|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 COGNITIVE SIGNIFICANCE AND COEXTENSIONALITY By SHAWN M BURTOFT A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPH Y UNIVERSITY OF FLORIDA 2011
2 2011 Shawn M Burtoft
3 To my Mom
4 ACKNOWLEDGMENTS I would like to thank the members of my committee, John Biro, Kirk Ludwig, Michael Jubien, and Greg Ray, for their very help helpful comments and suggestions. I would also like to thank William Butchard and Kevin Savage for their suggestions in the early stages of this project.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 ABSTRACT ................................ ................................ ................................ ..................... 6 INTRODUCTION ................................ ................................ ................................ ............. 8 ................................ ............................... 8 THE SEMANTIC APPROACH ................................ ................................ ...................... 12 The Semantic Metalinguistic Solution ................................ ................................ ..... 12 Description Theories ................................ ................................ ............................... 14 Objections to Description Theories ................................ ................................ ......... 16 The Generic Semantic Solution ................................ ................................ .............. 17 The Singularity Pro blem ................................ ................................ .......................... 18 The Variance Problem ................................ ................................ ............................ 19 The Semantic Problem Generalized ................................ ................................ ....... 21 The Problem of Ignorance ................................ ................................ ...................... 21 The Modal Problem Generalized ................................ ................................ ............ 22 THE NON SEMANTIC APPROACH ................................ ................................ ............. 26 The Non semantic Metalinguistic Solution ................................ .............................. 26 TOWARDS A NEW SOLUTION ................................ ................................ .................... 29 Refining the Explanandum ................................ ................................ ...................... 29 RELATED PUZZLES ................................ ................................ ................................ ..... 39 Empty Names and Fictional Discourse ................................ ................................ ... 39 Negati ve Existentials ................................ ................................ ............................... 43 Modal Contexts ................................ ................................ ................................ ....... 44 Belief Contexts ................................ ................................ ................................ ........ 46 ALTERNAT IVE NON SEMANTIC SOLUTIONS ................................ ........................... 50 CONCLUDING REMARKS ................................ ................................ ........................... 54 LIST OF REFERENCES ................................ ................................ ............................... 57 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 60
6 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CO GNITIVE SIGNIFICANCE AND COEXTENSIONALITY By Shawn M Burtoft May 2011 Chair: John Biro Major: Philosophy Cognitive significance, as Frege was thinking of it, is a value speakers assign to sentences based on their informativeness. He was particularly conc erned with explaining how sentences containing coextensional terms that, intuitively, express the same thought, or proposition, can differ in this respect. For example, sentences of the a = a entences of the form a = b question is how to explain this, given that both sentences (if true) say the same thing about the same object, namely, that a is identical to itself This has come to be known such sentences do not, in fact, say the same thing and, as a result, have sought to explain differences in cognitive significance semantically My aim is twofold. First, I raise a number of objections to the semantic approach, among them an argument that assumptions implicit in the construction of the puzzle straightforwardly rule out a semantic solution. Second, in light of this, I consider the virtues of a non semantic approach and ultimately outline a non semantic solution that does the explanatory work with the fewest assumptions. As an added bonus, I show how this account easily
7 extends to various related puzzles and problems, including empty names, negative existentials, and substitution in opaque contexts.
8 CHAPTER 1 INTRODUCTIO N Cognitive S It was once believed that the Morning Star, Hesperus, was distinct from the Evening Star, Phosphorus, but we have since d iscovered that, in fact, Hesperus is identical to Phosphorus. Perhaps only a philosopher would take issue with this sentence. After all, it is widely held or at least widely claimed that it took the advent of the telescope to conclusively determine that He sperus is (identical to) Phosphorus. But what did the telescope reveal exactly? That the identity relation holds between a pair of objects? The identity relation, by definition, holds between a single object and itself, and nothing else. Moreover, it does not take empirical evidence to tell us that us that each object bears this relation to itself. No one has ever doubted that Hesperus = Hesperus, or that Phosphorus = Phosphorus. So why are we inclined to speak as though we discovered that Hesperus = Phosp horus? Frege addressed this iss Identity challenges reflection through questions which are connected with it sentences of obviously different cognitive sign and according to Kant is to be called analytic, whereas sentences of the are not always grounded a priori. Cognitive significance, as Frege appears to be thinking of it, is a value that speakers assign to sentences based on what can be learned from them. But we need to be values we assign them is dependent on their semantic content. But it is not clear that this follows. In fact, there are compelling arguments for rejecting this claim, so it will be
9 outset. The following definitions are an attempt to do this in a neutral way. First, what a speaker can learn from a sentence is dependent on what follows from the sentence, if true, and what th e speaker believes at the time: (CS1) For any speaker, x, w hat x can learn at t from a sentence S (in language L) is what follows from S's being true (in L) and what x believes at t Given (CS1), we can spell out what it means for sentences to differ in co gnitive significance in terms of what can be learned from them: (CS2) For any sentences of L, S, S', for any x, for any time t S and S' differ in cognitive significance (in L) for x at t iff what x can learn at t from S (in L) differs from what x can lear n at t from S' (in L). Finally, given (CS2), we can say, more generally, what it means for sentences to differ in cognitive significance: (CS3) For any sentences of L, S, S', S and S' differ in cognitive significance (in L) iff it is possible for there to be an x and a time t such that what x can learn from S (in L) at t differs from what x can learn from S' (in L) at t 1 This definition is useful because it brings to light four points worth emphasizing about the puzzle. I. While Frege suggests that the puz zle is peculiar to identity sentences, it seems clear that the same puzzle arises for sentences containing coextensional terms 1 More relativi zation would be required for context sensitive sentences but, for simplicity, I ignore it here.
10 co refer can also differ in cognitive significance, leading to the same puzzle. II. T o say that two sentences differ in cognitive significance is not necessarily to say that one of the sentences is a priori and the other is not. For exa mple, two a posteriori sentences containing coextensional terms can differ in cognitive value : The Adventures of Huckleberry Finn Clemens is the author of The Adventures of Huckleberry Finn Thus, while properties like being a priori and being a posteriori may provide sufficient conditions for when two sentences differ in cognitive significance they do not provide necessary conditions. III. The cognitive value that a speaker assigns to a sentence is speaker relative Th is may be true even in less obvious cases, like those involving identity claims. For a priori and trivially true whereas the latter is often said to be a post eriori and informative. However, consider the epistemic status of the sentences for Mark Twain. Presumably, for Twain, they were equivalent in cognitive value. This suggests that cognitive significance should not be thought of as a property of sentences si mpliciter but as a relational property holding between sentences and speakers. IV. The puzzle is not peculiar to proper names. The puzzle arises with coextensional expressions in general, including descriptions, indexicals, common nouns, and predicates. There are contexts, for example, in which each of the following pairs of
11 sentences involve extensionally equivalent terms and differ in cognitive value for a particular speaker. (1a) The first Postmaster General is the first Postmaster General (1b) The first Po stmaster General is the inventor of bifocals (2a) If something is gold then it is gold (2b) If something is gold then it is Au (3a) All eye doctors are eye doctors (3b) All eye doctors are oculists (4a) This is this (4b) This is that (where the same th (I) (IV) suggest basic desiderata for a solution to the puzzle. A n adequate account should accommodate each of these points. ultimately, my aim is to understand how coextensional sentences of any type can differ in cognitive significance. M y hope is that coming to a puzzle will provide sufficient insight into the nature of the more general problem. I will begin by drawing a distinction between semantic and non semantic approaches to the puzzle, discussing common examples of each type and the difficulties they face. In the remainder of the project I will outline an account, a species of the non semantic variety, w hich avoids these difficulties. In addition, I will show that the proposal easily ext ends to a wide range of related and perhaps a few seemingly unrelated, philosophical problems, including: the problem of empty names, fictional discourse, negative existe ntials, and substitution in belief and modal contexts.
12 CHAPTER 2 THE SEMANTIC APPROACH can be seen as falling into two general categories, semantic and non semantic. The semantic approach seeks to account for differences in co gnitive significance by positing some difference in semantic content between the = differ in cognitive significance because they differ in semantic content. Non semantic solutions, as the name suggests, seek to explain differences in cognitive significance by appeal to non semantic factors. In this chapter, I focus on examples of semantic approaches, beginning with the theories of Frege and Russell and reviewing some of the standard objections. In the remainder of the chapter, I raise a number of objections against semantic solutions in general. The Semantic Metalinguis tic Solution Initially, Frege (ibid.) treated sentences of the form = as expressing a relation between proper names. dealing with those signs: a relation between them would be ass erted. If this is right, then the identity predicate in such cases does not express a relation between objects in the world; rather, it expresses a relation between names This suggests a semantic explanation of = = differ in cognitive value: the former literally co refer a seemingly a priori truth, 2 while the latter co refer which is something one comes to know a posteriori 3 2 In ordinary language there are counterexamples because a name may have multiple bearers. 3 There are apparent counterexamples to this as well; recall the exam ple mentioned earlier regarding the
13 On this view, the identity predicate in statements of the form = does not express the identity relation but the lingui stic relation has the same referent as ; the = It is a semantic solution in that it seeks to explain the difference in cognitive signifi = = by treating the sentences as having distinct semantic content. On this account, the type of information conveyed is linguistic and part of the meaning of the sentence. 4 The most common complaint against the metalinguistic solution is that it is counterintuitive. Frege, himself, later denounced it for this reason. As noted at the outset, that Hesperus is Phosphorus, e.g., is generally taken to be an astronomical discovery, and not just a fact about our use While I think this objection has some force, it seems to me that there is an underlying problem which should be addressed The problem is that it is not clear what is driving the intuition that identity statements in ordinary dis course are about objects in the world. It is an empirical discovery, but, again, what was discovered, exactly? Not that a relation holds between two objects, because, given that the sentence is true, t here is only one object in question. Nor can the sentence be said to express the proposition that Hesperus is identical to itself, as this is an a priori truth, not something that came to be known through empirical means. I will return to this question in the next section to see whether 4 There is a non semantic version of the metalinguistic solution which I will c onsider in the next chapter.
14 Another problem facing the (semantic) metalinguistic approach is that it is specifically aimed at cases involving identity claims. However, as noted earlier (in (I)) Hesperus is a planet if Hesperus is a planet then presu do, but the proposed solution clearly does not apply to such cases because it turns on a particular way of interpreting the identity predicate and, hence, works only for identity sentences. Description Theories = is to be understood as it traditionally has been as expressing the identity relation as opposed to Mi view of proper names, the view that the meaning of a proper name is exhausted by its referent. Th e argument can be formulated as follows. 1. = = differ in c ognitive significance. [premise ] 2. If two sentences differ in cognitive significanc e then they differ in semant ic content. [premise ] 3. = = differ in semantic content [1, 2] 4. If the semantic content s are exhausted by their referent s refer = = do not differ in semantic cont ent. [ premise ] 5. refer. [premise ] 6. The semantic content s not exhausted by their referent s [3, 4, 5]
15 Thus, it appears that a semantic solution is incompatible with the Millian view of proper n ames, which suggests that the mer heavily on its treatment of proper names. Bertrand Russell suggested that ordinary proper names are abbreviated definite descriptions, which are best represented as quantified noun phrases 5 On his view a sentence of the form F Hesperus as a constituent. Instead, it takes the form of an existential sentence: (1) ( x predicates a property (or conjunction of properties) said to be uniquely instantiated by the object (2) ( x) P x contributions to the sentence. Frege held that proper names refer, but not directl y. According to him, names refer by semantically expressing something else, a sense The sense of an expression is a conceptual representation of whatever object uniquely fits the representation. have the same referent but express different sens es, which are said to determine the same referent. Thus, like Russell, Frege sought to explain differences in cognitive significance by positing distinct 5 It is best construed as semantic considerations, but I will assume they are, as this seems to be the general consensus (and, in any case, one could hold such a view).
16 semantic content for coextensional terms. Such theori es are known as description theories because they treat proper names as referring via some descriptive content which speaker s commonly associates with the name. Objections to Description Theories There are, however, serious difficulties facing theories of this sort. The most obvious problem, I think, is that the puzzle we are interested is not limited to sentences involving proper names. It is unclear, for instance, how such considerations could be extended to account for examples involving common nouns, su ch as (3a) and (3b). Moreover, there are independent reasons for rejecting a solution to the puzzle that appeals to a description theory of proper names, because there are well known problems facing description theories in general. I will briefly mention two of the most commonly cited, each noted by Saul Kripke (1980). First, intuitively, proper names are rigid, i.e., they designate the same object in every possible world in which the object exists. Most d efinite descriptions, by contrast, are not rigid; brightest o might have designated an object other than Hesperus. Thus, the problem is that treating (non rigid) descriptions as giving the meaning of proper names clashes with our modal intuitions. This is known as the modal problem Second, a speaker may fail to associate descriptive content which uniquely determines the referent of a name, or may associate inaccurate content with a name, and yet still be a competent user of the name. For instance, suppose a spea ker associates only a celestial object According to the description theory, the speaker fails to refer to anything when a celestial object fai ls to
17 pick out Hesperus from among other celestial objects. Also, a speaker may associate descriptive content with a name that the referent of the name does not satisfy. For example, suppose one mistakenly associates the content the brightest star in the e vening sky with Hesperus (in fact this was the case, as Venus was initially thought to be a star). The description theory predicts, implausibly, that each time the speaker use s not to Venus but to Sirius. This is an example of what is known as the semantic problem. The Generic Semantic Solution istinct semantic content s for coextensi onal expressions. Presumably, as with descriptive theories of names, these content s will include distinct characteristics or properties associated with the object (s). However, since we do not want to limit our account to names, it will be useful to charact erize this feature, more generally, in terms of coextensional expressions (names common nouns, indexicals, etc.). The question we are interested in is how to understand differences in cognitive significance between two sentences that differ at most in hav ing distinct coextensional expressions; call them e 1 and e 2 The semantic theorist must posit some property or set of properties P 1 as part of the content of e 1 and s ome property or set of properties, P 2 as the content of e 2 such that P 1 P 2 The details the particular properties involved and how these contribute to the truth conditions of the sentence may vary, but for the purposes of the subsequent discussion the details are unimportant I will often allude to in my examples, because it i s widely known, but what will be said should apply to any semantic theory that has this feature
18 The Singularity Problem As I see it, there is a fundamental problem that undermines any semantic solution of the generic type, as well as any semantic solutio n that treats the identity sign as expressing a relation between two (or more) objects (such as semantic metalinguistic solution ) The instantiation of logical principles (the laws of identity, and inference rules ( modus ponens modus tollens reductio ad absurdum etc.) requires that for any name n the following two conditions be met. (R1) n designates a single object on each of occasion of use (i.e., n is a singular term) (R2) n designates t he same object on each occasion of use Without (R1) and (R2), we could not guarantee the truth of even seemingly a priori Fa & ~Fa) Fa v ~Fa A semantic account of cognitive significance is committed to the view t hat while expresses a non trivial truth. all true. And we can grant this only on the assumption that the terms designate one and the same obj straightforwardly semantic metalingui stic solution.
19 According to that refer. However, conditions (R1) and (R2) preclude this. co refer ing terms then whatever relation is co refer is a fact about two obje object. By parity of reasoning, this would also rule out any view that treats the identity predicate as expressing an irreflexive relation. On the other hand (R1) and (R2) also raise a problem for any semantic theory that treats the identity sign in the standard way, as signifying the identity relation. Given says that Hesperus bears the identity relation to itself, which is a priori and, hen ce, uninformative. But, then, it also co refer that the same a priori proposition. Thus, the only obvious a posteriori premise in the set up of the puzzle is co refer and, as we have just seen, this cannot be what is expressed by For this reason, it seems to me that the semantic approach is un equipped to explain the apparent inf ormativeness of sentences The Variance Problem If differences in cognitive significance w ere to be explained semantically, we would expect any competent speaker to acquire the same or at least very similar information in com ing across a sentence of the form = because such information
20 would be expressed by the sentence 6 However, there may be counterexamples the same star Copernicus came to believe that the names designated the same planet Presumably, prior to forming these beliefs, each man assigned different cognitive values However, given their background beliefs about the nature of the object (P ythagoras believed it was a star, Copernicus believed it was a planet) it could have been the case that the contents of their beliefs differed, respectively. For example, suppose Pythagoras c a me to believe (something like) that the brightest star in the mo rning and the brightest star in the evening are one and the same star while Copernicus came to believe (something like) that the brightest planet in the morning and the brightest planet in the evening are one and the same planet Such discrepancies should not be possible if the explanation of cognitive significance is based on differences in semantic content If the correct would expect Pythagoras and Copernicus to have formed the same (or a very similar) belief in discovering that the terms co refer but it seems possible that they did not. 7 One available move here is to treat semantic content as speaker rel ative Sainsbury (2002) suggests something along these lines, positing senses for expression tokens rather than types. But adopting this line runs the risk of yielding the sort of 6 At least, assuming, that sent ences do not vary in meaning for different speakers. 7 In any case, it is not difficult to imagine cases that have this peculiar result.
21 psychological theory that Frege, for one, was opposed to I will examine thi s option in greater detail in Chapter 5. The Semantic Problem Generalized cognitively significant for him for reasons that are inconsistent with what we now know about the referent of these names namely, Venus. For instance, in recognizing the truth evening was identi cal to the brightest star in the morning. To explain his epistemic position semantically we must assume that part of the semantic content of the terms of being a star, which seems counterintuitive because t he names pick out a planet, not a star. The Problem of Ignorance A further problem for the semantic approach is that there appear to be cases where the only suitable semantic content is metalinguistic. For example, suppose a competent English speaker wit h no knowledge of the Roman orator Cicero, nor any Tully Despite her ignorance, it is plausible to think that the sentence will differ in cognitive significance for her from the sentence Cicero is Cicero because, in encountering the former, she comes to form the non co refer If we take a semantic approach to explaining cognitive significance then we must say that this metalingui stic information constitutes the semantic cont Cicero is Tully because, by hypothesis, the speaker is not in possession of any additional
22 content This explanati on would commit the semantic theorist to the (semantic) metalinguistic view, which seems untenable for reasons we have already considered. The Modal Problem Generalized identity reconciling the view that such sentences are informative with the view that the identity predicate expresses a relation between objects in the world. According to Frege, the tr a posteriori because it expresses a proposition and it is not a priori that these two senses determine the same object. But then it woul d seem to follow, contrary to the identity thesis ((x)(y) if x = y then does not the brightest object in the morning sky the brightest object in the evening sky brightest object in the morning sky = the brightest object in the evening sky. But as Kripke pointed out, this is not a necessary truth because the descriptions flanking the identity sign are not rigid designators (terms that pick out the same object in every possible tuitively, the truth of a Kripke (1980, pp. 28 29) has made similar remarks: [We] see a star in the evening and it is in the morning an
23 it is identical with itself. This is something we discover ed. However, we have just seen that this view, while intuitively compelling, clashes with another intuitively compelling view: that true identity statements are necessarily true. It seems to me that any semantic approach which treats the identity sign as e xpressing the identity relation is going to face the same difficulty. Again, on any semantic view, the a = b a b erties of, or associated with, the object a / b But it is not clear that this can be done in a way that is consistent with the view that such sentences express necessary truths. On the Millian view of proper names, the explanation is straightforward: the na a b semantic content, object a a = b identity thesis, the sentence is necessarily true (assuming proper names are rigid designators). But if we posit non does not appear to informative. Take any such theory and let P a be the property or properties that makeup a b be the prope rty or properties that are b a and P b are essential properties properties had by a necessarily a b will not be rigid designators and, thus, it will not a = b arily true. There are a couple work around options for this problem. One method involves rigidifying the terms flanking the identity sign by invoking modifiers that in effect transform them into rigid designators. A common way of doing this is by prefixin g the
24 actual brightest brightest object visual in possible world, viz., Venus. If this is right, then the following expresses a necessary truth. (2) The actual brightest object visual in the evening sky = the actual brightest object visual in the morning sky. omething like (2) then we could develop a semantic theory of the generic type that yields necessarily true identity statements. The problem with this move, however, is that, in effect, it treats the expressions flanking the identity sign as contributing no thing more to the sentence than the unique object denoted by [(the F)], even though there is a term appearing in it which has an intension, and our grasping its intension is relevant to our understanding of what it picks out. In other words, to get the res ult we want that s entence (2) is necessarily true we have to understand it as express ing the same proposition expresses on the M illian view of proper names: that a particular object (Venus) is identical to itself This is an a priori truth, and one that holds irrespective of the expressions we use to pick out the object in question. Thus, while the rigidifier move a = b does not ultimately help us account for the intuition Another option is to subscribe to a dual content view, according to which, such sentences express two (or more) propositions. Bealer (1993), for example has argued ositions, a proposition that is knowable a priori and a proposition that is knowable a posteriori Perry (2001) has advanced a similar view. In each case the a priori proposition is said to be a singular proposition whereas the a posteriori proposition is said to include contingent
25 information about the referent which is tracked through conventional use of the terms that flank the identity sign. On the face of it, this line provides an answer to the modal ingular proposition explains the intuition that the sentence is necessarily true because it expresses the necessary truth that Hesperus/Phosphorus is self additional a posteriori proposition helps explai n the intuition that the sentence is informative, because, in addition to the trivial information that Hesperus is self identical, it also expresses non trivial information a posteriori information associated by convention among speakers of the language ab out the referent. Unfortunately, this move does not avoid the problem of variance, the semantic problem, or the problem of ignorance. 8 Moreover, I think it will become clear that such a move is unmotivated. Ultimately, we can account for differences in cog nitive significance without assuming that sentences are ambiguous in this way. 8 At least, not without the consequence that the a posteriori proposition expressed can vary from speaker to speaker.
26 CHAPTER 3 THE NON SEMANTIC APPROACH The Non semantic Metalinguistic Solution The non semantic approach seeks to explain the cognitive value a speaker assigns to a sentence by appealing to information which is independent from the information conveyed by the sentence. For example, one might endorse a non semantic version of the metalinguistic account, according to which differences in cognitive significance are the result of metalinguistic information that is posited as part of the = differ in cognitive signif icance from = ion with respect to co refer semantic however, this in formation is not part of the meaning of the sentence. For instance, does not co refer however, a speaker might infer this information upon encountering the sentence and there by come to form a belief that she would not come semantic explanation of why a speaker may assign different cognitive values to the The good news about the non semantic metalinguistic solution is that it avoids most of the difficulties raised against the accounts we have considered so far. First, it is compatible with the view that the sentences in question are about objects in the world, not names of objects, because it does not say anything about what is expressed by the
27 sentences. Second, it is not limited to identity statements; it can be easily extended to most of the cases mentioned in (IV) at the outset in each case, the difference in cognitive value is simply attributed to differences in the linguistic vehicle. Additionally, since it is neutral with respect to what constitutes the semantic content of the sentences in question including the semantic content of p roper names it does not objections to description theories. Furthermore, it provides us with an explanation of cases of ignorance. For example, consider again the Cicero case mentioned above. for a speaker who has no knowledge of Cicero because, in encountering the former, she may come to form a belief that she would not come to form in encountering the co refer The bad news is that this solution faces difficulties of its own. For one thing, like the previous solutions we considered, it is seemingly incompatible with the intuition that a posteriori truth about objects in the world. For whether one tries to explain the intuition semantically or non semantically, the metalinguistic approach is committed to an explanation which hinges on i nformation about the terms involved, not about the objects those terms refer to. I mentioned that the metalinguistic solution works for most of the example in (IV), however, one type of example it obviously cannot handle is a case like (1), where the demo nstratives flanking the identity sign demonstrate the same object (imagine, e.g., the speaker pointing to two very different photographs of the same object). (1) This = this
28 Intuitively, (1) could be informative in such a context. However, the metalinguis tic view does not offer a satisfactory explanation since its explanatory power turns on there being distinct expressions in play. 9 A further problem for the metalinguistic approach is that we can imagine cases where the puzzle arises even though the relev ant linguistic information is not available. Consider the following thought exper iment. Suppose that, prior to being introduced, Clark Kent and Lois Lane work as reporters for the Daily Planet newspaper in Metropolis. Lois has certain beliefs about Clark t hat he is a reporter, that he is mild mannered, and so on but, importantly, she does not know that he goes by the name enter a phone booth and emerge in tights and a cape, matching the description of the local super hero. It seems to me that there are three intuitively compelling assumptions one might make about the case described. 1) Lois comes to form a new non trivial belief. 2) If Lois was aware of the fact that her coworker goes would not choose to express this same belief with the differ for Lois in cognitive significance. However, it is clear that t he metalinguistic account (semantic or non semantic ) is of no help to us here. Call this the Lois Lane problem. 9 Assuming it hinges on types and not tokens. In any case, it is counterintuitive to think that the information conveyed, or inferred, in such a context is linguistic in nature.
29 CHAPTER 4 TOWARDS A NEW SOLUTI ON Refining the Explanandum Despite the problems facing the non semantic metali nguistic solution, it seems to me that it is on the right track. 10 More specifically, I think it gives us sufficient conditions for explaining why sentences may differ in cognitive significance (for example, the Cicero case) but that it does not give us nec essary conditions, as revealed by the Lois Lane problem and the demonstrative case. In my view, these problems arise because the metalinguistic solution is too narrow starts with the assumption that the ex planandum essentially involves linguistic expressions, but it is not clear that this is the right place to start. We represent things in a number of ways which often leads to the same sorts of issues. Questions arise over whether two photographs are of the same person, whether two works of art are of the same place or thing, whether two observations are of the same phenomenon, and so on. Of course, such examples can be, and often are, cast as puzzles about linguistic expressions we can use them to co nstruct = it does not follow from this that the solution directly involves facts about these sentences or their constituent term s. In fact, it might be that thinking of the puzzle in this way is running interference on developing the right solution. It might be that the putative examples supervene on this more general phenomenon of co representation. This is the line I plan to deve lop in the following chapters. 10 See Biro (1995) for a more detailed discussion of the virtues of the metal inguistic solution.
30 initially very broadly to include anything that intuitively is about, or picks out, some object for some particular individual. (In the broad sense I have in mind, most anything is apt to count as a representation, but typical examples include, words, books, photos, paintings, sculptures, sounds, mental images, represents Sherlock Holmes, the Venus de M ilo represents Aphrodite, and so on. ( I will return to the ontological question, as it bears on issues I wish to address later. ) A Representational Account The view I want to outline is a species of the non semantic variety which is consistent with desid erata (I) (IV) and avoids each of the difficulties mentioned in connection with the earlier solutions. To see how this approach differs from the ones we have Lois discovers something extraordinary about a certain male coworker And while she does not know the man by name, it seems reasonable to suppose that, if she did It is also reasonable to assume t hat the sentence would differ for her in cognitive witnessed an apparently ordinary man a ma n with apparently ordinary properties like being a reporter and being mild mannered enter a phone booth. Let K be the set of these properties. It seems safe to say that whatever other properties constitute K, properties like being faster than a speeding l ocomotive and being able to leap tall buildings in a single bound are not among them. Let S be the set of these properties
31 (i.e., the set of properties Lois associates with the name Lois believed that her coworker instantia ted the properties of K. Afterwards, she comes to believe that he instantiates the properties of K and S. 11 So the information she comes by, upon witnessing this man in suit and tie enter a phone booth at time t 0 and emerge i n tights and cape at t 1, can be characterized as a relation between a certain individual and two sets of properties, K and S. S o far there is nothing philosophically puzzling about the case described. The puzzle arises on the assumption that Lois would deem it appropriate to express her new belief or something very similar, if she knew his name ) and that this sentence would differ in cognitive significance for her from seems entirely plausible. We can imagine her asking around, learning his name rushing into as a shock to the editor and everyone else at the Daily Planet. By contrast, t he would not make headlines. So the question is is context and whether the explanation is best understood as semantic or no n semantic. In the present case we are assuming that Lois Lane would deem it appropriate to express he Clark Kent = S Clark Kent = Superman expresses something different tha n 11 Or some combination of properties from both sets, as some of them may be incompatible. For example, K might contain the property having human strength while S could contain the property having super human strength
32 Sup of K as part of the properties of S as part of the content of This would yield a semantic Clark Kent = Su differs Clark Kent = Clark Kent because the two sentences have distinct semantic content The problem, of course, is that this solution faces the problems covered in chapter 2. The metalinguistic approach (semanti c or non semantic ) does not work either. Fo r one thing, by hypothesis, Lois does not have (at t 1 ) the relevant linguistic What is more, even if she did have this inf ormation, it is counterintuitive to say that her co refer it seems more accurate to say that it would involve non linguistic properties like those in K and S A s I see it, the semantic approach, while intuitively compelling, demands too much and that the metalinguistic approach, while less demanding, is too narrow. My strategy is to incorporate virtues of both. Consider, again, how the scenario might play out onc would not ru with the same intention. This suggests that she does not believe that the editor realizes the two sentences are about the same individual, a reasonable assumption. Presumably the edito r, and most everyone else in Metropolis anyone who is not
33 position with respect to the matter would form a new belief on this basis (the belief may vary from one person to the next because personal representations may differ, a point wh ich will be accommodated by the positive proposal). Hence, for those people the sentences Clark Kent = Superman Superman = Superman differ in cognitive significance I think it is important, if possible, to preserve the intuition that often when we use identity sentences we intend to convey information abou t the extension(s) of the terms and not information about the terms themselves. It seems clear that Lois intends to convey information about Clark Kent not the name he point can be accommodated without semantic commitments. The main thing is recognizing that the puzzle in such cases is an instance of a phenomenon which is not peculi ar to language. The present case is a clear example because the puzzle only arises when we consider how Lois would express her new belief with a sentence very similar, if not identical, to what she learned at t 1 What is required is a way of spelling this out that is compatible with our intuitions about the case. Intuitively, Lois has two distinct ways of representing Clark Kent, the Clark Kent way and the Superman way Presumably, the former bears some relation to the properties of K, and the latter bears the same (type of) relation to the properties of S. From a theoretical standpoint, I do not think anything crucial hinges on how we understand this relation, so long as it captures what the speaker has in mind. It strikes me that the most economical approach
34 following Russellian fashion. At t 0 Lois believes that there is an x such that K x and a y such that S y and x y (where K and S predicate the properties of K and S, respectively) 12 At t 1 she believes that there is an x such that Kx and Sx. 13 Less idiomatically, she learns of a particular individual that, in addition to some ordinary characteristics (the properties of K), he exhibits some extraordinary characteristics (the properties of S). It is reasonable to suppose that this is what Lois intends to convey with associates conveying this descriptive information. In this respect, the suggestion i s similar to the idea behind descriptive theories. The crucial difference, however, is that, on the present proposal, all that follows is that the information is implicated ; it does not follow that it is expressed by the sentence. 14 The upshot is that we pr eserve the intuition that the without the semantic complications. Thus, according to this view, it is not the semantic content that does the captured by co predicate that includes properties that the speaker believes are instantiated by the object in question. Representational content is clearly speaker relative. Thus, the 12 For a full fledg ed Russellian treatment, we need an additional conjunct specifying that the predicates are uniquely satisfied, but I will omit this detail for the sake of simplicity. 13 Or some combination of the properties of K and S. 14 This leaves open the possibility th at the information is part of the content, in case there are independent reasons for thinking so.
35 account avoids the semantic problem and the problem of variance. Nonetheless, we should expect some overlap among speakers; typically, speakers associate a lot of the same properties with extensional expressions (e.g. having the ability to fly and being faster than a locom otive are properties most speakers associate with the name ordinarily have the same object in mind. puzzle does not pose a view of proper names. Recall the argument against the Millian theory : 1. = = differ in c ognitive significance. [premise ] 2. If two sentences differ in cognitive significance then they differ in semant ic content. [premise ] 3. = = differ in semantic content [1, 2] 4. If the semantic content s are exhausted by their referent s refer = = do not differ in semantic content. [ premise ] 5. refer. [premise ] 6. The semantic content s are not exhausted by their referent s [3, 4, 5] On the presen t view, premise (2) is false. Recall, the primary support for (2) is based on a posteriori truths; it is a posteriori Hes a priori truth, and this could not be the case unless the identity sentences can express a posteriori truths, but now I think we a re in a better po sition to se e why this should not be seen as evidence for premise (2). Frege
36 discovery but, again, it is unclear what this comes to. Consider the case of Pythagoras and Copernicus distinct stars; Copernicus initially believed that the names referred to distinct planets. We can imagine each man coming to form a new belief in recognizing the truth of the sentence they each learn that co refer then yes, but then this is the sort of view that the example is supposed to undermine. So, supposing Frege was right that a posteriori truth about some object in the conceptual representations of whatever object uni quely fits the representation. T his is supposed to explain why but, intuitively, it gets the wrong result in a case such as this Take Pythagoras, who initially believed t brightest star in the morning was distinct from the brightest star in the evening but that he has sinc e come to believe that they are one and the same star. If Frege was right, differe d for everyone else at the time who believed the same). We should be able to say brightest e brightest cannot say this for at least two
37 reasons. First, Venus does not meet either description because Venus is not a star (the semantic problem). Second, we can easily imagine cases in which the sentences different individuals (the problem of variance). Suppose that, for Copernicus, the sentences differed in cognitive significance because his associated belief was that the brightes t planet in the morning is distinct from the brightest planet in the evening. On a brightest brightest pl Pythagoras and Copern icus for different reasons, and a theory that treats the sentences as having distinct semantic content cannot explain this. The meta representational view, by contrast, handles these sorts of cases easily because, first, it is irrelevant on this view whe ther or not the representational content accurately characterize the object in question, and, second, unlike senses, it may (or may not) vary from one individual to the next. Initially, Pythagoras believes there are two stars, one that is only visible at n ight and one that is only visible in the morning. He has two distinct ways of representing Venus, one as a morning star, one as an evening star. the latter. Initially, he believes that there is an x such that Mx and a y such that Ey and that x His later epistemic position consists in the belief that there is a single object which instantiates the properties of M and E (or, more idiomatically: that there is an x
38 such that Mx and Ex). Note, the fact that Venus is not a star is irrelevant, on this proposal, because even if Pythagoras were to learn of his mistake it would not change the fact that he came to believe that a single object instantiated both sets of properties which is all that is required to explain why is c ognitively significant for him. We can, mutatis mutandis, explain Copernicus
39 CHAPTER 5 RELATED PUZZLES Empty Names and Fictional Discourse Another virtue of ( R) is that it helps explain certain intuitions about ordinary discourse with fewe r complications than a semantic theory. For example, consider the problem of empty names (names without referents) Intuitively, sentences containing especially problematic for the Millian view of proper names, because if the semantic content of a proper name is exhausted by its referent then such sentences are semantically incomplete and neither true nor false. I think the reason that sentences containing empty names are often treated as truth evaluative in ordinary discourse is that they are (or, at least, can be) cognitively sign ificant. An individual may come to form a new belief upon readi ng them or hearing them uttered. However, like the case of identity sentences, I do not think anything follows about the semantic content of the names involved. The representational framework can do the work of explaining these sorts of ca ses as well. Cons ider the Santa Claus has eight flying reindeer Believers and non believers both assent to this sentence. A child may assent to the sentence because she believes it is true that a man owns eight flying reindeer, while adults may assent to the se ntence because they believe it is true that according to a myth an individual owns eight flying reindeer (or something along those lines). Nonetheless, in some sense they think the sentence is about the same thing. I Santa C laus has eight flying reindeer because the sentence is true then we need to explain how a
40 sentence containing a name without a referent could be true or false. This has been tried, by appeal to non Millian theories of names, but these face difficulties al ready discussed. The general difficulty, as I see it, for any non Millian account of empty names is this. If there is more to the semantic content of a name than its referent then the additional content must, in some way, involve properties to explain why the sentence is true or false and this inevitably leads to counterintuitive results. Take any proper name, N. If N is non is not true because there is no object, content which is distinct from the referent. Presumably this will involve certain properties. Let P 1 P 2 P 3 P n be any such pro perties. It seems there are two options at properties P 1 P 2 P 3 P n or its instantiation is entailed by the instantiation of P 1 P 2 P 3 P n 15 The upshot is that an y true sentence containing an empty name will turn out analytic on such an account. 16 15 The sort of property entailment I have in mind is a necessary relation, e.g., instantiating the property being red entails instantiating the property being colored Predication is then the analogue of instantiation: 16 This only applies to theories that claim there are true sentences containing empty names. On some non Millian theories, such sentences are not true. For example, a descriptive theorist that appeals to a Russellian view of definite descriptions will tre at sentences containing empty names as existential sentences. As a result, any sentence containing an empty name will turn out false because nothing will satisfy the description. For this reason, I think such theories are in adequate as explanations for cog nitive significance again, presumably the reason certain sentences containing empty names appear true is because they a re cognitively significant, and hence descriptivist theories of this sort fail as explanations of cognitive significance because they d o not help us explain the intuition that the sentences are true. The only exception to this that I can see is a semantic account based on an ontological theory such as Alexi Meinong a sort it just so happens that such entities do not exist in the same way that ordinary spatiotemporal objects do. One might argue that such entities have certain properties contingently, in which case certain sentences ascribing properties to them would n ot count as such an entity and Pegasus does not essentially instantiate the property having wings
41 Take the sentence 1. Hamlet is indecisive If the sentence is true then the explanation must be that either the semantic content of being indecisive or the semantic content involves some property, or properties, the instantiation of which entails the instantiation of being indecisive In either case, the result is that (1) is analytic which seems false. 17 The meta repres entational theory do es not have this consequence because it is neutral with respect to the semantic content of such sentences. It does not commit one to the view that these sentences are truth evaluative; instead, it offers an explanation of why one might be inclined to say t hat they are. In fact, it is a matter of some dispute among Shakespearean scholars whether or not (1) is true. Both sides hold that the sentence is meaningful and truth evaluative but some have reasons for assenting to the sentence, while some have reasons for dissenting. Disagreements like this suggest that something is genuinely at issue in such cases, but no semantic theory of proper names alone can account for it. Any theory that treats the sentence as false (or untrue), such arly will not help. And any theory that treats the sentence as true appears to be committed to the view that the sentence is true solely in virtue of meaning, something presumably neither side would endorse. ory of objects is not a widely held view. 17 Perhaps there are examples involving empty names where this result is less worrisome. The French existence of being a planet of the way the name was introduced. However, the problem r emains when we consider apparently
42 Many philosophers have suggested introducing qu (P) According to myth, Pegasus has wings One obvious problem with this suggestion, as it stands, is that the empty name remains. analysis clearly does n ot help us explain the intuition that the speaker conveys Millian name, the semantic treatment alone may be enough to explain why the sentence appears to be true (suppose, for example, that we treat the name descriptively and part of the analysis like (P). So, it is unclear how this move, as it stands, could be of use. For another thing, we can easily imagine cases where the speaker wishes to convey something true with an empty name which cannot be captured with a general suppose a child that believes in Santa utters the se Santa Claus has eight flying reindeer. It seems clear that the child does not mean to convey something like (H). (H) According to myth, Santa Claus has eight flying reindeer.
43 Presumably, she means to convey whatever one would convey wi th that sen tence if the was not empty. This is a point accommodated by t he representational account because it does not presuppose that the speaker knows the name is empty. Negative Existentials raise additional problems. does not of it, the sentence is both true and about something. But if the sentence is true, it is not does not denote anything. S e mantic theories face the same dilemma here as with empty names. Any semantic theory that treats the sentence as false will not help us. And any theory that treats the sentence as true faces the same difficulty mentioned above of explaining how this could b e so without the implication that it is true solely in virtue of meaning. Furthermore, the qualifier move clearly does not work here, because, intuitively, it is false that, according to myth, Atlantis does not exist. On the representational view by cont rast, we can easily explain the apparent Atlantis does not exist accommodates apparent disagreements over to the truth value of sentences where it is
44 example, because some speakers will diss ent from it on grounds that there is, or was, such a place. Modal Contexts statue and Lumpl is the piece of clay from which the statue is made. At all times that Goliath and L umpl exist, it seems they share all the same intrinsic properties. And, in this sense, it seems that Goliath and Lumpl are one and the same physical object, and, hence, numerically identical. However, it also seems that there are certain properties which L umpl could instantiate but Goliath could not, and vice versa. For example, it seems Lumpl could take the shape of something other than a statue, but, presumably, Goliath could not. This problem is presented as a metaphysical one, but it can be cast as a pu zzle about reference. Consider the following sentences. (SC1) Goliath = Lumpl (SC2) Lumpl could survive being shaped into a sphere (SC3) Goliath could not survive being shaped into a sphere Notice, if proper names are Millian then the sentences are incom patible (assuming But if names are not Millian then it is not clear that this is the case. Suppose, instead, ir semantic content. For example, suppose they are descriptive Suppose clay chosen by X at t 0 is equivalent to t 1 case, (SC1) may be true, though only contingently, since the descriptions are not rigid designators. And (SC2) and (SC3) may both be true, because, for instance, while it seems possible that Lumpl could survive being shaped into a sphere it seems we
45 cannot say the same of Goliath. 18 On this line, the supp ort for (SC2) and (SC3) is due to the property being a statue then predicating a property of Goliath that is inconsistent with being a statue will yield a sentence tha t is not true. However, the problem as before, is that any such view of proper names will entail analytic sentences of the form which, arguably, are not analytic. 19 I think (R) provides a simple and intuitive solution here. It is clear from the w ay in which the case is presented that certain distinct refers to a statue, there is no support for the claim made in (SC3). For this reason, it is reasonable to think that what is driving the intuition that (SC2) is true while (SC3) is false is the result of thinking about what is possible with respect to what sorts of propertie s can be co instantiated. 20 The property being a piece of clay and the modal 18 There is a wide scope reading, on a Russellian analysis of definite descriptions, which seems true, t 1 ~Sz)). 19 The generalized semantic problem, for one, raises a worry here. Suppose, b efore unveiling the statue, latest work of art. This sparks some discussion among those in attendance over the exact nature of the referent involved in the discussion are competent enough with the name, but if being a statue were part of its content, the matter would be easily settled before the work was unveil ed. 20 Jubien (2001) endorses this line According to Jub ien, metaphysical problems such as the statue and the clay arise as the result of object fixation physical object, any truth about the boat is just a truth about the physical object (and vice versa). In somewhat different terms, we tend to think that a truth about an object qua boat is perforce a truth about the object simpliciter n we consider examples like qua statue and qua piece of clay are thoughts about the physical object simpliciter the result is a contradiction: the object does and does not instantiate the modal property possibly being squashed Jubien suggests that what i s really going on is this. In thinking about the object qua statue, our thoughts are not merely about the object, but also about the property being a statue and this explains our modal intuition that the object spherical, for example, because the object could not possib ly instantiate the property being a statue and some non statue shaped property.
46 property possibly being spherical can be co instantiated, whereas the properties being a statue and possibly being spherical cannot. Drawing on this, we can describe the case in a way that is consistent with (SC1) as follows. When one assents to (SC2) she has a belief of the form: (SC2*) Possibly: there is an x such that x is a piece of clay and x is spherical When one assents to (SC3) she has a belief of the form: (SC3*) ~Possibly : there is an x such that x is a statue and x is spherical Clearly, (SC2*) and (SC3*) are compatible with (SC1) since they yield true sentences for any name we plug in for x. At the same time this solution does not require additional semantic commitments. Belief Contexts The cases we have considered thus far show that there is often an asymmetry their beliefs. This is a point worth emphasizing because there are a number of puzzles in the philosophy of language that apparently hinge on it. More specifically, they hinge on whether the following claim is true, sometimes referred to as the disquotational principle (D). (in L) 21 Consider the following thought experiment proposed by Burge (1979). First, e because arthritis is a condition specific to joints, but the patient is unaware of this. 21 A more accurate expression of the principle would require adjusting it to accommodate for context sen sitivity (e.g., indexicals and tense); for simplicity I ignore this complication.
47 Now, imagine the very same patient, with all the same intrinsic properties and out, not arthritis, but a rheumatoid condition which is not specific to joints but also counterfactual case, when the patient reports rue. Burge argues that we cannot attribute the same beliefs to the patient in the initial case and his counterfactual counterpart. In particular, while the content case rather, in the counterfactual case the patient believes that he has tharthritis in his thigh. If this is right, then it follows that some men tal contents fail to supervene on intrinsic properties, and, hence, that internalism about mental content is false, since, by hypothesis, the patient and his counterfactual counterpart share all the same intrinsic properties. The argument has roughly the f ollowing form. 1. The patient in the first case and the counterfactual case have all the same intrinsic properties. 2. The content (in L) provides 3. The content (in L) does not 4. Hence, the patient in the first case and the counterfactual case do not share all the same belief contents, and, hence, do not share all the same mental contents. (2 3 ) 5. Hence, sameness of intrinsic properties does not guarantee sameness of mental content (i.e., internalism about men tal content is not true.). (1, 4 ) This argument implicitly appeals to a n assumption regarding the relation between what speakers report, or assent to, and the contents of their beliefs. According to the thought
48 experiment, the patient in the initial case, call him Patient A, reports ient in the counterfactual case, call him Patient C, develops the disposition to assent to ) do not directly follow from this; the argument requires a further premise to the effect that what compe tent speakers report, or assent to, provides the content of their beliefs. In other words, Burge is implicitly appealing to something like (D). Kripke (1979) asks us to imagine the case of Pierre a native French speaker, who, having heard about London whi Pierre later moves to a borough of London, which he finds unattractive and eventually picks up enough English to dissent from is prett does not realize that London is the same city he heard about Kripke insists there is a genuine puzzle here if we accept (D), because it would s eem to follow that Pierre has a pair of contradictory beliefs about London: that London is pretty and not pretty. semantic, we may have good reasons for rejecting (D). On a non puzzle may provide counterexamples to (D). The suggestion that a content clause like presupposes that the content. However, since the non semantic approach is neutral with respect to semantic content, this need not be the case, and, arguably, is not the case Intuitively, when one
49 mind. But suppose names are Millian, i.e., the meaning of a name is exhausted by its ess an a posteriori arguably provide counterexamples to (D).
50 CHAPTER 6 ALTERNATIVE NON SEMANTIC SOLUTIONS Before closing, I want to mention a few solutions which I thin k are similar in spirit to the meta representational solution. Like (R), t hese are non semantic theories They seek to understand differences in cognitive significance by recognizing and explaining ent expressed by the linguistic vehicle. Nathan Salmon (1986) proposed a way of explaining differences in cognitive significance by treating belief as a ternary relation between believers, propositions, and a third relatum that varies according to the wa y in which believers may be familiar with a particular proposition. For example, when we say that a speaker does not believe that Hesperus = Phosphorus (though she believes that Hesperus = Hesperus), we mean to assert a relation between the speaker, the pr oposition that Hesperus = Phosophorus, speaker does not believe that Hesperus = Phosphorus w hen she understands this information in the way she does when it is presented to her through the sentence does not strike me as an ideal solution for a few reasons. First, the account is based on singular propositi ons. That is, it treats the propositional content expressed by a sentence containing a proper name as partly consisting of the referent Sorcrates and the property being wise) I think Salmon makes a fairly compelling case for thinking of propositions in this way, but it seems to me, as I hope is clear by this
51 point, that we can explain cognitive significance without taking on board a commitment to a specific theory of pr opositions, (or a commitment to propositions for that matter). Second, it is x relatum, as Salmon (ibid. p. 120), himself, points out. x such that A, p, and x stand in [this] relation, what is the extra something x ? Is it a way of taking the proposition? Is it a mode of presentation of the proposition? Is it What sort of thing is it, a nd how are such things individuated? Salmon leaves the question open, admitting that, on his account, part of the puzzle remains. I am inclined to think that the represetational view provides an answer here. It seems to me that representations as I have characterized them, would fit the bill. But then I think representations can do the work without reconstruing belief as a three place relation in fact, without any presupposition about the nature of belief. s singular propositions, empty names present an obstacle to a general account of cognitive significance. Consider a competent English speaker who assigns different cognitive values to the sentences ain a case like this on proposition. For an identity sentence containing a non empty name (or names), the singular proposition expressed contains the r eferent of the name( s) but in cases like this, where the names do not refer, no singular proposition is expressed thus, one of the relata is missing. character as the third relatum. Character is a function, as sociated with expressions by convention, which takes contextual elements, such as speaker, place, and time as arguments and yields
52 difficulty. It is questionable whether cha racters are fine grained enough to handle all the problem cases. If characters are associated with expressions by convention then it should follow that different speakers will stand in the same belief relation to the proposition expressed by a sentence, bu t there appear to be counterexamples to this, as the problem of variance suggests. Intuitively, Pythagoras and Copernicus did not d different properties with the referent (Pythagoras believed it was a star; Copernicus believed it was a planet). Another way to understand substitution failure in attitude contexts is to explain it as the result of differences in standard conversationa l implacata. Kirk Ludwig (1996) imilar position, associating implicate something informative even if the sentence, itself, does not convey this ontent that he calls the the content can be represented by definite descriptions, which may or may not denote an object so described. 22 This so lution is more in line with (R) insofar as it makes no assumptions about what exactly the speaker has in mind. Both accounts avoid the 22
53 variance problem, which seems to raise difficulties for the previous non semantic solutions.
54 CHAPTER 7 CONCLUDING REMARKS Cognitive significa nce is a property of linguistic expressions that can vary as a function of their informativeness.The puzzling cases are the ones in which coextensional expressions differ in this way. But there are two ways of understanding this. Do the expressions differ because they differ in semantic content, or is the difference merely the result of speakers drawing different inferences from them? Frege (ibid) distinguished between the sense of an expression and the image a particular individual associates with the refe rent. Someone observes the moon through a telescope. The moon is comparable with the referent; it is the object of observation which is mediated through the real image projected by the object lens into the interior of the telescope, and through the retina l image of the observer. The first may be comapred with the sense, the second with the presentation (or image in the psychological sense). The real image inside the telescope, however, is relative; it depends upon the standpoint, yet, it is objective in th at it can serve several observers. Arrangements could be made such that several observers could utilize it. But every one of them would have only his own retinal image. Because of the different structures of the eyes not even geometrical congruance could b e attained; a real conicidence would in any case be impossible (p. 201). The analogy is intended to show that the sense of an expression, unlike the associated image, has an objective character, which would help explain how it could be that two or more ind ividuals could come to associate the same sense with the same expression. This is a key distinction for Frege, and anyone trying to develop a semantic solution to the puzzle, because it avoids the view that the meanings of our expressions are speaker relat ive The problem, however, is that this distinction opens the door to a host of other puzzles and problems. For one, it requires a commitment to semantic content that is incompatible with the Millian view of proper names, so the semantic theorist must deve
55 semantic and modal objections pose considerable problems for any semantic theory, not just description theories. In addition, there is the problem of variance, which seems t o undermine the distinction Frege wanted to draw; if differences in cognitive significance can vary between speakers, which seems undeniable, and these differences are the direct result of differences in semantic content, then it seems the conclusion that meaning varies between speakers is unavoidable. Perhaps the biggest I am right, there are certain assumptions about referring terms that we must make in order to constru semantic approach. For Frege, there was the subjective mental content an individual associates with the object, and some objective semantic content generally associated with the expr ession used to pick out the object. He then appealed to the latter, the semantic content (senses), to explain differences in cognitive significance. In my view, this is a mistake for two reasons. First, it runs into the aformentioned problems, which seem t o me insurmountable. Second, it is not clear that we need to draw the distinction in the or, at least, like them in relevant respects in away the puzzles that arise for coextensional sentences without making additional assumptions about the semantic content of the sentences or expressions involved. The subjective content can do the work, and with fewer complications. 23 Perhaps it is 23 I want to stress here that I do not ineffable or inaccessible to anyone else; it is subjective merely in the sense that it occurs in the mind of the individual. Representations, while subjective in this sense, are nonetheless objective in the sense that different speakers may have representations tha t share the same properties.
56 assum ed by the semantic theorist that, because we are dealing with linguistic items, the problem calls for a semantic solution. But this does not follow, and, as the examples we have looked at suggest, semantic analyses are not fine grained enough to account fo r differences in cognitive significance which meets desiderata (I IV), avoids each of the objections to the other solutions we covered and does so without additional sema ntic commitments.
57 LIST OF REFERENCES Ayer, A. J. 1952. Language Truth & Logic New York: Dover Publications, Inc. Bealer, G. 1993. A solution to Frege's Puzzle. Philosophical Perspectives 7: 17 60. Beaney, M. 1997. The Frege Reader Oxford: Blackwel l. Bertolet, R. 1984. Reference, Fiction, and Fictions. Synthes e 60: 413 437. Black, M. 1952. The Identity of Indiscernibles. Mind 61 : 153 164. Biro, J. The Neo Fregean Argument. In (eds.) Biro and Kotako, 1995: Frege: Sense and Reference: One Hundred Years Later New York: Springer. at the Karlovy Vary Symposium on Meaning, Czech Republic. Burge, T. 1979. Individualism and the Mental. I n (eds.) Pessin & Goldber g, 1996: The Twin Earth Chronicles 125 142, Armonk, NY: M. E. Sharpe. Donnellan, K. 1983. Kripke and Putnam on Natural Kind Terms I n (eds.) Ginet & Shoemaker Knowledge and Mind Oxford University Press, New York. Frege, G. 1892. On Sense and Reference I n A.P. Martinich (ed.) 2001: The Philosophy of Language New York: Oxford University Press. Jubien, M. 2009. Possibility Oxford University Press. Kaplan, D. 1977. Demonstratives I n (eds.) Almog, Perry & Wettstein, Themes From Kaplan Oxford Univer sity Press. Kaplan, D. 1978. Dthat I n (ed.) Cole 1978: Syntax and Semantics 9: Pragmatics, 221 53, New York: Academic. Kripke, S. 1979. A Puzzle about Belief In (eds.) Salmon & Soames, 1988: Propositions and Attitudes 102 48, Oxford: Oxford University Press. Kripke, S. 1979. In (eds.) French, Uehling & Wettstein, Contemporary Perspectives in the Philosophy of Language Minneapolis: University of Minnesota Press, 1979. Kripke, S. 1980. Naming and Necessity Cambridge, MA: Harvard University Press. Lepore, E. and Ludwig, K. 2000. The Semantics and Pragmatics of Complex Demonstratives Mind 109 : 199 240.
58 Lewis, D. 1979. Attitudes De Dicto and De Re Philosophical Review 87 : 513 543. Lewis, D. 1981. What P uzzling Pierre Believes Australian Journal of Philosophy 59: 283 289. Ludwig, K. 1996. Singular Thought an d the Cartesian Theory of Mind. Nous 30 : 434 460. May, R. 2006a. Frege on Indexicals Philosophical Review 115 : 487 516. May, R. 2006b. The Inva riance of Sense The Journal of Philosophy 103: 111 144. Mates, B. 1950. Synonymity. I n (ed.) Linsky, 1952: Semantics and the Philosophy of Language University of Illinois press. McGinn, C. 1981. The Mechanism of Reference Synthese 49: 157 86. Parsons T. 1982. Fregean Theories of Fictional Objects Topoi 1 : 81 87. Perry, J. 1977. Frege on Demonstratives. Philosophical Review 86 : 474 497. Perry, J. 2001. Reference and Reflexivity Stanford: CLSI Publications Putnam, H. 1954. Synonymity, and the Ana lysis of Belief Sentences Analysis 14 : 114 122. Putnam, H. 1975. I n (eds.) Pessin & Goldberg, 1996: The Twin Earth Chronicles 3 53, Armonk, NY: M. E. Sharpe. Quine, W. V. O. 1953. From a Logical Point of View New York: Harper a nd Row. Ray, G. 2005. On the Independence of What Our Thoughts are About. Presented at LOGOS Origins of Reference Conference, Barcelona. Richard, M. 1983. Direct Refe rence and Ascriptions of Belief. Journal of Philosophical Logic 12 : 425 452. Russell, B 1905. On Denoting Mind 14: 479 493. Russell, B. 1931. The Problems of Philosophy New York: Oxford University Press. Salmon, N. 1981. Reference and Essence Princeton University Press. Salmon, N. 1986. Cambridge: The MIT Press. Sear le, J. 1983. Intentionality Cambridge University Press.
59 Smiley, T. 1981. Frege and Russell. Epistemologica 4: 53 8. Soames, S. 2003. Philosophical Analysis in the Twentieth Century Princeton, NJ: Princeton University Press. Soames, S. 2005. Reference and Description : The Case Against Two Dimensionalism Princeton, NJ: Princeton University Press. Tye, M. 1978. The Puzzle of Hesperus and Phosphor us. Australasian Journal of Philosophy 56: 219 224. Wiggins, D. orning Star and th e Evening Star. In (ed.) Schirn, 1976: Studies on Frege II: Logic and the Philosophy of Language Stuttgart: Bad Canstatt. Wittgenstein, L. 1922. Tractatus Logico Philosophicus New York: Routledge.
60 BIOGRAPHICAL SKETCH Shawn Burtoft received his Master s of Arts in philosophy in 2007 and his Doctor of Philosophy in 2010 at the University of Florida. His general areas of interest are philosophy of language and metaphysics. In particular, he is interested in reference, time, and ontology.