to captur e this b y r epr esenting the contrast betw een extrinsic and
inher ent possessiv es r oughly as follo ws:
() a. J J ohn ’ s mountainK = lx[ mountain(x)^R( J ohn,x)]
b . J J ohn ’ s motherK = lx[ mother( J ohn,x)]
w er eR in ( a ) is a fr ee occurr ence of a v ariable for the possessiv e r elation. F r om this, he would
conclude that the S emantic G uidance thesis that all pr enominal possessiv es ar e semantically
under determined is false.
e pr oblem with this line of thought is that no matter ho w str ong our intuitions ar e about
a giv en inher ent possessiv e, ther e always will be communicativ e situations in which the r elation
contributed b y the head is o v erridden b y another , mor e r elev ant, locally determined r elation. ¹⁶
F or consider the follo wing case:
¹⁶ is idea is w ell captur ed b y P eters ’ & W esterståhl ’ s P rinciple F
3
:
( F
3
) F or any possessiv e NP , ho w ev er pr edictable and semantically describable its usual possessor
r elation is, cir cumstances can always be found wher e that same possessiv e NP is used with an-
other possessor r elation, not deriv able fr om grammatical or lexical information, but pr o vided
only b y the context of utterance. ( P eters & W esterståhl , , p . ).
J ’ M
T w o photography students, J ohn and Bob , ar e the nalists for a national awar d com-
petition. As their nal pr oject, they hav e been assigned to por tray a mother looking
at her ne wly born bab y . After car eful ev aluation of the por traits, the judges announce
their nal decision:
() J ohn ’ s mother is mor e beautiful than Bob ’ s mother .
I n this scenario, it ’ s clear that the judges ’ utterances of ‘J ohn ’ s mother ’ and ‘Bob ’ s mother ’ ar e
intended to pick out the photographs taken b y J ohn and Bob r espectiv ely , not the women who
gav e bir th to them (or adopted them). I n general, it ’ s not difficult to imagine similar scenarios
for any inher ent pr enominal possessiv e. Communicativ e situations like this challenge the thesis
that the linguistic meanings of inher ent possessiv es x their possessiv e r elations. ¹⁷
ese cases can be explained smoothly in the S emantic G uidance frame wor k. A ccor ding to
it, both inher ent and extrinsic possessiv es ar e semantically under determined, i.e., their linguis-
tic meanings don ’ t x the possessiv e r elation in any communicativ e situation. H o w ev er , since
the linguistic meaning of ‘ mother ’ imposes a r elational global constraint on its candidate r ef-
er ents, the linguistic meaning of ‘J ohn ’ s mother ’ will determine a constant acceptability range
that includes only entities that satisfy one of the two positions of the mother hood r elation. P ut
¹⁷ e S emantic D eterminist might be tempted to r eply b y saying that inher ent possessiv e ar e, in fact, ambiguous
betw een inher ent and extrinsic r eadings. e pr oblem with this r eply is that since for ev er y contextc , ther e is an
open-ended list of possible extrinsic r elations betw een the possessor and the possessed inc , accepting the S emantic
D eterminist ’ s alternativ e ultimately commits y ou to the claim that pr enominal possessiv es ar e massiv ely ambiguous.
is r esult b y itself suggests that the S emantic D eterminist is missing a generalization.
differ ently , the linguistic meaning of ‘J ohn ’ s mother ’ will pr o vide an asser tion checklist appr o x-
imately like this:
A C r e individual this expr ession is being used to talk about bears the mother-
hood r elation to another individual.
r e individual this expr ession is being used to talk about is signicantly
r elated to J ohn.
is is as far as the linguistic meaning of ‘J ohn ’ s mother ’ can take us with r egar ds to the identity
of the possessiv e r elation. e r est will be done b y local constraints imposed b y par ticular
communicativ e situations. I f the communicativ e situation is cooperativ e enough, then all the
candidate r efer ents included in its global acceptability range that don ’ t ser v e the purposes of
the conv ersation will be lter ed out automatically leaving just the intended asser ted content.
Communicativ e situations such as J ’ M normally play this r ole optimally , as the
judges ’ utterance of ( ) wouldn ’ t cause confusion among par ticipants. ¹⁸
ings ar e differ ent in non-cooperativ e situations—e.g., a situation in which a passerb y
accidentally o v erhears an asser tiv e utterance of ( ). I n those cases, speakers will assume the
r e lation denoted b y a r elational head noun to be the r elev ant possessiv e r elation. is is so
because, ev en in standar d cooperativ e situations, that r elation will occupy b y default the top
position in or der ed acceptability ranges locally determined. is is evident b y the fact that I
¹⁸ e astute r eader may hav e noticed that A C doesn ’ t r eally captur e the asser ted content
w e ar e looking for , as ‘J ohn ’ s mother ’ is used in the situation described to r efer to a photograph, rather than a
person. I will make the necessar y adjustments to handle this kind of cases in section .. .
can use ‘J ohn ’ s mother ’ with the default interpr etation without pr eviously setting the stage, and
kno w my hear ers will take the r elev ant r elation to be the default one. H o w ev er , since ev er y
or der ed acceptability range is defeasible, the default r elation can be o v err uled b y enriching the
communicativ e situation.
.. e U niqueness Constraint
V er y often, singular pr enominal possessiv es ar e used to pick out unique objects. F or example,
an utterance of ‘J ohn ’ s mother ’ at a giv en communicativ e situation s conv eys that ther e is a
unique x r elev ant in s such that x bears R to J ohn. U tterances of extrinsic po ssessiv es exhibit
the same pr oper ty . F or instance, felicitous utterances of ‘J ohn ’ s mountain ’ normally conv ey that
the r e lev ant mountain J ohn bears R to is one and only one.
I n addition, plural pr enominal possessiv es ar e naturally taken to denote maximal sets of
objects. ¹⁹ F or example, if I asser tiv ely utter ‘J ohn ’ s childr en ar e smar t ’ (inher ent), I will nor-
mally asser t that all of J ohn ’ s childr en ar e smar t. And if I asser tiv ely utter ‘J ohn ’ s cars ar e r ed ’
(extrinsic), I will normally asser t that all of J ohn ’ s cars ar e r ed.
e intuition that pr enominal possessiv es conv ey uniqueness has motiv ated many authors
to assume that they ar e semantically equiv alent to (R ussellian) denite descriptions. F or in-
stance, when D onnellan ( ) intr oduces his w ell-kno wn distinction betw een r efer ential and
attributiv e uses of denite descriptions, the v er y rst constr uction he uses to illustrate his point
is not a canonical denite description but a pr enominal possessiv e in subject position:
¹⁹ S ee B ar ker ( ).
() S mith ’ s mur der er is insane.
is assumption seems to be shar ed b y Kripke ( ) who, despite his criticisms to D onnellan ’ s
distinction, also tr eats ‘S mith ’ s mur der er ’ as a denite description, and b y N eale ( ) who
r epr esents ( ) as ( ):
() [ the x : x mur der ed S mith] insane x
P ar tee ( ) and P ar tee & Borschev ( ) pr opose analyses of pr enominal possessiv es that
also incorporate deniteness:
() a. J J ohn ’ s mountainK = x[ mountain(x)^R( J ohn,x)]
b . J J ohn ’ s motherK = x[ mother( J ohn,x)]
S hould w e conclude, then, that pr enominal possessiv es ar e semantically equiv alent to denite
descriptions? I n par ticular , do they impose a uniqueness constraint on asser ted contents? I
don ’ t think so . As sev eral authors hav e sho wn, ²⁰ many uses of pr enominal possessiv es don ’ t
conv ey uniqueness. F or example, consider the follo wing sentence:
() J ohn ’ s nger was hur t.
F elicitous utterances of ( ) can be made without asser ting that J ohn ’ s only nger was hur t
or that only one of his ngers was hur t. e asymmetr y betw een pr enominal possessiv es and
denite descriptions can also be seen in the follo wing case:
²⁰ S ee, for instance, B ar ker ( , ), and P eters & W esterståhl ( ).
S ’ M
S mith has been foully mur der ed b y two individuals who ar e no w under arr est. O ne
of them, J ones, is an intr o v er ted standar d-looking bank cler k. e other , M anson, is
a violent homicidal maniac. F r om M anson ’ s bizarr e appearance and odd behavior , a
police officer ex claims:
( ) S mith ’ s mur der er is insane.
I n the scenario described, the utterance of ( ) might asser t a pr oposition about M anson, ev en if
the speaker and ev er y conv ersational par ticipant kno w that S mith was killed b y two individuals.
B y contrast, he can ’ t felicitously r efer to M anson b y uttering ‘e mur der er of S mith is insane ’.
is suggests that the uniqueness constraint isn ’ t par t of the global asser tion checklist giv en
b y the linguistic meaning of ( ), but it may be par t of some asser tion checklists locally deter-
mined. is explains why sometimes w e can use the v er y same pr enominal possessiv e constr uc-
tion to conv ey uniqueness in some communicativ e situations and non-uniqueness in others.
.. T ying U p S ome Loose E nds
S o far , I hav e been urging that instead of looking for a semantic r ule capable of fully deter-
mining the possessiv e r elations that can be asser ted fr om situation to situation b y literal uses
of pr enominal possessiv es, w e should adopt a semantic theor y on which the linguistic meaning
of a sentence S containing a pr enominal possessiv e mer ely giv es us a non-exhaustiv e asser-
tion checklist imposing global conditions on the pr opositions S can be used to asser t in those
situations. B eing non-exhaustiv e, such an asser tion chec klist would guide conv ersational par-
ticipants to war ds their asser ted contents, without xing one possessiv e r elation. Armed with it,
conv ersational par ticipants will hav e to appeal to conv ersational maxims and shar ed backgr ound
pr esuppositions to infer the missing r elation. As an example of ho w an asser tion checklist for
‘J ohn ’ s book ’ would r oughly go, I suggested the follo wing:
A C r e individual this expr ession is being used to talk about is a book.
r e individual this expr ession is being used to talk about is signicantly
r elated to J ohn.
H o w ev er , one might think that this chec klist is inadequate. T ake a look at the rst checkbo x.
I t r equir es the r efer ent of ‘J ohn ’ s book ’ to be a book. H o w ev er , ‘J ohn ’ s book ’ can be used to talk
about something that isn ’ t a book. F or suppose that J ohn is an ar tist who has painted a book
fr om his imagination. I n such a case, Alice could asser tiv ely utter ( ) to say , of J ohn ’ s painting ,
that it ’ s color ful:
() J ohn ’ s book is color ful.
I n such situation, her utterance of ( ) wouldn ’ t be infelicitous, ev en if the object painted
doesn ’ t ev en look like a book (suppose that the painting is C ubist).
O ne might think that this also applies sentences containing inher ent possessiv es. R ecall our
sentence:
( ) J ohn ’ s mother is mor e beautiful than Bob ’ s mother .
As w e saw , this sentence can be used in J ’ M to compar e two photographs; one
taken b y J ohn and the other b y Bob . at is, it can be used to asser t that J ohn ’ s photo of a
mother is mor e beautiful than Bob ’ s photo . N ev er theless, this fact seems to be inconsistent with
the rst checkbo x in the asser tion checklist pr oposed for ‘J ohn ’ s mother ’,
A C r e individual this expr ession is being used to talk about bears the mother-
hood r elation to another individual.
r e individual this expr ession is being used to talk about is signicantly
r elated to J ohn.
Ar e these counter examples to the pr oposed asser tion checklists? I don ’ t think so . I believ e that
cases like these can be explained b y r eecting on the natur e of asser tion.
Asser tion is a speech act go v erned b y social conv entions to the effect that b y per forming it
one is committing oneself to the tr uth of what is asser ted. ²¹ Both the speaker and the hear er
kno w this and kno w that they kno w it. S o, when a cooperativ e, sincer e, r eectiv e speaker wants
to asser t p she will exploit (i) the hear er ’ s expectations that she will be committed to the tr uth
of what she asser ts, (ii) the extra-linguistic r esour ces at hand, and (iii) the asser tion checklist
determined b y the linguistic meaning of her wor ds, to enable the hear er to get at p .
I n addition, cooperativ e, sincer e, r eectiv e speakers will tr y to do this in the most efficient
way , i.e., in a way that economiz es both the speaker ’ s and hear er ’ s effor ts to get atp . S tandar dly ,
²¹ S ome people hav e argued that y ou ar e also committing y ourself to kno wing what is asser ted. S ee, W illiamson
( , chapter ).
this means that the richer cooperativ e communicativ e situation is, the poor er the linguistically
encoded information will need to be. S o, when it ’ s clear to ev er y par ty inv olv ed in the conv ersa-
tion that the topic of conv ersation is J ohn ’ s painting of a book, a speaker might asser t something
tr ue b y pr edicating a pr oper ty to it, ev en if his choice of wor ds would nev er pick out a painting
in a differ ent communicativ e situation.
When Alice asser tiv ely utters ( ) in a communicativ e situation in which the topic of con-
v ersation is the book J ohn painted , she will succeed in pr edicating the pr oper ty of being color ful
of it, because the pr oper ty of being a book isn ’ t under discussion. A similar stor y might be giv en
to account for the use of ( ) in the photography contest. S ince the topic of conv ersation in
the r elev ant communicativ e situation is deciding which of the photos of a mother taken b y
J ohn and Bob is the winner of the photography competition, the conv ersational par ticipants ’
attention will be focused on the r elation pr edicated of them b y ( ), not on whether the r elata
hav e the pr oper ty of being a mother . I n that scenario, the use of ‘J ohn ’ s mother ’ and ‘Bob ’ s
mother ’ is good enough for the audience to successfully identify the r elata .
. Conclusion
M y main purpose in this chapter has been to challenge the or thodo x vie w in semantics about
context-sensitivity that I hav e called ‘S emantic D eterminism, ’ and pr opose a ne w pictur e of the
r elation betw een the linguistic meaning of a context-sensitiv e expr ession and the contents it can
be used to asser t. I hav e called this ne w pictur e ‘S emantic G uidance ’. A ccor ding to S emantic
G uidance, the linguistic meaning of an unambiguous, context-sensitiv e, declarativ e sentence
doesn ’ t x the pr opositions it can be used to asser t acr oss communicativ e situations. Rather , it
mer ely pr o vides speakers with an asser tion checklist that pr o vides some (incomplete) guidance
to war ds those pr opositions.
I hav e illustrated these claims b y offering an account of pr enominal possessiv es. e strat-
egy emplo y ed to explain these constr uctions has been to put the missing r elations betw een
the denotations of their constituents out of r each of their linguistic meanings, and leav e to
conv ersational par ticipants the task of inferring them using the r esour ces av ailable to them in
their r espectiv e communicativ e situations. I n doing so, I hav e tried to sho w the superiority of
S emantic G uidance o v er S emantic D eterminism.
I f this pictur e is on the right track, one would normally expect it to apply to other linguistic
constr uctions that shar e some of the most salient pr oper ties w e hav e seen pr enominal possessiv es
exhibit. I believ e these expectations ar e justied. I n the next chapter , I will explor e compound
nominals and argue that S emantic G uidance can handle them.
Chapter Compound N ominals
. I ntr oduction
R oughly , compound nominals ar e sequences of two or mor e wor ds that function as a single
noun, e.g., ‘ v acuum cleaner ’, ‘ candy stor e ’, ‘ pie char t ’, ‘ tear gas ’, ‘ crime inv estigation ’, ‘ v olume
feeding management success formula awar d ’, ‘ air craft bomb bay door actuating cylinders ’, etc. ¹
e most common kind of compound nominal in E nglish consists of two lex emes, the meaning
of which is a function of the meaning of the second constituent (the head) modied in some
impor tant way b y the rst constituent (the modier).
Although compounding is a v er y common pr oductiv e pr ocess in E nglish and other natural
languages, giving a uniform account of the contents compound nominals can be used to asser t
¹ ese constr uctions go also under the follo wing labels in the r elev ant literatur e: complex nominals , compounds ,
compound nouns , nominal compounds , noun compounds , noun pr emodiers , noun sequences , noun-noun compounds ,
and noun+noun compounds among others.
in differ ent occasions r emains a challenge. F or although it ’ s clear that the meaning of a com-
pound depends on the meaning of its constituents, what is asser ted b y uses of it doesn ’ t depend
ex clusiv ely on those meanings. is fact poses a challenge to the friend of S emantic D etermin-
ism that is v er y similar to that posed b y pr enominal possessiv e constr uctions. F or compar e the
follo wing expr essions: ‘ steel knife ’, ‘br ead knife ’, and ‘kitchen knife ’. Although their sur face
str uctur e is identical (N + ‘knife ’), they can be used to conv ey differ ent r elations betw een
the r efer ents of their constituent nouns: ‘ steel knife ’ may be used to r efer to a knife made fr om
steel; ‘br ead knife ’ may be used to talk about a knife for cutting br ead; and ‘kitchen knife ’ may
be used to r efer to a knife used in a kitchen.
An additional difficulty is that a single compound nominal can be felicitously used to r efer
to differ ent things in differ ent contexts. F or instance, ‘ child mur der er ’ could be used to r efer
to a person who mur ders childr en or to a child who mur ders; ‘ woman doctor ’ could be used
to r efer to a female doctor or a doctor whose patients ar e female; ‘ chocolate bo x ’ could be used
to r efer to a bo x with chocolates inside or to a bo x made out of chocolate; ‘ chemical purier
factor y ’ ( S ainsbur y ) could be used to r efer to a factor y that makes chemical puriers, a
factor y that uses chemical puriers, a factor y that puries chemicals, or a factor y that puries
with chemicals.
e main purpose of this chapter is to apply the S emantic G uidance frame wor k dev eloped
in the pr evious chapter to account for the contents compound nominals ar e literally used to
asser t. e plan for the chapter is as follo ws. F irst, I will star t b y identifying some of the most
impor tant featur es of compound nominals in E nglish. is will giv e us a better pictur e of what
an adequate account of them would need to explain. S econd, I will discuss some impor tant
semantic accounts of compound nominals in the literatur e. e upshot of this discussion will
be that they ar e inadequate and that this inadequacy is due to their implicit commitment to
S emantic D eterminism. F inally , I will sho w that S emantic G uidance has the r esour ces to deal
with the puzzling behavior compound nominals exhibit.
. Compound N ominals: S ome F eatur es
D espite appearances to the contrar y , specifying exactly what a compound nominal is, is still a
matter of debate among r esear chers. F or consider the wor d ‘honeymoon ’. I f y our characteriza-
tion of a compound nominal r equir es that its constituents be or thographically separated, then
‘honeymoon ’ doesn ’ t qualify as one. H o w ev er , it seems clear that this wor d is composed of
‘honey ’ and ‘ moon ’. S o, it seems that the or thogr aphic criterion is too r estrictiv e. is pr oblem
is made ev en worse b y the fact that many prima facie compounds in E nglish can hav e differ ent
spellings, i.e., they can appear or thographically separated, connected with a hyphen, or con-
catenated; e.g., ‘health car e ’, ‘health-car e ’, or ‘healthcar e ’. M or eo v er , so me authors ev en allo w
complex str uctur es like ‘ state-of-the-ar t ’, ‘ par t-of-speech ’ or ‘ o v er-the-counter ey e dr op ’ in the
compound nominal categor y . ² H ence, as a condition for compoundhood, the or thographic
criterion seems to be inadequate.
O ther authors—e.g. Chomsky & H alle ( )—opt for a phonological criterion for com-
poundhood in E nglish. A ccor ding to this criterion, if a wor dw pr eceding a nounn r eceiv es the
² S ee N ako v ( , ).
primar y str ess, then w forms a compound with n . F or example, since ‘black ’ gets the primar y
str ess in ‘blackboar d ’, ‘blackboar d ’ qualies as a compound. A ccor ding to this criterion, ‘black
boar d ’ doesn ’ t qualify as a compound because its components get equal str ess. As a general
description of the facts about E nglish, the phonological criterion seems to be r easonably good.
H o w e v er , it isn ’ t completely r eliable as ther e ar e phrases in E nglish that would intuitiv ely qual-
ify as compound nominals, but that ar e str essed on the right-hand component. E xamples of
this kind of constr uctions include Boston már athon , mor ning páper , summer níght , aluminum
fóil , steal brídge , etc. F ur thermor e, a competent E nglish speaker might v ar y her pr onunciation
of a giv en compound acr oss contexts in or der to denote differ ent things in those contexts. F or
example, a speaker might utter tó y factor y to denote a factor y that makes to ys and to y fáctor y to
denote a factor y that is a to y . us, as a tool for detecting compoundhood, the phonological
criterion is neither necessar y nor sufficient.
I f instead one chooses an ev olutionar y criterion for compounding, i.e., one that character-
iz es a compound nominal as the combination of two pr eviously giv en wor ds, then wor ds like
‘honeymoon ’ would qualify . O ne difficulty with this criterion is that it is way too inclusiv e,
i.e., many simple wor ds in contemporar y E nglish star ted as compound nominals long time ago,
but w er e pr ogr essiv ely lexicaliz ed to the point of changing their original spellings. F or example,
although ‘M onday ’ would be naturally consider ed as a simple wor d in contemporar y E nglish, it
would count as compound accor ding to the ev olutionar y criterion since its origins can be traced
back to the O ld N orse compound ‘M ōnandæg ’ meaning ‘ the day of the moon ’. Another dif-
culty is that the ev olutionar y criterion leav es out a lot of compounds whose constituents ar e
not wor ds but stems, i.e. uninected par ts of wor ds that ar e not themselv es wor ds. Although,
E nglish lacks this kind of expr essions, these compounds ar e pr etty common in inectional lan-
guages such as R ussian, C z ech, and S lo v ak. ³
I t isn ’ t my purpose her e to attempt to solv e these pr oblems, but to giv e a taste of some of
the immediate complications any complete account of compound nominals would hav e to face.
ese pr oblems ar e ev en worse for the friend of S emantic D eterminism whose ultimate task is to
pr o vide a mechanism that assigns a unique content to each and ev er y meaning ful expr ession of
the language. B ut w e don ’ t need to go that far to challenge S emantic D eterminism. F or ev en if
w e r estrict our attention to just the uncontr o v ersial cases of compound nominals in E nglish, w e
will nd data that is v er y har d to account for b y the S emantic D eterministic model. T o illustrate
these pr oblems, I will focus on compound nominals whose constituents ar e or thographically
separated nouns. ⁴
.. D ir ectionality
D espite the exibility that compounding displays in E nglish, ther e ar e syntactic r estrictions
go v erning it. is is par ticularly clear in compound nominals that hav e mor e than two con-
stituents. F or example, consider the compound ‘ colon cancer tumor suppr essor pr otein ’. W e
³ S ee Lieber & Š tekauer ( , ).
⁴ N ot all compounds in E nglish ar e noun+ noun compounds. er e ar e also adjectiv e+ noun compounds (‘hot
dog ’, ‘ white collar ’), v erb+ noun compounds (‘ pickpocket ’, ‘ cutthr oat ’), pr eposition+ noun compounds (‘ counter-
attack ’, ‘ upper-class ’), noun+ adjectiv e compounds (‘ army str ong ’, ‘ dog tir ed ’), adjectiv e+ adjectiv e compounds
(‘ dar k-gr een ’, ‘ dr y-clean ’), etc.
cannot r efer to this par ticular pr otein b y means of ‘ pr otein cancer colon suppr essor tumor ’ or
‘ cancer tumor pr otein suppr essor colon ’.
er e ar e morphosyntactic constraints on the pr oper wor d or dering or dir ectionality in E n-
glish compound nominals. I n most cases, compound nominals in E nglish ar e right-headed
and their constituents ar e or thographically separated. ⁵ is isn ’ t so in other natural languages
such as G erman, S w edish, or D anish in which compound nominals ar e almost ex clusiv ely
concatenated—i.e. their constituents ar e not or thographically separated—and in R omance
languages in which compound nominals ar e rar e and normally left-headed (e.g., ‘ perr o policia ’
and ‘ hombr e rana ’).
.. E ndocentr ic, E x ocentr ic, and Copulativ e Compounds
A common distinction one nds in the literatur e on compound nominals is that betw een endo-
centric compounds and ex ocentric compounds . An endocentric compound is a compound nom-
inal whose head determines the denotation of the whole compound and whose modiers at-
tribute some pr oper ty to it. D iffer ently put, the denotation of an endocentric compound is a
subclass of t he denotation of its head. ⁶ F or example, ‘ oliv e oil ’ and ‘ night club ’ ar e endocentric
because they denote a kind of oil and a kind of club r espectiv ely . e v ast majority of E nglish
compound nominals ar e endocentric.
⁵ er e ar e v er y fe w ex ceptions to the right-headed r ule in E nglish. O ne such ex ception is ‘ notar y public ’.
⁶ is pr oper ty of r elation betw een the denotation of the head and the denotation of the whole compound is
kno wn as hyponymy . us, w e say that a compound is endocentric iff it ’ s a hyponym of its head.
B y contrast, ex ocentric compounds ar e expr essions whose r efer ents ar en ’ t denoted b y any
of their constituents. F or example, ‘bo x spring ’ is ex ocentric because it ’ s used to denote an
entity that is neither a bo x nor a spring. er e ar e basically two ways describing this fact:
either ex ocentric compounds (i) lack heads, or (ii) hav e unpr onounced heads. E ither way , what
matter to us is that, as a r esult, the meaning of ex ocentric compounds can ’ t be usually wor ked
out fr om the meanings of their par ts. ⁷ S o, many of these compounds ar e highly lexicaliz ed and
idiomatic. is categor y includes compounds such as ‘bir d brain ’, ‘ ey e candy ’, ‘ poker face ’, ‘head
count ’, ‘ sugar daddy ’, ‘ir on maiden ’, ‘head case ’, and the concatenated compounds ‘ skinhead ’,
‘hunchback ’, ‘ pickpocket ’, and ‘bighead ’.
A thir d kind of compounds ar e the copulativ e compounds (also kno wn a s appositional or
co-or dinate compounds). Copulativ e compounds ar e compounds whose constituents might be
thought of as equally sharing head-like characteristics. F or instance, ‘ pr oducer-dir ector ’ is used
to r efer to someone who is both a pr oducer and a dir ector . O ther examples include ‘ o wner-
builder ’, ‘ secr etar y-lo v er ’, ‘br oker-analyst ’, etc.
I n what follo ws, I will focus on endocentric compound nominals whose constituents ar e
or thographically separated nouns. e question I want to addr ess is the follo wing: what is the
natur e of the r elationship betw een the linguistic meaning of an endocentric compound nominal
of the form⌜ N
1
N
2
⌝ and the contents it can be literally used to asser t in differ ent occasions?
I n the next section, I will r evie w some accounts giv en in the literatur e.
⁷ is is not to say that the meaning of an ex ocentric compound has nothing to do with the meanings of the
par ts of the compound. I n most cases, the meaning of an ex ocentric compound is r elated to that of its par ts.
. e Argument fr om Ambiguity
A common diagnostic of en docentric compound nominals is that they ar e systematically am-
biguous. O n this vie w , compounds of the form⌜ N
1
N
2
⌝ would be ambiguous betw een the
differ ent ways the denotations of N
1
and N
2
can be r elated. As in the case of pr enominal
possessiv es, the friend of S emantic D eterminism might appeal to this vie w to argue that his
theor etical commitments ar en ’ t incompatible with the data.
A classic v ersion of this vie w is defended b y Levi ( ). A ccor ding to her , ev er y endocentric
compound nominal is such that either (i) the head noun is a dev erbal nominalization and its
modier is interpr eted as an argument of the r elated v erb (e.g., ‘ swor d swallo w er ’ and ‘ chur ch
goer ’); or (ii) the two nouns ar e r elated b y one of the follo wing R eco v er ably D eletable P r edicates
(RDP s): CA USE
1
(e.g., ‘u vir us ’ denotes a vir us that causes the u), CA USE
2
(e.g., ‘ sno w
blindness ’ denotes a momentar y blindness caused b y exposur e to the ultraviolet rays r eected
fr om sno w or ice), HA VE
1
(e.g., ‘ college to wn ’ denotes a to wn that has a college or univ ersity the
pr esence of which per v ades its economic and social life), HA VE
2
(e.g., ‘ company assets ’ denotes
assets had b y a company), MAKE
1
(e.g., ‘honey bee ’ denotes a bee that makes honey), MAKE
2
(e.g., ‘ daisy chain ’ denotes a chain made out of daisy o w ers) ⁸ , USE (e.g., ‘ steam ir on ’ denotes
an ir on that uses steam to wor k), BE (e.g., ‘ woman driv er ’ denotes a driv er that is a woman), IN
(e.g., ‘ mountain lodge ’ denotes a lodge that is in a mountain), FOR (e.g., ‘headache pill ’ denotes
⁸ CA USE, HA VE, and MAKE hav e two v ariants each: one corr esponding to the activ e v oice and the other
corr esponding to the passiv e v oice.
a pill for headaches), FR OM (e.g., ‘bacon gr ease ’ denotes gr ease fr om bacon), and ABOUT (e.g.,
‘ adv entur e stor y ’ denotes a stor y about some adv entur e).
O n Levi ’ s account, ev er y r elation expr essed b y a compound nominal is deriv ed fr om one of
these tw elv e RDP s. A ccor ding to her , this degr ee of ambiguity is exible enough to account for
the seemingly semantic v ariability of compound nominals, but at the same time it ’ s r easonably
r e stricted for a hear er to identify the r elation intended b y a speaker when uttering them. As
any other case of ambiguity , this one will be r esolv ed pragmatically .
er e ar e w ell-kno wn difficulties with vie ws like this that pr opose multiple specic mean-
ings for a giv en compound. F irst, ther e ar e many r elations that one could asser t in a giv en
context b y an utterance of a giv en compound nominal that don ’ t t any of the pr oposed cat-
egories. F or example, giv en the r elev ant conv ersational backgr ound, pr esuppositions, etc. a
speaker could successfully use ‘ cinnamon leather ’ to r efer to a piece of leather that smells like
cinnamon or ‘ man hands ’ to r efer to hands that look like those of a man. M or e challenging
examples include compounds such as ‘ pumpkin carriage ’ (a carriage that turns into a pumpkin
at midnight), ‘ thalidomide par ent ’ (a par ent of a child born with deformities caused b y the
mother ’ s ingestion of thalidomide), and ‘ cranberr y morpheme ’ (a kind of morpheme ex empli-
ed b y ‘ cran ’ in ‘ cranberr y ’ that doesn ’ t hav e a meaning on its o wn). I n fact, ther e is no good
r e ason to r estrict the number of r elations that can be asser ted and conv ey ed b y a giv en com-
pound nominal. As D o wning puts it, “e existence of numer ous no v el compounds like these
guarantees the futility of any attempt to enumerate an absolute and nite class of compounding
r e lationships ” ( D o wning , ).
S econd, ev en if w e w er e able to accommodate all the possible r elations that can be asser ted
b y an utterance of a compound nominal into one of these RDP s, w e would still be at a lost
about the meanings of many of them. F or consider ‘ sleeping pill ’, ‘headache pill ’, and ‘ dog
pill ’. Although the r elations they ar e used to conv ey could be said to fall under FOR, the
identities of those r elations ar e quite differ ent. I f all y ou kne w about ‘ sleeping pill ’ is that it
denotes a pill for sleeping, ‘headache pill ’ that it denotes a pill for headaches, and ‘ dog pill ’
that it denotes a pill for dogs, y ou wouldn ’ t understand ho w they differ in the r elations they
ar e used to conv ey and asser t. e pr oblem with Levi ’ s inv entor y is that it ’ s too general to
adequately captur e these specic r elations; kno wing the categor y under which a giv en r elation
falls is insufficient to determine the content asser ted b y an utterance of a compound.
ir d, a major difficulty for this appr oach is that many compound nominals can be cate-
goriz ed under mor e than one RDP ev en when their uses don ’ t normally v ar y acr oss contexts.
F or example, ‘ car factor y ’ can be equally classied as a factor y for pr oducing cars (FOR), a
factor y that causes cars to be cr eated (CA USE), a factor y in which cars ar e pr oduced (IN), and
a factor y fr om which cars originate (FR OM). ⁹ D epending on ho w w e describe the r elation in
question, what aspect of it w e privilege, etc. the same compound nominal will be classied
differ ently . us, as a theor etical tool for determining the meanings of compound nominals,
Levi ’ s appr oach is unsatisfactor y .
⁹ I borr o w this example fr om Ó S éaghdha ( , ).
I t ’ s note wor thy that the objections raised against the Argument fr om Ambiguity don ’ t tar-
get the claim that some compound nominals ar e str uctur ally ambiguous. F or example, the
compound ‘brain stem cell ’ that is ambiguous betw een ( a ) and ( b ):
() a. [[ brain stem] cell]
b . [ brain [ stem cell]]
H o w ev er , the ambiguity in question isn ’ t due to the differ ent r elations these constr uctions can
conv ey , but due to the differ ent ways their heads can be constituted. us, if the objections
formulated in this section ar e corr ect, they sho w that the Argument fr om Ambiguity fails to
account for the contents compound nominals ar e used to asser t.
. e Argument fr om E xistential G eneralization
An alternativ e account of compound nominals that is compatible with S emantic D eterminism
is the follo wing: instead of thinking of compound nominals as semantically encoding a specic
r elation, one could think of them as expr essing that ther e is some r elation holding betw een
the denotations of their constituents. O n this vie w , sentences containing compound nominals
semantically encode full-edged pr opositions that ar e nonetheless general. P r esumably , then,
conv ersational par ticipants would pragmatically r eplace this existentially generaliz ed content
with one that contains the specic r elation R r equir ed in the r elev ant context.
O ne such pr oposal is offer ed b y B auer ( ), who r epr esents compound nominals as hav-
ing each an abstract pr o-v erb ¹⁰ among their constituents that could be interpr eted as ‘ ther e is
a connection betw een lex eme A and lex eme B in a compound of the form AB such as can be
pr edicted b y the speaker/hear er par tially on the basis of her kno wledge of the semantic make-up
of the lex emes inv olv ed and par tially on the basis of other pragmatic factors. ’ ( B auer , )
M or e r ecently , S ainsbur y ( ) has defended a similar vie w . H e defends what he calls
‘ the unspecic account ’ of compound nominals. O n this vie w , the meaning of a giv en com-
pound nominal is ‘ an unspecic but denite, unambiguous and complete meaning ’ ( S ainsbur y
, ). H e writes: ‘O n the unspecic account, the determinate meaning does not supply
specic information about ho w the satisers of the nouns ar e r elated ’ ( S ainsbur y , ).
S o, although S ainsbur y doesn ’ t explicitly say that the meanings of compound nominals inv olv e
existential generalization, it seems to follo w fr om two of the pr oper ties he assigns to them that
this is what he has in mind: unspecicity and completeness. F ollo wing W eiskopf ( ), w e
can r epr esent this pr oposal as follo ws:
() J N
1
N
2
K = lx[ N
′
2
(x)^9R9y[ N
′
1
(y)^R(x,y)]]
F or instance, on this account, ‘ queen bee ’ would be analyz ed like this:
() J queen beeK = lx[ B ee
′
(x)^9R9y[ Q ueen
′
(y)^R(x,y)]]
I nformally put, the meaning of ‘ queen bee ’ would be r oughly bee someho w r elated to a queen .
is pr oposal has the vir tue that it av oids the massiv e ambiguity that other vie ws r esult in and,
¹⁰ A pr o-v erb i s an expr ession that substitutes a v erb or a v erb phrase (just as a pr onoun substitutes a noun or a
noun phrase). A common pr o-v erb is ‘ do ’ as used in: Who has my notebook? I do [hav e it].
instead, understands the r ole of semantics as that of capturing what is common to asser ted
contents. N ev er theless, it has the same aws that its counterpar t in the case of pr enominal
possessiv es has.
e main pr oblem for this vie w is it makes the semantic content of compound nominals
utterly trivial. F or all it takes for ( ) to be satised is the existence anywher e in the univ erse
of at least one bee and one queen: no matter wher e they ar e, ther e will be a r elation holding
betw een them. is same point can be made b y appealing to negation. S uppose that I point at
a wor ker bee and utter ‘at bee is not a queen bee ’. I f the meaning of ‘ queen bee ’ is giv en b y
( ), then the pr oposition expr essed b y the sentence in question is blatantly false, i.e. it ’ s false
that that bee isn ’ t someho w r elated to a queen.
N o w , the natural r esponse to this worr y is to appeal to contextual domain r estriction. S o,
idea would be to say that the existential quantier pr o vided b y the meaning of ‘ queen bee ’
ranges only o v er those r elations that ar e contextually r elev ant. Although the in tuition behind
this pr oposal is on the right track, I don ’ t think that postulating an existentially generaliz ed
content is the best way to captur e it. As I noted in the pr evious chapter , such a content is
too w eak to ser v e any inter esting purpose other than keeping S tr ong P r opositionalism. S o, if
semantic content is so w eak that has to be systematically r eplaced b y a mor e specic content,
w e might just do away with it.
. e Argument F r om Co v er t I ndexicality
e pr evious accounts take the semantic contents of sentences containing compound nominals
to be both complete and semantically context- in sensitiv e. A ccor ding to the indexicalist, this is
wr ong. F or him, the best way to account for compound nominals without giving up any of
the pr opositionalist theses is to claim that they ar e indexical. M or e specically , he claims that
compound nominals contain among their constituents a syntactically r eal aphonic r elational
v ariable or hidden indexical expr ession the v alue of which is to be supplied b y the context of
utterance. A r ecent v ersion of this appr oach is defended b y W eiskopf ( ), who formulates
the follo wing semantic r ules specifying the meanings of compound nominals:
I. J N
1
N
2
K = lx[ N
′
2
(x)^R
⋆
(x, N
′
1
)]
II. J N
1
N
2
K = lx[ N
′
2
(x)^Qy( N
′
1
(y)^R
⋆
(x,y))]
III. J N
1
N
2
K = lx[ N
′
2
(x)^typically
′
(R
⋆
(x, N
′
1
))]
wher e ‘R
⋆
’ is an co v er t indexical expr ession whose character picks out ‘ the most r elev ant and
conv ersationally sensible r elation obtaining betw een N
1
and N
2
’ ( W eiskopf , ), and
‘Q ’ in R ule II is a contextually specied quantier .
H er e is ho w these r ules wor k. R ule I applies to compound nominals that ar e used to asser t a
r elation betw een either (i) an individual N
2
and the class of N
1
s, or (ii) the class of N
2
and the
class of N
1
s. E xamples include ‘ dog house ’ (a par ticular dog ’ s house or a house intended for use
b y dogs in general) and ‘r e depar tment ’ (a depar tment in charge of managing and extinguish-
ing r es generally). I t also handles (modulo the application of some type-shifting principles)
compounds that contain pr oper names as modiers, e.g., ‘Chomsky hierar chy ’, ‘M ontague
grammar ’, and ‘B ush doctrine ’.
R ule II handles compound nominals that ar e used to asser t a quantied r elation betw een
the denotations of N
2
and N
1
. W eiskopf giv es the follo wing examples: ‘hamburger plate ’ (used
to denote a plate that has some hamburgers on it), and ‘ mur der w eapon ’ (used to denote the
w eapon used in a par ticular mur der). I n the rst case w e hav e that Q is the existential quantier
andR
⋆
is the r elation of having on top of . us, accor ding to R ule II, ‘hamburger plate ’ denotes
a function that maps an entity x to tr uth iff x is a plate and ther e is some hamburger y such
that x has on top of it y . I n the second example, Q is the existential quantier and R
⋆
is the
r elation of being used in . R ule II, then, will say that ‘ mur der w eapon ’ denotes a function that
maps an entity x to tr uth iff x is a w eapon and ther e is a mur der y such that x is used in y .
O ther quantiers that can ll in for Q ar e ‘ a ’, ‘ many ’, ‘ most ’, etc.
R ule III applies to compound nominals that ar e used to asser t habitual, r egular or r ecurr ent
r elations that hold betw een the denotations of N
2
and N
1
. etypically
′
operator—W eiskopf
claims—is a ‘ pr opositional operator that giv es a sentence a habitual r eading ’ ( W eiskopf ,
). F or example, ‘ garbage man ’ is used to denote a man who habitually or typically collects
and r emo v es the garbage. H ence, R ule III is designed to handle generic uses of compound
nominals. ¹¹
¹¹ W eiskopf ackno wledges that these r ules could be r educed to just one: R ule I. H o w ev er , he thinks that such
a r eduction would obscur e the fact that quanticational and habitual r eadings of compound nominals ar e quite
distinct fr om each other and fr om nonce uses of compounds.
W eiskopf ’ s pr oposal is superior to the pr evious alternativ es at least in two r espects. F irst,
instead of engaging in a futile attempt to enumerate a complete list of possible specic r elations
that can be asser ted b y a use of a compound nominal, it classies all of them in just thr ee general
categories. S econd, although R ules I-III ar e general characterizations o f the possible asser table
r elations, they ar en ’ t as general as the existentially quantied pr oposal. S o, b y constraining
their generality in specic ways, W eiskopf makes his r ules mor e informativ e than S ainsbur y ’ s
pr oposal.
H o w ev er , this appr oach suffers fr om similar pr oblems as the indexicalist account of pr enom-
inal possessiv es. O ne of those pr oblems is r elated to the intr oduction of a hidden r elational
v ariable R
⋆
into LF . As any other v ariable, occurr ences of R
⋆
should be able to be bound.
H o w ev er , just like in the indexicalist account of pr enominal possessiv es, they don ’ t seem to be
bindable. W eiskopf ackno wledges this fact. ¹² S o, he suggest that R
⋆
might be understood as
a constant indexical. Although he ultimately doesn ’ t settle the issue one way or the other , he
makes it clear that R
⋆
is a r eal constituent with an indexical character . S o, what makes him
think that compound nominals hav e a co v er t constituent with an indexical character? W eiskopf
giv es the follo wing:
(CS) Context-S ensitivity: Compound nominals, like indexicals, expr ess differ ent
contents in differ ent contexts.
I ntuitiv ely , this is right. W e can use compound nominals to asser t differ ent r elations betw een
their constituents in differ ent contexts. I s this evidence that ther e is a hidden element in syntax
¹² S ee W eiskopf ( , –).
for the missing r elation? I don ’ t think so . As I will a rgue later on, the need to explain compound
nominals ’ context-sensitivity b y positing a co v er t constituent in their str uctur e is a consequence
of an implicit—and, I also think, unjustied—commitment to S emantic D eterminism.
(SI) S emantic I ncompleteness: Compound nominals, like indexicals, when taken
out of context ar e semantically incomplete.
W eiskopf is right that both indexicals and compound nominals ar e semantically incomplete
when taken out of context. H o w ev er , this fact pr o vides no evidence in suppor t of the claim
that compound nominals hav e co v er t indexicals at LF . All it sho ws is that the contents asser ted
b y utterances of compound nominals hav e constituents pr o vided b y sour ces other than their
heads and modiers. As I will sho w later on, w e could giv e an account that takes this pr oper ty
seriously and, at the same time, obser v es the fact that uses of these expr essions asser t complete
contents.
(DR) D ir ect R efer ence: Compound nominals, like indexicals, r etain the content they
inherit fr om their context of utterance when embedded in intensional contexts.
T o illustrate this pr oper ty , W eiskopf invites us to consider the follo wing sentence:
() I am closer to having enough bus money .
S u ppose that ( ) is utter ed in two differ ent contexts—c
1
andc
2
—b y Alice and Bob r espectiv ely .
S u ppose that while c
1
is a context in which Alice utters ( ) to asser t that she is saving money
to buy a bus,c
2
is a context in which Bob utters ( ) to asser t that he almost has enough money
to ride the bus. W eiskopf claims that while an asser tiv e utterance of ( ) in c
1
expr esses the
pr oposition that Alice is closer to having enough money to buy a bus, an asser tiv e utterance of
( ) inc
2
expr esses the pr oposition that Bob is closer to having enough money to buy a ride on
the bus. N ext, he invites us to consider ( ):
() E v er y day I am closer to having enough bus money . ( tempor al shift )
S uppose that Alice and Bob asser tiv ely utter ( ) inc
1
andc
2
r espectiv ely . W eiskopf claims that
despite of being embedded in an intensional context, asser tiv e utterances of ( ) in c
1
and c
2
asser t the v er y same r elations they w er e originally used to asser t when Alice and Bob asser tiv ely
utter ed ( ) in those contexts. I n other wor ds, since accor ding to W eiskopf (i) Alice ’ s utterance
of ( ) in c
1
expr esses the pr oposition that ev er y day Alice is closer to having enough money
to buy a bus r elativ e to c
1
, and (ii) Bob ’ s utterance of ( ) in c
2
expr esses the pr oposition that
ev er y day Bob is closer to having enough money to buy a ride on the bus r elativ e toc
2
, w e hav e
evidence in fav or of his indexicalist appr oach.
I don ’ t think that W eiskopf ’ s example giv es evidence that compound nominals ar e dir ectly
r e fer e ntial. A ccor ding to his analysis, the contents of ‘bus money ’ r elativ e to the contexts de-
scribed ar e the follo wing:
() a. J bus moneyK
c
1
= lx[ M oney
′
(x)^ B uy
′
(x, B us
′
)]
b . J bus moneyK
c
2
= lx[ M oney
′
(x)^ Ride
′
(x, B us
′
)]
N otice that, accor ding to this account, the denotations of ‘bus money ’ r elativ e to c
1
and c
2
ar e functions, not specic amounts of money: r elativ e to c
1
, it ’ s a function which maps an
individual x to tr uth iff x is money to buy a bus; and r elativ e toc
2
, it ’ s a function which maps
an individualx to tr uth iffx is m oney to ride a bus. S ince the amount of money to buy or ride
a bus will differ fr om world to world, ‘bus money ’ will not be rigid on W eiskopf ’ s account. And
since ev er y dir ectly r efer ential term is rigid, it follo ws that ‘bus money ’ isn ’ t dir ectly r efer ential.
S o, what explains the fact that the intentional operator ‘ ev er y day ’ doesn ’ t shift the r elations
asser ted b y utterances of ‘bus money ’ inc
1
andc
2
r espectiv ely? e explanation is that, being
a temporal operator , ‘ ev er y day ’ can only shift the time of ev aluation of its sentential argument,
not the r elations asser ted b y a compound nominal occurring in it. I f this is right and indexicals
ar e d ir ectly r efer ential (as W eiskopf thinks), then compound nominals ar en ’ t indexicals. ¹³
(DA) D eictic and Anaphor ic U ses: Compound nominals, like indexicals, hav e deictic
and anaphoric uses.
T o illustrate the ability of compound nominals to hav e deictic uses, W eiskopf giv es the follo wing
example (slightly modied her e):
S S Alice and Bob ar e walking on the woods when suddenly both see a squirr el standing
upright on top of a stump . Alice utters :
¹³ H er e it might be adequ ate to use C&L ’ s Collectiv e D escription T est (see chapter ) to v erify whether ‘bus
money ’ is indexical. (As I argue in chapter , C&L ’ s tests ar e good only for indexicals, not for context-sensitiv e
expr essions in general). A ccor ding to the test, ‘bus money ’ is indexical only if y ou can ’ t infer fr om Alice ’ s and
Bob ’ s utterance of ‘I hav e enough bus money ’ in c
1
and c
2
r espectiv ely , the sentence ‘Both Alice and Bob hav e
enough bus money ’. B ut y ou can! I f a contextc
3
is such that (i) a bus ticket costs one dollar , (ii) buses sell for one
hundr ed dollars, and (iii) Alice and Bob hav e a hundr ed and one dollars each, the coll ectiv e description would be
tr ue r elativ e toc
3
.
() Look at that stump squirr el.
I n this scenario, it seems ob vious that the r elation asser ted b y Alice ’ s use of ‘ stump squirr el ’ is
standing upright on top of . What determines this r elation and not others? W eiskopf suggests
that it ’ s Alice ’ s intention to draw Bob ’ s attention to a per ceptually salient featur e of the squirr el.
Cases like this lead W eiskopf to think that compound nominals, like demonstrativ es, hav e
deictic uses. As for anaphoric uses, W eiskopf giv es the follo wing examples:
() e US P ostal S er vice is issuing a ne w stamp bearing an image of the older ,
fatter E lvis. e E lvis stamp is not expected to sell v er y w ell.
() e curators of G raceland ar e auctioning a stamp that is believ ed to hav e been
licked b y E lvis himself. e E lvis stamp is expected to fetch a high price at
auction.
H er e—W eiskopf claims—the occurr ences of ‘E lvis stamp ’ get a differ ent contents depending
on the sentence that pr ecedes it. S ince compound nominals hav e both deictic and anaphoric
uses, W eiskopf concludes that they must contain a co v er t indexical at LF .
I don ’ t think that W eiskopf ’ s examples sho w what he thinks they sho w . e rst thing to
notice is that the occurr ence of ‘ stump squirr el ’ in ( ) is par t of the complex demonstrativ e ‘ that
stump squirr el ’. S o, the fact that ‘ stump squirr el ’ in ( ) has a deictic av or doesn ’ t pr o vide any
evidence that it, as used b y Alice, is deictic. I n fact, par t of the r eason w e hav e str ong intuitions
about the identity of the asser ted r elation betw een a squirr el and a stump in W eiskopf ’ s example
seems to be the fact that it has been made salient b y its being mentioned in the description of
the scenario in which Alice utters ( ). F or suppose w e changed the scenario as follo ws:
S S Alice and Bob ar e biologist looking for a v er y rar e species of squirr el that only liv es
ar ound stumps. After spending w eeks in the woods looking for it, they nally nd one
standing upright on top of a stump . rilled, Alice utters:
( ) Look at that stump squirr el.
D id Alice asser t the standing upright on top of r elation? G iv en the backgr ound pr esuppositions
of the scenario described, it doesn ’ t seem so . B ut, if ‘ stump squirr el ’ w er e indexical, this shouldn ’ t
be possible, as the squirr el in question was in fact standing upright on top of a stump .
N otice that in S S , Bob might be able to pick out the right squirr el with-
out kno wing exactly what r elation betw een the squirr el and the stump Alice was intending to
conv ey . I n fact, he might be able to identify the squirr el Alice is talking about ev en if he didn ’ t
kno w the meaning of the wor d ‘ stump ’. As long as he kno ws the meanings of ‘ that ’ and ‘ squir-
r e l ’ , and ther e is a per ceptually salient squirr el in the context, he would be normally expected
to succeed in identifying the right squirr el.
As for W eiskopf ’ s discussion of ( ) and ( ), notice that the compound in question is par t of
the incomplete denite description ‘e E lvis stamp ’. S ince incomplete denite descriptions ar e
context-sensitiv e and hav e anaphoric uses, it shouldn ’ t be surprising that W eiskopf ’ s examples
trigger the intuitions he appeals to . I n fact, w e can get the same r esults if instead of ( ) and ( )
w e hav e the follo wing:
() e US P ostal S er vice is issuing a ne w stamp bearing an image of the older ,
fatter E lvis. e stamp is not expected to sell v er y w ell.
() e curators of G raceland ar e auctioning a stamp that is believ ed to hav e been
licked b y E lvis himself. e stamp is expected to fetch a high price at auction.
D o these examples sho w that the incomplete denite description ‘ the stamp ’ has a co v er t in-
dexical in its str uctur e? I don ’ t think so . O nly if y ou ar e committed to S emantic D eterminism,
y ou might be tempted b y this alternativ e. I n the next section, I am going to offer an account
that dispenses with S emantic D eterminism.
. e S emantic G uidance A ppr oach
A v ast subclass of endocentric compound nominals in E nglish wor ks v er y similarly to pr enom-
inal possessiv es. I n fact, in many cases ther e is no semantic distinction betw een this kind of
compounds and pr enominal possessiv es. F or example, although the possessiv e mar ker is used
in ‘hen ’ s egg ’ and ‘bir d ’ s egg ’, it isn ’ t in ‘ ostrich egg ’, ‘ goose egg ’, and ‘sh egg ’. ¹⁴
I n addition, many compound nominals in E nglish a r e literally translated into other natural
languages in the same way as E nglish pr enominal possessiv es ar e. is is par ticularly the case
when ther e is no syntactic differ ence betw een them in the language in question or the sets
¹⁴ S ee T aylor ( , chapter ).
of constr uctions that fall under each categor y don ’ t coincide with their E nglish counterpar ts.
F or instance, both the compound nominal ‘br ead knife ’ and the pr enominal possessiv e ‘br ead ’ s
cr ust ’ would be translated into S panish constr uctions of the same post-nominal possessiv e form,
i.e., ‘ cuchillo de pan ’ and ‘ cor te za de pan ’ r espectiv ely . is suggests that whatev er our account
of endocentric compound nominals is it should be compatible with our account of pr enominal
possessiv es. is is pr ecisely what I intend to do her e.
S emantic G uidance, the appr oach w e used to account for pr enominal possessiv es in the
pr evious chapter , can be easily extended to endocentric compounds. Consider the follo wing
sentence:
() e br occoli politician is inv olv ed in a scandal.
D epending on the communicativ e situation, this sentence can be used to asser t a pr oposition
about a politician who gr o ws br occoli, pr omotes the consumption of br occoli, defends the
inter ests of br occoli farmers, expor ts br occoli, has pr oposed a bill to eradicate br occoli, looks
like a br occoli, etc. S o, an adequate asser tion checklist for ( ) would determine a global
acceptability range that includes these and all the contents ( ) can be used to asser t acr oss
communicativ e situations. H o w would such a checklist go? W ell, w e could build our checklist
along the follo wing lines:
A C r is sentence is being used to pr edicate the pr oper ty of being inv olv ed in a
scandal of the individual the sentence is used to talk about.
r e individual this sentence is being used to talk about is a politician.
r e individual this sentence is being used to talk about is signicantly r elated
to the pr oper ty of being a br occoli .
I f my use of ( ) is such that it complies with each and ev er y item in this checklist, I will succeed
in asser ting a singular pr oposition. As in the case of pr enominal possessiv es, this checklist
doesn ’ t hav e a checkbo x for the identity of the r elation betw een a politician and the pr oper ty
of being a br occoli . H o w ev er , it tells me that, if I want to asser t a pr oposition, the politician I’ m
talking about must be signicantly r elated to the pr oper ty of being a br occoli .
O n the semantic G uidance pictur e, whether a giv en r elation R satises this constraint isn ’ t
something the linguistic meaning of ( ) car es about. I n fact, it doesn ’ t ev en tell us what exactly
is to be signicantly r elated to the pr oper ty of being a br occoli . ese ar e issues that ar e to be
locally determined in accor dance with shar ed backgr ound pr esuppositions, common goals, and
mutual expectations of conv ersational par ticipants.
.. G lobal Constraints
Let ’ s focus on the global constraints imposed b y the meaning of endocentric compound nom-
inals.
e rst one is imposed b y dir ectionality and headedness. As w e saw mentioned in section
.. , endocentric compound nominals in E nglish ar e normally right-headed. is explains,
for instance, why ‘br occoli politician ’ is used to denote a politician (or someone who r esembles
one) and not a br occoli. e modiers, on the other hand, will pr o vide pr oper ties that will
narr o w do wn the pool of candidate asser ted contents of the compound as a whole. I n the
case of ‘br occoli politician ’, only those politicians (or politician-like) signicantly r elated to the
pr oper ty of being a br occoli (accor ding to some locally determined or dering of signicance)
will be members of the global acceptability range determined b y its asser tion checklist.
is leads us to the second global constraint: that the denotations of the two constituents
must be signicantly r elated. is constraint is imposed b y the v er y natur e of modier-head
constr uctions: the syntactic str uctur e of this kind of constr uctions imposes the constraint that
the denotations of the head and modiers must be r elated. Kno wing this, is a r equir ement for
using modier-head constr uctions competently . is fact explains why cooperativ e speakers
will not normally use ‘br occoli politician ’ to asser t politician someho w r elated to a br occoli ¹⁵ or
to asser t politician not r elated in any signicant way to a br occoli .
is constraint is cr ucial for assembling the contents asser ted b y utterances of compound
nominals. F or it ’ s in vir tue of this constraint that linguistically competent conv ersational par-
ticipants will look for a r elev ant r elation upon hearing an utterance of a compound nominal.
.. Local Constraints
J ust kno wing that the contents asser ted b y uses of ( ) must satisfy A C will not help anybody to identify the one a speaker intended b y her utterance of it. F or the
acceptability range it determines is too big to be useful for asser tion. S o, in or der to narr o w
do wn this set conv ersational par ticipants will need to appeal to local constraints imposed b y
¹⁵ is might also explain the superuity of the Argument fr om E xistential G eneralization.
par ticular communicativ e situations and the maxims and principles of conv ersation and rational
interaction.
e rst task will be to determine what exactly counts as signicantly r elated to the pr oper ty
of being a br occoli in a giv en communicativ e situation. er e ar e differ ent ways for a r elation to
gain signicance. S ome hav e to do with the spatio-temporal setting itself , and others to course
and purposes of the conv ersation. F or instance, imagine the follo wing scenario:
B P
Alice and Bob ar e in a H allo w een par ty hosted b y Alfr ed, a w ell-kno wn politician.
Alfr ed is dr essed as a giant br occoli. H e has been all o v er the ne ws lately because of the
inv estigation he ’ s conducting into a possible corr uption case inv olving B er t, a fello w
high-pr ole politician. B er t has been identied as a suspect in an illegal transaction
inv olving w ealthy br occoli expor ters. N oticing that Alice has been talking to Alfr ed,
Bob says to her:
() I just saw y ou talking to the br occoli politician. H e seems to be a v er y nice guy
…unlike the r eal br occoli politician, who seems to be a cr ook. I don ’ t like that
br occoli politician. I bet he ’ s lying when he says that he has nothing to do with
the whole br occoli scandal. I just hope w e ’ ll kno w the tr uth about the br occoli
politician pr etty soon.
I ntuitiv ely , the rst occurr ence of ‘ the br occoli politician ’ in ( ) picks out Alfr ed, the politician
dr essed as a giant br occoli, wher eas the last occurr ence picks out B er t, the politician inv olv ed
in the br occoli scandal.
e r easons the rst o ccurr ence picks out Alfr ed ar e differ ent fr om the r easons the last
occurr ence picks out B er t. e rst occurr ence picks out Alfr ed because Alice and Bob kno w
he is a politician, kno w that Alice has just been talking to him, and ar e able to see that he is
w earing a br occoli costume. ese facts, together with the asser tion checklist for ‘ the br occoli
politician ’ guide them dir ectly to war ds Alfr ed. e last occurr ence picks out B er t because he
was raised to salience in Bob ’ s monologue, Alice and Bob kno w he is a politician, and also kno w
he is inv olv ed in a scandal inv olving br occoli farmers. ese facts, together with the asser tion
checklist for ‘ the br occoli politician ’ guide them dir ectly to war ds B er t.
e fact that the r elev ant r elation betw een the denotations of a compound nominal ’ s con-
stituents can be pr o vided either b y the spatio-temporal setting of the conv ersation or b y the
discourse itself explains the intuition behind W eiskopf ’ s claim that compound nominals hav e
deictic as w ell as anaphoric uses. is doesn ’ t pr o vide evidence that ther e is an indexical hidden
in deep str uctur e. I t just sho ws that the meaning of a compound nominal instr uct us to look for
a contextually signicant r elation. Rational, w ell-informed agents will follo w this instr uction
b y grabbing—so to speak—the most ob vious, r easonable, salient r elation betw een the pr oper-
ties denoted b y the constituents of the compound nominal. S ometimes that r elation will be
pr o vided b y the wor ds utter ed in the surr ounding discourse and sometimes it will be pr o vided
b y the communicativ e situation. Whether a giv en R will do the job , is something conv ersa-
tional par ticipants will hav e to decide on the y based on the conv entions go v erning asser tion,
conv ersational maxims, and general principles of rationality .
N o w , not ev er y endocentric compound can be used as fr eely as ‘br occoli politician ’ can.
S ome of them hav e pr etty standar d uses. As an example, consider the compound ‘bug spray ’.
S tandar dly , this compound is used to r efer to a spray for killing bugs. H o w ev er , nothing pr e-
v ents us fr om using it in par ticular communicativ e situations to r efer to a spray made fr om bugs,
a spray for attr acting bugs, or ev en a spray pr oduced b y bugs. I s ther e something in the linguis-
tic meaning of ‘bug spray ’ that fav ors the killing -r elation o v er the others? I s ther e a hidden
typically
′
operator suggesting it to us? N ot at all. e killing -r elation is normally the default
one, simply because it ’ s the r elation ‘bug spray ’ is most fr equently used to asser t. H o w ev er , in
ex ceptional communicativ e situations, the other r elations could be giv en priority o v er it.
S imilarly for the compound ‘beach r esor t ’. N ormally , it ’ s used to r efer to a r esor t located
on or near a beach. D oes it mean that its linguistic meaning determines the location -r elation?
A ccor ding to S emantic G uidance, it doesn ’ t. H o w ev er , it differs fr om ‘br occoli politician ’ in
that the modier ‘beach ’ denotes a location, which puts the location -r elation in the top position
of its or der ed acceptability ranges. S ince, in addition, w e kno w that it ’ s commonly used to
denote to a r esor t located on or near a beach, w e will take it as the default asser ted content. As
in the case of inher ent possessiv es, this default can be o v err uled b y enriching the communicativ e
situation.
Another sour ce of signicant r elations ar e r elational nouns. F or instance, the default asser ted
content of ‘ duck feet ’ will be feet of a duck , simply because nouns denoting body par ts ar e r ela-
tional, i.e., they ar e dened in r elation to another entity . O ther examples include ‘bus driv er ’,
‘je w elr y thief ’, ‘ dog lo v er ’, ‘ univ ersity pr esident ’, ‘ giraffe o wner ’, etc. N otice that all of these
compounds simply pr o vide default r elations that can be pragmatically o v erridden. S imilarly for
nouns that denote tools or utensils. I f the head noun denotes a tool, and the modier denotes
something r elev antly r elated to its function, then that function might pr o vide a default r ela-
tion. is would explain default interpr etations of ‘ salad plate ’, ‘br ead knife ’ and, W eiskopf ’ s
example, ‘hamburger plate ’. Again, w e don ’ t n eed to write a special (quanticational) semantic
r ule to handle these cases.
.. F ilter ing and O r der ing
I n highly cooperativ e situations, conv ersational par ticipants will immediately kno w whether
a giv en r elation is signicant or not. H o w so? W ell, cooperativ e situations ar e situations in
which conv ersational par ticipants shar e a lot. S ome of the things they shar e ar e conv ersational
goals and inter ests. S ince, in conv ersation, signicance is a gradable r elation betw een pieces
of information and shar ed conv ersational goals and inter ests, conv ersational par ticipants in
highly cooperativ e situations will shar e r oughly the same ranking of signicance. is shar ed
ranking of signicance is what allo ws them to constitute local acceptability ranges b y enabling
them to lter out members of global acceptability ranges that don ’ t ser v e the purposes of their
conv ersation.
H o w ev er , in less cooperativ e situations—e.g., situations in which ther e isn ’ t a shar ed ranking
of signicance—ltering will leav e too many equally good candidates for asser ted content. I n
those situations, conv ersational par ticipants will hav e to hypothesiz e a ranking of signicance
based on the information they hav e. W e call this pr ocess ‘ or dering ’. F or instance, imagine that
both Alfr ed and B er t ar e in the H allo wing par ty described in B P . S uppose
that ev er ybody in the par ty kno ws that both ar e politicians, and that B er t is inv olv ed in a scandal
r elated to some br occoli expor ters. S uppose also that B er t, who is dr essed as a giant br occoli, is
having a heated discussion with Alfr ed. N o w suppose that Bob says to Alice:
() e br occoli politician o w es me some money .
I n that scenario, Alice might not be in a position to identify the pr oposition asser ted. Assuming
that she didn ’ t kno w that either of the politicians o w ed Bob some money , she will hav e to rank
the options based on the information and beliefs she alr eady has. F or example, she might think
that giv en that B er t is inv olv ed in some shady business, it ’ s highly pr obable that he doesn ’ t honor
his debts either . us, she might rank thex -is-inv olv ed-in-a-scandal-inv olving-y r elation a s the
most likely candidate for the asser ted r elation. B y contrast, if she thinks that Bob would nev er
lend money to someone as untr ustwor thy as B er t, she might put the x -is-dr essed-as-y r elation
at the top of her list.
.. U nspecic R elations
An immediate adv antage of adopting S emantic G uidance instead of S emantic D eterministic
is that, b y appealing to general principles go v erning rational interaction and conv ersation to
explain the data, it simplies considerably the semantics. is is good ne ws if y ou think of
linguistic meaning as that which must be kno wn, in addition to our kno wledge about the
world and the par ticular conv ersational situation, in or der to competently use and understand
the expr essions of the language. F or the mor e abstract, sophisticated, and computationally
demanding y our semantics is, the less likely it captur es what w e must kno w to pr operly use and
understand an expr ession. ¹⁶
B ut ther e is another impor tant adv antage that I would like to point out. A ccor ding to
S emantic D eterminism, giv en a set of parametersc , the unique content asser ted b y a literal use
of a r elativ e to c is x ed b y a ’ s linguistic meaning. H o w ev er , many times w e deliberately use
compound nominals without having any specic r elation in mind. F or consider the follo wing
case:
L W
Alice is telling Bob a stor y her father told her about his teens. e stor y is about a
beautiful woman her father was in lo v e with. I n the middle of the stor y , Alice pauses
and says:
() N o w that I think of it, I don ’ t kno w much about that woman. I’ m not sur e
whether my dad met her in the librar y , or whether she wor ked in a librar y , or
whether she lo v ed libraries, or perhaps …. Anyho w , the libr ar y woman looked
liked one of those H ollywood div as of the s…
¹⁶ F or mor e on the division of labor betw een semantics and pragmatics, see chapter .
What r elation betw een a woman and a librar y did Alice asser ted when she utter ed ‘librar y
woman ’? M any specic ones? A v ague one? I t isn ’ t clear . is is ne for us. F or ther e is a
v er y natural explanation of these uses in the S emantic G uidance frame wor k dev eloped her e.
A ccor ding to it, linguistic meaning just leads us to war ds a set of contents fr om which w e will
hav e to choose accor ding to our common backgr ound pr esuppositions, shar ed goals, mutual
expectations, and principles of rationality .
S ince for the purposes of Alice ’ s stor y it isn ’ t impor tant to choose a specic r elation betw een a
woman and a librar y , neither the global co nstraints imposed b y the meaning of ‘librar y woman ’
nor the local constraints imposed b y her communicativ e situation will choose one for us. As
long as both Alice and Bob hav e some set of r elations in common, they will be able to use and
understand the compound ‘librar y woman ’.
. Conclusion
I n this chapter , I hav e offer ed an account of the contents asser ted b y literal uses of endocentric
compound nominals based on the S emantic G uidance frame wor k dev eloped in the pr evious
chapter . As in the case of pr enominal possessiv es, S emantic G uidance assigns non-exhaustiv e
asser tion checklist to endocentric compound nominals that don ’ t specify any r elation betw een
the denotations of their constituents. O n this pictur e, the linguistic meaning of a compound
nominal guides conv ersational par ticipants to war ds a set of acceptable r elations fr om which they
will hav e to choose one or mor e depending on their common inter ests and goals, backgr ound
pr esuppositions, mutual expectations, and general principles of rational interaction. F inally , I
hav e argued that this appr oach has r eal adv antages o v er S emantic D eterministic theories.
H o w about indexicals? ey look like the home territor y of S emantic D eterminism. I
believ e that S emantic G uidance has the r esour ces to not only account for the data that con-
temporar y theories of indexicals committed to S emantic D eterminism handle best, but also
account for a wider range of data. I aim to sho w this in the next chapter .
Chapter I ndexicals
. I ntr oduction
What is the natur e of the r elationship betw een the linguistic meaning of a context-sensitiv e ex-
pr essiona and the informationa is used to literally asser t? P r ompted b y this question, philoso-
phers hav e traditionally turned to indexicals for an answ er . F or example, if I tell y ou ‘I’ m
talking to y ou ’, I would asser t a pr oposition that would be differ ent fr om the pr oposition y ou
would asser t if y ou w er e to utter it. H o w ev er , the r elation betw een these pr opositions and the
linguistic meaning of the sentence seems to r emain the same: the linguistic meaning of ‘I am
talking to y ou ’ fully deter mines , for ev er y context, what is asser ted b y literal utterances of it.
O r consider the sentence ‘ T oday is my bir thday . ’ B y uttering it on a F riday , I asser t, of that
F riday , that it ’ s my bir thday , and b y doing so on a W ednesday , I asser t, of that W ednesday ,
that it ’ s my bir thday . As befor e, the r elation betw een the linguistic meaning of the sentence
and the pr opositions asser ted acr oss contexts seems to be the same: its linguistic meaning fully
deter mines its asser ted contents r elativ e to those contexts.
S ince indexicals ar e commonly taken as the paradigmatic set of context-sensitiv e expr es-
sions, cases like these hav e motiv ated many philosophers and linguists to pr esuppose S emantic
D eterminism—r oughly , the thesis accor ding to which, giv en the information contributed b y a
contextc , the pr oposition asser ted b y a literal utterance of an unambiguous, context-sensitiv e,
declarativ e sentenceS is fully deter mined b y the linguistic meaning ofS inc . U nfor tunately for
S emantic D eterminism, semantic theories committed to it hav e r eal difficulties accounting for
a signicant range of data. As I hav e argued in chapters and , theories of meaning commit-
ted to S emantic D eterminism don ’ t giv e us a satisfactor y account of the pr opositions asser ted
b y literal uses of context-sensitiv e sentences containing pr enominal possessiv es and compound
nominals.
B ased on these data, I hav e urged that w e should adopt a differ ent conception of the r elation
betw een linguistic meaning and asser ted content. is conception is the basis of an alternativ e
appr oach I c all ‘S emantic G uidance ’. I n a nutshell, S emantic G uidance holds that the linguistic
meaning of an unambiguous, context-sensitiv e, declarativ e sentenceS is a set of constr aints that
a pr oposition must satisfy in or der to qualify as a candidate asser ted content of a literal utter-
ance of S . e set of constraints determined b y the linguistic meanings of all the meaning ful
constituents of S together with the constraints imposed b y S ’ s syntactic str uctur e constitute
what I call ‘S ’ s asser tion checklist ’. O n this pictur e, S ’ s asser tion checklist is non-exhaustiv e,
i.e., it will leav e, in ev er y communicativ e situation, multiple candidate asser ted contents for
S as liv e options. C r ucially for this account, this is as far as linguistic meaning can take us in
the pr ocess of identifying S ’ s asser ted content. O n this vie w , linguistic meaning simply guides
conv ersational par ticipants to war ds asser ted contents without xing them. Armed with this
information, conv ersational par ticipants will hav e to narr o w do wn the set of candidate asser ted
contents appealing to their shar ed backgr ound pr esuppositions, mutual expectations, and pur-
poses of their communicativ e ex changes, together with conv ersational maxims and principles
of rational interaction.
e purpose of this chapter is to extend S emantic G uidance to the home territor y of S e-
mantic D eterminism: indexicals. I aim to sho w that S emantic G uidance can offer a smooth
explanation of or dinar y uses of indexicals that pose signicant difficulties to the or thodo x S e-
mantic D eterministic pictur e.
. S emantic G uidance & S emantic D eter minism
S uppose y ou hav e to explain the linguistic meaning of the personal pr onoun ‘ she ’ to someone
who doesn ’ t kno w it. What would y ou say? P r esumably , y ou would say that ‘ she ’ is used b y
speakers to r efer to a single female individual who is neither the speaker nor the addr essee. I f
asked for an example, y ou might say something along the follo wing lines. e sentence,
() S he is a philosopher
can be used b y me at a giv en moment to asser t that Alice is a philosopher (if Alice is some-
ho w made sufficiently r elev ant to our conv ersation) and, at a differ ent moment, to asser t that
Amanda is a philosopher (if Amanda is someho w made sufficiently r elev ant to our conv ersa-
tion).
is little stor y captur es what is essential about the linguistic meaning and deictic use of the
wor d ‘ she ’. N otice that it doesn ’ t mention contexts, demonstrations or demonstr ata , contents
r elativ e to contexts, or tr uth in a model. O f course, it could hav e appealed to them, but it
didn ’ t hav e to . What is cr ucial for our stor y to sho w is that the linguistic meaning of ( ) itself
doesn ’ t giv e us a pr oposition. I t just imposes some impor tant conditions that literal deictic uses
of ( ) must satisfy in or der to asser t v arious pr opositions at differ ent times. What pictur e of
linguistic meaning best captur es these obser v ations? H er e is one that I like: y ou can think of
the linguistic meaning of ( ) as pr o viding y ou with the follo wing asser tion checklist:
A C r is sentence is being used to pr edicate the pr oper ty of being a philosopher
of the individual the sentence is used to talk about.
r e individual this sentence is being used to talk about is singular .
r e individual this sentence is being used to talk about is female.
r e individual this sentence is being used to talk about is neither its utter er
nor the utter er ’ s addr essee.
I f I am able to check each and ev er y item in this checklist, I will succeed in asser ting a singular
pr oposition. Which pr oposition would I asser t? e checklist doesn ’ t say . All that it giv es me
is a set of constr aints that my literal deictic uses of ( ) must satisfy if I want to successfully asser t
such a pr oposition. ¹ O f these constraints, the rst is stricter than the r est as it determines the
identity of the main pr oper ty pr edicated of the sentence ’ s subject. S o, if my use fails to satisfy
that constraint, I will likely fail to asser t anything at all. H o w ev er , if it fails to satisfy only one
of the r emaining constraints, depending on the communicativ e situation, my hear er might be
able to gur e out what I meant. I n such a scenario, although my asser tion might be defectiv e,
it might be intelligible and ev en tr ue.
Asser tion checklists play two r oles. F irst, they ar e per for mance assessment tools : they help
speakers to ensur e their asser tions ar e in good shape. S econd, they ar e interpr etation guiding
tools : they help hear ers to narr o w do wn the set of possible pr opositions that a speaker can
literally asser t b y uttering a sentence at a giv en moment. I call this pictur e of the r elation
betw een linguistic meaning and asser ted content ‘S emantic G uidance ’:
S G :
I f S is an unambiguous, context-sensitiv e, declarativ e sentence and p is a pr oposition
asser ted (without conv ersational implicatur es) b y a literal utterance ofS , thenp satises
the non-exhaustiv e set of constr aints imposed b y the linguistic meaning ofS .
A ccor ding to S emantic G uidance, the linguistic meaning of S just giv es the backbone ar ound
which asser ted contents ar e to be constr ucted. e r est of the component par ts needed to
assemble them will come fr om discursiv e and non-discursiv e sour ces. I n highly cooperativ e
communicativ e situations, these sour ces will be ob vious and r eadily av ailable to ev er y conv ersa-
¹ I borr o w the notion of linguistic meaning as set of constr aints fr om S oames ( , , , ).
tional par ticipant. I n less cooperativ e situations, conv ersational par ticipants will hav e to sear ch
for them.
e pictur e w e get fr om S emantic G uidance is simple and intuitiv e. N otwithstanding, it
isn ’ t the pictur e fav or ed b y most contemporar y philosophers of language. ey pr efer a pic-
tur e in which linguistic meaning takes a mor e pr ominent r ole in the determination of asser ted
content. A pictur e in which not just speakers, but ev en linguistic expr essions themselv es say
things. ² T o this purpose, they built the technical notion of content r elativ e to context and they
thought of the linguistic meaning of a context-sensitiv e expr ession as a r ule that fully deter mines
that content. I call this pictur e ‘S emantic D eterminism ’:
S D :
I f S is an unambiguous, context-sensitiv e, declarativ e sentence and p is a pr oposition
asser ted (without conv ersational implicatur es) b y a literal utterance ofS in a contextc ,
then p is fully deter mined b y the linguistic meaning ofS inc .
R oughly , her e is ho w S emantic D eterminism would be typically used to explain the asser ted
content of ( ). F irst, w e obser v e that whenev er J ohn utters ( ) while gesturing to war ds Alice, he
successfully asser ts the pr oposition that Alice is a philosopher , and whenev er he utters it while
gesturing to war ds Amanda, he successfully asser ts the pr oposition that Amanda is a philosopher .
N ext, w e ask ourselv es: What is the r elation betw een the linguistic meaning of ( ) and
those pr opositions? I mpr essed b y the intensional systems dev eloped b y philosopher-logicians,
² e widespr ead use of the phrase ‘ what is said ’ to describe semantic content is r ev ealing. I discuss this use of
‘ what is said ’ in chapter .
w e think of linguistic meaning as a function that maps packages of entities to contents. O ne of
those packages will contain Alice and the other will contain Amanda. W e call those packages
‘ contexts ’, their members ‘ contents ’, and the function fr om contexts to contents ‘ character ’.
S ince w e also notice that J ohn ’ s gestur es ar e playing an impor tant r ole in xing the r efer ents of
‘ she ’, w e come up with a formulation of the character of ‘ she ’ along the follo wing lines:
() I n any context c , an occurr ence of the term ‘ she ’ r efers to the demonstrated
individual inc .
F inally , w e apply the same strategy and formulate the character of ( ) r oughly as follo ws:
() I n any contextc , the content of ( ) is the singular pr oposition about the demon-
stratumo assigned b yc to the occurr ence of ‘ she ’; thato is a philosopher .
Although this account is arguably mor e complicated and less intuitiv e than the pr evious one,
at least in the simplest cases it arriv es at the same r esults. S o, it may seem to the r eader that
S e mantic G uidance doesn ’ t add any r eal benet to those alr eady br ought b y S emantic D eter-
minism. I disagr ee. I believ e that theories committed to S emantic D eterminism hav e r eal
difficulties accounting for a wider range of data. I n fact, I think that a commitment to S eman-
tic D eterminism might ev en obstr uct our understanding of indexicality and context-sensitivity
in natural language. T o fully appr eciate this point, let ’ s r evie w some impor tant tenets of the
S tandar d S emantic D eterministic pictur e of indexicals.
. e S tandar d S emantic D eter ministic P ictur e
e no wadays pr edominant appr oach to indexicality and context-sensitivity in general has its
r oots in the intensional systems dev eloped b y philosophers-logicians in during the sixties and
sev enties. I n this section, I will briey intr oduced some of the key ideas belonging to this
tradition.
.. T wo G oals
T o understand the S tandar d S emantic D eterministic P ictur e of indexicality w e need to under-
stand two of its main goals:
(G
1
) F ormulating a logical calculus for indexicals.
(G
2
) Capturing what a user of indexicals must kno w to be competent with them.
B u ilding a theor y of indexicals capable of achieving these goals wasn ’ t easy , as they seem to pull
apar t fr om each other . (G
1
) r equir es the theor y to abstract away fr om the niceties of natural
language and r econstr uct a w ell-behav ed r egimented v ersion of it. S ince building a logic is
an a priori enterprise of considerable difficulty , it allo ws for the intr oduction of sophisticated
techniques and formal machiner y of arbitrar y complexity . G iv en the intensional frame wor k
pr e supposed at the time, (G
1
) r equir es a characterization of the notions of logical tr uth and log-
ical consequence using the methodology dev eloped to characteriz e them in modal and temporal
logic.
B y contrast, (G
2
) r equir es car eful empirical examination of ho w normal speakers use index-
icals in ev er y day language and what they kno w when they kno w the meaning of an indexical.
A theor y satisfying (G
2
) would be able to giv e an account of the information conv entionally
encoded b y indexical expr essions such that it would enable us to make empirical pr edictions
about their differ ent uses. And giv en that competent users of indexicals include philosophically
unsophisticated individuals with limited computational abilities, w e can ’ t expect them to kno w
ev er ything that is allo w ed to achiev e (G
1
).
e most successful attempt to build a theor y aiming at these goals was Kaplan ’ s Logic of
D emonstr ativ es (LD). I n it, Kaplan intr oduces the notion of char acter which pr o vided a pr omis-
ing means to achiev e both (G
1
) and (G
2
). T o see ho w it did so, let ’ s take a brief look at Kaplan ’ s
theor y of indexicals.
.. I ndexicals & I ntensional O perators
Kaplan ( b ) distinguishes two kinds of indexical expr essions: (i) pur e indexicals and (ii) tr ue
demonstrativ es. P ur e indexicals ar e expr essions whose meanings ar e sufficient to fully determine
their contents for each context. ese expr essions include personal pr onouns such as ‘I’ and
‘ my ’, adv erbs such as ‘her e ’, ‘ no w ’, ‘ tomorr o w ’, ‘ y ester day ’, and adjectiv es such as ‘ past ’ and
‘ pr esent ’. B y contrast, tr ue demonstrativ es—wor ds such as ‘ this ’, ‘ that ’, ‘he ’, ‘ she ’, ‘it ’, and their
plurals—ar e expr essions whose linguistic meanings ar en ’ t sufficient for getting to their contents
fr om the information pr o vided b y the context of utterance. ey r equir e some extra help fr om
speakers to pick out their r efer ents: an o v er t display of their r efer ential intentions. Kaplan calls
this additional help ‘ demonstrations ’.
Among the pr oper ties indexicals exhibit, one that ’ s especially r elev ant fr om a compositional
point of vie w is the way they interact with intensional operators. T o appr eciate this point,
consider the follo wing sentence ( Kaplan b , ):
() I t is possible that in P akistan, in v e y ears, only those who ar e actually her e no w
ar e envied.
er e ar e thr ee occurr ences of thr ee differ ent intensional operators in ( ): a modal (‘it is possible
that ’), a locativ e (‘in P akistan ’), and a temporal (‘in v e y ears ’). A ccor ding to the intensional
frame wor k in which Kaplan ’ s system is built, an utterance of ( ) in Los Angeles, says
that ther e is a world w such that in P akistan, in , only those who ar e in w in P akistan
in ar e envied. Call this pr oposition ‘ q ’. Clearly , q isn ’ t what competent speakers would
systematically succeed in asser ting b y uttering ( ) in the y ear . P ut differ ently , it isn ’ t what
is said b y an asser tiv e utterance of ( ) in Los Angeles, .
I nstead, what they asser t and ar e expected to ass er t is the pr oposition that ther e is a world
w such that in P akistan, in , only those who ar e in the actual world, in Los Angeles, ar e envied. Call this pr oposition ‘p ’. S ince (i) what is said b y asser tiv e utterances of ( ) in
the context described is p , not q , and (ii) a semantics for E nglish that assigns q rather than
p to ( ) r elativ e to the context at hand is empirically inadequate, it follo ws that traditional
intensional systems ar e inadequate tools for capturing what is said b y sentences (r elativ e to
contexts) containing indexicals occurring under the scope of intensional operators. S o, what
would a theor y that assigns p to ( ) look like?
Kaplan star ts b y obser ving that the denotations of ‘ actually ’, ‘her e ’, and ‘ no w ’ don ’ t shift
when they occur under the scope of intensional operators—something quite ex ceptional as
intensional operators shift the point of ev aluation of their pr ejacents. H e also notices that this
same pr oper ty is shar ed b y lexical items such as ‘I’, ‘ y ou ’, ‘he ’, ‘ she ’, ‘it ’, ‘ that ’, ‘ this ’, ‘ tomorr o w ’,
and ‘ y ester day ’. F or instance, take the rst person pr onoun ‘I’. Whether or not it occurs under
the scope of an intensional operator seems to hav e no effect whatsoev er on the fact that it ’ s
naturally understood as simply picking out the speaker of the context. F or example, compar e
the follo wing sentences:
() a. N ecessarily , the speaker of this utterance is O bama.
b . N ecessarily , I am O bama.
I f O bama w er e to utter these sentences, only his utterance of the second will asser t a tr uth. e
rst one would expr ess falsities at ev er y possible world in which O bama isn ’ t the speaker of that
utterance. us, despite appearances, ‘I’ and ‘ the speaker of this utterance ’ ar en ’ t synonymous.
is kind of obser v ations motiv ates Kaplan ’ s major insight: indexicals ar e dir ectly r efer ential ,
while denite descriptions ar en ’ t:
(DR) D ir ect R efer ence: An expr ession is dir ectly r efer ential iff its content r elativ e to
a context (and assignment of v alues to v ariables) just is its r efer ent r elativ e to the
context (and assignment).
I ndexicals dir ectly r efer to a featur e of the context without the mediation of F r egean senses or
any pr opositional component. ³ ey simply pick out their r efer ents without including any
kind of information about the way those r efer ents w er e determined. I n shor t, the semantic
contribution of an indexical in ev er y cir cumstance just is its r efer ent r elativ e to the context
of utterance. ⁴ D ir ect r efer ence is closely linked to a second pr oper ty of indexicals: obstinate
rigidity:
(OR) O bstinate Rigidity: An expr ession a is obstinately rigid iff whenev er a r efers
to an object o in c , then a r efers to o (and to nothing else) with r espect to c
plus ev er y cir cumstance of ev aluation irr espectiv ely of whethero exist at any of
them.
is is a consequence of the pr evious pr oper ty . S ince indexicals ar e dir ectly r efer ential, an
indexical ’ s r efer ent is loaded, so to speak, into the pr oposition expr essed befor e the pr oposition
gets ev aluated at any cir cumstance or point o f ev aluation (including those at which the r efer ent
doesn ’ t exist). I n or der to captur e these pr oper ties in his formal system, Kaplan adopts Kamp ’ s
( ) double-indexing technique.
³ E x ceptions to this claim ar e ‘ actually ’-rigidied descriptions such as ‘ the man who actually won the pr esidential election ’.
⁴ S trictly speaking, this isn ’ t what one nds in Kaplan ’ s formal system. F or accor ding to it, the content of an
indexical is a constant function fr om indices to the r elev ant element of the context. N onetheless, Kaplan declar es
his pr efer ence for the str uctur ed pr opositions pictur e in which an indexical ’ s semantic contribution is simply a
unique object fr om the context.
.. D ouble I ndexing
S ince intensional operators quantify o v er parameters such as world and time, and indexicals get
their r efer ents fr om the r elev ant featur es of the context of utterance irr espectiv ely of whether
they occur under the scope of intensional operators, Kaplan argues for the need to divide the
index or points of r efer ence in to two: one x ed and the other shiftable. e former will be
ex clusiv ely r esponsible for xing the what-is-said of indexicals, while the latter for pr o viding
points of ev aluation o v er which intensional operators will operate.
e index r esponsible for the rst task will be called ‘ context of use ’, and the one r esponsible
for that second task will be called ‘ cir cumstances of ev aluation ’. S ince Kaplan ’ s system includes
modal and temporal operators, the cir cumstances of ev aluation will contain a world parameter
and a time parameter . O n the other hand, since what is said b y the set of indexical expr essions
Kaplan is inter ested in is determined b y either a world (‘ actually ’), a time (‘ no w ’), an a gent (‘I’),
or a location (‘her e ’), the context of use will contain a parameter for each of these featur es. e
technique that keeps the context and the cir cumstances as two separate indices is kno wn as
‘ double indexing ’. e adoption of the double-indexing technique is a huge step to war ds (G
1
).
Let me explain why .
.. Contexts as I ndices
e wor d ‘ context ’ is or dinarily used to r efer to differ ent sor ts of things. I t can be used to r efer
to the par ticular r egion of spacetime in which a speech act takes place, the text or discourse
surr ounding it, the cir cumstances—including the beliefs, goals, and intentions pr esupposed
—in which it was per formed, etc. N one of these intuitiv e ways of understanding the notion of
context, though, exactly corr esponds to the one emplo y ed b y Kaplan.
H is notion of context comes fr om M ontague ( , ). M ontague generaliz ed Car-
napian intensions b y tr eating them as functions fr om possible worlds to extensions. is mo v e
gr eatly enriched the expr essiv e po w er of his semantics, for it allo w ed to account for expr essions
whose semantic contents v ar y acr oss possible worlds. N onetheless, the adoption of intensions
didn ’ t help much with what he called ‘ pragmatic languages ’, i.e., formal languages that include
indexical expr essions.
e pr oblem that the intr oduction of indexicals posed is that the information they ar e used
to asser t is sensitiv e to things mor e ne-grained than possible worlds. F or example, if I said in
Lima ‘I t har dly ev er rains her e ’, I would say something tr ue, wher eas if y ou said it in London
y ou would say something false. is differ ence r emains unaffected ev en if the utterances in
question take place in the actual world. S o, sensitivity to possible worlds isn ’ t enough to captur e
the meaning of ‘her e ’; w e need to make it sensitiv e to a specic location in a possible world.
S omething similar must be done to captur e the information expr essed b y uses of other indexicals
such as ‘I’, ‘ no w ’, ‘ tomorr o w ’, and ‘ that ’.
M ontague ’ s solution consisted in making the meanings of indexicals sensitiv e to elements
such as times, locations, agents, etc. F ormally , the trick was done b y the intr oduction of se-
quences of those entities called ‘ parameters ’ whose unique r ole was to supply extensions to those
expr essions. us, the intensions of indexicals ar e r elativiz ed to those parameters. F or example,
the interpr etation of ‘I’ r equir ed an agent-parameter , ‘ no w ’ r equir ed a time-parameter , ‘her e ’ a
location-parameter , etc. M ontague called these sequences of parameters ‘indices ’ or ‘ points of
r efer ence ’.
Kaplan exploits M ontague ’ s notion of index but also incorporates Kamp ’ s double-indexing
technique. S o, he models cir cumstances of ev aluation as pairs ⟨w,t⟩ and contexts of use as
quadr uples⟨w
c
,t
c
,a
c
,l
c
⟩ consisting of a world w
c
, a time t
c
, an agent a
c
, and a location l
c
,
and claims that an adequate semantics for an indexical language (with modal and temporal
operators) r equir es each sentence to be interpr eted with r espect to those two indices. What ’ s
key for this formal notion of context is the fact that the sole purpose of a context is to pr o vide
contents to occurr ences of indexical expr essions, while being out of the r each of the intensional
operators.
A color ful way of putting this might be the follo wing. Y ou might think of Kaplan ’ s system
as a two-channel ster eo system. ese ster eo sound systems hav e two independent audio signal
channels, each of which is in charge of r epr oducing distinct sounds. P ut together these sounds
cr eate a close v ersion of the original sound sour ce. S imilarly , Kaplan ’ s system contains two
independent indices, each of which is ex clusiv ely in charge of pr o viding entities to differ ent
expr essions, with the purpose that together they be able r epr oduce the tr uth-conditions of
sentences containing occurr ences of indexicals. Context, on this pictur e, is one of those indices.
I t ’ s r ole in the system just is xing the contents of indexicals. Consider , for instance, the agent-
parameter . I ts r ole is exhausted as soon as it pr o vides a r efer ent to an occurr ence of the wor d ‘I’.
T o achiev e its purpose the agent-parameter doesn ’ t hav e to include any of the pr oper ties (other
than self-identity) instantiated b y the speaker . e interpr etation function doesn ’ t car e about
what the speaker is doing or what her mental states ar e like: the agent-parameter doesn ’ t ev en
hav e to captur e the fact that the speaker is rational, or that she is capable of understanding the
wor d ‘I’.
is pr oper ty of contexts is cr ucial for (G
1
). Kaplan and his colleagues thought of contexts
as these auster e n -tuples of parameters because that ’ s all they needed to trace logical conse-
quence r elations betw een indexical sentences-in-contexts and dene ‘logical tr uth ’ for indexical
languages. I n the intensional frame wor k they wor k ed on, this could be done only if ev er y
sentence-context pair giv es us a fully deter minate content that can be ev aluated for tr uth or fal-
sity , independently of the speaker ’ s intentions, pr esuppositions, communicativ e purposes, or
whether the indexical sentences in questions will ev er get utter ed.
.. Logical T r uth & Logical Consequence for LD
Let me elaborate a little bit on the pr evious point. An inter esting pr oper ty of indexical languages
is that the P rinciple of N ecessitation (PN) fails for them:
(PN) I f ϕ is v alid, then□ϕ is also v alid.
T o illustrate this point, consider the follo wing sentence:
() I exist.
er e is no context in which an agent would say something false b y asser ting it. ( ) is tr ue
in ev er y context of utterance c
⋆
in which the agent of c
⋆
utters it. us, the pair⟨ ( ), c
⋆
⟩ is
intuitiv ely v alid. H o w ev er , it doesn ’ t follo w fr om this that ( ) must a lso be tr ue:
() □[ I exist]
F or ther e ar e innitely many ways the world might hav e been in which I don ’ t exist. M y
existence is contingent. S o, in or der to achiev e (G
1
) the system had better meet two conditions:
(i) it must captur e the fact that ( ) is v alid, and (ii) it mus t pr edict that ( ) is false. An intensional
system with just one index is incapable of meeting these conditions. F or consider the follo wing:
() J I existK
⟨w,t,a,l⟩
I f indexicals w er e index-sensitiv e, ( ) would expr ess falsities at vir tually ev er y quadr uple⟨w,t,a,l⟩
—e.g., whenw = @,t = 10
5
seconds after the B ig B ang, a = Aristotle, andl = the ninth
moon of S aturn. D ouble indexin g, b y contrast, get us what w e ar e looking for . Although both
the context and the cir cumstances ar e n -tuples, only contexts ar e x ed (unshiftable) n -tuples
of parameters that bear the follo wing r elation: in the wor ldw and timet , the agenta is located at
locationl . N -tuples satisfying this r elation ar e called ‘ pr oper contexts ’. us, the rst condition
is met: ( ) is tr ue in ev er y pr oper context.
H o w about ( )? e necessity operator instr ucts us to ev aluate its pr ejacent r elativ e to ev er y
index, pr oper or not. S o, while ‘I’ dir ectly r efers to the agent of the context of utterance, the
necessity operator takes the pr oposition expr essed b y the pr ejacent at the context of use and
checks at each world-time pair⟨w,t⟩ whether it would be tr ue if w and t w er e actual. ese
unr estricted pairs ⟨w,t⟩ ar e the cir cumstances of ev aluation. B y sharply distinguishing the
context fr om the cir cumstances, the second condition is also satised.
is allo ws Kaplan to dene ‘logical tr uth ’ for LD in the follo wing way . F irst, he denes
a str uctur eA for LD consisting of nonempty sets of worlds, times, contexts, individuals, and
positions (or locations). Each str uctur e contains an interpr etation function that assigns an
intension to each pr edicate and functor in the str uctur e. en, he denes tr uth at a context and
cir cumstance ,
( T ) ϕ is tr ue inc andi iff8g[c2A! JϕK
c,i,g
= tr ue]
wher ec = context,i = cir cumstances of ev aluation, andg = assignment function. is allo ws
him to characteriz e logical tr uth for LD as follo ws:
(L T ) ϕ is logically tr ue in LD iff8A8c8g[c2A! JϕK
c,g
= tr ue]
is is, without doubt, a fundamental r esult for achieving (G
1
) for it makes possible to char-
acteriz e a notion of logical consequence for LD:
(L C) ϕ is a logical consequence of y in LD iff8A8c8g[c 2 A ! [JyK
c,g
= tr ue
! JϕK
c,g
= tr ue]]
N otice that these denitions pr esuppose that the indexical sentences to which they apply ex-
pr ess full-edged pr opositions r elativ e to contexts. I t also pr esupposes that whatev er determines
those pr opositions, does so r eliably , i.e., giv en a str uctur eA , context c , and assignment g , if
ϕ expr esses p r elativ e toA , c , and g , then ϕ can ’ t expr ess p
′
(wher e p ̸= p
′
) r elativ e toA , c ,
andg . e featur e of the system in charge of ensuring that the pr opositions LD r equir es ar e in
good shape is char acter .
.. Character: D eter minacy & Kno wability
B efor e Kaplan ’ s theor y of indexicals it was common to identify the meaning of an expr essiona
with its intension—wher e the intension ofa is a function fr om cir cumstances toa ’ s extension.
Although this seemed at the time to be a good appr o ximation to the meaning of an arbitrar y
a , Kaplan sho w ed that it fails to adequately captur e the meaning of indexicals—and, thus,
linguistic meaning in general.
I f the intension of an expr essiona w er e its meaning, then ignoring the intension ofa would
amount to ignoring its meaning. H o w ev er , this needn ’ t be so . F or suppose y ou nd a note on
a cocktail napkin lying on the oor at a bar that r eads as follo ws:
() I’ ll be at the main entrance of the D isney Concer t H all.
S u ppose y ou don ’ t kno w who wr ote the note, to whom or when it was written. W ithout that
information y ou won ’ t be able to kno w the intension of ( ), i.e. , the pr oposition it expr esses
r elativ e to the context described. As a consequence, y ou won ’ t kno w its extension either—in this
case, its tr uth-v alue. W ould it follo w that y ou would fail to understand what ( ) means? N ot
at all! As a competent E nglish speaker , y ou would understand the sentence ’ s meaning. H o w
is that possible? W ell, that ’ s possible—Kaplan argues—because y ou, as a competent E nglish
speaker , kno w the r ule that determines, for ev er y context, what is said b y ( ) in ev er y context.
F or instance, take the occurr ence of the wor d ‘I’ in ( ). N ot kno wing its intension r elativ e
to the context in question doesn ’ t pr ev ent us fr om understanding its meaning. All w e need to
kno w to be linguistically competent with ‘I’ is that it picks out its utter er , whoev er it is. I n
other wor ds, what one needs to kno w in or der to be linguistically competent with ‘I’ might be
formulated as the follo wing r ule of use:
(Ch
I
) I n any contextc , an occurr ence of the term ‘I’ designates its utter er inc .
I n fact, this is what most competent E nglish speakers would identify as the linguistic meaning
of ‘I’. Kaplan notes that similar r ules can be formulated for ‘ no w ’, ‘her e ’, ‘ tomorr o w ’, and the
r est of pur e indexicals. H ence, leti be the cir cumstances of ev aluation andc the context of use.
I n Kaplan ’ s system, indexicals hav e lexical entries along the follo wing lines: ⁵
() a. J IK
c,i
= a
c
b . J H er eK
c,i
= l
c
c. J A ctuallyϕK
c,⟨w,t⟩
= JϕK
c,⟨w
c
,t⟩
d. J N o wϕK
c,⟨w,t⟩
= JϕK
c,⟨w,t
c
⟩
e. J Y ester dayϕK
c,⟨w,t⟩
= JϕK
c,⟨w,t
c1
⟩
N otice that indexicals include both singular terms (e.g., ‘I’ and ‘her e ’) and operators (e.g., ‘ ac-
tually ’, ‘ no w ’, and ‘ y ester day ’). ese lexical entries can be thought of as formal r epr esentations
of the r ules determining what is said b y the indexicals in question in differ ent (pr oper) contexts.
A t this point, the generalization seems to be ob vious: the linguistic meaning of an indexical
expr essiona is a r ule for getting to what is said b ya in a giv en context fr om the infor mation pr o vided
b y that context . H e calls such a r ule ‘ character ’ and the what-is-said ‘ content ’. What kind of r ule
is character? W ell, as w e hav e seen, the task imposed b y (G
1
) r equir es it to be a r ule capable of
⁵ I ignor e the assignment parameter for simplicity .
fully determining a unique content r elativ e to a context. is is why Kaplan models character as
a function that maps contexts into contents, and content as a function that maps cir cumstances
into extensions. is characterization of content and character can be r epr esented as follo ws:
() a. Content of a inc = li.JaK
c,i
b . Character of a inc = lc.li.JaK
c,i
us, her e is a way to describe a major pr oper ty of character:
( D ) D : e character ofa is a r ule that fully determines, for each con-
textc , a unique content for a inc .
e distinction betw een content and character allo ws Kaplan to giv e a mor e pr ecise pictur e of
context-sensitivity:
(CS) Context-S ensitivity: An expr ession is context-sensitiv e iff its character fully de-
termines differ ent (unique) contents in differ ent contexts.
S i nce the character of a complex expr ession is a function of the characters of its par ts, w e could
formulate the character of ( ) r oughly as follo ws:
() I n any context c , the content of sentence ( ) is the singular pr oposition about
the agent ofc and the D isney Concer t H all, that the former will be at the main
entrance of the latter .
e distinction betw een content and character plays a cr ucial r ole in Kaplan ’ s o v erall theor y of
meaning. N ot only they ar e technical devices to get the tr uth-conditions of indexical sentences
right, and thus, meet (G
1
). ey ar e also assigned impor tant epistemic r oles in language use:
while the content ofS is the thought expr essed (the what-is-said) b y it inc , the character ofS is
what a competent speaker has to kno w in or der to be linguistically competent with S . us, a
speaker ’ s kno wledge of the character ofS is what allo ws her to effectiv ely compute the thought
expr essed (the what-is-said) b y S in c . is is the idea behind the second major pr oper ty of
character:
( K ) K : e character of a is what is kno wn b y a competent user of a .
I f character w er e kno wable in the sense of ( K ), then Kaplan ’ s system would achiev e (G
2
). I n
the simplest cases, character seems to hav e ( D ) and ( K ). H o w ev er , if w e look bey ond them, it
becomes incr easingly difficult for character to satisfy ( D ) and ( K ) at the same time. I n the next
section, I will illustrate this claim.
. e M ultiple O ccurr ences P r oblem
Consider the follo wing scenario . S uppose that Alice and Bob ar e walking on Bond S tr eet in
London when suddenly they see M ick J agger giving a cigar ette to a beggar . S uppose that Alice
asser tiv ely utters:
() H e is rich and famous, but he is not.
I ntuitiv ely , in the scenario described, Alice has asser ted something tr ue and nothing false. er e
is nothing bizarr e going on in this case. H o w ev er , as harmless as it may look, this scenario poses
a serious challenge to the friend of the S emantic D eterminism. T o appr eciate the pr oblem, let ’ s
explicitly state the conict:
. Character is a pr oper ty of expr ession-types.
. Content is a pr oper ty fully determined b y character r elativ e to a context.
. us, content is a pr oper ty of expr ession-types.
. I f the character of an expr ession-type a assigns to it content k r elativ e to c , then ev er y
occurr ence of a inc will getk .
. B ut, the intuitiv e content asser ted b y an utterance of ( ) in the context described r e-
quir es each occurr ence of ‘he ’ to get a differ ent content.
e satisfaction of (G
1
) r equir es (), (), and () to be tr ue. Kaplan is explicit about ():
F irst, it is impor tant to distinguish an utter ance fr om a sentence-in-a-context . e former
notion is fr om the theor y of speech acts, the latter fr om semantics. U tterances take time,
and utterances of distinct sentences cannot be simultaneous (i.e., in the same context).
B ut to dev elop a logic of demonstrativ es it seems most natural to be able to ev aluate
sev eral pr emises and a conclusion all in the same context. us the notion of ϕ being
tr ue inc andA does not r equir e an utterance ofϕ . I n par ticular ,c
A
need not be uttering
ϕ inc
W
atc
T
. ( Kaplan b , )
() is simply ( D ): the mechanical computation of tr uth-v alues in LD r equir es the assignment
of unique contents to each expr ession of the language. As w e hav e seen, in LD, character is in
charge of this task. H o w does it do it? W ell, b y assigning the same content to ev er y occurr ence
of an indexical in a context c —which is pr ecisely what () says. S o far , so good. e conict
arises when w e also attempt to achiev e (G
2
), as it r equir es the obser v ation of (). H o w does
the friend of the S tandar d P ictur e handle this conict? Kaplan ’ s o wn solution is to deny a key
pr esupposition of the pr oblem, i.e., that ther e is one expr ession-type ‘he ’ occurring twice in
( ). I n LD, each occurr ence of ‘he ’ is r eplaced with a complex expr ession composed of the
wor d ‘dthat ’ plus a singular term. I t is this complex expr ession—not the bar e demonstrativ e
‘he ’—that has a character that in turn fully determines a content r elativ e to a context. T o see
ho w this mo v e deals with the M ultiple O ccurr ences P r oblem, let ’ s turn to Kaplan ’ s tr eatment
of demonstrativ es.
.. D emonstrativ es & Dthat
A ccor ding to Kaplan, bar e demonstrativ es, i.e., tr ue demonstrativ es without an associated demon-
stration, ar e expr essions whose characters ar e insufficient to fully determine their contents for
each context. ey r equir e an associated demonstration to pick out their r efer ents. G iv en Ka-
plan ’ s o wn characterization of them, demonstrativ es should hav e raised a r ed ag about the
whole frame wor k, as they seem to lack ( D ). N ev er theless, they didn ’ t. I nstead, Kaplan tw eaked
the morphosyntactic str uctur e of demonstrativ es to make them t into his S emantic D etermin-
istic frame wo r k. B riey , her e is ho w he does it. H is analysis of demonstrativ es has to obser v e
two featur es:
(D
a
) Complete demonstrativ es ar e context-sensitiv e, dir ectly r efer ential, and obsti-
nately rigid.
(D
b
) B ar e demonstrativ es ar e incomplete expr essions.
S ince demonstrations ar e extra-linguistic, they ar en ’ t par t of Kaplan ’ s indexical language. S o, he
r eplaces them with denite descriptions and devises a formal surr ogate of bar e demonstrativ es
labeled ‘dthat ’. us, a lexical entr y for adthat -expr ession will hav e the follo wing form (wher e
a is a denite description or singular term):
() Jdthat aK
c,⟨w,t⟩
= JaK
c,⟨w
c
,t
c
⟩
I n other wor ds,⌜dthat a⌝ dir ectly r efers to the content of a r elativ e toc . N otice that Kaplan
doesn ’ t giv e a lexical entr y for ‘dthat ’ alone. e r eason is that ‘dthat ’ isn ’ t a syntactically
complete expr ession b y itself and, thus, lacks meaning in isolation. I t r equir es an associated
singular terma to get a character and a content. S ince the system doesn ’ t impose any r estrictions
ona , it could r efer to an object that can ’ t be demonstrated in the context of utterance (nor ev en
its pictur e or any other sor t of r epr esentation for that matter). F or example, the expr ession
‘dthat[ the smallest pebble in Alpha Centauri] ’ is a singular term that dir ectly r efers to the
smallest pebble in Alpha Centauri, if ther e is one (other wise it is undened).
E vidently , ther e is a signicant disanalogy her e betw een dthat -expr essions and natural lan-
guage demonstrativ es. B ut ev en if w e w er e to r estrict a such that it picks out only objects that
can effectiv ely be demonstrated in a context, ther e is still something ar ticial about this way
of modeling natural language demonstrativ es. F or w e kno w what the content and character of
a singular term ar e. B ut what is the content and character of a pointing or a nod? I dealizing
considerably , Kaplan tr eats demonstrations as having the general form of a denite description
such as ‘ the individual that has appearance A fr om her e no w ’, and ‘dthat ’ as a rigidier of such
a description. S o, on this pictur e, while the content of a demonstration r elativ e to a context
will be a complex pr oper ty of the sor t encoded b y a denite description (mor e specically , an
indexical denite description), its character will be the set of conv entions go v erning gestur es,
nods, and other ostensiv e acts. ⁶
Complete demonstrativ es, d[d] , will shar e the same semantic pr oper ties with the pur e
indexicals—i.e., context-sensitivity , dir ect r efer ence, and obstinate rigidity . eir contents r el-
ativ e to contexts will be unique individuals and their characters will hav e the follo wing form:
(D
ch
) I n any context c , d[d] is a dir ectly r efer ential term that designates the demon-
stratum, if any , of d inc , and that other wise designates nothing.
As in the case of sentences containing pur e indexicals, sentences containing complete demon-
strativ es will semantically expr ess full-edged pr opositions r elativ e to contexts. ese pr opositi-
ons—the what-is-said of those sentences r elativ e to contexts—will be fully determined b y the
characters of their par ts together with the application of compositional r ules. us, b y including
demonstrations (or , mor e pr ecisely , denite descriptions) into the morphosyntactic str uctur e
of complete demonstrativ es, Kaplan manages to giv e a unied account of pur e indexicals and
(complete) tr ue demonstrativ es.
N o w , let ’ s go back to ( ). O n the analysis of demonstrativ es just sketched, each occurr ence
of ‘he ’ will be r eplaced b y differ ent expr ession-types with differ ent characters. F or instance, ( )
might be analyz ed as follo ws:
⁶ is is the original vie w defended in D emonstr ativ es . H o w ev er , in After thoughts , Kaplan seems to change his
mind. I n this later wor k, Kaplan says: ‘N o w accor ding to my ne w vie w of what determines the r efer ent of a
demonstrativ e, the demonstration (her e, the description) is ther e only to help conv ey an intention and plays no
semantical r ole at all. ’ Kaplan ( a , –).
() dthat[ the lead v ocalist of e R olling S tones] is rich and famous, butdthat[ the
beggar I see fr om her e no w] is not.
is analysis of ( ) av oids the multiple occurr ences pr oblem, since ther e ar en ’ t two occurr ences
of the same expr ession-type in ( ), but two differ ent dthat -expr ession-types. Although this
is a per fectly legitimate way to formaliz e ( ) if y ou want to build a logic of demonstrativ es, it
fails to captur e the fact that ‘he ’ is a meaning ful E nglish expr ession in its o wn right—differ ent
fr om other meaning ful expr essions such as ‘ she ’, ‘it ’, ‘ this ’, ‘ that ’, ‘ they ’, etc. F or r ecall that, on
the Kaplanian pictur e, all of them ar e uniformly r eplaced b y ‘dthat ’, which is meaningless on
its o wn.
F ar fr om being a minor technical issue, this pr oblem thr eatens the plausibility of the whole
pr oject. R ecall that the system was built to achiev e both (G
1
) and (G
2
). Although it arguably
achiev es (G
1
), b y failing to captur e the empirical fact that bar e indexicals in E nglish hav e mean-
ings, it fails to achiev e (G
2
). I n addition, the multiple occurr ences pr oblem is par ticularly
pr essing because in ev er y day conv ersations it ’ s v er y common to use the same expr ession multi-
ple times in the same discourse or sentence to r efer to differ ent individuals of the same context,
and do it successfully . us, unless the friend of S emantic D eterminism giv es an adequate solu-
tion to this pr oblem, he can ’ t claim to hav e achiev ed (G
2
)—a goal that, b y the way , ev er y theor y
of mea ning is expected to achiev e. us, the multiple occurr ences pr oblem should be par tic-
ularly worrisome for the friend of S emantic D eterminism. I n the next section, I will discuss
thr ee solutions that hav e been pr oposed b y scholars wor king within the S emantic D eterministic
frame wor k.
.. r ee U nsatisfactor y S olutions to the M ultiple O ccurr ences P r oblem
e friend of S emantic D eterminism might think that the multiple occurr ences pr oblem can
be o v er come without giving up ( D ) and ( K ). H er e I pr esent thr ee inuential solutions that
attempt to do just that but, I argue, ar e ultimately unsatisfactor y .
e Context S hifting eor y
A ccor ding to the Context S hifting eor y ( B raun ), what explains our intuitions about
( ) is a shift of context in mid-sentence. R ecall that being context-sensitiv e means for ‘he ’
that it can hav e differ ent contents in differ ent contexts. S entence ( ) poses a challenge to the
Kaplanian pictur e because, in or der to get at its asser ted content in the context described, one
seems to be for ced to assign differ ent contents to occurr ences of the same expr ession-type in
the v er y same context. B raun ’ s solution to this pr oblem is to deny the assumption that ther e
is only one context inv olv ed in the scenario described. I nstead, the Context S hifting eor y
claims that the scenario described abo v e inv olv es two contexts: one supplying M ick J agger to
the rst occurr ence of ‘he ’, and the other supplying the beggar to the second occurr ence of ‘he ’.
I n general, accor ding to this theor y , if a sentence or discourse contains n occurr ences of the
v er y same indexical and each of them gets a differ ent content, then ther e must be n contexts
supplying those contents.
e main pr oblem with this theor y is that it r ev erses the or der of explanation. I nstead
of telling us ho w to get at the asser ted content of ( ) fr om the information av ailable in the
scenario described, it tells us which contexts ar e inv olv ed fr om the pr oposition asser ted b y the
utterance of ( ) in the scenario described. I n other wor ds, instead of pr edicting the asser ted
content of ( ) in the scenario described, the Context S hifting eor y pr edicts ho w many
contexts would make it tr ue. S o, it turns out that the Context S hifting eor y isn ’ t a semantic
theor y of indexicals after all, but a rather unintuitiv e theor y of context-detection.
N otice that the Context S hifting eor y can ’ t do without pr esupposing fr om the outset the
asser ted content of the sentences it ’ s expected to account for . F or suppose it didn ’ t. S uppose
that it simply stipulated that if a sentence or discourse containsn occurr ences of an indexical,
then ther e must ben contexts inv olv ed. W ould this help the theor y? Clearly , it wouldn ’ t. F or
it would fail to captur e the contents asser ted b y utterances of sentences containing multiple
co-r efer ential occurr ences of the same indexical. I n fact, b y blindly assigning differ ent contents
to differ ent occurr ence of indexicals, the theor y would fail to captur e a lot of intuitiv ely v alid
infer ences such as ( ):
() is is a hand. I f this is a hand, then I am not a brain in a v at. er efor e, I am
not a brain in a v at. ⁷
F or these r easons, the Context S hifting eor y doesn ’ t pr o vide us with a satisfactor y solution
to the M ultiple O ccurr ences P r oblem.
P utting D emonstrations in Contexts
I n D emonstr ating and N ecessity ( ), N athan S almon offers a r evised v ersion of Kaplan ’ s the-
or y of demonstrativ es that aims at xing its shor tcomings. S almon ’ s pr oposal consists in taking
⁷ is point is raised b y B raun ( ).
demonstrations out of the language and r elocating them to the context of utterance. S o, in ad-
dition to a world, time, agent, and location, contexts will hav e an assignment of demonstrations
to occurr ences of bar e demonstrativ es in a sentence. H er e is what he says:
M y pr oposal is that a context of use be r egar ded as sometimes including a demonstration
among its featur es, along with an agent, a time, a place, and a possible world. N ot the
bar e demonstratum, but the demonstration with all its r epr esentational content.
B etter y et, since the same demonstrativ e may r ecur within a single sentence or str etch of
discourse, each time accompanied b y a differ ent demonstration (‘at one goes betw een
that one and that one ’), the context should include an assignment of a demonstration
for each syntactic occurr ence of a demonstrativ e in a sentence—the rst occurr ence,
the second, and so on. ( S almon , )
Let ’ s tr y to wor k this out. R econsider our pr oblematic sentence:
( ) H e is rich and famous, but he is not.
Let d
1
be the pointing of M ick J agger and d
2
the pointing of the beggar . Also, let ’ s distinguish
the occurr ences of ‘he ’ in ( ) b y attaching a numerical subscript to them. S ince contexts
contain assignments of demonstrations to occurr ences of bar e demonstrativ es, let ’ s formulate
the follo wing two alternativ e assignment functions, f and g :
f( he
1
) = d
1
g( he
1
) = d
1
f( he
2
) = d
1
g( he
2
) = d
2
W e can no w distinguish two contextsc
1
andc
2
, wher ec
1
=⟨w,t,a,l, f⟩ andc
2
=⟨w,t,a,l,g⟩ .
I n addition, since on S almon ’ s account, bar e demonstrativ es ar e complete, meaning ful expr es-
sions of the language, ‘he ’ will be assigned a character that w e could formulate as follo ws:
(Ch
he
) I n any context c , an occurr ence of ‘he ’ designates the demonstratum, if any ,
of the demonstration assigned b y c to that occurr ence; other wise it designates
nothing.
F ormally speaking, the character of ‘he ’ is a function fr om contexts to individuals demonstrated
b y the demonstration assigned b y the context. N o w w e ar e in a position to formulate the
character of ( ) which will be r oughly this:
(Ch
(13)
) I n any contextc , the content of the sentence ‘H e is rich and famous, but he is
not ’ is the singular pr oposition about the demonstrata of the demonstrations
assigned b y c to the rst and second syntactic occurr ences of ‘he ’; that the
rst is rich and famous, but the second is not.
F inally , w e plug inc
1
andc
2
into ( ) and get the follo wing demonstration-specic descriptions
of the content of ( ):
(M
1
) e content of the sentence ‘H e is rich and famous, but he is not ’ r elativ e to
c
1
is the singular pr oposition about the demonstrata of d
1
that it is rich and
famous, but it is not.
(M
2
) e content of the sentence ‘H e is rich and famous, but he is not ’ r elativ e toc
2
is the singular pr oposition about the demonstrata of d
1
and d
2
that the rst is
rich and famous, but the second is not.
N o w , what enables a speakers to grasp the pr oposition encoded b y ( ) r elativ e to eitherc
1
or
c
2
? A ccor ding to this account, what enables s to grasp it ar e (i) s ’ s awar eness of the r elev ant
context—in this case, whether it is c
1
or c
2
, (ii) s ’ s kno wledge of the conv entions go v erning
ngerpointings, and (iii) s ’ s grasping of either (M
1
) or (M
2
).
S almon ’ s stor y is superior to Kaplan ’ s account of demonstrativ es in thr ee r espects. F irst,
while in Kaplan ’ s system bar e demonstrativ es ar e syncategor ematic (characterless) incomplete
expr essions, on S almon ’ s account they hav e characters. ⁸ S econd, while Kaplan intr oduces
demonstrations (b y pr o xy) into the str uctur e of language, S almon places them in a less im-
plausible spot, namely , the context. ir d, S almon ’ s stor y all o ws occurr ences of the same bar e
demonstrativ e conv ey differ ent contents r elativ e to the same context, wher eas Kaplan ’ s doesn ’ t.
ese ar e impr o v ements on Kaplan ’ s system. H o w ev er , a fe w signicant difficulties r emain.
F irst, on S almon ’ s account, pur e indexicals as w ell as demonstrativ es ar e complete, mean-
ing ful, expr essions. S o, in this r espect, they ar en ’ t differ ent. e differ ence—S almon claims—is
in the context. H e says,
e demonstrativ e itself is a complete expr ession, fully assembled and r eady to go .
S trictly speaking, it is the context that is incomplete. O r if y ou pr efer , it is the oc-
curr ence of the demonstrativ e in the defectiv e context that is incomplete, because of
a contextual deciency . I t is like the use of ‘ no w ’ in a timeless univ erse (“befor e ” the
B i g B ang?), or the use of ‘ ther e ’ in O akland, California—fully complete expr essions
occurring in defectiv e contexts. ( S almon , )
O n S almon ’ s vie w , a context that doesn ’ t assign a demonstration to an occurr ence of a demon-
strativ e is someho w defectiv e. is is a quite str ong claim (the parallel with an occurr ence of
‘ no w ’ in a timeless univ erse is striking!) B ut, is it tr ue? I don ’ t think so . F or example, consider
the follo wing sentences:
⁸ Although, fr om an empirical point of vie w , the character S almon ’ s theor y assigns to demonstrativ es doesn ’ t
seem to be adequate.
() is is a multi-ethnic society .
() ese ar e difficult times.
() at was w eir d! [ said right after a talking donkey cr osses the r oom ]
() is is simply astonishing! [ said upon seeing M achu P icchu for the rst time ]
() H e ob viously had plenty of money . [ said walking thr ough the T aj M ahal ] ( Levin-
son )
N one of the occurr ences of the demonstrativ es in these sentences r equir es an accompanying
demonstration. I n fact, in some cases no demonstration would be ev en adequate. e fact that
w e hav e no pr oblem identifying the r efer ents of these occurr ences suggests that ther e is nothing
defectiv e about their contexts.
P erhaps, S almon ’ s notion of demonstration is br oad enough to include r efer ential inten-
tions. I f so, then the idea would be that a context that failed to assign r efer ential intentions
to occurr ences of demonstrativ es would be defectiv e. As pr omising as this idea might sound,
I believ e that adding intentions to contexts doesn ’ t solv e the pr oblem. I n fact, I think that it
brings in mor e difficulties.
F irst, the implementation of this idea poses a signicant challenge to the friend of S emantic
D eterminism. I n or der for this strategy to wor k, it has to pr o vide a mechanism for discriminat-
ing among speakers ’ intentions. F or no matter ho w sincer e the r efer ential intentions of a speaker
s ar e in asser ting p in c b y uttering S , if the linguistic meaning of S pr ecludes the deriv ation
of p , then putting s ’ s intentions in c would be otiose. F or instance, if the amnesiac R udolph
Lingens, who is lost in the S tanfor d Librar y , ass er tiv ely utters ‘is is the D oheny Librar y ’, he
will asser t, of the S tanfor d Librar y , that it ’ s the D oheny Librar y , ev en if he is absolutely cer tain
he is in the D oheny Librar y . S o, the challenge for the friend of S emantic D eterminism would
be to design a selection algorithm capable of putting only the right intentions in c . B ut, ho w
is he going to do it without looking rst each time at the asser ted content of S inc ? I don ’ t see
ho w . And, as w e said in the case of the Context S hifting eor y , if he can only kno w which
intention to put inc b y appealing at the outset to the asser ted content the theor y was designed
to pr edict, then w e hav en ’ t made much pr ogr ess.
S econd, ther e is an impor tant r eason not to include speakers ’ r efer ential intentions into the
context of utterance and, at the same time, assume that linguistic meaning is character . e
pr oblem is that b y doing so, one misses the impor tant G ricean distinction betw een linguistic
meaning and speaker ’ s meaning . R oughly , the distinction is this: wher eas linguistic meaning is
the publicly av ailable information conv entionally encoded b y expr ession-types, speaker ’ s mean-
ing is the information communicated (not encoded) b y expr essions-token in par ticular contexts
as a r esult of the audience ’ s r ecognition of the speaker ’ s intentions. I f (i) character is a func-
tion fr om contexts to contents, and (ii) contexts include intentions (or assignments of them),
then the character of a demonstrativ e would be a function fr om intentions (or assignments of
intentions) to unique objects. B ut, of course, this coincides with a natural characterization of
speaker ’ s meaning. S o, unless S almon wants to deny the distinction betw een speaker meaning
and linguistic meaning or pr opose a differ ent way to make the distinction, he should deny that
demonstrations include intentions or that linguistic meaning is character .
ir d, an additional pr oblem with this alternativ e is that it makes it har d to explain why
only demonstrativ es should be sensitiv e to speaker ’ s intentions and not ev er y other lexical item
in the language, say , pr oper names. F or sur ely when I utter ‘D avid Kaplan ’ I intend to r efer to
D avid Kaplan. O f course, including in the context assignments of intentions to names would
be semantically otiose for a M illian like S almon. B ut then, why would the context contain
assignments of intentions to ev er y occurr ence of a demonstrativ e and not to occurr ences of
other singular terms?
F inally , r ecall that S emantic D eterminism was designed with (G
1
) in mind. B ut in or der
achiev e (G
1
), one should be able to ev aluate indexical sentences and arguments independently
of the mental states of their users. I n fact, one should be able to do it ev en r elativ e to contexts
that don ’ t contain any living cr eatur e. ⁹ e intr oduction of intentions into contexts seems
inter fer e with this goal. er efor e, unless w e get adequate r esponses to these worries, w e should
continue looking for an adequate solution to the M ultiple O ccurr ences P r oblem.
Analyzing P r onouns as V ar iables
e nal r esponse to the M ultiple O ccurr ences P r oblem in the S emantic D eterministic line of
thought I want to consider is based on a theor y defended b y H eim & Kratz er ( ). ¹⁰ e
basic idea is to tr eat pr onouns as v ariables and let contexts determine assignments of objects to
⁹ is point is raised against S almon ’ s account in S oames ( , -).
¹⁰ D isclaimer: I’ m not assuming that H eim & Kratz er (hencefor th ‘H&K’) ar e concerned at all with the M ultiple
O ccurr ences P r oblem or that they ought to . Rather , my purpose in this section is to imagine the friend of the
S tandar d S emantic D eterministic pictur e of indexicals appealing to H&K’ s system to look for a solution to the
M ultiple O ccurr ences P r oblem, while keeping his philosophical commitments .
occurr ences of them. H er e is ho w the theor y wor ks. F irst, each occurr ence of a pr onoun or a
trace will hav e a numerical index at LF . us, if a is a pr onoun or a trace and i a numeral, an
occurr ence of that pr onoun will be r epr esented as a
i
at LF .
C r ucial to this account is the idea that the determination of the semantic content of a fr ee
occurr ence of a pr onoun r equir es an assignment function fr om numerical indices to individuals
in the domain. I n addition, the interpr etation of fr ee occurr ences of pr onouns and traces will
be taken car e of b y the application of the T r aces and P r onouns R ule :
( T P) I fa is a pr onoun or a trace,a is a v ariable assignment, andi2 dom(a) ,Ja
i
K
a
=
g(i) . ( H eim & Kratz er , )
O n this account, the LF of our pr oblematic sentence ( ) would be r oughly the follo wing:
() H e
7
is rich and famous, but he
5
is not.
F ollo wing the analogy betw een pr onouns and v ariables, the interpr etation of ‘he
5
’ and ‘he
7
’
will r equir e an assignment function that maps individuals fr om the domain into the indices ‘’
and ‘’ r espectiv ely . S o, let f and g be assignment functions such that:
f( ) = J ohn g( ) = J oe
f( ) = Bob g( ) = J agger
F r om these assignment functions and ( TP) w e get that ( ) expr esses under f the pr oposition
that Bob is rich and famous, but J ohn is not, and under g the pr oposition that J agger is rich
and famous, but J oe is not.
N o w the theor y has to allo w us to r ule out assignment f and leav e only g (assume that the
beggar ’ s name is ‘J oe ’). F or w e want the assignment function to r eect the speaker ’ s r efer ential
intentions in the r elev ant context. H&K’ s system attempts to captur e this r esult b y adding
an assignment parameter only to appr opriate contexts (AP) and formulating tr uth-conditions
r elativ e to them ( T C):
(AP) A contextc is appr opriate for an LFϕ only ifc determines a v ariable assignment
g
c
whose domain includes ev er y index which has a fr ee occurr ence in ϕ .
( T C) I fϕ is utter ed inc andc is appr opriate forϕ , then the utterance ofϕ inc is tr ue
r elativ e toc if JϕK
g
c
= 1 and false if JϕK
g
c
= 0 . ( H eim & Kratz er , ) ¹¹
O n this pictur e, then, index ed pr onouns of the forma
i
just like fr ee occurr ences of v ariables ar e
assignment-sensitiv e. N o w notice that (AP) is mor e liberal than one might want it to be. F or it
allo ws contexts in whichg
c
assigns a male or a plurality of individuals to an index ed occurr ence
of ‘ she ’. S o, this account needs to nd a way to scr een out those contexts. I n par ticular , it has
to nd a way to constrain the assignments of v alues to fr ee occurr ences of pr onouns. H&K’ s
solution is to appeal to the so-called ‘ϕ -featur es ’.
Constraining Assignments: F -F eatur es
Although on H&K’ s account pr onouns ar e v ariable-like expr essions, they ar en ’ t just v ariables.
O ther wise, a fr ee occurr ence of ‘ she ’ could be assigned a contextually salient male or a fr ee
occurr ence of ‘he ’ a contextually salient female. A war e of this pr oblem, H&K appeal to the
¹¹ N otice that ( T C) is dened for utterances, not sentences. is is alr eady a major deviation fr om Kaplan ’ s
pictur e.
so-called ‘ϕ -featur es ’, i.e., the formal pr oper ties of NP s that typically enter into syntactic agr ee-
ment: person, number and gender . ¹² us, whenev er the ϕ -featur es of an expr ession a ar e
pr esent, H&K’ s system will adjoin them to the index ed node in the follo wing way: ¹³
() [Person[Number[Gender[a
i
]]]]
I n H&K’ s system, each featur e denotes a par tial identity function of type⟨e,e⟩ . F or example,
gender is analyz ed as follo ws:
() G ender
a. JMascK = lx:x is male.x
b . JFemK = lx:x is female.x
I nformally put,Masc denotes a function that maps individualsi fr om the domain to themselv es
pr o vided thati is male (other wise it is undened). S imilarly , Fem denotes a function that maps
individualsi fr om the domain to themselv es pr o vided thati is female (other wise it is undened).
is tr eatment of gender featur es is extended in H eim ( ) to person and number . Let
s
c
be ‘ the speaker ofc ’ andh
c
‘ the addr essee ofc ’. H ence,
() N umber
a. JSingK = lx:x is an atom.x
b . JPlK = lx:x is a plurality.x
¹² e identication of ϕ -featur es as the sour ce of syntactic agr eement can be traced back to Chomsky ( ).
¹³ e or der of the featur es doesn ’ t matter .
() P erson
a. J1
st
K = lx:x includess
c
.x
b . J2
nd
K = lx:x includesh
c
and ex cludess
c
.x
c. J3
rd
K = lx:x ex cludess
c
.x
P utting these elements together , this account pr edicts the follo wing r efer ence conditions for
occurr ences of non-r eexiv e pr onouns:
() a. J he
5
K
g
= g( ) if g( ) is a male atom ex cluding s
c
; undened other-
wise.
b . J she
7
K
g
= g( ) if g( ) is a female atom ex cluding s
c
; undened other-
wise.
e friend of S emantic D eterminism might nd in H&K’ s system the tools he is looking for
to tackle the M ultiple O ccurr ences P r oblem. A ccor ding to the account w e ar e examining, the
LF of ( ) would be the follo wing:
() [3
rd
[Sing[Masc[ he
7
]]]] is rich and famous, but [3
rd
[Sing[Masc[ he
5
]]]] is
not.
As y ou may r ecall, in the original scenario described, ( ) was used to asser t of J agger that he
is rich and famous and of J oe that he is not. H&K’ s system attempts to captur e this intuition
b y r epr esenting the context of utterancec as determining an assignment function g
c
mapping
the index ‘’ to J agger and ‘’ to J oe. F r om the point of vie w of H&K’ s theor y , the M ultiple
O ccurr ences P r oblem is a pr oblem only is y ou assume that the job of semantics is to interpr et
sentences. I t isn ’ t. O n their account, semantics is in charge of interpr eting LFs. And, as ( )
sho ws, each occurr ence of ‘he ’ carries a differ ent index.
H as the M ultiple O ccurr ences P r oblem been S olv ed?
I don ’ t think so . U nfor tunately , this appr oach doesn ’ t far e much better than the Kaplanian
use of ‘dthat ’. F or , accor ding to H&K’ s account, what gets interpr eted isn ’ t an occurr ence of
the E nglish wor d ‘he ’, but an index ed occurr ence of it, i.e., a complex expr ession composed
of the occurr ences of two letters (‘h ’, ‘ e ’), and a numerical subscript. F ur thermor e, it ’ s just the
numerical subscript that is doing all the semantic wor k as H&K’ s semantic r ules just see it, not
the pr onoun. As they put it:
W e ar e not assuming her e that the index is par t of the pr onoun as a lexical item. I f
pr onouns ar e listed in the lexicon at all, they ar e listed ther e without an index and as
semantically v acuous items. N o semantic r ule sees the pr onoun itself […] ( H eim &
Kratz er , , n. ).
J ust as in the case of Kaplan ’ s ‘dthat ’, on this pictur e, bar e ( un index ed) E nglish pr onouns ar e
meaningless.
N o w , an immediate r eply to this charge might be that this is per fectly ne, as ther e is
no such thing as an unindex ed E nglish pr onoun in LF . And since semantics interpr ets LFs
not sentences, the meaninglessness of unindex ed occurr ences of pr onouns isn ’ t a defect of the
system. is r eply isn ’ t good for the friend of S emantic D eterminism. R ecall that S emantic
D eterminism is a pictur e of the natur e of linguistic meaning and its r elationship with asser ted
content. A ccor ding to this pictur e, meaning is that which must be kno wn b y a speaker to be a
competent user of the language, and its r elation to asser ted content (r elativ e to context) is one
of full determination.
B ut, what would be the context-inv ariant meaning (character) that a speaker must kno w
in or der to be a competent user of an indexical expr ession , if only its index ed occurr ences hav e
meanings? Can it be, perhaps, the T r aces and P r onouns R ule ( TP)? Clearly not. F or in or der
to understand and use ( TP) one only needs to kno w some basic set theor y . S o, in principle, I
could competently use ( TP) (and the other r ules) without kno wing what ‘ she ’, ‘he ’, ‘it ’, ‘ that ’,
‘ this ’, etc. mean and ho w their meanings differ fr om each other (r ecall that in H&K’ s system
ϕ -featur es ar e semantic pr esuppositions, and not par ts of meaning).
I n addition, this r eply leav es the friend of S emantic D eterminism in the same position
as the pr evious attempts to solv e the M ultiple O ccurr ences P r oblem—namely , this account
pr esupposes what is expected to pr edict. F or , ho w do H&K kno w that ( ) is the right LF
for ( )? H o w do H&K kno w that the two occurr ences of ‘he ’ in the LF of ( ) ar en ’ t co-
index ed? O r , ho w do they kno w that the context determines an assignment of J agger to ‘’ and
J oe to ‘’ and not the other way r ound? ey kno w these things because they kno w the asser ted
content of ( ) in the scenario described. B ut, if S emantic D eterminism was supposed to offer
an explanation of ho w asser ted contents ar e determined b y linguistic meaning together with the
information pr o vided b y context, that ’ s something the theor y should pr edict, not pr esuppose. I f
these worries ar e on the right track, then w e should look for a differ ent solution to the M ultiple
O ccurr ences P r oblem.
. e S emantic G uidance P ictur e
I n this section I want to pr opose an account of pur e indexicals and tr ue demonstrativ es that
giv es an intuitiv e pictur e of the r elation betw een their meanings and asser ted contents, and
that o v er comes the M ultiple O ccurr ences P r oblem. M y pr oposal is based on a conception of
linguistic meaning r ecently defended b y Scott S oames in a number of places. ¹⁴ A ccor ding to
this conception, the meaning of an indexical sentenceS is a set of constr aints that a pr oposition is
expected to satisfy , if it ’ s to count as a candidate asser ted content ofS in a giv en communicativ e
situation.
I n a nutshell, the account is this. I follo w S oames ’ s pr oposal and formulate the linguistic
meanings of indexical expr essions as combinations of specic constr aints on the range of the
possible r efer ents they can take. S o, instead of fully determining a unique content for a giv en
indexical a , the meaning of a just tells us which pr oper ties its content ought to instantiate.
D epending on the communicativ e situation in which a gets utter ed, these constraints will be
satised b y one or mor e contents. Whenev er ther e is mor e than one candidate asser ted content,
conv ersational par ticipants will hav e to appeal to the extralinguistic information av ailable in
their communicativ e situation to narr o w do wn the number of candidates till they ar e able
to identify the asser ted content(s). B ut it will be them, not linguistic meaning plus a set of
parameters, who will do the heavyw eight lifting in pinning do wn asser ted content. A ccor ding
to this account, the r ole of linguistic meaning isn ’ t to x asser ted contents but to simply guide
conv ersational par ticipants to war ds them.
¹⁴ S ee S oames ( , , , ).
.. What ’ s O ur G oal?
What ar e w e looking for when w e build a theor y of meaning for a languageL ? W e ar e looking
for an account of the information that an agent must master in or der to be a competent user
of L . H o w do w e kno w whether a giv en theor y of meaning has achiev ed this goal? is is
something w e can ’ t kno w a priori. All w e can do is test empirically the theor y b y checking
whether its pr edictions ar e compatible with the data. F or instance, if theor yT assigns meaning
m to a , but it turns out that m fails to captur e a widely acceptable, literal use of a , then w e
kno w thatT is inadequate.
is isn ’ t the case for logical systems. T o borr o w G raham P riest ’ s wor ds, ‘ the point of logic
is to giv e an account of the notion of v alidity: what follo ws fr om what ’ ( P riest , ). I n
or der to determine what follo ws fr om what, y ou don ’ t need to do any empirical wor k, especially
since normal competent speakers ar e pr one to err or and faulty patterns of r easoning. M or eo v er ,
whetherS follo ws fr omS
′
doesn ’ t ev en depend on their meanings. Rather , it depends on their
forms. us, if v alidity is the point of logic and v alidity is a formal pr oper ty , then the pr oject
of building a logic is quite differ ent fr om the pr oject of building a theor y of meaning.
S o, ho w come logicians call some of their denitions of logical notions such as v alidity , logi-
cal consequence, consistency , etc. ‘ semantic ’? I sn ’ t semantics the study of meaning? W ell, when
logicians do semantics, they don ’ t build empirically testable theories of meaning. Rather , they
specify an interpr etation of the nonlogical v ocabular y of a formaliz ed language, wher e ‘ specifying
an interpr etation ’ means specifying a domain of entities of some kind or another , an interpr e-
tation function that maps the nonlogical v ocabular y into the elements of that domain, and
applying the notion of tr uth-in-a-model.
e specication of an interpr etation of an object language L and the application of the
notion of tr uth-in-a-model allo ws logicians to determine—among other things—what follo ws
fr om what inL . is appr oach is called ‘ semantic ’ to differ entiate it fr om a syntactic or pr oof-
theor etic characterization of the logical notions. e differ ence betw een the two is r oughly
this. Wher eas syntactic or pr oof-theor etic characterizations of the logical notions appeal just
to some pr ocedur e that makes r efer ence only to the symbols of the object language, semantic
characterizations appeal to cer tain r elations betw een expr essions of the object language and the
entities fr om the domain those expr essions ar e about.
S emantic D eterminism is the or thodo x vie w of the r elation betw een the meaning of a
context-sensitiv e expr ession and the information that expr ession is used to asser t that w e inher-
ited fr om Kaplan ’ s LD. What ’ s inter esting about it is that Kaplan ’ s LD is a logic that attempts
to wor k as a theor y of meaning for indexicals. As w e hav e seen, it attempts to achiev e (G
1
)
and (G
2
) at the same time. e pr oblem with this is that these goals impose v er y differ ent
tasks, which in turn r equir e the application of differ ent methodologies. at ’ s why although
his intr oduction of Dthat -expr essions into LD is a ne solution to the M ultiple O ccurr ences
P r oblem fr om a logician point of vie w , it ’ s empirically inadequate as an account of the meaning
and morphology of demonstrativ es in E nglish.
B ut, why should w e be bound to look for solutions only fr om within the S emantic D eter-
ministic frame wor k? I believ e that if w e want to giv e a theor y of meaning for indexicals that
captur es what must be kno wn b y competent users of them, and that is capable of pr edicting
the contents competent speakers literally asser t b y their use, w e should abandon S emantic D e-
terminism. I n what follo ws, I will sho w ho w S emantic G uidance can giv e us a natural account
of the meaning of indexicals in E nglish that o v er comes the M ultiple O ccurr ences P r oblem.
.. e M eanings of I ndexicals
Let ’ s look at the pr oposal in detail. e key idea is to understand the linguistic meaning of an
indexical not as a r ule that fully determines its asser ted content r elativ e to a contextc , but as a
set of constraints on the contents it can be used to asser t in a giv en communicativ e situation.
Constraints differ fr om deterministic r ules in that they simply tell us which pr oper ties a content
ought to instantiate if it ’ s to belong in the set of candidate asser ted contents of a giv en expr ession
a . us, b y themselv es, they don ’ t x a unique asser ted content for a .
O n the S emantic G uidance frame wor k, C is a constraint on asser ted content iff C is a
condition that ev er y acceptable asser ted content of a giv en expr ession must satisfy . W e can
think ofC as a pair⟨t,P⟩ wher et is a meaning ful constituent ofS andP is a pr oper ty that ev er y
acceptable candidate asser ted content oft ought to hav e. e set of constraints determined b y
the linguistic meanings of all the meaning ful constituents of S together with the constraints
imposed b yS ’ s syntactic str uctur e is called ‘S ’ s asser tion checklist ’.
H ence, an asser tion checklist for S is a r epr esentation of the context-inv ariant informa-
tion conv entionally associated with S that a speaker would hav e to kno w , o v er and abo v e her
kno wledge of the world and surr ounding cir cumstances, in or der to make and understand lit-
eral asser tions of S . O n this pictur e, each item in a asser tion checklist will determine a set of
acceptable asser ted contents called an ‘ acceptability range ’. F or instance, if a constraintC of a
giv en asser tion checklist is satised b y a set G , thenG is the acceptability rang e determined b y
C .
J udging fr om the kinds of acceptability range they determine, constraints can be catego-
riz ed into two main gr oups: global and local. G lobal constraints ar e constraints that determine
constant acceptability ranges acr oss possible communicativ e situations. ey do so b y imposing
conditions that must be satised b y ev er y acceptable asser ted content of a giv en expr ession a ir-
r espectiv ely of wher e, when, ho w or for what purposea is used. O n this pictur e, the constraints
imposed b y syntax and semantics ar e global. B y contrast, local constraints a r e constraints that
determine acceptability ranges that v ar y acr oss possible communicativ e situations. ey ar e
conditions imposed b y the specic information av ailable in par ticular communicativ e situa-
tions that is r elev ant to the identication of asser ted content.
An additional essential ingr edient of S emantic G uidance is the notion of communicativ e
situation. As y ou may hav e noticed, I hav e used the phrase ‘ communicativ e situation ’ wher e
the friend of S emantic D eterminism would hav e used ‘ context ’. M y choice of wor ds isn ’ t gra-
tuitous. R ecall that, on the S emantic D eterministic frame wor k, contexts ar e simplyn -tuples of
parameters whose r ole is just to supply contents to indexicals. is auster e notion of context
is good enough for the friend of S emantic D eterminism because he thinks of meaning as a
function designed to see ex clusiv ely the right parameters.
B y contrast, accor ding to S emantic G uidance, conv ersational par ticipants—not meanings
—ar e in charge of getting at asser ted contents. S ince the asser ted contents of expr essions in
actual conv ersations don ’ t come in sequential or der , with numerical subscripts, or with any
label indicating their identity , conv ersational par ticipants will hav e to appeal to ev er y r esour ce
av ailable to them at the time of utterance to gur e out asser ted contents. B eing this the sit-
uation, our account needs a richer notion of context that includes not just r efer ents, but also
all the pieces of information that a rational, competent speaker would need to hav e access to
at a giv en time of utterance to be able to get at the asser ted content. e body of informa-
tion r equir ed would hav e to include items such as shar ed backgr ound pr esuppositions, shar ed
communicativ e purposes, mutual expectations, etc. is richer notion of context is what I call
‘ communicativ e situation ’.
us, on the S emantic G uidance pictur e, giving a theor y of meaning for an indexical ex-
pr ession a inv olv es identifying a set of global constraints on its asser ted contents that allo ws
us to formulate an adequate asser tion checklist for it. Wher e could w e star t looking for those
constraints? A good star ting point is to look closely at ϕ -featur es such as gender , number , and
person. H o w ev er , gender , number , and person ar en ’ t sufficient to account for the contents one
can asser t b y uttering indexicals. A mor e accurate pictur e of the constraints imposed b y the lin-
guistic meanings of indexicals would include also pr oper ties r elated to perspectiv e, dimension,
and par ticipant r ole. us, on the S emantic G uidance pictur e, the meaning of an indexical
pr e scribes that the asser ted contents of its occurr ences be objects (or set of objects) instantiating
suitable combinations (to be specied after war ds) of the follo wing pr oper ties:
() I ndividuation
a. S: e pr oper ty of being a singleton
b . P: e pr oper ty of being a plurality
() P ar ticipant R ole
a. A: e pr oper ty of including the asser ter
b . A: e pr oper ty of including the asser ter ’ s addr essee
() G ender
a. M: e pr oper ty of being male
b . F : e pr oper ty of being female
() Kind
a. A: e pr oper ty of being animate
() D imension
a. S: e pr oper ty of being a r egion of space
b . T: e pr oper ty of being a r egion of time
() P erspectiv e
a. P : e pr oper ty of being pr o ximal to the asser ter
b . D: e pr oper ty of being distal to the asser ter
U sing these pr oper ties, w e can formulate the complex pr oper ties that the individual(s) con-
tributed b y an occurr ence of an indexical ought to instantiate either b y itself or perhaps along
with fur ther information locally determined:
() a. ‘I’ = S^ A
b . ‘ Y ou ’ = S^ A ¹⁵
c. ‘ Y ou ’ = P^ A^: A^ A
d. ‘H e ’ = S^ M^: A^: A
e. ‘S he ’ = S^ F ^: A^: A
f . ‘I t ’ = S^: A
g. ‘ W e ’ = P^ A^ A
h. ‘ey ’ = P^: A^: A
i. ‘H er e ’ = S^ P^ place of asser tion
j. ‘er e ’ = S^ D^: place of asser tion
k. ‘N o w ’ = T^ P^ time of asser tion
l. ‘en ’ = T^ D^: time of asser tion
m. ‘is ’ = S^ P
n. ⌜ is F⌝ = S^ P^ F-
o . ‘ese ’ = P^ P
p . ‘at ’ = S^ D
q. ⌜ at F⌝ = S^ D^ F-
¹⁵ I’ m focusing just on the personal use of ‘ y ou ’. I n its impersonal use, ‘ y ou ’ is close to the impersonal ‘ one ’.
r . ‘ose ’ = P^ D
s. ‘ T oday ’ = T^ P^ day of asser tion
t. ‘ T omorr o w ’ = T^ day after today
u. ‘ Y ester day ’ = T^ day befor e today
As w e can see, on this account, the linguist ic meaning of an indexical simply r estricts the set
of r efer ents it can take at a giv en communicativ e situation, without necessarily singling out
any one. is featur e of the account captur es the exibility that par ticular uses of indexicals
exhibit. F or on the S emantic G uidance frame wor k, speakers will normally need to appeal to the
extra-linguistic information av ailable in the communicativ e situation to dir ect their audience
to war ds their intended r efer ents.
H er e is a simple example. S uppose that Alice and Bob ar e having a conv ersation at a cof-
feehouse, when two women—Claudia and D iana—seat at the table behind Bob . U pon r ecog-
nizing one of them, Alice tells Bob:
() S he is a pr ominent physician.
Bob looks ar ound and counts a total of six people: Alice, J ohn, E mily (who is seating behind
Alice), Claudia, D iana, and himself . H o w does he get at the pr oposition Alice asser ted? A c-
cor ding to S emantic G uidance, the rst step is for him to check as as many checkbo x es as he
can in the asser tion checklist determined b y the linguistic meaning and syntactic str uctur e of
( ):
A C r is sentence is being used to pr edicate the pr oper ty of being a pr ominent
physician of the individual the sentence is used to talk about.
r e individual this sentence is being used to talk about is a singleton.
r e individual this sentence is being used to talk about is neither the asser ter
nor the asser ter ’ s addr essee.
r e individual this sentence is being used to talk about is female.
D oing so will allo w him to automatically r ule out J ohn, Alice, and himself fr om the acceptability
range determined b y the meaning of ‘ she ’. H o w ev er , in the scenario described, this will leav e
thr ee pr opositions as acceptable candidate asser ted contents for Alice ’ s utterance of ( ):
() a. ⟨ Claudia, P P ⟩
b . ⟨ D iana, P P ⟩
c. ⟨ E mily , P P ⟩
Which of these pr oposition is the one Alice asser ted? A ccor ding to S emantic G uidance, this is a
question that neither syntax nor semantics can answ er . is is a pr oblem that Bob will hav e to
solv e b y appealing to all the r elev ant information he can get fr om the local constraints imposed
b y his communicativ e situation. F or instance, the av ailable extra-linguistic information might
include the follo wing facts:
(i) Alice ’ s gesturing to war ds Claudia together with Bob ’ s kno wledge of the conv en-
tions go v erning those par ticular gestur es
(ii) Claudia ’ s w earing of a white coat together with Bob ’ s kno wledge of that physicians
usually w ear white coats
(iii) E mily ’ s being out of Alice ’ s sight together with Bob ’ s belief that Alice might not
been awar e that E mily is at the coffeehouse
(iv) D iana ’ s still being a teenager together with Bob ’ s kno wledge that only adults ar e
allo w ed to practice medicine
I f the extra-linguistic information av ailable to Bob is rich enough, he will be able to grasp the
pr oposition Alice asser ted b y uttering ( ). I f it isn ’ t, he might need to ask Alice for fur ther
information. F or instance, he might ask ‘I s she the one w earing glasses?’ or perhaps ‘ Ar e y ou
talking about the woman drinking an I rish coffee?’
I t ’ s wor th emphasizing that, just as in Kaplan ’ s system, the pr oper ties determined b y the
linguistic meanings of indexicals don ’ t enter into the semantic contents of the sentences of
which they ar e par t. Rather , they r estrict the set of singular pr opositions that a speaker can
asser t b y uttering an indexical sentence. us, just like the S tandar d S emantic D eterministic
P i ctur e, this account builds in dir ect r efer ence and obstinate rigidity . H o w ev er , unlike it, those
contents ar en ’ t semantic but mer ely asser ted. is is so because, on the S emantic G uidance
pictur e, semantic contents ar en ’ t r elativiz ed to contexts. I n addition, S emantic G uidance allo ws
descriptiv ely enriched pr opositions as possible asser ted contents of indexical sentences, i.e., it
allo ws for pr opositions containing the r efer ent of the indexical in question plus some fur ther
pr oper ty pr edicated of it. F or instance, if in the conv ersation betw een Alice and Bob at the
coffeehouse they hav e been r eferring to Claudia and D iana as ‘ the woman w earing a white coat ’
and ‘ the w earing glasses ’ r espectiv ely , the local permissibility range corr esponding to Alice ’ s
asser tion of ( ) might include the follo wing set of pr opositions:
() a. ⟨ Claudia, W W C , P P ⟩
b . ⟨ D iana, W G , P P ⟩
Among the adv antages of adopting S emantic G uidance instead of S emantic D eterminism, one
is immediate. O n the S emantic G uidance account, bar e demonstrativ es ar e meaning ful, syn-
tactically complete lexical items. S o, unlike some S emantic D eterministic accounts, S emantic
G uidance has no pr oblem accounting for the felicity and tr uth of asser tions of sentences such as
( )–( ) since it doesn ’ t r equir e, as a matter of linguistic meaning, accompanying demonstra-
tions for any occurr ence of a demonstrativ e. D emonstrations ar e only locally r equir ed whenev er
the speaker believ es that without them the communicativ e situation might not contain enough
information to allo w conv ersational par ticipants to narr o w do wn the set of candidate pr opo-
sitions to the one asser ted. is might happen, for example, if both Claudia and D iana ar e
physicians w earing white coats.
H o w does S emantic G uidance far e with r espect to the M ultiple O ccurr ences P r oblem? I n
the next section, I will offer an answ er to this.
.. R evisiting the M ultiple O ccurr ences P r oblem
R ecall our pr oblematic sentence:
( ) H e is rich and famous, but he is not.
e pr oblem posed b y this sentence in the communicativ e situation described is that in or der
to get at its asser ted content—call it ‘p ’—w e need to be able to assign differ ent contents to
each occurr ence of ‘he ’. W e saw that S emantic D eterminism has a har d time coming up with
a notion of linguistic meaning (character) that satises both ( D ) and ( K ):
( D ) D : e character ofa is a r ule that fully determines, for each con-
textc , a unique content for a inc .
( K ) K : e character of a is what is kno wn b y a competent user of a .
I n contrast, S emantic G uidance ’ s explicit r ejection of ( D ) puts it in a better position to handle
cases like this. S o, what ’ s S emantic G uidance ’ s take on ( )? I t ’ s actually v er y straightfor war d.
F or simplicity , let ’ s just focus on the demonstrativ e ‘he ’. R ecall that, accor ding to S emantic
G uidance, the linguistic meaning of ‘he ’ simply tells us that the asser ted content of an occur-
r ence of it must be a male singleton ex cluding the asser ter and his addr essee. Armed with this
information, Bob will hav e to get at p r elying on the extra-linguistic information av ailable to
him in the communicativ e situation in which the asser tion of p takes place. is means that
he will hav e to r ule out thr ee pr opositions fr om ( )’ s local acceptability range:
() a. M ick J agger is rich and famous, but M ick J agger is not.
b . J oe is rich and famous, but M ick J agger is not.
c. M ick J agger is rich and famous, but J oe is not.
d. J oe is rich and famous, but J oe is not.
e rst step in this endeav or is to lter out those pr opositions that don ’ t ser v e the purposes of
Alice ’ s and Bob ’ s communicativ e situation. Assuming that Alice is being cooperativ e and fol-
lo wing the conv ersational maxims, ( a ) and ( d ) will be lter ed out , as they ar e clear violations
of the G ricean M axim of Q uality .
F iltering—as I call this pr ocess—will leav e two liv e options. I f Bob kno ws something about
M ick J agger and the life of beggars, he will able to easily get at ( c ). I f he doesn ’ t, he will need
to r ely on other pieces of information such as Alice ’ s demonstrations, their physical appearances,
their behavior , the behavior of other people to war ds them, etc. I f none of these ar e clear enough
for Bob to get at p , he might need to ask Alice about the asser ted content of her utterance.
As y ou can see, the M ultiple O ccurr ences P r oblem pr esents no especial difficulty to S eman-
tic G uidance. And this is as it should be. F or no r easonably w ell-informed, competent speaker
of E nglish would nd the M ultiple O ccurr ences P r oblem par ticularly challenging. F r om the
S emantic G uidance vie wpoint, the fact that S emantic D eterminism stumbles with sentences
such as ( ) is fur ther evidence that it ’ s missing a pr etty basic fact about the actual r elationship
betw een linguistic meaning and asser ted content.
. Conclusion
Why ar e philosophers so inter ested in indexicals? er e ar e at least a couple of r easons. F irst,
indexicals constitute a class of natural language expr essions that ar e uncontr o v er tibly context-
sensitiv e—-a pr oper ty that distinguishes natural language fr om mathematical languages. And
second, the application of vir tually the same tools and techniques dev eloped for the study of
the logical pr oper ties of mathematical languages deliv er ed pr etty accurate pr edictions about
the information many of their uses asser ted in par ticular contexts. ese facts gav e people
hopes that soon w e would be able to use this methodology to giv e a unied account of context-
sensitivity in natural language.
I n this chapter I hav e argued that, despite its vir tues, this appr oach has serious limitations.
I hav e argued that the o v erall pictur e of the r elation betw een linguistic meaning and asser ted
content that Kaplan ’ s methodology pr esupposes r esponds to two potentially conicting goals:
characterizing logical pr oper ties such as logical consequence and logical tr uth, and capturing
what a speaker must kno w in or der to become linguistically competent.
I n addition, I hav e argued that this tension explains the system ’ s difficulties in handling
sentences containing multiple occurr ences of indexicals. S o, rather than attempting to o v er-
come this impasse b y complicating the system, I hav e used the S emantic G uidance frame wor k
dev eloped in pr evious chapters to offer a solution. I n doing so, I hav e offer ed an account of
the linguistic meanings of indexicals accor ding to which they ar e distinctiv e congurations of
pr oper ties that ev er y candid ate asser ted content o ught to instantiate. O n this pictur e, these
pr oper ties don ’ t x the contents an indexical sentence is literally used to asser t on a giv en oc-
casion, but simply giv es us some cues as to what those contents must look like. I hav e argued
that the S emantic G uidance account is not only as good as S emantic D eterministic accounts
on the data the latter do best with, but can offer an intuitiv e account of the data S emantic
D eterminism has difficulties with.
Chapter e S emantics/P ragmatics D istinction
. I ntr oduction
H er e is an intuitiv e pictur e of the r elation betw een the linguistic meaning of a context-sensitiv e
expr ession S and the contents S can be used to literally asser t on differ ent occasions. e
meaning of a wor d is the conv entionally encoded information that one who learns a language
must associate with it in or der to use it corr ectly . e meaning of a phrase or sentence is the
conv entionally encoded information that r esults fr om the combination of the meanings of the
wor ds occurring in it. I f S is a context-sensitiv e expr ession, it can be literally used to asser t
differ ent pr opositions fr om one occasion to another . S o, when a competent speaker uses S
to literally asser t p , p will be the r esult of combining the meaning of S with some r elev ant
extra-linguistic information av ailable on the occasion of use.
M ost philosophers of language agr ee with this general pictur e. N onetheless, they disagr ee
on ho w rich the conv entionally encoded information is, and ho w much of p is contributed b y
something other than that. is is the hear t of the semantics/pragmatics disputes. T o a rst
appr o ximation, semantics can be thought of as the study of the information conv entionally
encoded b y S , and pragmatics as the study of the information asser ted and communicated b y
an utterance ofS that isn ’ t conv entionally encoded. O n this issue, one can see a continuum of
possible positions one could take ranging fr om the claim that nearly ev er ything that one can
asser t b y uttering a declarativ e sentence is conv entionally encoded to the claim that v er y little
of it is. e purpose of this chapter is to situate S emantic G uidance in this continuum and
compar e it with some w ell-kno wn alternativ e ways of drawing the line betw een semantics and
pragmatics.
. S emantics vs. P ragmatics: r ee Views
I n this section, I want to characteriz e thr ee positions one can take about the division of la-
bor betw een semantics and pragmatics in the explanation of the information a giv en linguistic
expr ession can be literally used to asser t. M y characterization of them is meant to be general
enough so as to encompass differ ent theories within each of them.
S emanticism
O ne of the positions one can take in the semantics/pragmatics disputes might be called ‘S e-
manticism ’. S emanticism can be characteriz ed as the conjunction of thr ee theses:
. e semantic content of a declarativ e sentenceS is always asser ted b y literal uses
ofS .
. e contribution of context to semantic content is simply to supply r efer ents to
occurr ences of context-sensitiv e expr essions.
. S emantics alone is in charge of laying do wn strict r ules for matching up each
expr ession of the language with a unique semantic content.
A ccor ding to thesis (), ther e is a v er y close r elationship betw een the semantic content S and
the asser ted content of a literal use S . F or most S emanticist theories, that r elationship just is
identity . F or others, S ’ s semantic content might be one of the pr opositions asser ted. I n any
ev ent, since what is asser ted is always a pr oposition, thesis () entails S tr ong P r opositionalism ,
i.e. the thesis that the semantic contents (r elativ e to contexts) of declarativ e sentences ar e always
pr opositions.
esis () is meant to ensur e the tr uth of thesis () when S contains context-sensitiv e expr es-
sions. S ince (i) sentences containing context-sensitiv e expr essions don ’ t expr ess pr opositions b y
themselv es, and (ii) con text-sensitiv e expr essions get their contents dir ectly fr om the context of
use, the adoption of thesis () r equir es the adoption of thesis (). G iv en that, on this pictur e,
the sole r ole of context is to pr o vide r efer ents to occurr ences of context-sensitiv e expr essions,
S emanticism models contexts asn -tuples of parameters.
F inally , thesis () together with the pr eceding theses entail that if a declarativ e sentence
S is literally used to asser t p , then it ’ s the job of semantics to account for p . is means
that, accor ding to S emanticism, pragmatics can enter into the pictur e only after semantics has
nished its job . e object of pragmatics is to explain speaker ’ s meaning , i.e., the information
that speakers might conv ey or communicate—o v er and abo v e p —b y their uses ofS .
P ragmaticism
A second position o ne can adopt in the territorial disputes betw een semantics and pragmatics
might be named ‘P ragmaticism ’. P ragmaticism is committed to the follo wing theses:
. e semantic content ofS is nev er asser ted.
. e contribution of context to the pr opositions asser ted b y literal uses ofS is mor e
than just supplying r efer ents to occurr ences of context-sensitiv e expr essions.
. P ragmatics alone is in charge of explaining asser ted content.
P ragmaticism is at the other end of the spectr um fr om S emanticism. I t holds that the semantic
contents of the sentences in natural language either (i) play no r ole whatsoev er in asser tion, or
(ii) ar e so poor that play a minimal, negligible r ole in asser tion. I n any case, semantic contents
ar e nev er pr opositional and, thus, nev er asser ted. us, the P ragmaticist distinguishes betw een
semantic content and asser ted content.
P ragmaticism also differs fr om S emanticism in that context plays a mor e r obust r ole in
the determination of asser ted content . Rather than simply pr o viding r efer ents to occurr ences
of context-sensitiv e expr essions, context plays a cr ucial r ole in disambiguation, saturation of
sub-pr opositional contents, and optional or non-linguistically mandated enrichment of pr opo-
sitions. C r ucial for P ragmaticism is the idea that context isn ’ t an external speech situation, but
an internal set of mental r epr esentations that each conv ersational par ticipant has. ¹
A ccor ding to P ragmaticism, a pr oper theor y of the information asser ted b y our literal uses of
declarativ e sentences should focus on speaker ’ s meaning and on the cognitiv e pr ocesses behind
it. S ince for the P ragmaticist the pr opositions asser ted b y our literal utterances ar en ’ t encoded
but inferr ed, a pr oper study of them should look closely at the infer ential pr ocesses that make
them possible. And since pragmatics is in charge of the non-encoded information communi-
cated b y our utterances, ifS is literally used to asser t p , it ’ s the job of pragmatics to account for
p .
Cooperativism
e nal position I want to consider might be called ‘Cooperativism ’. Cooperativism occupies
the middle gr ound betw een S emanticism and P ragmaticism. I t holds the follo wing thr ee theses:
. e semantic content ofS is sometimes, but not always, asser ted b y literal uses of
S .
. e contribution of context to the pr opositions asser ted b y literal uses ofS is mor e
than just supplying r efer ents to occurr ences of context-sensitiv e expr essions.
. I f S is context-sensitiv e, semantics together with pragmatics ar e in charge of ex-
plainingS ’ s asser ted content.
¹ As I will explain later on in the chapter , R elev ance eorists—the paradigmatic P ragmaticists—explain the
pr ocess of interpr etation as the pr ocess of incorporating ne w information into y our existing set of beliefs.
Cooperativism agr ees with P ragmaticism that semantic content and asser ted content ar e two
differ ent things. H o w ev er , it differs fr om it in that Cooperativism holds that semantic content
can sometimes be asser ted. A ccor ding to Cooperativism, the conv entions go v erning asser tion
differ fr om the r ules go v erning semantic content and, thus, the fact thatS can be literally used
(without conv ersational implicatur es) to asser t p on a giv en occasion, doesn ’ t guarantee that p
is the semantic content ofS .
I n addition, Cooperativism agr ees with P ragmaticism that context does mor e than just sup-
plying r efer ents to occurr ences of context-sensitiv e expr essions. H o w ev er , it disagr ees with it
in that contexts ar en ’ t internal psy chological entities, but external communicativ e situations
accessible to conv ersational par ticipants. F or Cooperativism, contexts ar e bodies of public in-
formation that include the shar ed backgr ound pr esuppositions, comm on purposes, and mutual
expectations of a gr oup of conv ersational par ticipants.
F inally , despite of wor king together to ultimately achiev e the same goal, semantics and prag-
matics ar e seen b y Cooperativism as two differ ent endeav ors with their o wn distinctiv e objects
and methodologies. F r om a cooperativist point of vie w , the job of semantics is to account
for the common cor e, the shar ed, inv ariant pieces of information carried b y expr ession-types
acr oss their uses. P ragmatics has the complementar y job of accounting for the bits of informa-
tion conv ey ed and communicated b y utterances of S that may v ar y fr om one occasion of use
to another . is means that if S is context-sensitiv e or under determined and is literally used
to asser t p , semantics will be in charge of explaining only a pr oper par t of p . P ragmatics will
enter into the pictur e to giv e an account of the r emaining elements needed to constitute p .
I n the pr eceding chapters, I hav e defended a pictur e of the r elation betw een the linguistic
meaning and asser ted content that I ha v e called ‘S emantic G uidance ’. S emantic G uidance is
committed to Cooperativism. B efor e mounting my defense of Cooperativism, it will be useful
to say something about the impor tance of this debate.
. e P r oblem of U nder deter mination
Why thr ee differ ent ways of drawing the line betw een semantics and pragmatics and not just
one? An impor tant phenomenon that pr ompted the r evision of the traditional division of labor
betw een semantics and pragmatics was under determination. e purpose of this section is to
focus on the challenge posed b y it.
e study of language is traditionally divided in thr ee branches: syntax, semantics, and
pragmatics. ² R oughly , they ar e characteriz ed as follo ws: syntax is concerned with the gener-
ation of expr essions and the r elations betw een them; semantics with the r elation betw een ex-
pr e ssions and the objects they ar e about; and pragmatics with the use of expr essions b y speakers
in differ ent occasions.
e traditional division of labor pr esents a clean pictur e of the study of language in which
syntax pr o vides the input to semantics, and semantics the input to pragmatics. O n this pic-
tur e, syntax tells us which strings of symbols ar e w ell-formed expr essions of a language, and
ho w they combine to compose complex ones. S emantics takes those expr essions and pr o vides
² is is the classic trichotomy pr oposed b y Charles W . M orris (see M orris ). N o wadays, w e must add at
least phonology .
interpr etations of them b y assigning entities of one kind or another to expr essions ending up
with the assignment of pr opositions to sentences. F inally , pragmatics takes the pr opositions
assigned b y semantics to the sentences of the language, and pr o vides an explanation of fur ther
pr opositions one might be able to conv ey and communicate in differ ent occasions of use.
I n r ecent y ears, many scholars hav e contested this description of the eld. I n par ticular ,
sev eral authors hav e challenged the autonomy of semantics fr om pragmatics arguing that the
pr opositions many sentences ar e literally used to asser t can ’ t be explained b y appealing only to
the traditional techniques used in semantics. F or example, consider the follo wing sentence:
() T ipper is r eady .
I ntuitiv ely , if Alice w er e to literally utter ( ), she would asser t that T ipper is r eady for some-
thing or to do something. H o w ev er , ( ) itself doesn ’ t giv e us any clue about what T ipper is
r e ady for/to . I f semantics is in charge of assigning pr opositions to w ell-formed sentences of the
language and to do so b y combining only the information encoded b y their constituents, ( )
pr e sents a challenge to semantics. A similar challenge is posed b y the follo wing sentence,
() I t ’ s raining.
I ntuitiv ely , if Bob w er e to literally utter ( ), he would normally asser t that it ’ s raining at some
specic location. B ut, wher e does the information about that location come fr om? Can seman-
tics alone account for the pr oposition Bob asser ted or do w e need to appeal also to pragmatics?
Wher e should w e draw the line betw een semantics and pragmatics?
ese ar e questions many philosophers str uggle with. e phenomenon that ( ) and ( )
illustrate is commonly kno wn as ‘ under determination ’. T o a rst appr o ximation, let ’ s say that S
is under determined if the pr opositions intuitiv ely communicated b y literal asser tiv e utterances
ofS ar en ’ t standar d G ricean implicatur es, and at least one of the follo wing conditions ho lds: (i)
the pr opositions intuitiv ely communicated b y literal asser tiv e utterances ofS hav e constituents
not contributed b y any of the wor ds or phrases occurring inS , or (ii) the contents contributed
b y the wor ds or phrases occurring in S plus the way they ar e combined ar e insufficient to
constitute a pr oposition.
U nder determination poses a serious challenge to the traditional way of doing semantics and
pragmatics. F or if w e make use of our traditional indexical semantics to explain under determi-
nation, then all w e need to nd is a constituent (mor e pr ecisely , an indexical) in the str uctur e
of each of ( ) and ( ) whose character makes it pick out the missing content fr om the context
of use. e pr oblem with this alternativ e is that ther e seems to be no such constituent in ( ) or
( ). Alternativ ely , if w e use our traditional G ricean pragmatics, then w e might think that the
missing contents ar e implicated b y utterances of ( ) and ( ). e pr oblem with this alternativ e
is that the calculation of implicatur es r equir es tr uth-ev aluable content as a star ting point. B ut
that is pr ecisely what ( ) and ( ) lack without those contents.
e pr oblem posed b y under determination is especially signicant because under determi-
nation seems to be ubiquitous in natural language. H er e is a small sample:
(A) S entences containing P r enominal P ossessiv es
() J ohn ’ s book is the pr oduct of ten y ears of r esear ch.
() Diana ’ s pictur e was taken in .
(B) S entences containing Compound N ominals
() er e is a br ead knife among the br ead sculptur es .
() e br occoli politician is inv olv ed in a scandal.
(C) S entences containing R elational T er ms
() J ohn visited a local bar .
() Alice pr efers for eign beer .
(D) S entences with I mplicit T ime/Location
() I t ’ s sno wing.
() I hav e eaten br eakfast.
(E) M iscellaneous
() Al has nished.
() J ohn has had enough.
() As usual. [ said upon seeing Bob arriving late for wor k ]
All of these sentences seem to be such that the pr opositions they ar e literally used to asser t on
any giv en occasion will always hav e constituents that ar en ’ t the contents of any of the wor ds or
phrases occurring in them. Literal uses of the sentences in (A) and (B) will asser t pr opositions
that contain specic r elations betw een the lexical items that compose the pr enominal possessiv es
and compound nominals occurring in them. Literal uses of the sentences in (C) will asser t
pr opositions containing information about a r efer ence location that will anchor the asser ted
contents of ‘local ’ and ‘ for eign ’. Literal uses of the sentences in (D) will asser t information
about a location or a time that is no wher e to be found in the sentences themselv es. And in
literally using the sentences in (E) one will asser t pr opositions containing information about
actions and ev ents whose identities ar en ’ t pr o vided b y the sentences in question.
H o w do S emanticism, P ragmaticism, and Cooperativism deal with cases like these? Which
one should be pr eferr ed?
. I nterlude: T wo P otentially P r oblematic N otions
B efor e attempting to answ er these questions, let ’ s tr y to get clear about the fr equently used, but
potentially pr oblematic, notions of thought and what is said , and why I think that most appeals
to them b y competing theories in the semantics/pragmatics debates don ’ t help us choosing
betw een S emanticism, P ragmaticism, and Cooperativism.
.. ought
Why do w e want a theor y of meaning? H er e is a plausible answ er: w e want a theor y of meaning
because w e ar e inter ested in the natur e of thought, and the thoughts w e manage to expr ess and
communicate b y our utterances depend on the meanings of the expr essions w e utter . B ased on
this idea, one might formulate the ultimate task of a theor y of meaning r oughly as follo ws:
U ltimate T ask of a eor y of M eaning
e ultimate task of a theor y of meaning is to explain the natur e of the r elationship
betw een thought and language.
e pr oblem with this characterization of the ultimate task of a theor y of meaning is that
the wor d ‘ thought ’ can be used in two differ ent ways: it can be used to r efer to (i) the thinking
pr ocess, and (ii) the object of thinking. U nsurprisingly , ther e ar e competing theories of meaning
based on these two ways of understanding the term. T o av oid confusions, let ’ s distinguish the
theories in dispute as follo ws:
O bjectivism
O bjectivist theories use the term ‘ thought ’ as shor t for ‘ thought-content ’. ought-
contents ar e objectiv e units of information that differ ent agents can hav e access to,
shar e, and hav e attitudes about. M eaning, on the other hand, is a r elation that holds
betw een w ell-formed expr ession-types and those objectiv e units of information. us,
accor ding to objectivism, the ultimate task of a theor y of meaning is to account for
the natur e of that r elationship betw een w ell-formed expr ession-types and the objectiv e
information speakers asser t and communicate b y their use of them.
S ubjectivism
S u bjectivist theories use the wor d ‘ thought ’ as shor t for ‘ thought-pr ocessing ’. A ccor ding
to S ubjectivism, the ultimate task of a t heor y of meaning is to model the cognitiv e
mechanisms inv olv ed in linguistic information pr ocessing—i.e., the mental operations
that allo w competent speakers to use language to successfully ex change information
with each other . Consequently , subjectivists see their theories of meaning as abstract
descriptions of linguistic competence.
U nder determination poses differ ent challenges to objectivism and subjectivism. F or the objec-
tivist, the challenge is to constr uct a theor y that bridges the gap that seems to exist betw een
the information encoded b y an under determined sentence S and the information asser ted b y
its literal uses. F or the subjectivist, the challenge is to explain ho w our minds effor tlessly grasp
complete pr opositions upon hearing an utterance ofS , ev en when the input they r eceiv e fr om
S seems to be sub-pr opositional.
ese ar e two differ ent theor etical pr ojects. Wher eas objectivists ultimately want a theor y
of information, subjectivists want a theor y of information-pr ocessing. us, whenev er a the-
or etician talks about meaning and r elates it to thought, w e must make sur e w e kno w exactly
what is being talked about.
.. What is S aid
A second notion that is constantly appealed to and w e need to be clear about is the notion
of what is said. I n Logic and Conv ersation ( a ), G rice intr oduces his famous distinction
betw een what is said and what is implicated . H e claims that together , what is said and what
is implicated, exhaust what is communicated b y a giv en asser tiv e utterance of a sentence. B ut,
ho w does G rice characteriz e what is said? H er e is a w ell-kno wn passage:
I n the sense in which I am using the wor d say , I intend what someone has said to be
closely r elated to the conv entional meaning of the wor ds (the sentence) he has utter ed.
( G rice a , )
B ased on this quote, many authors hav e simply assumed tha t G rice ’ s notion of what is said just
is our notion of linguistic meaning. N ev er theless, this interpr etation of what is said is at odds
with G rice ’ s characterization of it in his U tter er ’ s M eaning and I ntentions ( b ). I n that paper ,
he analyz es the phrase ‘U (utter er) said p ’ as ‘U did somethingx ’, which in turn is analyz ed as
follo ws:
U did something x () b y which U meant that p
() which is an occurr ence of an utterance type S (sentence) such that
() S means ‘p ’
() S consists of a sequence of elements (such as wor ds) in a way lice-
nsed b y a system of r ules (syntactical r ules)
() S means ‘p ’ in vir tue of the par ticular meanings of the elements
ofS , their or der , and their syntactical character . ( G rice b , )
Conditions (), (), (), and () seem to suggest that what is said is simply the r esult of the
compositional combination of the contents contributed b y the wor ds and phrases occurring in
S . H o w ev er , G rice ’ s intr oduction of () seems to go against it. Condition () states that what
is said must be meant.
U nder determination destabiliz es this pictur e of what is said. F or under determination pr e-
sumably occurs when ther e is a mismatch betw een the compositionally generated content en-
coded b y an unambiguous, declarativ e sentence S and the non-implicated pr oposition meant
b y a speaker in asser ting S literally . I n par ticular , under determination seems to challenge the
idea that a unique kind of entity can satisfy conditions () and () at the same time.
As Le wis ( ) points out, the phrase ‘ what is said ’ can be used to r efer to differ ent things.
S aying is a speech act; mor e pr ecisely , a locutionar y act. S o, strictly speaking, agents—not
expr essions —say things. I t ’ s only in a metaphorical sense that w e say that a piece of language
says something. H o w ev er , this metaphorical use can get easily out of hand, as the expr ession
‘ what is said ’ can be used to r efer to differ ent things. F or instance, it can be used to r efer to:
(a) e wor ds themselv es used in the per formance of a phatic act ³
(b) e information associated with an utterance on a giv en occasion of use
(c) e information conv entionally associated with an expr ession-type
F or example, suppose that after Bob asks Alice wher e her car is, Alice says:
() I’ m par ked in the P ar king S tr uctur e Z.
I f ‘ what is said ’ is be used to r efer to (a), then what is said is “I’ m par ked in the P ar king S tr uctur e
Z. ” I f it ’ s used to r efer to (b), then what is said is that A lice ’ s car is par ked in the P ar king S tr uctur e
Z. And if it ’ s used to r efer to (c), then what is said is that A lice is par ked in the P ar king S tr uctur e
Z.
N one of the par ties in the semantics-pragmatics debates is inter ested in (a). ⁴ S o, w e can
leav e it out of our discussion. W e cannot say the same about (b) and (c). M any of the
³ A phatic act is ‘ the uttering of cer tain v ocables or wor ds, i.e. noises of cer tain types, belonging t o and as
belonging to, a cer tain v ocabular y , conforming to and as conforming to a cer tain grammar . ’ ( A ustin , ).
⁴ H o w ev er , w e cannot assume that competent speakers will discar d this interpr etation of ‘ what is said ’. S o, it
is a potential sour ce of noise in the no wadays common use of empirical tests for detecting semantic content and
context-sensitivity . S ee, for example, Cappelen & Lepor e ( ).
semantics-pragmatics territorial disputes switch rapidly—and sometimes car elessly—fr om one
to the other .
E xpr ession-tokens or utterances ar e spatiotemporal entities that r esult fr om our speech acts.
ey can be per ceiv ed, r ecor ded, r epr oduced, and manipulated. J ust like any other spatiotem-
poral entity , utterances ar e go v erned b y physical laws. is means that they ar e r estricted in
v er y specic ways. F or instance, utterances can ’ t be innitely long, eternal or hav e eternal pr op-
er ties, occur in mor e than one place at the same time, etc. I n fact, their v er y existence depends
on the prior per formance of an uttering act.
B y contrast, expr ession-types ar e abstract theor etical objects. ey can ’ t be per ceiv ed, r ecor ded,
r epr oduced, or manipulated in any way other than in thought. ⁵ J ust like numbers, sets, or func-
tions they ar en ’ t go v erned or limited b y physical laws. us, they can be innitely long, eternal,
hav e eternal pr oper ties, don ’ t hav e to occur some wher e, etc. I n fact, they don ’ t ev en need to be
utter ed to exist.
N o w , if y our appr oach to semantics is subjectivist, y ou will want to focus on the informa-
tion carried b y expr ession-tokens or utterances. S ince utterances ar e the r esult of speech acts,
communicativ e intentions and competent speaker ’ s intuitions about the information carried b y
their utterances will be cr ucial in determining what is said. B y contrast, if y ou hold an objectivist
conception of semantics, y ou will be inter ested in the information carried b y expr ession-types.
I f so, then communicativ e intentions and competent speaker ’ s intuitions about what is said will
play a secondar y r ole—if any—in y our semantics.
⁵ O f course, w e can instantiate them b y writing or uttering them. H o w ev er , in doing so, w e ar en ’ t altering any
of their intrinsic pr oper ties.
O nce w e distinguish these two ways of understanding ‘ what is said ’ it becomes clear why
the same entity doesn ’ t always satisfy () and (). ⁶ I f ‘ what is said ’ is meant to pick out the
information associated with an utterance on a giv en occasion of use, then it clearly satises ().
N o w , what do normal, competent speakers mean when they asser tiv ely utter an unambiguous,
declarativ e sentenceS ? W ell, they mean to say a pr oposition. at is what declarativ e sentences
ar e for . N ev er theless, if what is said satises (), then it will often fail to satisfy (). F or if S is
under determined, then the pr opositions speakers mean to say b y asser tiv e utterances of them
hav e elements not contributed b y any of the wor ds or phrases occurring inS . is is pr ecisely
what under determination amounts to . B ut if so, then what is said violates ().
Conv ersely , if ‘ what is said ’ is used to r efer to the information conv entionally associated
with an expr ession-type, then satisfying () might entail failing to satisfy (). e r eason is
simple. I fS is under determined, then arguably it will lack a constituent the content of which is
needed forS to expr ess a complete pr oposition. N otwithstanding this fact, competent speakers
will normally succeed in communicating complete pr opositions b y asser tiv ely utteringS . S ince
the utter ers ofS would mean those pr opositions, ther e would be elements in the pr opositions
meant that wouldn ’ t corr espond to what is said b yS . is means that what is said would violate
().
I f this is right, ther e ar e thr ee lessons w e can draw fr om it. F irst, consider sentences of
the form ‘S he said that S . ’ What an utterance of a sentence of this form standar dly r epor ts is
the information asser ted b y a tokening of a sentence-type S . Although our theories of mean-
⁶ M y discussion her e is indebted to B ach ( ).
ings ought to help us get an better understanding of this piece of information, b y itself the
information asser ted b y a tokening of S doesn ’ t tell us ho w much of it is determined b y S ’ s
linguistic meaning, and ho w much b y some extra-linguistic sour ce. S econd, asking competent
speakers about their pr e-theor etical intuitions about what is said will nev er settle the semantics-
pragmatics debates on under determination. is is so because (i) the phrase ‘ what is said ’ in
or dinar y E nglish picks out theo r etically distinct kinds of info rmation and, thus, pr e-theor etical
intuitions will likely track all of them indiscriminately , and (ii) instr ucting speakers about the
technical sense of ‘ what is said ’ w e ar e looking for would likely bias their r esponses making
them useless for adjudicating the debates. ir d, if in fact () and () ar e inconsistent, no sin-
gle notion of what is said would be able to play all of the numer ous r oles it is expected to play
in semantic and pragmatic theories. us, unless ‘ what is said ’ is explicitly dened within a
theor etical frame wor k, appealing to it fr eely to draw conclusions about linguistic meaning and
semantic content is deeply mistaken.
S o much for the interlude. W ith these distinctions in mind, let ’ s turn our attention to the
positions described in section . and ho w they handle under determination.
. r ee W ays of D ealing with U nder deter mination
e aim of this section is to examine some of the strategies emplo y ed b y S emanticism, P rag-
maticism, and Cooperativism r espectiv ely to account for under determination.
.. S emanticism on U nder deter mination
S emanticism encompasses a v ariety of theories that might emplo y differ ent techniques and
methodologies to achiev e their shar ed goals. O ne r espect in which they differ is the set of
expr essions they count as context-sensitiv e. While some hold that the set of context-sensitiv e
expr essions in natural language is v er y small (e.g., S emantic M inimalism), others think that
ther e is mor e to context-sensitivity than meets the ey e. S ince w e hav e alr eady discussed S emantic
M inimalism in chapter , let ’ s briey focus on a vie w that allo ws mor e context-sensitivity than it
and see ho w it handles under determination; a vie w that might be called ‘e Co v er t I ndexicality
A ppr oach ’ (CIA).
e Co v er t I ndexicality A ppr oach
A ccor ding to CIA, semantics is in charge of giving an account of the pr opositions asser ted
b y literal utterances of sentences—including under determined ones. O f course, not any ac-
count would do . F or CIA-theorists, only compositional accounts of semantic content count
as legitimately semantic, i.e., accounts that assign semantic contents to complex expr essions
based ex clusiv ely on the information contributed b y their syntactic str uctur es and the semantic
contents of their par ts.
U nder determination poses a challenge to CIA pr ecisely because the pr opositions under de-
termined sentences ar e used to literally asser t contain information that can ’ t be traced to any
wor d or phrase occurring in them. is means that the task for CIA is to come up with an
account of those pr opositions that is strictly compositional in the str ong sense described abo v e
and, at the same time, is capable of capturing our intuitions about them.
H o w could this task be fullled? H er e is an idea. I f the missing constituents ar e some wher e
in the syntactic str uctur e of an under determined sentenceS , then the pr opositions asser ted b y
literal utterances ofS would be simply the r esult of the compositional combination of the con-
tents assigned toS ’ s constituents. is is pr ecisely what CIA-theorists do . e key idea behind
this appr oach is that instead of wor king with the o v er t syntactic str uctur es of under determined
sentences, CIA-theorists constr uct the kind of str uctur es generativist linguists wor k with and
claim that the missing constituents ar e to be found ther e.
What ar e those str uctur es like? Linguists wor king on the generativist frame wor k tend to see
syntax as an autonomous generativ e mental component that pr o vides abstract disambiguated
str uctur es to the also psy chologically r eal semantic component. ese str uctur es ar e called
‘Logical F orms ’ (LFs) and constitute a lev el of syntactic r epr esentation that lies betw een sur-
face str uctur es and semantic interpr etation. O n this frame wor k, the r ole of natural language
semantics is to interpr et LFs, not sur face str uctur es.
CIA-theorists shar e this conception of semantics. F or them, the fact that under determined
sentences ar e missing some constituents in their sur face str uctur es that gets assigned a semantic
v alue b y the semantic component isn ’ t a r eal pr oblem. What r eally matters for them is that
the missing constituents be in their LFs. What do these co v er t c onstituents look like? S ince
the evidence points at non-linguistic contexts as the sour ce of the missing pieces of information
r equir ed to complete the pr opositions asser ted, these LFs ar e expected to hav e some constituents
capable of catching those contextual bits of information. CIA-theorists like to think of those
constituents as v ariables of some sor t or another .
I t ’ s impor tant to emphasiz e that although this vie w allo ws context to affect the semantic
contents of under determined sentences, it imposes v er y strict r estrictions on what context can
do: context is only allo w ed to pr o vide v alues to the fr ee occurr ences of v ariables in the LFs of
under determined sentences. King & S tanley ( ) call this r estricted contribution of context
‘ w e ak pragmatic effects ’ on semantic content:
A w eak pragmatic effect on what is communicated b y an utterance is a case in which
context (including speaker intentions) determines interpr etation of a lexical item in
accor d with the standing meaning of that lexical item. ( King & S tanley , )
S tr ong pragmatic effects on semantic content, i.e., contributions of context that go o v er and
abo v e mer ely supplying r efer ents to the constituents of sentences, ar e banned. us, in the CIA
frame wor k, contexts play the same r ole they play in Kaplan ’ s LD.
N o w , simply putting fr ee occurr ences of v ariables wher ev er a sentence is missing a con-
stituent would har dly qualify as an account of under determination. What y ou need is inde-
pendent evidence that those v ariables occur wher e y our account of under determination needs
them to be. S o, what evidence CIA-theorists appeal to? O ne inuential pr oposal is put for war d
b y S tanley ( , ), who appeals to the fact that occurr ences of v ariables can be bound
b y v ariable-binding operators. H is pr oposal is kno wn as the argument fr om binding . R oughly ,
her e is ho w it wor ks:
(i) T ake a declarativ e sentence S whose sur face str uctur e is missing a constituent a
the content of which is r equir ed forS to expr ess a pr oposition (i.e,S is under de-
termined).
(ii) P utS under the scope of a suitable v ariable-binding operatorQ .
(iii) I f ⌜Q(S)⌝ is a sentence that can be literally used to asser t p in c , and p is the
r esult of the compositional semantic interpr etation of a LF in which Q binds an
occurr ence of a v ariable v wher e a would be other wise expected, then v must
occur fr ee inS ’ s LF wher ea would be expected.
H er e is an example. R econsider our sentence ( ),
( ) I t ’ s raining.
N o w , put a QP in fr ont of it:
() E v er y time J ohn lights a cigar ette, it ’ s raining.
I magine a context in which an asser tiv e utterance of ( ) intuitiv ely communicates what ( )
does:
() F or ev er y timet at which J ohn lights a cigar ette, it ’ s raining att at the location
in which J ohn lights a cigar ette att .
S i nce in ( ) the location in question v aries with the choices of times, this pr o vides evidence
—S tanley thinks—that ther e is a location/time v ariable that is bound b y the QP . F r om this,
S tanley concludes that the v erb ‘rain ’ always comes with a hidden location/time v ariable which,
in the case of ( ), r eceiv es its v alue fr om context. ⁷
N ot satised with this r esult, S tanley boldly claims that ev er y under determined sentence
has bound r eadings whenev er w e stick a suitable QP in fr ont of them, and that this constitute
evidence that they hav e fr ee occurr ences of v ariables at LF . us, he concludes:
[…] all tr uth-conditional effects o f context ar e traceable to logical form. […] E xtra-
linguistic context only can affect what is expr essed in a linguistic asser tion if its contri-
bution can be traced to a constituent in the expr ession utter ed. ( S tanley , )
S ome W orr ies
I hav e two kinds of worr y about CIA. e rst targets some claims and r esults that per tain to
CIA but not necessarily to other S emanticists theories, and the second targets some featur es of
CIA that it inherits fr om its commitment to S emanticism. Let ’ s star t with the rst one. What ’ s
the r eal goal of CIA? I t ’ s not just giving an account of a fe w pr oblematic sentences. e r eal
goal is to offer a strategy to deal with under determination in natural language. e pr oblem
with CIA ’ s strategy is that it ’ s not generalizable. As I argue in chapters and , it doesn ’ t giv e
us an adequate account of sentences with pr enominal possessiv es or compound nominals.
R ecall that the challenge tha t these constr uctions pose is that in or der to be able to assign
r efer ents to their occurr ences, w e need to identify a r elation betw een the denotations of their
par ts. H o w ev er , no such r elation is to be found in the expr essions themselv es. F or instance,
⁷ S tanley thinks that ( ) can be captur ed in either of two ways. e v erb ‘rain ’ intr oduces either (i) a hidden
situation or ev ent v ariable, or (ii) a v ariable ‘f(i) ’, wher e ‘i ’ is an object v ariable and ‘f ’ a function v ariable—which
in this case will get assigned a function fr om times to locations. S ee S tanley ( , –).
in or der to interpr et ‘J ohn ’ s book ’ one needs to kno w the identity of the r elation betw een J ohn
and a book. e r eason the argument fr om binding doesn ’ t pr o vide evidence in fav or of the
existence of a r elational v ariable in ‘J ohn ’ s book ’ is that pr enominal possessiv es don ’ t hav e bound
r eadings. ⁸ I f this is right, as a general account of under determination, CIA fails.
A second worr y for CIA is the follo wing. T ake sentence ( ). A ccor ding to S tanley , the
fact that it can be used to literally asser t something that has the same tr uth conditions as ( ) is
evidence that the LFs of ( ) and ( ) hav e occurr ences of a location/time v ariable that r eceiv e
their v alues fr om context. B ut, suppose that y ou actually utter ( ). P r esumably , the occurr ence
of ‘raining ’ in ( ) must also come with an occurr ence of the same location/time v ariable. e
pr oblem with this is that that occurr ence of the v ariable would be completely unnecessar y . F or
w e alr eady hav e all the information needed to interpr et ( ) in the phrase ‘ at t at the location
in which J ohn lights a cigar ette at t ’, which occurs in the sur face str uctur e of ( ). Why isn ’ t
this fact evidence that no such v ariable occurs in the LFs of ( )? O f course, this worr y doesn ’ t
apply just to ( ). F or ev er y under determined sentenceS , w e can constr uct a parallel sentence
S
′
containing all the implicit information ofS in its sur face str uctur e. I f so, then putting co v er t
v ariables in S
′
’ s LF would be useless. ⁹ I f w e can giv e an account of under determination that
explains the data without multiplying entities bey ond necessity , w e should pr efer it o v er CIA.
⁸ S ee chapters and for details.
⁹ V ersions of this worr y hav e been v oiced b y R ett ( ) and S oames ( ).
ese worries lead me to some foundational concerns. W hat motiv ates CIA-theorists to
look for co v er t elements in syntax to account for under determination? I think that it ’ s their
commitment to S emantic D eterminism:
S D :
I f S is an unambiguous, context-sensitiv e, declarativ e sentence and p is a pr oposition
asser ted (without conv ersational implicatur es) b y a literal utterance ofS in a contextc ,
then p is fully deter mined b y the linguistic meaning ofS inc .
I f S emantic D eterminism is tr ue, then the pr opositions asser ted b y our literal uses of under-
determined sentences must be fully determined b y their linguistic meanings together with the
information pr o vided b y context. S ince, accor ding to CIA, (i) contexts can only hav e w eak
pragmatic effects on those pr opositions, and (ii) linguistic meaning must fully determine them
compositionally , the CIA-theorist has no other option than encoding the missing constituents
some wher e in syntax. is would for ce her to believ e that under determination is mer ely a
supercial phenomenon.
Why is S emantic D eterminism so widespr ead among S emanticists theories? W ell, it ’ s popu-
lar among them because it ’ s the natural vie w to adopt if y ou hold theses ()–() of S emanticism.
I f y ou (i) don ’ t make a distinction betw een semantic content and asser ted content, (ii) believ e
that the r ole of context is just to supply r efer ents to occurr ences of context-sensitiv e expr essions,
and (iii) believ e that semantics alone is in charge of laying do wn strict r ules for matching up
each expr ession of the language with a unique semantic content, then S emantic D eterminism
is for y ou. H o w ev er , as I hav e argued in the pr eceding chapters, w e don ’ t hav e to accept S e-
mantic D eterminism. I n fact, I think that the widespr ead adoption of S emantic D eterminism
b y most contemporar y semantic theories has pr ev ented us fr om getting a better understanding
of context-sensitivity and under determination.
I f that is the case and S emantic D eterminism depends on theses ()–() of S emanticism,
what is wr ong with those theses? P erhaps a little bit of histor y might help us getting clear about
this. S emanticism has its origins in the wor k of philosopher-logicians whose wor k focused
on the formaliz ed languages of logic and mathematics. ese languages lack the pr oper ties
that make under determination possible in natural language. F irst, they ar en ’ t context-sensitiv e:
mathematical and logical sentences expr ess the same pr opositions and hav e the same tr uth-
v alues acr oss ev er y possible context. S econd, they don ’ t make a distinction betw een semantic
content and asser ted content: the pr opositions asser ted b y mathematical and logical sentences
just ar e their semantic contents. S ince they lack these pr oper ties, building a model for such a
language—i.e., giving a semantics for it—simply consists in dening a domain, formulating an
assignment function mapping v alues fr om the domain to ev er y w ell-formed expr ession, dening
tr uth-in-a-model, and assigning pr opositions to sentences.
Kaplan ( b ) inherits this frame wor k and, building on the wor k of M ontague ( ;
) and Kamp ( ), constr ucts a logic of indexicals. e no v elty of his system was his
tr eatment of indexicals as dir ectly r efer ential terms, the distinction betw een content and char-
acter , and the implementation of the double-indexing technique. ¹⁰ H o w ev er , his idea of giving
a semantics for it was fundamentally the old one. e differ ence was that instead of interpr et-
¹⁰ F or details, see chapter .
ing sentences b y themselv es, his system interpr ets sentences-in-contexts, wher e contexts ar e
thought of as sequences of parameters whose r ole is simply to supply contents to occurr ences of
indexicals. B y focusing on sentences-in-contexts and not simply on sentences, Kaplan blocks
the possibility of assigning something less than pr opositions to indexical sentences.
is is the same idea behind CIA. H o w ev er , things ar e mor e complicated for CIA because
it ’ s much mor e ambitious than Kaplan ’ s LD. F irst, CIA doesn ’ t r estrict its account to a small set
of simple expr essions as Kaplan ’ s LD does. S econd, CIA aims at making empirical pr edictions
about the contents asser ted b y literal uses of those expr essions, wher eas the main motiv ation be-
hind Kaplan ’ s LD was the study logical tr uth and logical consequence in an indexical language.
ir d, wher eas Kaplan ’ s LD stipulativ ely denes its language ’ s grammar , CIA wants to wor k
with the (empirically disco v er ed) grammar of E nglish. And for th, CIA wants to achiev e these
goals using vir tually the same machiner y Kaplan used in giving his semantics. S ince the em-
plo yment of this machiner y for ces CIA-theorists to postulate syntactic str uctur e the existence
of which they lack solid evidence about, the CIA-theorist faces a dilemma: either (i) stick to the
evidence and distr ust the S emanticist methodology , or (ii) stick to the S emanticist methodol-
ogy and distr ust the evidence. Although I believ e the right choice should be (i), CIA-theorists
seem to disagr ee.
S emantic D eterminism might be a useful vie w to hold if y ou want to build a logical system
and study its logical pr oper ties. H o w ev er , if w e r eally want to understand the natur e of the
r elation betw een the meaning of an under determined sentence and the contents it can be used
to asser t, just adding contextual parameters and mapping them into co v er t constituents doesn ’ t
help us.
.. P ragmaticism on U nder deter mination
P ragmaticism emerged as a r eaction against S emanticism. I t holds that ev er y sentence in natural
language is under determined. M or eo v er , it maintains that in or der to gain a r eal understand-
ing of under determination, one must take into account the cognitiv e pr ocesses inv olv ed in the
interpr etation of utterances of sentences. Although ther e ar e differ ent implementations of P rag-
maticism in the literatur e, I want to focus on a pr ominent school of thought in the P ragmaticist
camp: R elev ance eor y (R T ).
R elev ance eor y
R T -theorists think that the information conv entionally encoded b y a sentence is so minimal
that it ’ s incapable of playing the r ole that philosophers and linguist hav e traditionally assigned
to it, namely , determining a pr oposition. ey think that a r ealistic account of the pr opositions
asser ted and communicated b y our uses of sentences should focus on speaker ’ s meaning, not
linguistic meaning. M or eo v er , they think that linguistic meaning is r educible to speaker ’ s mean-
ing, which in turn should be studied using the methods and techniques emplo y ed in cognitiv e
science. F or them, the goal of a theor y of speaker ’ s meaning is to offer a causal mechanistic
account of the mental systems inv olv ed in v erbal and non-v erbal communication.
R T -theorists defend what they call ‘ an infer ential model of communication ’ or ‘infer ential
pragmatics ’. S ince the conv entionally encoded information v astly under determines asser ted
content, the interpr etation of utterances of sentences inv olv es v er y little semantic decoding and
a gr eat deal of pragmatic infer ence. e key idea of R T is that, in the majority of cases, prag-
matic infer ence isn ’ t syntactically constrained, but led b y the utterance ’ s expected implications.
A ccor ding to S perber & W ilson ( ), ‘ the central claim of r elev ance theor y is that the ex-
pectations of r elev ance raised b y an utterance ar e pr ecise and pr edictable enough to guide the
hear er to war d the speaker ’ s meaning. ’ V er y r oughly , her e is an example of what they hav e in
mind:
(i) T alking to Bob , Alice asser tiv ely utters a declarativ e sentenceS .
(ii) Bob r ecogniz es her communicativ e intention which, in turn, raises Bob ’ s expec-
tations of r elev ance.
(iii) Bob hears the sounds coming out of Alice ’ s mouth and his linguistic decoding
system (a module of his mind) transforms those sounds into a logical for m in the
language of thought.
(iv) S ince that logical form is minimally informativ e, Bob ’ s pragmatic pr ocessor or
parser (another mental module) will look for the utterance ’ s optimally r elev ant
interpr etation .
(v) F ollo wing a path of least effor t, Bob will fr eely enrich the logical form until the
r esulting interpr etation meets his expectations of r elev ance.
(vi) O nce Bob ’ s expectations of r elev ance ar e satised, he will ha v e the most plausible
hypothesis, giv en his evidence, about what Alice meant.
A wor d of caution is in or der . I n R T , a logical form isn ’ t an LF of the kind inv oked b y generativist
syntacticians. Rather , it ’ s a conceptual r epr esentation in the language of thought pr oduced b y
the automatic decoding of linguistic signals. N o w , notice the r ole that Bob ’ s expectations of
r elev ance play in this account: they trigger the whole infer ential pr ocess of interpr etation and
also tell him when to stop .
B ut, what ar e Bob ’ s expectations of r elev ance? A ccor ding to R T , they ar e Bob ’ s pr esumption
that Alice ’ s speech act carries information that might yield a positiv e cognitiv e effect on him.
e idea is that, since b y uttering a sentence Alice is imposing a cognitiv e task on Bob , he
will pr esuppose that Alice has the intention of giving him information that is r elev ant to his
inter ests at that moment, e.g., it might answ er a question he had in mind at that moment.
O f course, the question no w is why w e should assume that Bob has that pr esumption. R T -
theorists claim that he has that pr esumption because that ’ s the way human cognitiv e systems
hav e ev olv ed. A ccor ding to them, that ’ s a br ute fact r esulting fr om natural selection. us R T
v er y rst principle:
Cognitiv e P r inciple of R elev ance
H uman cognition tends to be gear ed to the maximization of r elev ance.
R T ’ s conception of pragmatics clearly differs fr om the G ricean conpection of it. O ne of the
differ ences t hat inter ests us her e is the follo wing. O n the traditional G ricean appr oach, the
cooperativ e principle and conv ersational maxims ar e inv oked only to deal with implicatur es,
not literal explicit content. is is so because traditional G riceans tend to pr esuppose that the
pr opositions asser ted b y literal uses ofS ar e linguistically encoded, and that pragmatic infer ence
comes into play only to deriv e additional pr opositions a speaker might communicate on a giv en
occasion of use. B y contrast, R T holds that pragmatic infer ence enters into the pictur e fr om
the v er y beginning, go v erning the hear er ’ s r eco v er y of explicit content or “literal meaning ”. ¹¹
S ome W orr ies
R T is clearly a subjectivist theor y , i.e., it ’ s a theor y about thought-pr ocessing, not thought-
content. As such, one would expect to nd a substantial body of empirical evidence in fav or
of R T ’ s main tenets; evidence that would include psy chological experiments, brain scans, and
computational models. H o w ev er , that ’ s not what one nds. Rather , R T -theorists like to tell us
that R T impr o v es on other ’ s theories of meaning and communication.
e pr oblem with this line of argument is that R T ’ s targets ar e usually objectivist theories,
i.e., theories about thought-contents, not thought-pr ocessing. F or example, one of their targets
is G rice. R T -theorists (e.g. S perber & W ilson , ) suggest that R T impr o v es on G rice ’ s
account because, while G rice thinks of his Cooperativ e P rinciple and maxims as conv entional
r ules that ar e expected to be obey ed b y speakers, R T appeals to an all-purpose basic featur e of
human cognition: the sear ch for r elev ance. As they put it:
A ccor ding to r elev ance theor y , utterances raise expectations of r elev ance not because
speakers ar e expected to obey a Co-operativ e P rinciple and maxims or some other specif-
ically communicativ e conv ention, but because the sear ch for r elev ance is a basic featur e
of human cognition, which communicators may exploit. ( S perber & W ilson )
e main charge against G rice seems to be that his theor y is psy chologically implausible. What
S perber & W ilson seem to pr esuppose is that G rice was tr ying to describe the actual cognitiv e
¹¹ e pr ocesses inv olv ed in this rst stage of interpr etation ar e called ‘ explicatur es ’.
pr ocesses inv olv ed in the interpr etation of utterances. I n other wor ds, they seem to pr esuppose
that G rice aimed at giving a theor y about ho w the mind wor ks. is pr esupposition is mis-
taken. As many scholars hav e argued, ¹² G rice ’ s goal was to pr o vide a r ational r econstr uction ¹³
of communication—i.e., an idealiz ed model of what fully rational, w ell-informed, cooperativ e
speakers would do to communicate their thought-contents to each other in an efficient way .
is brings me to the to the issue of what subjectivist theories can and can ’ t do . S ubjectivist
theories can tell us ho w w e pr ocess linguistic information, what kind of strategies w e actually
use in extracting information fr om a giv en utterance, ho w the acquisition of ne w information
affects stor ed information, ho w w e infer implicit content fr om explicit one, etc. O f course, one
would expect a substantial body of empirical evidence backing up their claims. N onetheless,
what they can ’ t do is use their ndings to conclude that the pr oject of building an objectivist
semantics has little or no v alue. ¹⁴
I n general, y ou can ’ t justiably say that an objectivist theor y is no good because it doesn ’ t
achiev e subjectivist goals (or vice v ersa). M or eo v er , if y our theor y is about thought-pr ocessing,
¹² F or instance, S aul ( ) and S oames ( ).
¹³ e notion of rational r econstr uction emplo y ed her e was intr oduced b y Carnap ( ). ir ty v e y ears
later , in his intellectual autobiogr aphy , Carnap describes it as follo ws:
Although I was guided in my pr ocedur e b y the psy chological facts concerning the formation of concepts
of material things out of per ceptions, my r eal aim was not the description of this genetic pr ocess, but
rather its rational r econstr uction—i.e., a schematiz ed description of an imaginar y pr ocedur e, consisting
of rationally pr escribed steps, which would lead to essentially the same r esults as the actual psy chological
pr ocess. ( Carnap , )
¹⁴ Although not a R T -theorist himself , Chomsky seems to shar e R T -theorists ’ attitude to war ds objectivist se-
mantics. H e says: ‘insofar as w e understand language use, the argument for a r efer ence-based semantics (apar t
fr om an internalist syntactic v ersion) seems to me w eak. I t is possible that natural language has only syntax and
pragmatics. ’ ( Chomsky , )
y ou should be v er y car eful about what y ou take y our theor y to sho w about thought-contents.
J ust because linguistic meanings and semantic contents don ’ t play an impor tant r ole in y our
description of the actual cognitiv e pr ocesses inv olv ed in the interpr etation of utterances, it
doesn ’ t mean that they can ’ t play an impor tant r ole in a theor y of the information language is
used to expr ess, asser t, conv ey and communicate in differ ent occasions.
U nfor tunately , this is a mistake R T -theorists ar e pr one to make. C r ucial for R T is the idea
that pragmatic pr ocesses can make contributions to asser ted content that ar e neither syntacti-
cally nor semantically constrained. e only r eal constraint of an explicatur e is the satisfaction
of expectations of r elev ance. I t ’ s har d to tell whether they ar e right about this, in par t because
it ’ s unclear ho w one could test this claim. H o w ev er , ev en if w e grant that they ar e, it doesn ’ t
follo w that syntax and semantics don ’ t impose impor tant constraints on the information a giv en
sentence can be literally used to asser t. F or no matter ho w r elev ant it might be, an utterance of
‘ T ipper lo v es himself ’ can ’ t be used to asser t that T ipper lo v es his mother .
us, if the task is to come up with an account of asser ted content that sheds light on
under determination, R T doesn ’ t hav e much to tell us.
.. Cooperativism on U nder deter mination
e nal appr oach to the division of labor betw een semantics and pragmatics I want to con-
sider is Cooperativism. A ccor ding to Cooperativism, if a context-sensitiv e or under determined
declarativ e sentence S is used literally to asser t p (without conv ersational implicatur es), then
p is to be accounted for b y both semantics and pragmatics. S emantic G uidance—the vie w I
hav e been defending in this disser tation—is committed to Cooperativism. I n what follo ws, I
will explain why I believ e this is vie w one should adopt.
What is T ipper R eady F or?
Let ’ s go back to sentence ( ). Consider the follo wing case:
S
After months of inv estigation, the FBI has nally been able to pinpoint Corleone ’ s safe
house. S pecial agent T ipper will go under co v er to tr y to collect specic information
about the place. O nce they nish wiring him, his fello w detectiv e J ones calls headquar-
ters and says:
( ) T ipper is r eady .
What did J ones asser t? P r esumably , he asser ted one or mor e pr opositions in a set that might
include the follo wing: that T ipper is r eady to go to the safe house, that he is r eady to collect
the information he was asked to collect, that he is r eady for the mission, that he is r eady to
be deplo y ed to Corleone ’ s hiding place, that T ipper is alr eady wir ed and that the equipment is
wor king as expected, etc. H o w do w e kno w that J ones asser ted one (or mor e) of these pr opo-
sitions and not, for instance, that T ipper is r eady to r un the Boston M arathon? W ell, that ’ s
something w e can infer fr om the linguistic meaning of ( ) together with the information w e
get fr om the situation described. S ince being r eady to r un the Boston M arathon is irr elev ant
to the expected purpose of J ones ’ s utterance, w e can safely r ule it out—ev en if w e kno w that
T ipper has been training har d during the last four months for the Boston M arathon.
T wo featur es of this situation ar e wor th noticing: (i) ev en if w e w er e to x all the r elev ant
parameters of the context, the set of candidate asser ted contents of ( ) would still contain mor e
than one pr oposition; and (ii) for any one pr oposition p in this set that one might choose as
the asser ted content of J ones ’ s utterance, p could be o v erridden b y the addition of ne w infor-
mation r egar ding the par ticular situation. e rst featur e is conrmed b y the fact that J ones
could felicitously utter ( ) without having a specic completion in mind. F or instance, J ones
could utter it to simply asser t that they hav e nished their par t of the job—i.e., wiring T ipper-
—without having a clue about why T ipper had to be wir ed. e second featur e is conrmed
b y the fact that—on the face of it—for any possible completion k of ( ), J ones can clarify his
utterance b y adding ‘but not k ’ without contradicting himself . F or example, if someone in
headquar ters r eplies “G ood! S o, T ipper is r eady to collect information fr om Corleone ’ s safe
house, ” J ones could say without contradiction “ W ell, T ipper isn ’ t r eady for that y et. H e ’ s just
r eady to test the equipment with y ou guys. ”
F or S emantic G uidance, these featur es str ongly suggest that the corr ect account of the as-
ser ted content(s) of J ones ’ s utterance of ( ) must appeal to both semantics and pragmatics.
S emantic G uidance
A ccor ding to S emantic G uidance, ifS is a context- in sensitiv e sentence, its meaning or semantic
content will be a full-edge pr oposition. B eing context- in sensitiv e,S ’ s semantic content will be
the pr oposition that literal, implicatur e-fr ee utterances ofS will asser t on ev er y occasion of use.
H o w ev er , ifS is context-sensitiv e, its meaning or semantic content will be a set of constr aints on
the information that its literal, implicatur e-fr ee uses can asser t in ev er y communicativ e situa-
tion. S ince these constraints ar e to be satised acr oss all communicativ e situations in which S
is used, w e call them ‘ global constraints ’. S ’ s global constraints will simply state some, but not
all, pr oper ties a candidate asser ted content forS should hav e.
S o, the rst step to war ds an account of the pr oposition(s) asser ted b y J ones ’ s utterance is to
build an adequate asser tion checklist for ( ). I n or der to do so, w e need to extract the informa-
tion w e get fr om its syntactic str uctur e and the meaning of its constituents. Ov ersimplifying a
bit, the information w e get fr om them is that the dyadic pr oper ty of being r eady is being pr edi-
cated of an individual denoted b y the pr oper name ‘ T ipper ’. I n addition, the linguistic meaning
of ‘is r eady ’ tell us that the pr oper ty it denotes is to be instantiated b y pairs of individuals and
ev ents. W ith this information in mind, w e might formulate the follo wing asser tion checklist:
A C r is sentence is being used to pr edicate the dyadic pr oper ty of being r eady
of an individual and an ev ent.
r e individual the pr edicate is being applied to is T ipper .
As y ou may immediately notice, this asser tion checklist doesn ’ t tell us what T ipper is r eady for .
I t just imposes the constraint that, whatev er it is, it must be an ev ent. S ince this leav es an innite
set of candidate contents as liv e options—what w e call its ‘ global acceptability range ’—w e need
to supplement this asser tion checklist with local constraints imposed b y the communicativ e
situation in which J ones ’ s utterance of ( ) takes place.
H er e it ’ s impor tant to r ecall that communicativ e situations ar en ’ t sets of parameters. Rather ,
they ar e bodies of information av ailable to conv ersational par ticipants that include their com-
mon backgr ound pr esuppositions, shar ed purposes, and mutual expectations. H ence, commu-
nicativ e situations ar e external r esour ces that a gr oup of conv ersational par ticipants can equally
exploit guided b y conv ersational maxims and general principles of rational interaction to get at
asser ted contents. ¹⁵
What local constraints ar e imposed b y S on the ev ente T ipper is r eady for? W ell, it might
impose constraints such as these:
(i) e either is or facilitates the collection of information fr om Corleone ’ s safe house.
(ii) e r equir es T ipper ’ s being wir ed.
(iii) e must be in conformity with the standar d pr ocedur es on missions of this kind.
(iv) e is something J ones might kno w about.
(v) e is something headquar ters might kno w about.
W ith this information, conv ersational par ticipants will hav e to look for the identity ofe . N otice
than an immediate consequence of incorporating local constraints in our account is that it
dramatically r educes the set of candidate asser ted contents. P ossible ev ents that don ’ t satisfy
any local constraint will be automatically r uled out. is explains our intuition that no matter
ho w salient it might be, r unning the Boston M arathon is not what J ones asser ted that T ipper
¹⁵ is notion of communicativ e situation is closer to S talnaker ’ s ( ) notion of context set than to Kaplan ’ s
( b ) notion of context. H o w ev er , it differs fr om S talnaker ’ s notion in that communicativ e situations include
only information that is accessible to ev er y conv ersational par ticipant on a giv en occasion. B y contrast, S talnaker ’ s
pictur e allo ws for individual conv ersational par ticipants to hav e each their o wn context set in the same situation.
is r eady for . I n our frame wor k, w e call this pr ocess ‘ltering ’ and what r esults fr om it a ‘local
acceptability range ’.
I t ’ s wor th noticing that although ltering is cr ucial for getting at asser ted contents, local
acceptability ranges often hav e mor e than one member . F or example, in our case, e could be
one of the follo wing: r eceiving ne w or ders fr om headquar ters, testing the equipment, driving
to war ds the safe house, etc. I f ther e isn ’ t fur ther information av ailable to the conv ersational
par ticipants and the stakes ar en ’ t too high, this might be enough for them. I n that case, it
would be clear that J ones determinately asser ted a pr oposition in the local acceptability range
ev en though, for each pr oposition in that set, it would be indeterminate that he asser ted it.
H o w ev er , if the stakes ar e high and the par ties in the conv ersation need to identifye , ltering
will not be enough. Assuming that they can ’ t get information about the identity ofe fr om some
other sour ce, their task will be to determine the best candidate fr om the local acceptability
range. W e call the pr ocess of ranking the members of a local acceptability range ‘ or dering ’. F or
instance, a plausible or dering for S might be this:
. T ipper is r eady to r eceiv e instr uctions fr om headquar ters.
. T ipper is r eady to driv e to war ds Corleone ’ s safe house.
. T ipper is r eady to test the wiring equipment.
. T ipper is r eady to collect information fr om Corleone ’ s safe house.
I f the av ailable evidence corr oborates it, it would be appr opriate for them to believ e that their
top candidate is what J ones asser ted. I f not, they will either go with the next pr oposition in the
list or build a ne w one.
S ome K ey F eatur es of O ur A ccount
What pictur e of the division of labor betw een semantics and pragmatics does S emantic G uid-
ance r ecommend? I n this section, I would like to addr ess this question.
e rst thing to notice is that S emantic G uidance makes a clear distinction betw een seman-
tic content and asser ted content. A ccor ding to it, wher eas the semantic content of a context-
sensitiv e sentence S is a fully determined set of constraints, its asser ted content(s) will be un-
der determined b y its semantic content. B eing under determined b y its semantic content means
thatS ’ s semantic content will determine only a pr oper par t of all the pr opositions it can be used
to asser t. H ence,S ’ s semantic content will not be asser ted. N o w , ifS isn ’ t context-sensitiv e, its
semantic content will be pr opositional. H o w ev er , this will not guarantee b y itself that it will al-
ways be asser ted. is is so, because being asser ted takes mor e than simply being a pr opositional
semantic content. I t takes, in addition, at least the follo wing:
Asser ted Content
p is the asser ted content of a sentence S in a context c only if b y uttering S (i) p
is made av ailable to ev er y rational, competent conv ersational par ticipant in c , (ii) p is
inferable b y ev er y rational, competent conv ersational par ticipant inc , and (iii) it would
be appr opriate to tr eat the utter er ofS as committed to the tr uth of p .
S econd, accor ding to S emantic G uidance, semantics is in charge of formulating adequate as-
ser tion checklists for the sentences of the language. B ut, what counts as an adequate asser tion
checklist? An asser tion checklistA for a sentenceS is adequate iff (i)A contains each and ev er y
constraint contributed b y the linguistic meaning of each and ev er y meaning ful constituent of
S , and (ii) ev er y pr oposition asser table b y a literal utterance ofS satises ev er y constraint in A .
T wo things ar e wor th noticing her e. F irst, on this conception of adequacy , what ’ s not
pr ohibited b y A is permitted b y it. is means that asser ted contents that satisfy a g iv en A
might include information that goes bey ond what minimally satises A ’ s constraints. S econd,
adequate asser tion checklists tell us what ought to be the case, not what is actually the case.
er e ar e two r easons for this: (i) some global constraints ar e stricter than others, and (ii)
conv ersational par ticipants might succeed in asser ting a pr oposition ev en if they violate some
of the less strict global constraints fr om a giv en asser tion checklist. F or example, although one of
the constraints imposed b y the meaning of ( ) is that the subject is T ipper , if the conv ersational
situation is cooperativ e enough, one might succeed in asser ting that Ripper is r eady b y literally
uttering ( ). is, ho w ev er , will not normally wor k for violations of the constraint imposed b y
the pr edicate. F or no matter ho w cooperativ e the communicativ e situation is, a literal use of
( ) will not succeed in asser ting, say , that T ipper is not a painter (although it might implicate
it).
O n the S emantic G uidance pictur e, this is as far as semantics can take us. e r est will
be taken car e of b y pragmatics. T o do so, S emantic G uidance appeals to the notion of com-
municativ e situation. e notion of communicativ e situation is richer than the S emanticist
notion of context, but poor er than the P ragmaticist conception of it. e S emanticist notion
of context is that of a sequence of objectiv e parameters. As w e hav e seen, this notion of context
r e sponds to S emanticism ’ s commitment to S emantic D eterminism. B y contrast, the P ragmati-
cist notion of context is that of an internal networ k of mental r epr esentations. is notion of
context r esponds to P ragmaticism ’ s commitment to S ubjectivism.
S emantic G uidance is an objectivist vie w not committed to S emantic D eterminism. B eing
objectivist, S emantic G uidance is concerned with the natur e of linguistic information, not lin-
guistic information-pr ocessing. is means that its account of the r elation betw een linguistic
meaning and asser ted content will appeal only to publicly av ailable pieces of information. I n
addition, since S emantic G uidance r ejects S emantic D eterminism, its account of this r elation
will hav e to include all the public information fully rational conv ersational par ticipants would
need to communicate and understand asser ted contents. is is why the notion of communica-
tiv e situation includes things like common backgr ound pr esuppositions, shar ed conv ersational
goals, and mutual expectations.
I n sum, wher eas semantics is in charge of studying the principles needed to generate global
constraints, pragmatics is in charge of studying the principles r equir ed to identify and exploit
the constraints imposed b y par ticular communicativ e situations. What is the identity of these
principles and ho w do they exactly wor k? is is a question S emantic G uidance leav es open.
O n this vie w , they should be pr o vided b y our best objectivist theories of syntax, semantics, and
pragmatics.
A cceptable Constraint Violations
A key notion in the S emantic G uidance frame wor k is that of constraint. A constraint—w e
hav e said—is a r estriction on the set of possible contents a context-sensitiv e expr essiona can be
used to asser t on a giv en occasion. What do w e mean b y ‘r estriction ’? A ccor ding to S emantic
G uidance, a r estriction ona ’ s asser ted contents is a pr oper ty or set of pr oper ties that the asser ted
contents ofa must instantiate. F or example, the constraints imposed b y the linguistic meaning
of ‘ she ’ can be thought of as the complex pr oper ty of being a female singleton not including the
asser ter or addr essee . S ince this is a pr oper ty that must be instantiated b y ev er y candidate asser ted
content of ‘ she ’ acr oss possible communicativ e situations, w e take it to be a global constraint.
O ne might worr y that this pictur e of linguistic meaning doesn ’ t get the facts right. Ar-
guably , one could corr ectly use ‘ she ’ to r efer to a male individual whose appearance is that of
a ster eotypical female. I f this is so, then it would seem that the meaning of ‘ she ’ doesn ’ t im-
pose a constraint on the gender of its possible r efer ents—contrar y to what S emantic G uidance
claims. O r consider the follo wing case. S uppose that w e ar e walking in a lonely str eet and upon
noticing that someone is follo wing us, I tell y ou:
() S he is follo wing us.
I f in fact someone is follo wing us, I might count as having asser ted something tr ue ev en if the
individual in question turns out to be male. Again, this might suggest to some people that the
account of the linguistic meaning of ‘ she ’ put for war d b y S emantic G uidance isn ’ t quite right. ¹⁶
D o these cases sho w that our semantic account o f ‘ she ’ is wr ong? I don ’ t think so . What
they sho w is that some constraint violations can be tolerated in some (but not all) communica-
tiv e situations. R ecall that, accor ding to S emantic G uidance, linguistic meaning isn ’ t the only
sour ce of constraints on asser ted content. Communicativ e situations as w ell as general princi-
ples of rational interaction impose local constraints that ar e cr ucial for getting at what ’ s asser ted.
¹⁶ Cases like these seem to motiv ate H&K’ s pr esuppositional tr eatment of gender . S ee chapter for details.
S o, whenev er ther e ar e inconsistent competing constraints in place—wher e two constraints ar e
inconsistent just in case the satisfaction of one entails the violation of the other—cooperativ e
communicativ e situations will pr o vide an or dering of them accor ding to ho w much their satis-
faction would contribute to the shar ed goals and inter ests of the conv ersational par ticipants in
that situation.
F or example, in a communicativ e situation in which w e feel thr eaten b y the person follo wing
us, my utterance of ( ) might succeed in asser ting something ev en if the individual is male,
because in such situation it would be clear to y ou that the purpose of my asser tion is to call y our
attention to the fact that someone is follo wing us, while mistakenly identifying that perso n as a
woman. I n that case, the content of my asser tion would be tr ue, because being female is no par t
of it. H o w ev er , ther e will be other communicativ e situations in which such a violation would
r esult in the asser tion of a falsity—for instance, if w e ar e expecting M ar y to be follo wing us,
but w e r ealiz e that the only person follo wing us turns out to be male. As in the pr evious case, I
would succeed in asser ting something, because y ou would be in a position to determine what
I’ m intending to asser t despite my mistaken belief about the identity of the person follo wing
us. B ut in this case, what ’ s asser ted would include the asser tion of a falsehood because it would
be clear to y ou that I’ m pr esupposing that the person is M ar y—ther eb y leading to the false
asser tion that M ar y is follo wing us.
S imilarly for a use of ‘ she ’ to pick out a male individual that looks like a ster eotypical female.
What the linguistic meaning of ‘ she ’ tells us is that its r efer ent should be look for in the r elev ant
set of females. is isn ’ t a deterministic r ule. Rather , it ’ s a piece of guidance, a clue pr o vided
b y the meaning of ‘ she ’ that cooperativ e speakers exploit to help their hear ers gur e out the
content of their asser tions. W ith this piece of information conv ersational par ticipants will
sear ch for the best av ailable candidate content (if any), not necessarily for the unique one. I f
the communicativ e situation contains enough information, an utterance of ‘ she ’ might pick out
a r elev ant male individual. I f not, such a use of ‘ she ’ might giv e the audience gr ounds to think
that the utter er believ es or that he wants them to believ e something false. ese considerations
suggest the follo wing general principle:
A cceptable Constraint Violation
A violation of a constraintC is acceptable in a communicativ e situation s iff ther e is
another constraintC
′
ins such that (i)C
′
is inconsistent withC , and (ii) conv ersational
par ticipants in s ar e disposed to believ e (or , at least, ar e disposed to believ e the utter er
believ es) that satisfyingC
′
ser v es better their shar ed goals and inter ests than satisfying
C does.
is principle might help us to account for some mor e inter esting constraint violations. F or
instance, consider the follo wing sentence:
() I f a student fails the math exam, they should talk to their teacher .
S ince the linguistic meaning of ‘ they ’ (or its inected forms, ‘ them ’ or ‘ their ’) imposes the
constraint that its r efer ent must instantiate the pr oper ty of being a plur ality not including the
asser ter or addr essee , then an utterance of ( ) to asser t what ( ) does constitutes a violation of
the number-constraint imposed b y linguistic meaning:
() I f a student fails the math exam, that student should talk to that student ’ s o wn
teacher .
A ccor ding to the A cceptable Constraint V iolation principle, in or der for an utterance of ( ) to
asser t what ( ) does and, thus, make the violation of the number-constraint acceptable, ther e
must be a constraintC
′
imposed b y the r elev ant communicativ e situations that is inconsistent
with the number-constraint imposed b y the meanings of ‘ they ’ and ‘ their ’, but whose satisfac-
tion conv ersational par ticipants ar e disposed to believ e (or , at least, ar e disposed to believ e the
utter er believ es) ser v es better their shar ed goals and inter ests in s . As it turns out, ther e ar e
many communicativ e situations that impose such an o v erriding constraint.
e singular use of ‘ they ’ (or of its inected forms, ‘ them ’ or ‘ their ’) illustrated b y ( ) is
called ‘ epicene they ’. ¹⁷ U ses of epicene ‘ they ’ ar e meant to av oid imposing gender-constraints
that would be other wise imposed if instead of it w e used ‘he ’ or ‘ she ’ (or their inected forms).
is use of ‘ they ’ star ted to become par ticularly popular during the sixties as a r esult of the
raise of the feminist mo v ement and its disappr o v al of the use of generic ‘he ’ in the E nglish
language. S o, in communicativ e situations in which generic uses of gender-specic pr onouns
ar e highly disappr o v ed of , ther e will be a social constraint against those uses that will o v erride
the number-constraint imposed b y the meaning of ‘ they ’.
Although the choice of ‘ they ’ isn ’ t optimal, it ’ s the least pr oblematic of the v e E nglish
personal pr onouns whose meanings don ’ t impose gender-constraints: ‘I’, ‘ w e ’, ‘ y ou ’ (singular
and plural), ‘it ’ and ‘ they ’. B riey , the pr oblem with the rst and second person pr onouns
¹⁷ S ee B ar on ( ) and B alhorn ( ).
is that they impose person-constraints that make it impossible for them to co-r efer with the
generic antecedent ‘ a student ’. is leav es us only two options: ‘it ’ and ‘ they ’. ¹⁸ Each of them
intr oduces a difficulty of its o wn. e difficulty intr oduced b y ‘it ’ is that its linguistic meaning
imposes the constraint that its r efer ent must be inanimate. e difficulty intr oduced b y ‘ they ’
is that its meaning imposes the constraint that its r efer ent must be a plurality . us, if one
wants the anaphoric pr onoun to be a gender-neutral, then one must choose which constraints
of either ‘it ’ or ‘ they ’ would be less costly to violate. I ntuitiv ely , violating the number-constraint
imposed b y ‘ they ’ is less costly than violating the inanimate-constraint imposed b y ‘it ’, as the
use of the latter might suggest a denial of animacy of its r efer ent—a suggestion that might be
ev en mor e offensiv e than the suggestion of sexism that uses of generic ‘he ’ might make. ¹⁹
As the A cceptable Constraint V iolation principle states, this kind of constraint violations
ar e tolerated in communicativ e situations in which conv ersational par ticipants ar e disposed to
believ e ther e is an o v erriding r eason for such a violation—for instance, in communicativ e situa-
tions in which conv ersational par ticipants shar e the belief that violating a constraint imposed b y
political corr ectness might be mor e serious than violating a semantic constraint. ²⁰ is would
¹⁸ O f course, ther e is a thir d option I’ m leaving aside in this discussion: the disjunctiv e ‘he or she ’.
¹⁹ is isn ’ t a pr oblem for pr o-dr op languages such as S panish. F or in S panish, the translation of the bound r ead-
ing of ( ) leav es the anaphoric pr onoun tacit, and intr oduces the thir d-person, gender-neutral singular personal
pr onoun ‘ su ’ to stand for epicene ‘ their ’:
() S i un estudiante r epr ueba el examen de matemáticas, [pr o] debería hablar con su pr ofesor .
H o w ev er , ( ) isn ’ t gender-neutral either , because ‘ un estudiante ’ is masculine.
²⁰ is might be the case, for example, if ther e is a shar ed belief in s that using generic ‘he ’ might offend
some conv ersational par ticipants and, thus, be an obstacle to the other wise smooth and efficient communication
betw een them.
explain the widespr ead intuition that although a use of ‘ they ’ to r efer to an entity that is not
plural constitutes a kind of violation, no mistake is being made b y such a use.
A S emanticist O bjection?
A ccor ding to S emantic G uidance, context-sensitiv e expr essions such as pr enominal possessiv es,
compound nominals, and indexicals ar e semantically under determined. O n this pictur e, the
pr ecise determination of asser ted contents is a task left to conv ersational par ticipants who will
carr y it out b y appealing to the information at hand in their communicativ e situations together
with the application of general principles of rational interaction.
S emanticists ar en ’ t happy about this pictur e. ey pr efer a pictur e in which linguistic mean-
ing gets the lion ’ s shar e of the cr edit for the determination of asser ted content. S o, her e is an
objection a S emanticist might raise against S emantic G uidance. Consider a sentence of the form
‘J ohn ’ s mother is F ’. A ccor ding to S emantic G uidance, the meaning of this sentence will de-
termine an asser tion checklist that lacks a checkbo x for the r elev ant possessiv e r elation betw een
J ohn and a mother . H o w ev er , that asser tion checklist will hav e a checkbo x for the constraint
that the mother must be signicantly r elated to J ohn. is might suggest to the S emanticist
that the sentence ‘J ohn ’ s mother is F ’ is synonymous with the sentence ‘e mother that is
signicantly r elated to J ohn is F ’. S o the S emanticist might claim that b y giving a semantics
for the latter w e ar e ther eb y giving a semantics for the former .
F ur thermor e, the S emanticist might ev en claim that ‘J ohn ’ s mother is F ’ and ‘e mother
that is signicantly r elated to J ohn is F ’ shar e the same LF and, thus, that they differ only in
their sur face str uctur e. S ince for the S emanticist the job of semantics is to interpr et LFs and
not sur face str uctur es, and since it seems that w e can generate equiv alent sentences of this kind
for ev er y sentence S emantic G uidance tr eats as semantically under determined, the S emanticist
might conclude that w e can giv e a pur ely semantic account of the contents asser ted b y literal
utterances of context-sensitiv e constr uctions.
e S emanticist thinks that this mo v e giv es us semantically the same outputs that S emantic
G uidance giv es us non-semantically and, thus, constitutes a r efutation of S emantic G uidance. I
disagr ee. What the S emanticist ’ mo v e amounts to is putting in the object language a constraint
S emantic G uidance puts in the metalanguage. B ut this is no pr ogr ess whatsoev er . I f instead of
saying ‘J ohn ’ s mother isF ’ I simply said ‘e mother that is signicantly r elated to J ohn is F ’, my
audience would still need to appeal to the communicativ e situation to gur e out what signicant
r elation betw een J ohn and a mother is the one that matters for my asser tion. A dding the wor ds
‘ signicant ’ or ‘r elev ant ’ to the object language doesn ’ t spar e conv ersational par ticipants the
task of guring out what r elations ar e signicant or r elev ant in their communicativ e situations.
H ence, rather than being a r efutation of S emantic G uidance, the S emanticist ’ s mo v e nicely
illustrates the need to appeal to non-semantic r esour ces to get at asser ted content.
M or eo v er , ‘J ohn ’ s mother ’ isn ’ t synonymous with ‘ the mother that is signicantly r elated to
J ohn ’. F or one can imagine communicativ e situations in which an utterance of ( ) would be
informativ e while an utterance of ( ) wouldn ’ t be:
() e mother that is signicantly r elated to J ohn is J ohn ’ s mother .
() J ohn ’ s mother is J ohn ’ s mother .
is wouldn ’ t be the case, if ( ) and ( ) w er e synonymous. e fact that they ar en ’ t casts
serious doubts on the S emanticist ’ s alternativ e. I n fact, ( ) and ( ) hav e differ ent modal
pr oles, as sho wn b y the follo wing:
() N ecessarily , the mother that is signicantly r elated to J ohn is J ohn ’ s mother .
() N ecessarily , J ohn ’ s mother is J ohn ’ s mother .
Clearly , ( ) is tr ue, while ( ) isn ’ t. I f so, then ( ) and ( ) ar en ’ t synonymous. I conclude,
then, that this attempt to r efute S emantic G uidance fails.
What H av e W e G ained?
S emantic G uidance is cooperativist in that it sees semantics and pragmatics as making their
o wn distinctiv e contributions to war ds a common goal: giving a rational r econstr uction of ho w
asser ted content is assembled and understood. B ut, ar e ther e any adv antages of taking this
r oute o v er S emanticism or P ragmaticism? I think t her e ar e. Let ’ s star t b y comparing it with
P ragmaticism.
B eing objectivist, S emantic G uidance is neutral about the natur e of the cognitiv e mecha-
nisms inv olv ed in thought-pr ocessing. I nstead of engaging in armchair speculation about ho w
our mind wor ks, it offers an account of the n atur e of the information asser ted b y literal uses
of declarativ e sentences, ho w much of it is conv entionally encoded and ho w much of it has a
differ ent sour ce. I n doing so, it obser v es the division of labor betw een philosophy of language
and (empirically informed) psy chology .
S econd, although accor ding to S emantic G uidance the information conv entionally encoded
b y an under determined sentence S falls shor t of being a pr oposition, it ’ s the backbone of the
pr opositionsS can be used to asser t on differ ent occasions. is allo ws S emantic G uidance to
explain, for example, why a giv en literal use of ‘ T ipper lo v es himself ’ doesn ’ t asser t that T ipper
lo v es his mother , no matter ho w r elev ant that pr oposition might be.
S emantic G uidance has also some impor tant adv antages o v er S emanticism. F irst, b y dis-
tinguishing betw een semantic content and asser ted content as w ell as r ejecting S emantic D e-
terminism, S emantic G uidance isn ’ t under any pr essur e to postulate co v er t elements in syntax
just to explain under determination. is, of course, doesn ’ t mean that S emantic G uidance r e-
jects syntactic theories that appeal to co v er t str uctur e. I t simply r ejects theories like CIA that
postulate co v er t constituents without solid independent evidence.
S econd, S emantic G uidance allo ws for a mor e intuitiv e stor y about semantic competence
than the one allo w ed b y S emanticism—ev en though S emantic G uidance itself is neutral about
exactly ho w such a stor y would go . I ntuitiv ely , the linguistic meaning of an expr ession a is
that which must be learned in or der to pr operly use and understand a . S ince, accor ding to
S emanticism, the linguistic meaning of a is a r ule that fully determines, for ev er y context, a
content, being a competent user of a would r equir e kno wing that r ule. is imposes a signi-
cant cognitiv e task on speakers, for in or der to count as competent users of the language, they
will hav e to stor e in their “ semantic module ” a huge number of such deterministic r ules, one
for each context-sensitiv e lexical item. ²¹
²¹ I borr o w this point fr om Ray o ( for thcoming ).
B y contrast, S emantic G uidance imposes a much lighter bur den on language learners. All
it r equir es fr om them is to associate a small set of conditions with each lexical item that ev er y
content is expected to satisfy if it ’ s to count as a plausible candidate for asser tion. e addi-
tional information they will need to single out asser ted contents will come fr om non-linguistic
sour ces and will be accessible to them b y their application of general principles of rationality
and rational interaction.
Another adv antage of S emantic G uidance o v er S emanticism is that quite often ther e isn ’ t
one specic pr oposition asser ted r elativ e to ev er y context. S illustrates this point w ell enough.
Although it ’ s clear that J ones determinately asser ted a pr oposition in a cer tain set that doesn ’ t
include, for instance, the pr oposition that T ipper is r eady to r un the Boston M arathon, it might
be indeterminate which specic pr oposition in that set he asser ted. is cr eates a pr oblem
for S emanticism, whose commitment to S emantic D eterminism for ces it to choose a unique
pr oposition. As w e hav e seen, S emantic G uidance has the r esour ces to handle cases like these
smoothly .
. Conclusion
e aim of this chapter was to locate S emantic G uidance in the semantics-v ersus-pragmatics
debates, and offer some r easons as to why w e should pr efer it o v er alternativ e ways of dividing
the labor betw een semantics and pragmatics.
I star ted b y par titioning the continuum of positions one could take in these territorial de-
bates in thr ee: S emanticism, P ragmaticism, and Cooperativism. M y characterization of them
focused on thr ee main issues: (i) their conception of the r elation betw een semantic content and
asser ted content, (ii) the r ole they assign to context in the determination of asser ted content,
and (iii) the bur den they impose on semantics and pragmatics in the explanation of asser ted
content. After this, I explained the challenge posed b y the phenomenon of under determination
to the traditional division of labor betw een semantics and pragmatics, and why being able to
handle it matters.
en, I discussed two notions that appear pr ominently in the debates in question: the
notions of thought and what is said. I argued that ther e is mor e than one legitimate notion
of thought and what is said, and that failing to distinguish betw een them can lead to serious
misunderstandings about the natur e of the theor etical enterprise under taken. F inally , I used
the distinctions made to describe and assess thr ee par ticular theories that illustrate the basic
tenets of S emanticism, P ragmaticism, and Cooperativism r espectiv ely . I argued that S emantic
G uidance is cooperativist and that it has the r esour ces to handle under determination better
than the alternativ es consider ed.
B ibliography
A ustin, J. L. (). H o w to D o ings W ith W or ds . Oxfor d: Oxfor d U niv ersity P r ess.
B ach, K. (). Conv ersational I mplicitur e. M ind and Language , (), –.
B ach, K. (). e S emantics-P ragmatics D istinction: What it is and Why it M atters. L in-
guistische B erichte , , –. S pecial I ssue on P ragmatics.
B ach, K. (). Y ou D on ’ t S ay? S ynthése , (–), –.
B alhorn, M. (). e Rise of E picene ey . J our nal of E nglish L inguistics , (), –.
B ar ker , C. (). P ossessiv e D escriptions . S tanfor d: CSLI P ublications.
B ar ker , C. (). P ossessiv es and R elational N ouns. I n v . H. M aienborn, & P or tner (E ds.) S e-
mantics: A n I nter national H andbook of N atur al Language M eaning , v ol. . B erlin: de G r uyter .
B ar on, D. E. (). e E picene P r onoun: e W or d at F ailed. A merican S peech , (),
–.
B auer , L. (). O n the N eed for P ragmatics in the S tudy of N ominal Compounding. J our nal
of P r agmatics , (), –.
B raun, D. (). D emonstrativ es and their Linguistic M eanings. N oûs , (), –.
B raun, D. (). I ndexicals. I n E. N. Z alta (E d.) e S tanfor d E ncy clopedia of P hilosophy .
S tanfor d U niv ersity , summer ed.
Cappelen, H., & Lepor e, E. (). I nsensitiv e S emantics: A D efense of S emantic M inimalism
and S peech A ct P lur alism . Oxfor d: B lackw ell P ublishers.
Cappelen, H., & Lepor e, E. (a). R eply to C ritics. P hilosophy and P henomenological R e-
sear ch , ().
Cappelen, H., & Lepor e, E. (b). R esponse. M ind and Language , (), –.
Carnap , R. (). D er logische A ufbau der W elt . B erlin-Schlactensee: W eltkr eis-V erlag.
Carnap , R. (). M y P hilosophical D ev elopment. I n P . A. Schilpp (E d.) e P hilosophy of
R udolph C ar nap , (pp . –). LaS alle, I llinois: O pen Cour t P ublishing Company .
Chomsky , N. (). S yntactic S tr uctur es . e H ague: M outon.
Chomsky , N. (). Language and N atur e. M ind, N ew S eries , (), –.
Chomsky , N., & H alle, M. (). e Sound P atter n of E nglish . N e w Y or k: H arper and R o w .
Cohen, S. (). Kno wledge, Context, and S ocial S tandar ds. S ynthése , (), –.
Corbett, G. G. (E d.) (). H eads in G r ammatical eor y . Cambridge: Cambridge U niv ersity
P r ess.
D eR ose, K. (). Contextualism and Kno wledge A ttributions. P hilosophy and P henomeno-
logical R esear ch , (), –.
D onnellan, K. (). R efer ence and D enite D escriptions. P hilosophical R eview , , –.
D o wning, P . (). O n the C r eation and U se of E nglish Compound N ouns. Language , (),
–.
F ara, D. G. (). S hifting S ands: An I nter est-R elativ e eor y of V agueness. P hilosophical
T opics , , –. O riginally published under the name “D elia G raff”.
F r ege, G. (). O n S ense and R efer ence. I n P . G each, & M. B lack (E ds.) T r anslations fr om
the P hilosophical W ritings of G ottlob F r ege , (pp . –). Oxfor d: B asil B lackw ell. O riginally
published in as “Ü ber S inn und B edeutung. ” T ranslated b y M ax B lack.
G rice, H. P . (a). Logic and Conv ersation. I n S tudies in the W ay of W or ds , (pp . –).
Cambridge, M assachusetts: H ar v ar d U niv ersity P r ess.
G rice, H. P . (b). U tter er ’ s M eaning and I ntentions. I n S tudies in the W ay of W or ds , (pp .
–). Cambridge, M assachusetts: H ar v ar d U niv ersity P r ess.
H awthorne, J. (). T esting for Context-D ependence. P hilosophy and P henomenological
R esear ch , ().
H eim, I. (). F eatur es on Bound P r onouns. I n D. A. D aniel H arbour , & S. B éjar (E ds.) P hi
eor y: P hi-F eatur es acr oss M odules and I nter faces , (pp . –). Oxfor d: Oxfor d U niv ersity
P r ess.
H eim, I., & Kratz er , A. (). S emantics in G ener ativ e G r ammar . Oxfor d: B lackw ell P ub-
lishers.
Kamp , H. (). F ormal P r oper ties of ‘N o w ’. eoria , , –.
Kaplan, D. (a). After thoughts. I n J. Almog, J. P err y , & H. W ettstein (E ds.) emes fr om
Kaplan , (pp . –). Oxfor d: Oxfor d U niv ersity P r ess.
Kaplan, D. (b). D emonstrativ es. I n J. Almog, J. P err y , & H. W ettstein (E ds.) emes fr om
Kaplan , (pp . –). Oxfor d: Oxfor d U niv ersity P r ess.
King, J. C., & S tanley , J. (). S emantics, P ragmatics, and the R ole of S emantic Content.
I n Z. G. Szabó (E d.) S emantics v s. P r agmatics , (pp . –). Oxfor d: Oxfor d U niv ersity
P r ess.
Kripke, S. (). S peaker ’ s R efer ence and S emantic R efer ence. I n P . A. F r ench, T . Uehling,
Jr ., & H. K. W ettstein (E ds.) M idw est S tudies in P hilosophy , V olume II , (pp . –). M in-
neapolis: U niv ersity of M innesota P r ess. V olume : S tudies in S emantics.
Levi, J. N. (). e S yntax and S emantics of Complex N ominals . N e w Y or k: A cademic P r ess.
Levinson, S. C. (). D eixis. I n L. H orn, & G. W ar d (E ds.) e H andbook of P r agmatics ,
(pp . –). Oxfor d: B lackw ell.
Le wis, D. K. (). I ndex, Context, and Content. I n S. Kanger , & S. Ö hman (E ds.) P hilos-
ophy and G r ammar: P apers on the O ccasion of the Q uincentennial of U ppsala U niv ersity , (pp .
–). D or dr echt: D. R eidel P ublishing Co .
Le wis, D. K. (). E lusiv e Kno wledge. A ustr alasian J our nal of P hilosophy , ().
Lieber , R., & Š tekauer , P . (). e O xfor d H andbook of Compounding . N e w Y or k: Oxfor d
U niv ersity P r ess.
M ontague, R. (). P ragmatics. I n R. Klibansky (E d.) Contempor ar y P hilosophy: A S ur v ey .
I , (pp . –). F lor ence: La N uo v a I talia E ditrice. R eprinted in F or mal P hilosophy , b y
Richar d M ontague, Y ale U niv ersity P r ess, N e w H av en, CT , , pp . –.
M ontague, R. (). P ragmatics and I ntensional Logic. S ynthése , (–), –. R eprinted
in F or mal P hilosophy , b y Richar d M ontague, Y ale U niv ersity P r ess, N e w H av en, CT , ,
pp . –.
M orris, C. W . (). F oundations of the eor y of S igns . Chicago: U niv ersity of Chicago P r ess.
N ako v , P . I. (). U sing the W eb as an I mplicit T raining S et: A pplication to N oun Com-
pound S yntax and S emantics. T ech. R ep . UCB/EECS--.
N eale, S. (). D escriptions . Cambridge, M assachusetts: MIT P r ess.
N unberg, G. (). T ransfers of M eaning. J our nal of S emantics , , –.
Ó S éaghdha, D. (). Learning Compound N oun S emantics. T ech. R ep . UCAM-CL-TR-
, U niv ersity of Cambridge, Computer Laborator y .
URL http://www.cl.cam.ac.uk/techreports/UCAM- CL- TR- 735.pdf
P ar tee, B. H. (). G enitiv es – A Case S tudy . A ppendix to eo M.V . J anssen, ‘Composi-
tionality ’. I n J. v an B enthem, & A. ter M eulen (E ds.) e H andbook of Logic and Language ,
(pp . –). Amster dam: E lsevier .
P ar tee, B. H. (). M any Q uantiers. I n Compositionality in F or mal S emantics: S elected
P apers b y B arbar a H. P ar tee , (pp . –). Oxfor d: B lackw ell P ublishers.
P ar tee, B. H., & Borschev , V . (). I ntegrating Lexical and F ormal S emantics: G enitiv es,
R elational N ouns, and T ype-S hifting. I n R. Cooper , & T . G amkr elidz e (E ds.) P r oceedings of
the S econd Tbilisi S ymposium on Language , Logic, and Computation , (pp . –). Tbilisi:
Center on Language, Logic, S peech, Tbilisi S tate U niv ersity .
P err y , J. (a). I ndexicals and D emonstrativ es. I n C. W right, & B. H ale (E ds.) A Companion
to the P hilosophy of Language , (pp . –). Oxfor d: Oxfor d U niv ersity P r ess.
P err y , J. (b). ought W ithout R epr esentation. S upplementar y P r oceedings of the A ris-
totelian Society , (), –.
P eters, S., & W esterståhl, D. (). Q uantiers in Language and Logic . Oxfor d: Oxfor d
U niv ersity P r ess.
P riest, G. (). A n I ntr oduction to N on-Classical Logic: F r om I f to I s . Cambridge, E ngland:
Cambridge U niv ersity P r ess, ed.
Raffman, D. (). V agueness and Context-R elativity . P hilosophical S tudies , , –.
Ray o, A. (for thcoming). A P lea for S emantic Localism. N oûs .
R ett, J. (). Context, Compositionality and Calamity . M ind & language , (), –.
S ainsbur y , R. M. (). T wo W ays to S moke a Cigar ette. R atio , XIV , –.
S almon, N. (). D emonstrating and N ecessity . e P hilosophical R eview , (), –.
S aul, J. M. (). What I s S aid and P sy chological R eality: G rice ’ s P r oject and R elev ance
eorists ’ C riticisms. L inguistics and P hilosophy , (), –.
S oames, S. (). U nderstanding T r uth . Oxfor d: Oxfor d U niv ersity P r ess.
S oames, S. (). B ey ond Rigidity: e U nnished S emantic A genda of N aming and N ecessity .
Oxfor d: Oxfor d U niv ersity P r ess.
S oames, S. (). N aming and Asser ting. I n Z. G. Szabó (E d.) S emantics v ersus P r agmatics ,
(pp . –). Oxfor d: Oxfor d U niv ersity P r ess. R eprinted in P hilosophical E ssays , V ol. ,
b y Scott S oames, P rinceton U niv ersity P r ess, P rinceton, , pp . –.
S oames, S. (). A ctually . A ristotelian Society S upplementar y V olume , , –. R eprinted
in P hilosophical E ssays , V ol. , b y Scott S oames, P rinceton U niv ersity P r ess, P rinceton, ,
pp . –.
S oames, S. (). D rawing the Line betw een M eaning and I mplicatur e—and R elating both to
Asser tion. N oûs , (), –. R eprinted in P hilosophical E ssays , V ol. , b y Scott S oames,
P rinceton U niv ersity P r ess, P rinceton, , pp . –.
S oames, S. (). e G ap betw een M eaning and Asser tion: Why What W e Literally S ay
O ften D iffers fr om What our W or ds Literally M ean. I n P hilosophical E ssays , v ol. , (pp .
–). P rinceton: P rinceton U niv ersity P r ess.
S oames, S. (). P hilosophy of Language . P rinceton F oundations of Contemporar y P hiloso-
phy . P rinceton: P rinceton U niv ersity P r ess.
S perber , D., & W ilson, D. (). R elev ance: Communication and Cognition . Cambridge,
M assachusetts: H ar v ar d U niv ersity P r ess.
S perber , D., & W ilson, D. (). R elev ance eor y . I n L. R. H orn, & G. W ar d (E ds.) e
H andbook of P r agmatics , (pp . –). Oxfor d: B lackw ell P ublishers.
S talnaker , R. C. (). Asser tion. I n P . Cole (E d.) S yntax and S emantics : P r agmatics . A ca-
demic P r ess.
S tanley , J. (). Context and Logical F orm. L inguistics and P hilosophy , (), –.
S tanley , J. (). M aking I t Ar ticulated. M ind and Language , (–), –.
T aylor , J. R. (). P ossessiv es in E nglish: A n E xplor ation in Cognitiv e G r ammar . Oxfor d:
Clar endon P r ess.
W eiskopf , D. A. (). Compound N ominals, Context, and Compositionality . S ynthese ,
(), –.
W illiamson, T . (). Kno wledge and I ts L imits . Oxfor d: Oxfor d U niv ersity P r ess.
Abstract (if available)
Abstract
We utter declarative sentences to assert propositions. Many of the sentences we utter for this purpose are context-sensitive: the propositions they are literally used to assert can change from one context of utterance to another. For example, if I tell you ‘I am talking to you’, I assert a proposition that is different from the proposition you would assert if you uttered it. However, the relation between these propositions and the linguistic meaning of the sentence is the same: the linguistic meaning of ‘I am talking to you’ fully determines, for every context, what is asserted by a literal utterance of it. Cases like this have led many philosophers to believe that if S is an unambiguous, context-sensitive, declarative sentence and p is a proposition asserted (without conversational implicatures) by a literal utterance of S in a context c, then p is fully determined by the linguistic meaning of S in c. I call this belief ‘Semantic Determinism’. ❧ This dissertation has two main goals. The first is to challenge Semantic Determinism and the second is to propose an alternative theory of the relationship between the meanings of context-sensitive sentences and the contents they are literally used to assert. I show that semantic theories committed to Semantic Determinism have real difficulties accounting for ordinary uses of sentences containing linguistic constructions such as prenominal possessives, compound nominals, and indexicals. Thus, I propose an alternative view I call ‘Semantic Guidance’. According to Semantic Guidance, the linguistic meaning of an unambiguous, context-sensitive, declarative sentence S merely provides speakers with an assertion checklist, i.e., a non-exhaustive set of conditions that every literal use of S must satisfy in order to result in an assertion. Assertion checklists play two fundamental roles: they help speakers to ensure their literal uses of S result in assertions of the propositions they intend to communicate, and (ii) they guide hearers towards the asserted contents intended by speakers without fixing them. I show how this account not only can explain the data semantic theories committed to Semantic Determinism explain best, but also is able to handle a wider range of data. ❧ The dissertation proceeds as follows. In the first chapter, I examine a recent theory called ‘Semantic Minimalism’, which holds that the meanings of unambiguous indexical-free declarative sentences are always full-fledged propositions. It further maintains that since those propositions can be grasped and reported across different contexts, those sentences are context-insensitive. The few context-sensitive expressions in natural language are the Kaplanian indexicals. I argue against this view and show the different ways in which indexical-free sentences can be context-sensitive. Once this is seen, the Semantic Determinism presupposed by Semantic Minimalism falls by the wayside. ❧ The next two chapters are devoted to detailed analyzes of two linguistic constructions that pose serious difficulties to any theory committed to Semantic Determinism: prenominal possessives and compound nominals. Prenominal possessives are linguistic constructions such as ‘John’s book’ and ‘John’s mother’ that contain the morpheme ’s. Compound nominals are sequences of two or more nouns that function as a single noun, e.g., ‘steel knife’ and ‘pie chart’. Semantic Determinists are baffled by these constructions because competent speakers normally use them to assert a diversity of relations between their constituent parts that are nowhere to be found in the expressions themselves. For example, although ‘steel knife’ and ‘kitchen knife’ exhibit the same grammatical structure, the relations between knife and steel (e.g., knife made of steel), and knife and kitchen (e.g., knife used in the kitchen) are completely different. ❧ Moreover, typically different literal uses of a single construction can convey different relations depending on the context of use. For instance, the possessive phrase ‘John’s picture’ can be used to refer to a picture portraying John, a picture taken by John, a picture owned by John, etc. I argue that theorists have found these expressions deeply puzzling, because they have been assuming Semantic Determinism. I propose that the linguistic meanings of sentences containing them just give us assertion checklists that merely constrain the range of propositions that one can assert by literal uses of those sentences. I show how, as soon as we adopt this view, the puzzle disappears. ❧ The fourth chapter extends the Semantic Guidance framework to the home territory of Semantic Determinism: indexicals. I challenge the traditional treatment of indexicals as well as some influential alternative accounts in contemporary linguistics and philosophy by presenting some data that these theories fail to account. I argue that in order to account for these data, we should treat indexical sentences as semantically underdetermined, i.e., as expressions whose linguistic meanings determine, for every communicative situation, sets of propositions that one is allowed to assert by uttering them. This account leaves the job of identifying the intended asserted contents in those sets to conversational participants, who would carry out this task guided by conversational maxims and principles, their background presuppositions, and the purposes of their particular communicative situations. I show how, by adopting this view, the problematic data can be explained away. ❧ The last chapter attempts to situate Semantic Guidance in the semantics-vs-pragmatic debates. To do so, I characterize three positions one can take about the division of labor between semantics and pragmatics in the explanation of the information a given linguistic expression can be literally used to assert: Semanticism, Pragmaticism, and Cooperativism. My characterization of them focuses on three main issues: (i) their conception of the relation between semantic content and asserted content, (ii) the role they assign to context in the determination of asserted content, and (iii) the burden they impose on semantics and pragmatics in the explanation of asserted content. In addition, I explain the challenge posed by the phenomenon of underdetermination to the traditional division of labor between semantics and pragmatics, and why being able to handle it matters. Finally, I argue that the division of labor presupposed by Semantic Guidance (Cooperativism) has the resources to handle underdetermination better than the alternatives considered.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Conceptually permissive attitudes
PDF
Telling each other what to do: on imperative language
PDF
Meaningfulness, rules, and use-conditional semantics
PDF
Reasoning with uncertainty and epistemic modals
PDF
Narrowing the focus: experimental studies on exhaustivity and contrast
PDF
Superlative ambiguities: a comparative perspective
PDF
Comparative iIlusions at the syntax-semantics interface
PDF
Instantial terms, donkey anaphora, and individual concepts
PDF
Subjectivity, commitments and degrees: on Mandarin hen
PDF
Positivist realism
PDF
What "ought" ought to mean
PDF
Reference time in the dynamics of temporal dependency in Korean
PDF
The theory of empty noun in Chinese: with special reference to the right node raising construction
PDF
Contrastive reasons
PDF
Exploring the effects of Korean subject marking and action verbs’ repetition frequency: how they influence the discourse and the memory representations of entities and events
PDF
A reduplicative analysis of sentence modal adverbs in Spanish
PDF
A corpus-based discourse analysis of Korean grammatical constructions: Focus on the multifold functions and meanings of the pragmatic construction e kaciko
PDF
Rationality and the primacy of the occurrent
PDF
Towards a correlational law of language: three factors constraining judgement variation
PDF
The notion of topic-comment constructions and the meaning of the Korean topic marker '-(n)un'
Asset Metadata
Creator
Villanueva Chigne, Eduardo
(author)
Core Title
Constraining assertion: an account of context-sensitivity
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Philosophy
Publication Date
04/11/2012
Defense Date
03/05/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
assertion,compound nominals,constraint,context sensitivity,indexicals,meaning,OAI-PMH Harvest,philosophy of language,pragmatics,prenominal possessives,semantic minimalism,semantics
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Soames, Scott (
committee chair
), Schein, Barry (
committee member
), Schroeder, Mark (
committee member
)
Creator Email
edu.villanueva@gmail.com,evillanu@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-26882
Unique identifier
UC11288272
Identifier
usctheses-c3-26882 (legacy record id)
Legacy Identifier
etd-Villanueva-589.pdf
Dmrecord
26882
Document Type
Dissertation
Rights
Villanueva Chigne, Eduardo
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
Tags
assertion
compound nominals
constraint
context sensitivity
indexicals
meaning
philosophy of language
pragmatics
prenominal possessives
semantic minimalism
semantics