Jump to ContentJump to Main Navigation
Semantics versus Pragmatics$

Zoltan Gendler Szabo

Print publication date: 2005

Print ISBN-13: 9780199251520

Published to Oxford Scholarship Online: January 2011

DOI: 10.1093/acprof:oso/9780199251520.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 19 February 2018

Naming and Asserting

Naming and Asserting

Chapter:
(p.356) 9 Naming and Asserting
Source:
Semantics versus Pragmatics
Author(s):

Scott Soames (Contributor Webpage)

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199251520.003.0010

Abstract and Keywords

This chapter draws a distinction between the semantic content of a sentence, and the assertive content of a normal, literal utterance of it. Semantic content is a set of constraints that determines a class of structural enrichments of the minimal proposition, or proposition-matrix, satisfying the constraints. Pragmatic aspects of the context determine which propositions in this class are then asserted. Even when the constraints determine a complete minimal proposition (traditionally identified as the semantic content), it is not always asserted, and, whether or not it is, other propositions typically are. Examples involving linguistically simple, as well as partially descriptive, names are used to illustrate the view — which has the effect of reconciling certain Fregean intuitions about assertions and beliefs with a Millian semantics of names.

Keywords:   semantic content, assertive content, minimal proposition, partially descriptive name, Millian semantics, pragmatics

Many essays in semantics and the philosophy of language seem to proceed on the assumption that—special circumstances involving ironic, metaphorical, or other non-literal uses of language aside—the proposition asserted by an utterance of a sentence in a context is the proposition semantically expressed by the sentence in that context. At some level, of course, we all know that this is a fiction, since sometimes a single utterance may involve the assertion of more than one proposition. For example, the assertion of a conjunction involves the assertion of the conjuncts, too. Nevertheless, the fiction is often thought to be harmless, since it is standardly taken for granted that even in those cases that falsify it, the proposition p semantically expressed by a sentence that is assertively uttered is the primary proposition asserted, with other propositions counted as asserted only when they are particularly obvious, and relevant, consequences of p. In my book Beyond Rigidity I argued that this comfortable, but complacent, view is false.1 In many cases in which one speaks literally (i) the primary proposition q asserted by an utterance of S in a context C is not the proposition semantically expressed by S in C, and (ii) the proposition that is semantically expressed by S in C counts as asserted only because it is an obvious and relevant consequence of q. In the book I argued that these points are important, in part, because of the role they play in explaining how substitution of coreferential names in simple sentences may change the propositions asserted by utterances of those sentences, even if they don't change the propositions semantically expressed. Once this (p.357) lesson is learned, it can be applied to sentences that ascribe beliefs and assertions to others, as well. Sometimes substitution of coreferential names in these sentences changes the truth values of the propositions the sentences are used to assert, even if the propositions they semantically express remain the same. In short, I tried to show how a more sophisticated conception of the relationship between the semantic contents of sentences and the propositions they are used to assert can make an important contribution to solving Frege's puzzle.

But Beyond Rigidity was only the first step. The real relationship between the semantic contents of sentences and the propositions they are used to assert is even more indirect than indicated there. For one thing, semantic contents of grammatically complete sentences (relative to contexts) are not always complete propositions; sometimes they are incomplete propositional matrices, together with partial constraints on how contextual information may be used to complete them. Moreover, even when the semantic content of S (relative to a context) is—or, at any rate, fully determines—a complete proposition, this proposition is not always among those asserted by competent speakers who use S with its normal, literal meaning. For these reasons, the semantic content of a sentence in a context is often not something asserted by an utterance of the sentence in that context. Instead, its function is to constrain the candidates for assertion in certain ways, while allowing speakers and hearers a degree of freedom to operate within these constraints. In what follows, I will fill out this conception of the relationship between semantic content and assertion in more detail. First, I will review, and then pose certain problems for, the version of the conception presented in Beyond Rigidity. Next, I will indicate how those problems can be solved by making the conception more sophisticated in certain ways. Finally, I will illustrate the generality of the new conception of the relationship between semantic content and assertion by citing a few of the vast range of further examples that support it.

The View from Beyond Rigidity

We utter sentences to convey information. Some of this information is asserted, some is conversationally implicated, and some is merely suggested by virtue of idiosyncratic features of speakers and hearers in particular contexts. Because of this, the information carried by an assertive utterance of one and the same (unambiguous) indexical-free sentence S varies greatly from one context to another. Not all of this information is part of the meaning of S in the language in which it is being used. Suppose, as is usual, that the meaning of S is the proposition it (p.358) semantically expresses. It is natural to think that this proposition should consist of information that a competent speaker who assertively utters S asserts and intends to convey in any normal context in which S is used with its literal meaning (without irony, sarcasm, defeating conversational implicatures, and the like).2

With this in mind, consider the example:

  1. (1) Carl Hempel lived on Lake Lane in Princeton.

Suppose I were to assertively utter this sentence to a graduate student in the philosophy department at Princeton University. In this situation I would expect the student to have heard the name Carl Hempel before, to know that it refers to a well-known philosopher, to know that I know this, and to know that I expect all this of him. Because of this presumed common ground between us, I would intend my utterance to be understood as committing me to the claim that the well-known philosopher Carl Hempel lived on Lake Lane in Princeton. I would intend my utterance to convey this information, and, depending on the situation, I might even intend to assert it.3 Of course, it is clear that an utterance of this sentence wouldn't assert or convey the same information to every competent speaker of English. One doesn't have to know that Carl Hempel was a philosopher in order to understand and be a competent user of his name. So if I uttered the sentence to a boyhood friend of his in a conversation in which it was understood that the friend had known the young Carl Hempel by name, but lost touch with him at an early age before he went into philosophy, then I would convey, and perhaps even assert, different information.

What information would be asserted and conveyed to competent speakers by nearly all assertive utterances of (1) (interpreted as containing a name of the man to whom I here use it to refer)? Surely it would include little or no substantive descriptive information about Mr Hempel. Many speakers would know something about his philosophy, but others would not. Some would know what he looked like at a certain age, but not everyone would. Some would have knowledge of his family background, but many would not. Although most speakers who had enough familiarity with the name to be able to use it could be expected to possess some descriptive information about Mr Hempel, virtually none of it would be common to all such users.4 What would be common is that (p.359) the property of having lived on Lake Lane in Princeton would be ascribed to one and the same individual by utterances of the sentence by competent speakers in normal contexts. In light of this it is natural to take the singular proposition assertion of which predicates this property of Mr Hempel as the proposition semantically expressed by (1).

In Beyond Rigidity I used considerations like these to develop a model of the semantics and pragmatics of proper names, and sentences containing them. According to the view developed there, proper names come in two main types—linguistically simple names like Saul Kripke and Carl Hempel, and linguistically complex, partially descriptive names like Princeton University and Mount Washington. The meaning, or semantic content, of a linguistically simple name is its referent; the meaning, or content, of a linguistically complex name is its referent together with a partial description of it.5 For example, the content of Mount Washington is a complex consisting of the property of being a mountain, together with the mountain itself. Though for present purposes the details don't matter much, we may think of its content as being given by the description the x: [x is a mountain and x = y], relative to an assignment of the referent of the name to the variable y—where the range of the variables includes all individuals, past, present, and future, plus individuals that exist in other possible worlds, as well as those that exist in the actual world.6

In order to be a competent user of a linguistically simple name n of an object o two requirements must be met: (i) one must have acquired a referential intention that determines o as the referent of n. This may be done by picking up n and intending to use it with the standard meaning-reference it has already acquired in the language due to the baptisms, authoritative stipulations, and referential uses of others. (ii) One must realize that to assertively utter a simple sentence n is F (in a normal context) is to say of the referent of n that it “is F”.7 Analogous (p.360) conditions hold for linguistically complex, partially descriptive names. Here, I assume that to say of an object o that it “is F” is to assert the singular proposition that predicates F-hood of o. Hence, these competence conditions ensure that this singular proposition will be among the propositions asserted whenever a competent speaker assertively utters n is F in a normal context.

Putting all this together, we get a unified picture of the meanings of sentences containing names, plus the propositions those sentences are used to assert and convey in different contexts of utterance. The idea, roughly put, is (i) that assertive utterances of sentences containing names often result in the assertion of several propositions, (ii) that which propositions are asserted by utterances of such a sentence varies significantly from context to context, and (iii) that the meaning of an indexical-free sentence, the proposition it semantically expresses, is something which remains invariant from one context to another—i.e. it is a proposition that is asserted in all normal contexts in which the sentence is used by competent speakers with its literal meaning (without irony, sarcasm, defeating conversational implicatures, and the like).

One interesting feature of this model is the explanation it provides of how the sentences

  1. (2a) Peter Hempel is Carl Hempel

  2. (2b) Carl Hempel is Carl Hempel

can mean the same thing, even though assertive utterances of the former virtually always involve the assertion of propositions different from those asserted by utterances of the latter.8 This result is extended to cases in which the two sentences are embedded under attitudes verbs like believe and assert. In these cases, the attitude ascriptions have the same semantic content, even though assertive utterances of them may well involve the assertion of propositions with different truth values.

Problems of Extending the Beyond Rigidity Model

My aim here is to build on this semantic-pragmatic model by extending and revising it to handle problematic cases in the hope of producing a more widely applicable and acceptable framework. Two principles of the model that require (p.361) scrutiny are:

  1. (P1) In order to be a competent user of a name n for an object o, one must realize that to assertively utter n is F in a normal context (without irony, sarcasm, defeating conversational implicatures, and the like) is to say of the referent of n that it “is F”.

  2. (P2) The semantic content of an indexical-free sentence S is a proposition p which is asserted and conveyed by utterances of S in any normal context involving competent speakers in which S is used with its literal meaning (without irony, sarcasm, defeating conversational implicatures, and the like).

P1 is true as stated, where n is F is a simple, operator-free sentence. Although it remains true if it is generalized in limited ways, it appears to fail in some cases in which n is partially descriptive and n is F is replaced by a complex sentence … n … containing the name. P1 also seems to fail (though perhaps not as obviously) in these environments when n is linguistically simple. Because of this, it appears that P1 cannot be fully generalized. If this is so even when n is linguistically simple, we face an immediate problem with P2. If competent speakers may assertively utter … n … (in a normal context) without saying of its referent o that it “is an x such that . . . x …”, then it may very well be that they may do so without asserting the singular proposition expressed by … x … with respect to an assignment of o to ‘x’. Since this proposition is the proposition semantically expressed by … n … , we have a violation of P2. Something has to give.

Proposal for Revising P2

P2 is a prime suspect. Though the principle approximates the truth, we may be able to improve it and make it more plausible by revising it so that it no longer presupposes that the semantic content of a sentence is always a complete proposition, or at any rate one that is asserted and conveyed by normal utterances in every normal context. Instead, the semantic content of a sentence S may be viewed as something that constrains—but does not always completely determine—the propositions asserted and conveyed by utterances of S in normal contexts involving competent speakers.

For example, the semantic content of

  1. (3a) Peter Hempel isn't Carl Hempel

(p.362) may be seen as constraining normal assertive utterances of it in such a way that they are counted as assertions of propositions built around that content, without always counting as assertions of the content itself.9 One way of thinking of this is to view the semantic content of (3a) as the potentially gappy propositional matrix (4).10
  1. (4) 〈Neg 〈Identity 〈〈__, Mr Hempel〉, 〈 __, Mr Hempel〉〉〉〉

When (3a) is assertively uttered in a particular context, the gaps in the matrix may be filled with salient descriptive information associated with the corresponding names by speaker-hearers in the context. (If there is no such contextually salient information, then the proposition asserted is one that simply denies that Mr Hempel is Mr Hempel.) In a context in which (3a) is used to assert that the man, Peter Hempel, standing over there isn't the famous philosopher Carl Hempel, the gaps in (4) are filled in so as to produce (5a)—which may also be represented as the proposition expressed by (5b), relative to an assignment of Mr Hempel to ‘z’ and ‘w’.
  1. (5a) 〈Neg 〈Identity 〈〈Man standing over there, Mr Hempel〉, 〈Famous philosopher, Mr Hempel〉〉〉〉

  2. (5b) ∼([the x: (x is a man standing over there & x = z)] = [the y: y is a famous philosopher & y = w]).

Whereas someone who asserts (5a) asserts the proposition expressed by (5b) relative to an assignment of Mr Hempel to ‘z’ and ‘w’, it may not be a foregone conclusion that he thereby asserts the absurd proposition semantically expressed by
  1. (3b) Carl Hempel isn't Carl Hempel.

If it turns out that this proposition is not asserted, then we have a potential violation of P2; more precisely, we have a violation of the conjunction of P2 with a principle, P3, identifying the proposition semantically expressed by (3a) and (3b) with one that arises from propositional-matrix (4) by eliminating the gaps.
  1. (p.363) (P3) If the semantic content of S (relative to a context C and assignment f of values to variables) is a prepositional matrix like (4) (in which the constituents corresponding to names in S are pairs consisting of the referent of a name and a gap), then the proposition semantically expressed by S (relative to C and f) is the proposition that arises from the matrix by eliminating the gaps.

The point of the principle is to abstract away from the effects of contextual supplementation, thereby identifying a proposition entirely determined by the meaning of the sentence.

In what follows, I will argue that more complex examples in which sentences containing names are embedded under certain propositional attitude verbs provide serious challenges both to P2 (when conjoined with P3) and to unrestricted versions of P1. Particularly troubling are assertive utterances of examples like

  1. (6a) Mary doesn't know that Peter Hempel is Carl Hempel.

in contexts in which the speaker's primary intention is to assert and convey the information that Mary doesn't know that the man, Peter Hempel, standing over there is the famous philosopher Carl Hempel. In such a case, it would seem that the speaker may truly assert that Mary doesn't stand in the knowledge relation to the proposition
  1. (6b) 〈Identity 〈〈Man standing over there, Mr Hempel〉, 〈 Famous philosopher, Mr Hempel〉〉〉

without falsely asserting that Mary doesn't stand in the knowledge relation to the trivial proposition
  1. (6c) 〈Identity 〈 Mr Hempel, Mr Hempel〉〉

which is the proposition semantically expressed by the complement clause of (6a) (according both to Beyond Rigidity and to P3). Suppose, as I now believe, that this is right—i.e. (i) that the speaker does assert that Mary doesn't stand in the knowledge relation to proposition (6b), without asserting that she doesn't stand in the knowledge relation to proposition (6c), (ii) that proposition (6c) is the proposition semantically expressed by the complement clause of (6a), and (iii) that the proposition semantically expressed by (6a) characterizes Mary as not standing in the knowledge relation to the proposition expressed by its complement clause. It will then follow that the speaker does not assert the proposition (p.364) semantically expressed by (6a) in the context in which that sentence is assertively uttered. This is a challenge to P2. Unless the context can be shown to be one in which there is a conversational implicature canceling what would otherwise be the normal presumption that the proposition semantically expressed is asserted, we will have to reject P2 and find a replacement.

This is the path I will explore. In so doing, I will put aside the possibility of resolving the problem by finding a canceling implicature. I do this despite the fact that this alternative is not entirely without plausibility.11 After all, the claim that Mary does not know, of Mr Hempel, that he is who he is is an obvious falsehood that everyone in the conversation could be expected to recognize, if brought to their attention in the proper way. Moreover, in cases in which it appears to conversational participants that a speaker may have asserted something the falsity of which is utterly obvious, familiar Gricean principles motivate the search for an alternative interpretation of the speaker's remark. In such cases, the normal presumption that the speaker has asserted p (where p is the proposition semantically expressed by the sentence assertively uttered) is often conversationally canceled, and the speaker is taken to have asserted something else.12

However, there is a serious problem in applying this model to the present case: namely, that it would not appear to conversational participants in our imagined scenario that the speaker has said anything obviously false in assertively uttering (6a). Let us suppose that everyone in the conversation already knows, and is taken to know, that Carl Hempel and Peter Hempel are different names for the same man. Even then, they cannot be expected to recognize that the proposition semantically expressed by (6a) is an obvious falsehood. Since it would never occur to them that there is any presumption that the speaker should be taken to have expressed anything false, Gricean principles would not motivate them to search for an alternative interpretation. Thus, if there is a canceling conversational implicature, it can only be an implicature of some other sort—one that is quite distant from the (conscious or unconscious) thought processes of conversational participants. Since I am not sure precisely how genuine implicatures relate to the actual thought processes of agents, I do not wish to rule out the possibility that some basis for such a canceling implicature (p.365) might, ultimately, be found.13 However, neither do I wish to rely on this abstract possibility. Instead, I will sketch a different strategy for solving the problem.

In what follows I will propose a modification P2* of P2, based on the idea that the semantic content of a sentence (its meaning if it is indexical-free) is not always what is said (i.e. asserted) by normal utterances of it; rather, it is the skeleton of what is said—the inner structure around which speakers construct their assertions.

  1. (P2*) The meaning, or semantic content, of an indexical-free sentence S is a prepositional matrix pm, which is such that for any normal context involving competent speaker-hearers in which S is assertively uttered with its literal meaning (without irony, sarcasm, defeating conversational implicatures, and the like), the assertive utterance of S in the context is to be counted as an assertion of a proposition p which is an acceptable completion of pm. (When S contains linguistically simple names the bare singular proposition arising from pm by eliminating gaps corresponding to the names counts as an acceptable completion in contexts in which no descriptive enrichment takes place.)

Having argued for this revision of P2, I will indicate how the model might be extended to indexicals—which often can be regarded as constraining, rather than determining, contents in given contexts of utterance—and to sentences containing the possessive construction, the semantic contents of which are naturally seen not as complete propositions, but as propositional matrices including partially specified relations between possessor and possessed that require contextual supplementation.14

The end result is a conception of semantics and pragmatics in which the relationship between the semantic content of a sentence and what the sentence is (p.366) used to say, or assert, is looser than commonly thought. The former constrains and influences the latter, without always determining it, even when the sentence is used with its normal literal meaning, without canceling implicatures. The semantic contents of expressions provide the building blocks for assertions, and constrain how these blocks are assembled in normal contexts of use. But the rules of the language provide one with only a minimum common denominator. They facilitate communication and coordinate our linguistic activities, while allowing speakers considerable freedom to exploit the features of particular contexts to shape the information asserted and conveyed by their utterances.15

Apparent Failures of P1 to Fully Generalize: Partially Descriptive Names

In this section I will show that in a certain range of cases in which n is a partially descriptive name and S(n) is a sentence containing n, a speaker who assertively utters S(n), and thereby asserts the proposition semantically expressed by S(n), may nevertheless not assert the singular proposition expressed by S(v) relative to an assignment of the referent of n to the variable v (where S(v) arises from S(n) by replacing one or more occurrences of n by free occurrences of v). This result is based on an elementary observation: standardly, the fact that a particular utterance counts as the assertion of a certain proposition p in a conversational context C will, in and of itself, guarantee that the utterance also counts as the assertion of another proposition q only if q is both a necessary and a priori consequence of p (together with other obvious and salient shared background assumptions in C).16 For example, the fact that an utterance of a conjunctive (p.367) sentence counts as an assertion of the proposition that A and B guarantees that it also counts as an assertion of both the proposition that A and the proposition that B, each of which is a necessary and a priori consequence of the conjunctive proposition. Similarly, the fact that assertive utterances of (7a) and (8a) count as assertions of the propositions semantically expressed by (7b) and (8b) (relative to an assignment of the referent of the name to ‘y’), guarantees that these utterances also count as assertions of the associated propositions expressed by (7c) and (8c) (relative to the same assignment).

  1. (7a) Senator Clinton represents New York.

  2. (7b) [the x: x is a senator & x = y] represents New York.

  3. (7c) y represents New York.

  4. (8a) Ralph knows that Senator Clinton represents New York.

  5. (8b) Ralph knows that [the x: x is a senator & x = y] represents New York.

  6. (8c) Ralph knows that y represents New York.

To assert the proposition expressed by (7b) is, in effect, to assert of Hillary Clinton that she both is a senator and represents New York, which includes asserting that she represents New York. To assert the proposition expressed by (8b) is, in effect, to assert of Hillary that Ralph knows that she both is a senator and represents New York, which includes asserting that Ralph knows that she represents New York. In each of these cases, the proposition q (expressed by (7c) or (8c)) asserted as a result of asserting the proposition p (expressed by (7b) or (8b)) is a trivial, necessary, and a priori consequence of p.

In order for this pattern of reasoning to be universally applicable to assertive utterances of sentences containing partially descriptive names, it would have to be the case that for any sentence S(n) containing such a name n of an object o, if S(v) resulted from S(n) by replacing one or more occurrences of n with the variable v, then the proposition expressed by S(v) relative to an assignment of o to v would be a necessary and a priori consequence of the proposition semantically expressed by S(n). Of course, this is not so. Three illustrative classes for which this generalization fails are those in (9):

  1. (9a) negations of simple sentences containing n (within the scope of negation)

  2. (9b) negations of attitude ascriptions α says/believes/knows that … n …

  3. (9c) attitude ascriptions αjust learned/realized that … n …

Sentences of these types raise the possibility of exportation failure—i.e. of there being sentences S(n) and contexts involving competent speaker-hearers (p.368) in which one who assertively utters S(n) asserts the proposition it semantically expresses without asserting the proposition expressed by S(v), relative to an assignment of the referent o of n to v. Any sentences that do allow this constitute instances in which P1 fails to generalize. Our next task will be to examine sentences of the types (9a), (9b), and (9c) to determine whether assertive utterances of them do give rise to exportation failure, and hence whether P1 really does fail to generalize in these cases.

Negations of Simple Sentences: Examples of Type (9a)

We consider the case of Ralph, who has heard and seen pictures of Hillary Rodham Clinton, when she was First Lady, and who has read about Senator Clinton from New York, but who does not know that Senator Clinton is the former First Lady, or that Senator Clinton's name is Hillary Rodham Clinton. Ralph understands, accepts, and assertively utters sentence (10a), which semantically expresses the proposition indicated by (10b), and may roughly be paraphrased as (10c).17

  1. (10a) Hillary Rodham Clinton is not Senator Clinton.

  2. (10b) Hillary Rodham Clinton ≠ [the z: z is a senator & z = y] (relative to an assignment of H.R.C. to ‘y’).

  3. (10c) It is not the case that there is a unique person who is both a senator and identical with H.R.C., and who is identical with H.R.C.

The corresponding proposition (11b), semantically expressed by (11a),
  1. (11a) Hillary Rodham Clinton != Hillary Rodham Clinton

  2. (11b) 〈 Neg 〈 Identity, 〈HRC, HRC〉〉〉

is neither a necessary nor an a priori consequence of the proposition semantically expressed by (10a) (since the former is true in any circumstance in which Hillary Clinton is not a senator, whereas the latter is not). The question to be answered is whether in asserting, and expressing his belief in, the former (p.369) (contingently false) proposition, Ralph asserts, and expresses a belief in, the latter (necessarily false) proposition.

The details of the case are roughly as follows: When Ralph assertively utters (10a), he asserts the proposition it semantically expresses, which is necessarily equivalent to the singular proposition that Hillary is not a senator. It is because Ralph accepts the sentence Hillary Rodham Clinton is not a senator, and thereby believes this singular proposition, that he assertively utters (10a).18 However, he also believes of Hillary that she is a senator, since this is part of what is involved in his accepting the sentence Senator Clinton exists and believing the proposition it expresses. These facts, together with Ralph's assertive utterance of (10a), may make it clear to everyone in the conversation that he would accept (12) (when the first occurrence of Clinton is taken by him to be a linguistically simple name for the senator).

  1. (12) Clinton is a senator, but she is not Hillary Rodham Clinton.

If it is clear that Ralph would accept (12) while taking its second conjunct to express the singular proposition (11b), then he may very well believe, and be taken by conversation participants to believe, that proposition. Since all of this is pretty obvious, that proposition may even count as having been asserted by Ralph, in which case P1 can be generalized to include this example. On this account, the reason the singular proposition (11b) is asserted is not that it is an obvious consequence of the proposition the speaker was primarily interested in asserting (which in this case is also the proposition semantically expressed by the sentence uttered); rather, it counts as asserted because it is an obvious consequence of this together with everything else that is both relevant and obvious in the context of utterance. As is so often the case, there may be quibbles about what is both relevant and obvious enough to be included in what has been asserted by the speaker. For this reason, some might resist the conclusion that proposition (11b) is asserted in this case. Still, since neither view is entirely obvious, the most we can say is that this example does not provide a clear, uncontroversial instance of the failure of P1 to generalize. The same may be said of other sentences of type (9a).

(p.370) Negations of Certain Attitude Ascriptions: Examples of Type (9b)

Next, we continue the story of the previous section by considering the examples (13) and (14).

  1. (13a) Ralph doesn't know/believe that Senator Clinton is a former First Lady.

  2. (13b) Ralph doesn't know/believe that [the z: z is a senator & z = y] is a former First Lady (relative to an assignment of H.R.C. to the variable ‘y’).

  3. (14a) Ralph doesn't know/believe that Hillary Rodham Clinton is a former First Lady.

  4. (14b) 〈Neg 〈 knowledge/belief, 〈 Ralph, 〈Being a former First Lady, HRC〉〉〉〉.

The proposition (14b), semantically expressed by (14a), is neither a necessary nor an a priori consequence of the proposition indicated by (13b), which is semantically expressed by (13a). Since Ralph doesn't accept any sentence that characterizes someone as being both a senator and a former First Lady, he does not believe, and so does not know, the proposition semantically expressed by (15a).
  1. (15a) Senator Clinton is a former First Lady.

  2. (15b) Hillary Rodham Clinton is a former First Lady.

However, since he does understand and accept (15b), he believes—and may even know—the proposition semantically expressed by (15b). Hence (13a, b) are true even though (14a, b) are false. Ralph does, of course, believe, of Hillary (i) that she is a former First Lady (in virtue of understanding and accepting (15b)), and (ii) that she is a senator (in virtue of understanding and accepting Senator Clinton exists). However, he does not believe that she has the property P expressed by (16), or even that anyone does, which is, in effect, something he would have to believe if he were to believe the proposition semantically expressed by (15a).19
  1. (16) λx [x is a senator and x is a former First Lady].

Since (13a, b) are true, while (14a, b) are false, it would be quite surprising if asserting the true proposition semantically expressed by (13a, b) invariably involved (p.371) also asserting the false proposition (14b). However, I see no reason why it should. On the contrary, if Mary were to assertively utter (13a) she might say something true without thereby saying anything false, and in particular, without asserting the false (14b).

This result constitutes a failure of P1 to fully generalize, since Mary, who assertively utters (13a), does not assert the singular proposition expressed by

  1. (17) Ralph doesn't know/believe that y is a former First Lady

relative to an assignment of the referent, H.R.C., of the partially descriptive name Senator Clinton to ‘y’, and so does not say of H.R.C. that Ralph doesn't know or believe that she is a former First Lady.20 Of course this “failure” is not a bad thing; it simply means that we must be careful in formulating our theory to restrict P1 to the cases in which it is correct.

Related Attitude Ascriptions: Examples of Type (9c)

Our final example involving partially descriptive names is a simple extension of the example in the previous section. (It is included to show that the sentences we are interested in do not all contain the negation operator, or other overtly negative expressions.) Imagine that after long being ignorant about the matter, Ralph finally discovers that Senator Clinton is the former First Lady, Hillary Rodham Clinton. Learning of this, Mary reports Ralph's discovery by assertively uttering (18a), thereby asserting the true proposition semantically expressed by (18b), relative to an assignment of H.R.C. to ‘y’, without asserting the false proposition expressed by (18c), relative to an assignment of H.R.C. to ‘y’.

  1. (18a) Ralph just learned that Senator Clinton is the former First Lady.

  2. (18b) Ralph just learned that [the z: z is a senator & z = y] is a former First Lady.

  3. (18c) Ralph just learned that y is a former First Lady.

Here, I assume that the proposition semantically expressed by Ralph just learned that S is true with respect to a context of utterance C (and assignment A of values to variables free in S) if and only if (i) until just prior to the time of C, Ralph did not believe the proposition semantically expressed by S (with respect (p.372) to C and A), but (ii) at t Ralph has come to believe that proposition. In the scenario imagined, (18a) satisfies this condition, and so is true. By contrast, (18c) is false (relative to an assignment of H.R.C. to ‘y’), since Ralph has long believed of H.R.C. that she is a former First Lady. Again, this constitutes a failure of P1 to fully generalize when partially descriptive names are involved. As before, however, there is nothing wrong with this result. Surely, someone who assertively utters (18a) in the sort of case we have imagined may say something true without thereby saying anything false. Since we want a theory to be consistent with this fact, we want our theory to limit the range of P1.

Lessons for Linguistically Simple Names: Apparent Failures of P1 to Fully Generalize, and Counterexamples to P2

In this section I will explain how the results of the previous section may be extended to cases involving linguistically simple names. In particular, I will argue that in a certain range of examples in which a sentence S(n) contains a linguistically simple name n, a speaker who assertively utters S(n) with the primary intention of asserting a descriptively enriched proposition p—which results from filling in the gaps accompanying occurrences of the referent of n in the propositional matrix semantically associated with S(n)—may succeed in asserting p without asserting the singular proposition q expressed by S(v) relative to an assignment of the referent of n to the variable v. Since q is also the (unenriched) proposition semantically expressed by S(n) (according to P3), these cases constitute not only failures of P1 to fully generalize, and but also counterexamples to P2.

These results are obtained in cases in which the unenriched proposition q semantically expressed by S(n) is not a necessary and a priori consequence of either (i) the descriptively enriched proposition p which it is the speaker's primary intention to assert, or (ii) p together with other obvious and relevant shared background assumptions at the time of utterance. In these cases, the successful assertion of p does not guarantee that q has also been asserted. Absent an antecedent commitment to the idea that in normal contexts (without defeating conversational implicatures, and the like) the proposition semantically expressed by the sentence uttered is always asserted, we have no reason to think that anything else guarantees that q has been asserted. Since, as we shall see, it would (p.373) be counterintuitive to suppose that q has been asserted in the cases in question, we have positive reason to reject P2 in favor of P2*.

As with partially descriptive names, I will discuss examples of assertive utterances of sentences of the linguistic types indicated in (9b) and (9c), only this time the names will be linguistically simple. However, I will not discuss examples involving assertive utterances of sentences of type (9a)—negations of simple sentences containing an occurrence of a name n within the scope of negation. As we have already seen, when n is partially descriptive, sentences of this type do not provide clear instances of the failure of P1 to generalize. The same is true when n is linguistically simple; hence, these sentences don't provide clear counterexamples to P2, either. For this reason, we will concentrate on the interaction of propositional attitude constructions and linguistically simple names. In each case, it will be assumed that the speaker's primary intention in assertively uttering the attitude ascription (or its negation) is to assert a descriptively enriched proposition obtained by adding contextually salient descriptive information associated with the names to the propositional matrix semantically expressed by the sentence uttered.

Counterexamples to P2 Involving Linguistically Simple Names

First, an example of type (9b). The scene is the annual newcomers’ party at the Princeton philosophy department, where Mary has been introduced to a distinguished white-haired gentleman called ‘Peter Hempel’. Not knowing who he is, she responds a few minutes later to another newcomer's question “Is he Carl Hempel?”, by assertively uttering, No, Peter Hempel isn't Carl Hempel. The context of utterance is such that in assertively uttering

  1. (3a) Peter Hempel isn't Carl Hempel

Mary asserts that the man, Peter Hempel, standing over there, isn't the famous philosopher Carl Hempel. This proposition, which we may take to be represented by (5a), is semantically expressed by (5b), relative to an assignment of Mr Hempel to ‘z’ and ‘w’.
  1. (5a) 〈Neg 〈Identity, 〈〈Man standing over there, Mr Hempel〉, 〈Famous philosopher, Mr Hempel〉〉〉〉.

  2. (5b) ∼([the x: (x is a man standing over there & x = z)] = [the y: y is a famous philosopher & y = w]).

(p.374) This is the information that Mary primarily intends to get across; however, it is not the proposition (3c) semantically expressed by the sentence she uttered.
  1. (3c) 〈Neg 〈Identity, 〈 Mr Hempel, Mr Hempel 〉〉〉.

Although Mary believes the trivially true proposition
  1. (6c) 〈Identity 〈 Mr Hempel, Mr Hempel 〉〉

of which (3c) is the negation, she does not believe the true, but non-trivial proposition
  1. (6b) 〈Identity, 〈〈Man standing over there, Mr Hempel〉, 〈 Famous philosopher, Mr Hempel〉〉〉

of which (5a) is the negation.

Suppose, next, that someone else at the party who is in the know about Mr Hempel reports Mary's ignorance by assertively uttering

  1. (6a) Mary doesn't know that Peter Hempel is Carl Hempel

with the primary intention of asserting and conveying the information that Mary doesn't know that the man, Peter Hempel, standing over there, is the famous philosopher Carl Hempel. This speaker truly asserts that Mary doesn't stand in the knowledge relation to (6b). He does not also falsely assert that Mary doesn't stand in the knowledge relation to (6c); no ordinary conversational participant—not even those fully apprised of Peter Hempel's identity—would dream of accusing the speaker of falsely asserting that Mary doesn't know of the pair consisting of Mr Hempel and Mr Hempel that the former is the latter. But if the speaker doesn't assert that Mary doesn't know (6c), then the speaker doesn't assert the proposition semantically expressed by the sentence (6a) that he assertively utters.

Here, the proposition (6e), which is semantically expressed by (6a), is neither a necessary nor an a priori consequence of the descriptively enriched proposition (6d), which the speaker intended, successfully, to assert and convey.

  1. (6d) 〈 Neg 〈 Knowledge, 〈 Mary, 6b 〉〉〉.

  2. (6e) 〈 Neg 〈 Knowledge, 〈 Mary, 6c 〉〉〉.

As a result, the speaker's assertion of (6d) provides no reason, in and of itself, to suppose that (6e) has also been asserted. This means that one can retain principle P2 only if one can establish that there is a conversational implicature which defeats what would otherwise be a presumption that (6e) is asserted. Since I have (p.375) already indicated the difficulty inherent in such a strategy, I take this example to provide evidence of the failure of P1 to fully generalize, the falsity of P2, and the truth of P2*.

Our final example, of type (9c), is a simple extension of this case. Upon learning of Mary's ignorance, Paul sets her straight by assertively uttering (2a), Peter Hempel is Carl Hempel, thereby asserting and conveying the information that the man, Peter Hempel, standing over there is the famous philosopher Carl Hempel. Gil, overhearing the conversation, reports Mary's new-found knowledge by assertively uttering (19a) with the intention of asserting and conveying the descriptively enriched, true proposition (19c) expressed by (19b), relative to an assignment of Mr Hempel to the variables ‘z’ and ‘w’.

  1. (19a) Mary just learned that Peter Hempel is Carl Hempel.

  2. (19b) Mary just learned that [the x: (x is a man standing over there & x = z)] = [the y: y is a famous philosopher & y = w].

  3. (19c) 〈 Just learned, 〈 Mary, 6b 〉〉.

It was no part of Gil's intention to assert or convey the false proposition (19d), which characterizes Mary as having just learned, as opposed to having long known, that Mr Hempel is Mr Hempel.
  1. (19d) 〈 Just learned, 〈 Mary, 6c 〉〉.

Nor do his hearers, who know that (19d) is false, take him to have asserted it. Since this proposition is the semantic content of the sentence (19a), which Gil assertively uttered, we have another strongly intuitive counterexample to P2 that is fully consistent with the revised principle P2*. More precisely, we have this, absent the ability to establish what seems to me to be the questionable claim that there is a conversational implicature that cancels what would otherwise be the presumption that one who assertively utters (19a) should be taken as asserting (19d). In light of this, shouldn't we reject P2 in favor of P2*?

Well, yes, but I didn't always think so. Although I was aware of these problems with P2 when writing Beyond Rigidity, I was reluctant to abandon it. I was convinced that there must be some reasonably close relationship between the semantic content of a sentence and that which is asserted and conveyed by utterances of it in normal contexts. Since I hadn't yet formulated P2*, I couldn't see what that relationship might be, if it weren't the one expressed by P2. Now that a reasonable alternative has been put on the table, this consideration no longer has force, and our pre-theoretic intuitions about what is asserted and conveyed (p.376) by utterances of sentences like (6a) and (19a) can be given their proper weight. Hence, I now believe that P2 should be rejected in favor of P2*.

However, before resting with this conclusion, it is necessary to address one further argument. In Beyond Rigidity I was not content merely to insist on P2 despite apparent counterexamples like (19a). In addition, I offered an independent argument that in assertively uttering (19a) the speaker must assert the proposition (19d) that it semantically expresses, in addition to the descriptively enriched proposition that is the primary assertion made by the utterance. This argument must be disarmed before we can replace P2 with P2*.

The Argument from Beyond Rigidity

The argument from Beyond Rigidity is based on an example almost exactly like the one built around (19). In giving the argument I will use the formulation and the numbering from the book. In the example, Gil assertively utters (38) with the intention of asserting and conveying the descriptively enriched proposition expressed by (40).

  1. (38) Mary just learned that Peter Hempel is Carl Hempel.

  2. (40) Mary just learned that our colleague Peter Hempel, standing over there, is the famous philosopher Carl Hempel.

In Beyond Rigidity I argued that although Gil succeeds in asserting the true proposition (40), his utterance also counts as the assertion of a false proposition. Here is the argument for that point:

For example, suppose that shortly after his assertive utterance of (38) we were to ask “Is there some man such that Gil asserted that Mary has just learned that he is Carl Hempel?” I think that if Gil and his audience were confronted with this question, they would be inclined to agree with the following answer: “Yes, there is a man, Peter Hempel (standing right over there), such that Gil asserted that Mary just learned that he is Carl Hempel.” In this way, conversational participants could be brought to recognize the truth of (49).

  1. (49) ∃x (x = Peter Hempel & Gil asserted that Mary just learned that x = Carl Hempel)

Next we might ask, “But isn't it true that Mary has known for a long time that Carl Hempel is Carl Hempel?” Surely, the answer to this would be “Yes” We might follow this up with: “Isn't it therefore also true that there is a certain man, Carl Hempel, such that Mary has long believed that he is Carl (p.377) Hempel?” Here again, I think that if ordinary speakers and hearers were to reflect on the matter, they would be inclined to agree that Carl Hempel is such that Mary has long believed that he is Carl Hempel. In this way they could be brought to recognize the truth of (50).

  1. (50) ∃x (x = Carl Hempel & Mary has long believed that x = Carl Hempel)

Finally, we remind them of the truth of (51).

  1. (51) Peter Hempel is Carl Hempel

But surely, we would point out, (52) is a logical consequence of (49)–(51); therefore since they are true, it is true as well.

  1. (52) ∃x [x = Peter Hempel & x = Carl Hempel & (Gil asserted that Mary just learned that x = Carl Hempel) & (Mary has long believed that x = Carl Hempel)]

This means that there is a certain man such that Gil asserted that Mary just learned that he is Carl Hempel even though, in fact, Mary has long believed that he is Carl Hempel. But then, since Gil asserted that Mary has just learned that so and so, when in fact Mary has long believed (and even known) that so and so, it follows that at least one thing that Gil asserted is false.21

This argument can easily be extended, since if it is correct, then surely (53a) and (53b) must also be correct. (I here continue the sequence of numbering of the Beyond Rigidity examples.)

  1. (53a) ∃x ∃y [x = Peter Hempel & y = Carl Hempel & (Gil asserted that Mary just learned that x = y)].

  2. (53b) ∃x [x = Peter Hempel & x = Carl Hempel & (Gil asserted that Mary just learned that x ∃ x)].

But then, if the propositions semantically expressed by these sentences are true, Gil must have asserted the false proposition that is semantically expressed by the sentence he uttered. That is how I previously sought to defuse the objection to P2.

Although the argument can appear compelling, I no longer believe that it is successful. The natural response to it is, I think, to grant the force of each step, while retaining one's conviction that the conclusion cannot be correct. Gil (p.378) simply didn't say anything false when he assertively uttered (38). If this is right, then we should look for subtle confusions along the way that may lend one or more of the steps an initial, but misleading, appearance of validity. I now think that we can find two such confusions.

The first significant problem involves the move to (49). In order to assess this move, we have to determine what (49) is to be inferred from. The argument presupposes that conversational participants will grant the truth of the proposition expressed by what I will here number as (38a).

  1. (38a) Gil asserted that Mary just learned that Peter Hempel is Carl Hempel.

But which proposition is that? Are the conversational participants implicitly asked to endorse the proposition semantically expressed by (38a), or are they asked to endorse the descriptively enriched proposition
  1. (38b) Gil asserted that Mary just learned that our colleague Peter Hempel, standing over there, is the famous philosopher Carl Hempel

that they would primarily intend to assert were they to use (38a) to report Gil's remark? Although the argument fails to address this question, the answer to it is crucial.

Even though the truth of (49) follows from the truth of the proposition semantically expressed by (38a), it does not follow from the truth of (38b). Because of this, one bad, but nevertheless initially persuasive, reason for acquiescing in the move to (49) may be a failure to distinguish the austere proposition that sentence (38a) semantically expresses from the descriptively enriched proposition (38b) it would naturally be used to assert. A person who fails to make this distinction may focus on the latter proposition when judging Gil to have asserted a truth, while focusing on the former proposition when making the inference to (49)—wrongly concluding, thereby, that (49) is validly inferred from a true premise.

A second source of potential confusion involves (49) itself, or its English equivalent, (49a).

  1. (49a) There is a man, Peter Hempel, such that Gil asserted that Mary just learned that he is Carl Hempel.

In the previous paragraph I assumed that the proposition corresponding to these sentences in the argument from Beyond Rigidity is the proposition they semantically express. However, it is conceivable that these sentences might themselves be used to assert the further, descriptively enhanced, proposition indicated by (49b).
  1. (p.379) (49b) There is a man, Peter Hempel, such that Gil asserted that he, our colleague standing over there, is the famous philosopher Carl Hempel.

This proposition is clearly true, while being an obvious consequence of the descriptively enriched proposition (38b) that conversational participants would assert if they used (38a) to report Gil's remark.22 Since the conversational context is one in which participants are invited to endorse (49a) on the basis of an implicit inference from their characterization of what Gil asserted, it is conceivable that they might use the sentence (49a) to assert the proposition (49b). On this understanding, the move from (38b) to (49) is fully acceptable, as is an analogous move to (50). However, on this way of understanding sentences (49) and (50)—namely as sentences used to assert one or more descriptively enriched propositions—the false proposition semantically expressed by (52) does not follow from (49–51).23 That being so, we have no argument that in assertively uttering (38) Gil asserted anything false, let alone the false proposition semantically expressed by (38).

This completes my response to the argument from Beyond Rigidity. Although the example is a difficult one, I have come to believe that the account just presented is the best that can be given. Its virtues are: (i) that it preserves our strong pre-theoretic intuition that in assertively uttering (38) Gil did not say anything false, (ii) that it explains the plausibility of the individual steps in the argument by indicating the senses in which each is valid, and (iii) that it vindicates our sense that there is nevertheless something wrong with the argument as a whole by indicating why there is no single interpretation on which its premises are true and each of its steps are truth-preserving. I therefore reject the argument quoted from Beyond Rigidity, and I accept the conclusion that this example, and the others discussed in the previous section, provide evidence that P1 fails to fully generalize, that P2 is false, and that P2* should be adopted as a replacement of P2. I will close with a brief discussion of ways in which the idea behind the new principle might profitably be extended and applied to a broader range of cases.

(p.380) Semantic Incompleteness and Pragmatic Enrichment

As stated, P2* applies only to indexical-free sentences. There was no principled reason for this restriction; it was adopted solely to simplify the discussion. The idea of contextually determined, descriptive enrichment of singular propositions—involving either the mechanism of gappy propositional matrices or its equivalent mentioned in footnotes 10 and 14—can easily be extended to familiar treatments of indexicals, such as that of David Kaplan.24 More important than any such mechanical extension of P2*, however, is the way in which the guiding idea behind it meshes with certain kinds of indexicality.

The guiding idea is that the meaning of an expression constrains its contributions to the assertions made by normal, literal utterances of sentences containing it, without always fully determining those contributions. This is just what we find when we consider demonstratives like he, she, that, then, there, we, and now. A natural way of understanding these expressions is to see their meanings as constraining, but not fully determining, their referents in different contexts. Oversimplifying, and idealizing a bit, the referent of he is constrained to be male, the referent of she is constrained to be female, the referent of that may be any salient non-animate thing, the referent of then must be a time, the referent of there must be a place, the referent of we is a group that contains the agent (the speaker in cases in which we is uttered), and the referent of now is some stretch of time including the present moment.25 To know the meanings of these terms is, roughly, to know these constraints, and to know that when one uses the terms in simple sentences one says of their referents that they are so and so. Everything else used to determine the referents of the expressions in different contexts, and to fill out the assertions made by utterances of sentences containing them, is non-semantic, extra-linguistic, pragmatic information.

These demonstratives differ from proper names in two significant ways. First, there is a significant contrast between proper names and all indexicals—including both pure indexicals, like the first-person singular pronoun and certain temporal (p.381) indexicals like today, and the kinds of demonstratives we have just been discussing. Whereas there is a natural sense in which the semantics of a (disambiguated) sentence n is F by itself determines a complete proposition, independent of any contextual contribution (in the manner of P3 above), the semantics of i is F is never sufficient to determine a proposition without some contextual contribution to the content of the indexical i. Second, there is a contrast between the demonstratives mentioned above, on the one hand, and both proper names and pure indexicals, on the other, involving the greater extent and variety of contextual supplementation with demonstratives. With names and pure indexicals, reference is secured without pragmatic supplementation. In the case of (linguistically simple) names, their referents are their meanings. In the case of pure indexicals, their meanings together with fixed parameters of the context in which they are used—e.g. the agent, plus the time and place of the utterance—fully determine their referents, which may be taken to be their semantic contents relative to the contexts. In both cases, descriptive enrichment (based on speaker intentions and salient assumptions in the conversational background) may associate extra, pragmatic information with the referents that have already been determined semantically—thereby contributing to the propositions asserted by utterances of sentences containing names or pure indexicals. However, the referents themselves are fixed prior to any such pragmatic supplementation. This is not so in the case of demonstratives, where any descriptive enrichment of the sort that occurs with names and pure indexicals comes on top of a prior pragmatic completion that is needed to provide the referents of demonstratives in the first place.26 The lesson here is that the conception of meaning, and its relation to assertion, that stands behind P2*, and the analyses I have given of examples containing proper names, is part of a much broader and more pervasive picture of language and language use.

My final illustration of this picture involves possessive noun phrases. A striking feature of these noun phrases is the staggering variety of interpretations they may receive. Here is a sample: one may use Sam's sibling to talk about a person with the same parents as Sam; one may use Sam's leg to talk about a part of Sam, Mary's wish to talk about something she wishes for, Susie's property to talk about property she owns, Martin's watch to talk about a watch he owns, or a watch he wears; one may use Barbara's book to talk about a book she wrote, a book she (p.382) wants to write, a book she is reading, a book she owns, or a book she has requested from the reference desk; one may use Martha's party to talk about a party for Martha or a party she is giving; one may use Gopal's language to talk about a language he speaks, a language he has decided to study, or a language he invented; one may use Tuesday's meeting to talk about a meeting that will occur on Tuesday; one may use the argument's premises to talk about the starting points of the argument, and the argument's conclusion to talk about the proposition to be inferred from the premises of the argument; and one can use John's car to talk about a car he owns, a car he is driving, a car he has rented, a car he is riding in, the car he arrived in, or a car he wagered on in the Indianapolis 500.

This is not ambiguity; it is pragmatic contextual supplementation. Call it indexicality, if you like, but if so, recognize that it is of the demonstrative sort, with relatively few semantic constraints on the relationship between the nominal possessor and the nominally possessed.27 To know the meaning of the construction NP's N is, at a first approximation, to know that the referent of the possessor noun phrase is characterized as standing in some not too heavily constrained relation to something to which N applies. Which relation this is in any given case is not something for semantics to decide; it is determined by pragmatic features of the context of utterance. Accordingly, we should not view the semantic content of a sentence like John's car is a Corvette as being a complete proposition, let alone one that is asserted in all normal contexts in which the sentence is assertively uttered. Rather, the semantic content of this sentence is something more like a propositional matrix with a gap in it to be filled by a contextually determined relation (meeting relatively minimal semantic constraints) that the speaker is claiming to hold between John and a certain car. Thus, what is asserted by an utterance of this sentence is not the proposition the sentence semantically expresses, but a proposition that arises from the semantic content of the sentence by adding features of the right sort that are obvious and salient in the conversational background.

I have argued that essentially this process of contextual supplementation is also the key to understanding the relationship between the semantic content of sentences containing proper names and the propositions that those sentences are used to assert. If I am right, this process, relating semantic content to assertion, is a ubiquitous one in language. This, in my view, is the thing we most need to better understand, if we are to continue to make progress in natural language semantics, pragmatics, and the relationship between the two.

Notes:

(1) Scott Soames, Beyond Rigidity: The Unfinished Semantic Agenda of Naming and Necessity (New York: Oxford University Press, 2002).

(2) Normality is not a matter of statistical regularity. For more on normal contexts, see ibid. 58–60.

(3) The distinction between asserting (or intending to assert) p and merely conveying (or intending to convey) p is discussed in ibid. 72–86.

(4) In order to keep matters simple, I here ignore the possibility that some extremely general sortal might be predicated of Mr Hempel in all these situations.

(5) Here, and throughout this chapter, I put aside any consideration of proper names without referents. See Beyond Rigidity, ch. 3, app., for some relevant considerations.

(6) Possible worlds are really possible states of the world—ways the world could have been. To say that an individual o exists “in” another possible world w is to say that had w obtained (i.e. had the world been in state w), o would have existed. The actual world is the possible state that the world is actually in.

(7) I use bold face italics to play the role of corner quotes. Thus, (ii) in the text is to be understood as saying that for any linguistically simple name n and predicate F, one must realize that to assertively utter a simple sentence consisting of n followed by ‘is’ followed by F (in a normal context) is to say of the referent of n that it “is F”. The scare quotes are used to call attention to what I hope is a bit of comprehensible sloppiness. A more precise way of putting the point would be to substitute ‘to predicate the property expressed by F of the referent of n’ for to say of the referent of n that it “is F”. Here, say of is a variant of say, which, throughout, I use only in its indirect discourse sense. As I understand that sense, x says that S is roughly equivalent to x asserts that S.

(8) Beyond Rigidity, 67–72.

(9) In discussing this and related examples, I ignore tense.

(10) A different, perfectly acceptable, way of thinking of it would be to retain the view that the semantic content of (3a) is the singular proposition 〈Neg 〈Identity 〈 Mr Hempel, Mr Hempel 〉〉〉, but to allow the propositions asserted by utterances of (3a) to be enrichments of this proposition obtained by adding descriptively salient properties to the arguments of the identity relation. I adopt the description in the text primarily to highlight the process of contextual supplementation. The gaps in (4) aren't intended to have any special metaphysical status.

(11) I am indebted to my student Mike McGlone for helping me appreciate the force of this point.

(12) See Beyond Rigidity, 343–4, n. 7, for a discussion of how this works in other cases in which an utterance of a propositional attitude ascription does not result in the assertion of the proposition semantically expressed, because of a canceling conversational implicature.

(13) See ibid. 59–60 for a brief discussion of one natural way of understanding Gricean conversational implicatures that does not require the actual conscious or unconscious reasoning of speaker-hearers to match the idealized Gricean reasoning that explains the implicature. McGlone is exploring ways to develop this idea that would allow one to generate a canceling implicature in the case involving (6a).

(14) Following the thinking in n. 10, one could maintain the substance of this view while treating sentences containing linguistically simple names as expressing non-gappy propositions, provided one allowed for their enrichment by contextually salient properties. For example, one might define propositional matrices to include both explicitly gappy structures and ordinary singular propositions, while conceiving of enrichment (completion) as encompassing both filling in gaps and adding properties to individuals that occur as constituents of singular propositions. I leave it open which way of formulating the view is to be preferred.

(15) Important aspects of this general conception of the relationship between semantics and pragmatics can be found in a number of places in the recent literature. See in particular Kent Bach, ‘You Don't Say’, Synthese, 128 (2001), 15–44, for useful discussion and examples supporting the theses (i) that the semantic contents of non-indexical sentences are not always complete propositions, and (ii) that the (primary) propositions asserted by utterances of sentences are often contextual enrichments and completions of the semantic contents of the sentences uttered (even in some cases in which those contents themselves determine complete propositions). Other relevant literature includes id., ‘Conversational Impliciture’, Mind and Language, 9 (1994), 124–62; Robyn Carston, ‘Implicature, Explicature, and Truth-Theoretic Semantics’, in R. M. Kempson (ed.), Mental Representations: The Interface between Language and Reality (Cambridge: Cambridge University Press, 1988), 375–409; François Recanati, ‘The Pragmatics of What Is Said’, Mind and Language, 4 (1989), 294–328; and id., ‘What Is Said’, Synthese, 128 (2001), 15–44.

(16) Note, I here give only a necessary, not a sufficient, condition for when the assertion of p guarantees the assertion of q. It is not to be assumed that assertion is closed under the relation of necessary, a priori consequence.

(17) As before, I represent the partially descriptive name as a description, which I let occupy an argument place for an n-adic predicate. Although I leave open the possibility that these descriptions are generalized quantifiers, I do not presuppose this or any other particular analysis. The one thing I insist on in discussing (10), and the later examples built upon it, is that however the descriptions corresponding to partially descriptive names are analyzed, occurrences of them in the negative identity sentences that appear below are to be understood as within the scope of the negation operator in the clause. Take this to be part of the stipulation of the cases.

(18) In this and other sections where I examine whether sentences containing partially descriptive names provide cases of failure of P1 to generalize, I ignore possible descriptive enrichments corresponding to linguistically simple names in a sentence. The issue of whether such enrichments give rise to failures of P1 to generalize is taken up later.

(19) If we treat Senator Clinton as a generalized quantifier, then (on one familiar analysis) (15a) will semantically express the proposition that predicates the property of being a property of a unique individual who is both a senator and H.R.C. of the property of being a former First Lady. We may take it that believing this proposition involves believing that someone (in fact H.R.C.) is both a senator and a former First Lady. Although I don't presuppose any specific analysis of partially descriptive names, I do presuppose that any adequate analysis will give this result.

(20) Here I assume without argument that the reason she doesn't assert this is not that there is a conversational implicature which defeats some presumption that it is asserted.

(21) Beyond Rigidity, 233–5.

(22) Think of proposition (49b) as expressed by a sentence that comes from Gil asserted that Mary just learned that [the x: x is our colleague standing over there & x = z] = [the y: y is a famous philosopher & y = w] by existentially generalizing on ‘z’.

(23) Perhaps (52), or an English equivalent, could itself be used to assert a true descriptively enriched proposition in which different descriptive enrichments accompany the final two occurrences of the variable. That would be rather unnatural, but even if it is possible, it doesn't help the argument, since from such an understanding of (52) we don't get the conclusion that Gil has asserted anything false.

(24) D. Kaplan, ‘Demonstratives’, in J. Almog, J. Perry, and H. Wettstein (eds.), Themes from Kaplan (New York: Oxford University Press, 1989).

(25) There are, of course, extended uses of some of these demonstratives, as when one refers to a ship as she even though the ship is not female, or says of one's favorite team We won the game even though one is not really a member of the group that won the game. In addition, one may say, while listening to a piece of music “Here is where the tubas come in, and there is where the trumpets come in” without, arguably, referring to a spatial location. As for ‘now’, uses of it in narratives invoking the specious present need not refer to the time at which the narrative is given. I leave aside the question of how one accommodates such uses. (Thanks to the referees for bringing some of this material to my attention.)

(26) For this reason it is natural to deny that sentences containing these demonstratives semantically determine complete propositions, even relative to contexts. See Bach, ‘You Don't Say’, sect. 5.1, and Kenneth Taylor, ‘Sex, Breakfast, and Descriptus Interruptus’, Synthese, 128 (2001), 45–61, for discussions of this issue.

(27) For a discussion of some constraints that do appear to be imposed by the possessive construction, see Chris Baker, Possessive Descriptions (Stanford, Calif.: CSLI, 1995).