Brendan Lalor. Erkenntnis 47, 67-88, 1997 (pre-publication version)
Abstract. Kaplan (1977) proposes a neo-Fregean theory of demonstratives which, despite its departure from a certain problematic Fregean thesis, I argue, ultimately founders on account of its failure to give up the Fregean desideratum of a semantic theory that it provide an account of cognitive significance. I explain why Kaplan’s (1989) afterthoughts don’t remedy this defect. Finally, I sketch an alternative nonsolipsistic picture of demonstrative reference which idealizes away from an agent’s narrowly characterizable psychological state, and instead relies on the robust multiply realizable relation between the skilled agent and demonstrated object.
When the unclean spirit has gone out of a person, it wanders through waterless regions looking for a resting place, but it finds none. Then it says, ‘I will return to my house from which I came.’ When it comes, it finds it empty, swept, and put in order. Then it goes and brings along seven other spirits more evil than itself, and they enter and live there; and the last state of that person is worse than the first. So will it be also with this evil generation. (attributed to Jesus in Matthew 12:43-45, New Revised Standard Version of the Bible)
When you clean house, do it right the first time — or matters may eventually get worse. In the 1970s some truth spread among Fregeans, promulgated perhaps most notably by David Kaplan (1977). His criticisms challenged some core beliefs of Fregean semantical theorists, and he exorcised numerous Fregean demons from semantics. Still a Fregean at heart, however, he failed to recognize the corrupting nature of a key presupposition of Frege’s framework — that semantics must account for the ‘cognitive significance’ of language. Nor do his 1989 ‘Afterthoughts’ constitute a break with this fundamental assumption. In this paper, I first explain Kaplan’s allegiance to that presupposition; then I indicate why even his improvements of Frege’s theory do not satisfy what the presupposition requires. Finally, I explain why this presupposition is mistaken, and point in what I take to be a more promising direction which would help secure final exorcism.
On the Fregean conception, semanticists must account for the informativeness of sentences in which coreferring expressions flank the identity sign — that is, how it is that such expressions can differ in cognitive significance or meaning, from the standpoint of those who understand them. As Frege (1892a) put it, ”a = a’ and ‘a = b’ are sentences of obviously different cognitive significance’ (85). That is, the attitudes one holds toward them may well differ. Importantly, one might believe one and deny the other, and therefore behave differently depending on which of them one accepts. Hence, Frege held that the semantic accounts of the expressions ‘a’ and ‘b’ must differ, which shows that, contra the Millian account of proper names, the referents of expressions cannot exhaust their meanings. ‘If the sign ‘a’ differs from the sign ‘b’ … not in the manner in which it designates anything, then the cognitive significance of ‘a = a’ would be essentially the same as that of ‘a = b’, if ‘a = b’ is true’ (86, emphasis added). But their cognitive significance is not the same. Frege thus concluded that in such sentences the expressions ‘a’ and ‘b’ must present the referent in different manners, via different concepts, or Sinne (senses), with which the expressions are associated (86). Thus, loose talk about the ‘meaning’ of an expression is ambiguous between referring to the object or state of affairs picked out by the expression (i.e. the reference), and the manner in which it is picked out and presented (i.e. the sense).
It is worth noting that the phenomenon of intensionality here discussed is not unique to identity statements, and the same morals can be drawn from any cases in which substitutivity of coreferring singular terms fails — for example, in propositional attitude contexts. In general, Frege holds two expressions differ in cognitive significance just in case someone might affirm sentences involving the one while denying corresponding sentences involving the other. For example, someone might accept, ‘The current president of the United States is a saxophonist,’ but deny the corresponding sentence, ‘The former governor of Arkansas is a saxophonist.’ Thus, the expressions ‘The current president of the United States’ and ‘The former governor of Arkansas’ differ in cognitive significance, in spite of their common reference.
On this view, an expression becomes meaningful by being associated with a Sinn, which in turn both determines a referent, and is that which exhibits that referent via a cognitively significant mode of presentation. One core belief of Fregeanism is that it is only by means of such conceptual mediation that the intentionality of language works — only by this means that expressions can be about things. For example, it is only because the expression ‘the inventor of bifocals’ is associated with a Sinn which does the work of picking out Ben Franklin that I refer to him by using that expression. I’ll call this the Fregean Intentionality Thesis (FIT). Another core belief is that expressions which differ in cognitive significance do so because they are associated with different Sinne. Although the expression ‘the author of Poor Richard’s Almanack‘ also picks out Ben Franklin, the manner in which it presents him differs since it is associated with a different Sinn, and picks him out via a different conceptual route. Thus do the expressions differ in cognitive value. This I’ll call the Cognitive Significance Thesis (CST).
I now turn to a discussion of Kaplan’s rejection of FIT and inclination to accept CST. I will ultimately argue that both theses ought to be exorcised from semantics.
1. Cleaning house Kaplan-style
1.1. The Fregean Intentionality Thesis
Direct reference theorists, most centrally Kaplan, have shown, against FIT, that mediation via a Sinn is not a necessary condition of linguistic reference (cf. e.g. Kaplan 1989: 568). Consider a rather noncontroversial case in which an object becomes the topic of discourse without such conceptual determination. While waiting to use a fitting room in a clothing store, Bert peers at the feet of the current occupant under the stall door. He says to himself, ‘That man has a nice pair of shoes.’ His expression is about that man (call him Ernie), but not in virtue of Bert’s ability to conceptually individuate him, as FIT requires. It seems, rather, that Bert’s expression is about Ernie in virtue of Bert’s perceptual relatedness to Ernie. Thus, the expression ‘that man’ here does not refer to Ernie in virtue of association with a Sinn which picks him out uniquely, and in this sense, ‘that man’ refers to Ernie rather directly. For Kaplan, he himself becomes a constituent of the proposition Bert expressed. As Kaplan might say, the expression so refers because its use is governed by rules which specify that the referent is some indicated man in the context of utterance. Ernie happened to be that man in the context in question. Prima facie, this sort of example refutes what was said of Fregean Sinn, namely, that Sinn-as-manner-of-presentation is in all cases that which also fixes the referent as one thing rather than another. In the case here considered, the linguistic manner of presentation of Ernie to Bert via the expression, ‘that man’, is not what fixes Ernie as the referent at all.
Kaplan holds that although Frege was right that there are at least two kinds of meaning, he nevertheless confused two, more specific notions of meaning, in his overly general, over-burdened notion of Sinn. They can be disentangled by distinguishing an expression’s character, or linguistic meaning, from its content, which is what it expresses (cf. e.g. 1977: 523; 1989: 568). Character is expressible in terms of content-determining linguistic rules. It accounts for the manner of presentation, but is often not itself sufficient to fix reference. So it is not the kind of meaning that is truth-evaluable. As a function, it assigns the content of an expression, given a context. Content, however, is truth-evaluable. Therefore, a claim’s content is the best approximation to the traditional bearer of truth and falsity, the proposition expressed by the claim. As a function, content assigns extensions of terms and truth values of claims, for each counterfactual situation. Suppose Ernie says, ‘I’m hungry.’ The character of ‘I’ is expressible roughly by the general rule that the one referred to is the utterer. That rule, taken in the context in which Ernie utters ‘I’, fixes him as the content.
The distinction is most clearly relevant in the case of directly referential terms. To briefly extend one of Kaplan’s examples, even someone locked in the trunk of a moving car, disoriented upon awakening, and devoid of conceptual knowledge sufficient to individuate his or her spatiotemporal location, can refer to the current time and place by saying, ‘It is awful to be here now.’ The actual time of the utterance and the actual place of the utterance themselves enter as constituents into the proposition expressed. As Kaplan (1977) put it: ‘Ignorance of the referent does not defeat the directly referential character of indexicals’ (536). In the example, the expression ‘now’ is not associated by the speaker with a Sinn, such as the concept, January-2nd,-1996,-5:41:32-PM, which conceptually introduces that particular moment into the content. Rather, given the context, the character of the expression ‘now’ determines which time is in question, regardless of whether the utterer could identify that time by other means. The character of the expression ‘now’ is not part of what is said, i.e. the content, or what gets evaluated with respect to actual and counterfactual contexts (Kaplan 1977: 497f., 500, 520, 523). It is a matter of semantical significance which does not show up in content — part of what might be called ‘prepropositional semantics’ (1989: 582f.). It is ‘off the record,’ from the standpoint of truth-evaluable meaning.
This militates against FIT because it underscores a way in which certain expressions refer without being associated with an individuating Sinn. At least in cases of direct reference, these considerations show that the requirements of being linguistically competent are far less stringent than Fregeans think — because speakers do not need to associate their expressions with complete senses in order to be competent.
1.2. The Cognitive Significance Thesis
While Kaplan has effectively called into question the generality of FIT, he strongly sympathizes with the intuitions behind CST, and attempts to press his improved version of Frege’s theory into service to account for cognitive significance. Frege was right, Kaplan and his followers hold, to reject semantical accounts of an expression on which what the expression ‘means’ is an object which itself simply enters into propositions without any mediation. They agree that such accounts are insensitive to the observations about cognitive significance noted above in connection with the phenomenon of intensionality. However, they think we can respect CST and still allow an expression’s content to involve objects and states of affairs directly — as indexicals seem to — by offering an account on which the expression’s cognitive significance is explained in terms of its character.
The character, expressible as a rule which fixes the content in a context, is the primary meaning of the indexical (1989: 574). Since different characters can present the same content variously, in this sense, a referent is presentable via different meanings (cf. 1977: 524). This accounts for why character might be thought to explain cognitive significance. Consider an example. Suppose Ernie and Bert are identical twins raised under qualitatively identical conditions (cf. Kaplan 1977: 531 — and notice that if their upbringings are to be qualitatively identical, neither will be able to refer to the other by name so long as their names differ; so be it). Searching for one another in a symmetrical game not unlike hide-and-seek, they eventually meet, and each exclaims to the other, ‘I finally found you!’ Although the contents of their expressions differ (Ernie asserts that Ernie finally found Bert, but Bert asserts that Bert finally found Ernie), the characters are identical (each asserts that the speaker finally found the other person). In this latter sense, what they understand by knowing how to use those words is the same. Hence the isomorphic cognitive significance. If, however, for the first time their behaviors diverged and Bert instead said, ‘You finally found me!,’ the contents of their expressions would be the same while their characters differed. This difference in character correlates with a difference in cognitive significance. Now, not only is what each knows about himself by knowing that content different, but how it is known — its manner of presentation — differs: Ernie asserts that he himself has done the finding while Bert asserts that he himself has been found.
So far, it seems that we fix cognitive significance by fixing character, and likewise, we vary it by varying character. This sort of consideration is taken by Kaplan and other neo-Fregeans as a basis for the hope, and sometimes the claim, that cognitive significance varies counterfactually with character, and therefore differences in cognitive significance of coreferential expressions are explicable in terms of differences in character.[ 1 ]
Kaplan (1977) originally thought that what distinguishes a demonstrative, such as ‘that’, from pure indexicals such as ‘I’, ‘now’, and ‘here’, is that while facts about the context of utterance determine the reference of both, only the former additionally require something else for completion. A demonstrative requires a demonstration, such as an act of pointing, to determine a referent — while a pure indexical such as ‘today’ requires only that the value of a certain temporal parameter be given to determine the day referred to. He held that differences in demonstration explain differences in cognitive significance. Consider some cases in which this seems to be so. Suppose Ernie is victim to an accident and is lying in the back yard with an arrow through his head. Upon finding Ernie, Bert nervously wonders aloud, pointing first at the arrow fragment extending out from the left side of Ernie’s head and then to the one on the right, ‘Oh my! Is that arrow that arrow?’ (hoping this is just the old arrow-through-the-head trick). Notice that each occurrence of ‘that arrow’ has both the same linguistic character and the same content. So why the cognitive uncertainty if every kind of meaning is identical? In response to this kind of question, Kaplan points out that any two demonstrations might have demonstrated different objects. (This Kaplan took over from Frege’s theory of demonstrations.) Indeed, suppose Ernie did play the old arrow-through-the-head trick on Bert. Bert instead comes upon a scene qualitatively identical to the one described above, but this time, Ernie is happily playing dead, barely able to control his laughter since he hears Bert approaching. To Ernie’s delight, Bert issues the same nervous query regarding the two pieces of arrow.
Kaplan holds that Bert’s queries in both counterfactual situations involve the very same demonstratives and demonstrations. This can be so because demonstrations, like characters, are separable from their contexts. How so? Demonstrations are types, so ‘it is not an essential property of a given demonstration … that it demonstrate a given individual …’ (525). We can consider what referent an occurrence of ‘that’, along with its completing demonstration, would pick out in other contexts. While a demonstration’s actual time, place, and agent are inessential, what is essential is its manner of presenting a referent from a certain perspective: It presents its referent ‘as the individual that looks thusly from here now‘ (525). Kaplan holds that there is a form of demonstrations, definable at least largely in terms of a narrow psychological state characterization: Their form is, ‘The individual that has appearance A from here now,’ where an ‘appearance’ is ‘something like a picture with a little arrow pointing to the relevant subject’ (526).[ 2 ] By fixing the perspective we determine a content for each context, that is, what, if anything, would be demonstrated in various contexts, given that perspective. Since a demonstration in a context determines a content, Kaplan holds that we can associate with each demonstration a different character (527).
In general, then, how is it that we might consistently doubt an utterance of the form, ‘that1 is that2,’ even if it is true? ‘That1’ and ‘that2’ differ in cognitive significance because their different associated demonstrations have different characters — present their referents in different manners. Suppose we didn’t allow difference in demonstration to account for difference in cognitive significance. If ‘that1’ and ‘that2’ pick out the same referent via the same linguistic character (the character being very roughly, something like, ‘the object demonstrated in the context’), it would seem that the denial of the claim, ‘that1 is that2,’ is irrational or at least somehow inconsistent. After all, it asserts what it does by referring directly to the same object twice via the same character, and identifying the referent of the first with that of the second. But irrationality need not follow if part of the cognitive significance is accounted for by the different demonstrations. Bert might have come on the scene of the real accident thinking that Ernie was playing the old arrow-through-the-head trick, and falsely declare, ‘That1 is not that2!’ The view that differences in cognitive significance can be accounted for as differences in character breaks down here for an obvious reason: The linguistic character of the demonstrative alone does not suffice to account for the difference in cognitive significance between ‘that1’ and ‘that2’. So Kaplan concludes it is explained by something else about the demonstrative situation, namely, the demonstration-types themselves.[ 3 ]
If demonstrations are thusly formalizable, we might consider whether they are well-behaved enough, logically speaking, to be included in a logic of demonstratives. This is just what Kaplan has done. He notes that there are many similarities between the above-mentioned facts about demonstratives and his device, ‘dthat’, which converts singular terms such as definite descriptions into directly referential expressions. For instance, an actual-world utterance of ‘dthat [the inventor of bifocals]’ directly refers to Benjamin Franklin, even in counterfactual situations in which someone else is the inventor of bifocals. While demonstratives and dthat expressions are directly referential, both the demonstrations associated with demonstrative occurrences and the descriptions associated with occurrences of ‘dthat’ simply fix reference without becoming part of truth-evaluable content. Kaplan is so impressed by this parallel that he declares that ‘dthat’ represents ‘the general case of a demonstrative,’ and that we should add to the syntax of his formal logic of demonstratives (LD) a category of ‘nonlogical demonstration constants,’ analogous to the singular terms associated with uses of ‘dthat’ (527).
Note that the demonstration is no longer what we might have intuitively thought it would be — i.e. the performance of the pointing act. On Kaplan’s view, it is a certain kind of narrowly specifiable mental state through which an apprehender grasps a referent. Nevertheless, if we really can isolate a systematically semantically significant component of a total cognitive state associated with a demonstrative (the component being a demonstration), then we can hold that the character of that component, plus the linguistic character of the expression, fixes the cognitive significance of the demonstrative expression. This would enable us to explain the cognitive significance of such expressions in a Frege-like way — as differing just in case someone might affirm a sentence involving one such expression while denying a corresponding sentence involving another. The implicit premiss in this argument is that if the characters of the expressions and demonstrations are the same in such a set of corresponding sentences, then no reasonable person would affirm one while denying the other sentence. But is this picture right? I shall argue that it is not.
2. The return of the Fregean spirits
Kaplan’s earlier improvements on the Fregean theory are not radical enough to keep problematic Fregean demons at bay. Rejecting FIT is not enough. The CST-derived challenge which has arisen in numerous forms above is that for a semantical theory to succeed as a Fregean semantical theory, a way must be found to individuate characters finely enough to deny inappropriate substitutions into propositional attitude contexts, and broadly enough to permit the right ones. For example, we have seen that if character is to account for cognitive significance, the character of the first ‘that’ in the expression, ‘That arrow is the same as that arrow!,’ cannot be the same as that of the second. I shall argue that we have reasons to doubt that character can be adequately individuated in the case of demonstratives. I have already expressed hesitancy about its individuation even in the case of at least one pure indexical (see footnote 3). In other discussions, others have found that related difficulties attend attempts to explain the cognitive significance of proper names in terms of character (cf. e.g. Ackerman 1989: esp. p. 21), and also, famously problematic are natural kind terms (cf. e.g. Burge 1982), and terms whose meanings depend on convention (cf. e.g. Burge 1979).
I suggest that these difficulties are symptomatic of problems with the presuppositions of even Kaplan’s improved Fregean theory. While his story could turn out to be a good account of the semantics of some indexicals, such as ‘now’, I have a serious two-part doubt about its generality. The first part pertains to the propriety of the functional identification above between demonstrations and singular terms associated with ‘dthat’. What is presupposed in Kaplan’s anointing of his device, ‘dthat’, as ‘the general case of a demonstrative’? The idealized characterization of the behaviors of demonstrations and singular terms associated with ‘dthat’, as depicted in a formal system such as LD, obscures the significantly different ways in which reference works in their respective cases. Demonstratives work extra-linguistically and descriptive singular expressions linguistically. Wittgenstein (1953) showed us that linguistic competence often rests on extra-linguistic skills (Part I: sections 10, 33, 454). Knowing what to make of an act of pointing requires mastery of a set of skills which could easily astound alien anthropologists studying humans, since ‘we cannot specify any one bodily action which we call pointing to the shape (as opposed, for example, to the color)…’ (Wittgenstein 1953: Part I, section 36). Nor can we specify an exhaustive set of such actions. Ernie points to a vase and says, ‘That is beautiful.’ It is implausible to think that the character of the expression ‘that’, plus the character of the demonstration, will determine that the content of what he said was that the vase’s shape rather than color is beautiful. Thus, the determination of what he said — the content — is foisted on the extra-linguistic (or maybe super-linguistic?) skills involved in determining what the referent is, rather than on the character of the expression.
While there are not obvious parallels between such demonstrations and singular terms as used in association with ‘dthat’, it might be suggested that we have a phenomenon appropriately related to demonstrations in what Keith Donnellan (1966) calls the referential use of a description. Suppose Ernie and Bert are at a party, and, looking at a man with a martini glass, Ernie asks Bert, ‘Who’s the guy with the martini?’ Donnellan points out that even if it should turn out that there is only water in the glass, Ernie would still have succeeded in referring to the man with the martini glass — the one to whom he meant to refer. Here, the description was not used to refer to whoever did have the martini, which Donnellan calls an attributive use. It was used referentially, that is, to draw attention to the right person, which was determined, in this case, perceptually. Just as a demonstration completes a demonstrative, so the perception here ‘completes’ the referential use. Perhaps this affinity will help justify some kind of functional identification of demonstrations and dthat descriptions. But descriptions used referentially are not the sort usually associated with uses of ‘dthat’ by Kaplan.[ 4 ] Anyway, Donnellan’s referential use, like Kaplan’s account of how demonstrative reference works, is pestered by the fact that it too irreducibly relies on extra-linguistic skills. Thus, I suggest that the difficulties I will point out below for Kaplan raise similar questions about Donnellan’s notion.
Someone may accept the extra-linguistic / linguistic difference and hold that even though the means of fixing reference might differ nothing substantive is obscured or misrepresented as a result of treating dthat descriptions and demonstrations as functionally isomorphic. That is, the introduction of the nonlogical demonstration constants into the syntax of LD does not too quickly gloss any important details. After all, from the standpoint of truth-evaluable semantic content, such details are, as we said, off the record.[ 5 ] However, I claim the idea that we can introduce such constants into LD rests on a mistake. This brings us to the second phase of my doubt about this type of semantical theory.
Since the introduction of such constants presupposes a systematic relationship between the character of demonstrations and contents, it implies that the relation between perceptual perspectives (defined in terms of narrow psychological states) and objects which those perspectives could be of (if the context were suitable) is specifiable and systematic. To begin with, although this is not itself an argument, it is worth noting that cataloguing these relations, even in ostensibly simple cases, is intractable. Kaplan himself notes that ‘perhaps we should say that a single performance may involve distinct demonstrations from the perspective of distinct audiences’ (515). This is another point of disanalogy between dthat descriptions and demonstrations — demonstrations turn out to be far more fine-grained than dthat descriptions, which are not generally thought of as perspectival. Indeed, in principle, there may be as many demonstrations associated with a given demonstrative as there are perspectives on an event — technically, an infinite number. But, however impractical, someone might say, there still is a principled logic of demonstratives.
Two kinds of consideration suggest that there is not a systematic relationship to be found between the character of demonstrations and contents, that is, between narrowly defined perceptual perspectives and objects which those perspectives could be of. The import of these two sorts of consideration will be seen throughout the rest of the paper. First, the fact that a given referent is demonstrated is multiply realizable. An analogy will cast light on the significance of this. The fact that a certain radio station is received is multiply realizable, since there are innumerably many possible physical architectures which count as instantiations of radios. Thus, that fact does not admit of a reductionistic explanation in terms of local, lower level features idiosyncratic to any particular radio or even set of radios. For every set of possible radio architectures, we can come up with another, slightly different, which also instantiates a radio. Hence, our account must be in terms of causally relevant higher level properties, such as receives-such-and-such-a-range- of-radio-waves-of-modulated-frequency. Likewise, the fact that a certain object is demonstrated is multiply realizable, and thus not reductionistically explicable. For every purportedly exhaustive set of possible ways any given object could be demonstrated, we can come up with another way. So our account of it ought to be in terms of higher level properties such as stands-in-such-and-such-an-intentional-relation. There just is no adequate non-intentional characterization of demonstratives which maps demonstrations in a context onto contents. Of course, a demonstrative in a context usually will determine a content; what I am denying is that there is a function which determines contents on the basis of non-intentional features of the speaker reifiable as character (or, I wager, any specified set of contextual features of demonstrative situations in general), as Kaplan thinks there is.[ 6 ]
Second, what a given demonstration demonstrates is context dependent. To return to an earlier example, Bert can point to a vase and say, ‘That is beautiful,’ in the same manner on two occasions, and refer to its shape on one occasion, and its color on another — for instance, in the context of a discussion with Ernie about favorite colors. It must be that what Bert refers to depends on features of the act other than the utterance and demonstration components. Thus, we can fix character and vary cognitive significance. Or consider a science fictiony example. Ernie is showing off his new invention, which unnoticeably vaporizes whatever object is on a certain platform and replaces it with a qualitatively identical holographic projection. He says proudly, pointing at the object, ‘Bert, watch. That … is not that,’ where the second ‘that’ refers to the hologram. Although linguistic character and perceptual perspective are both fixed here, cognitive significance varies because Ernie and Bert know what the machine does.
On the basis of these reflections about the multiple realizability and context dependence of demonstrations, I suggest that no interesting rule-governed systematicities between demonstratives and contents emerge at the level of organization at which demonstrations are conceived as narrowly characterizable psychological states. That is, there is not a genuine function from demonstrative character to content. The factors that go together to determine content are largely idiosyncratic to particular occasions of use.[ 7 ] If this is right, the demonstration is not the key to content determination; it may even be ‘noise’ from the standpoint of semantics. Thus, the attempt to give a general account of cognitive significance in terms of a prepropositional semantics of character will not be generally successful. As indicated above, this suggests that we have to ‘go meta’ with respect to the level of organization in question and appeal to higher-level facts (such as the relation between the utterer of the demonstrative and the object the utterer is directed at) in order to come up with an account which captures what we want.
This applies not just to demonstratives, but also to the pure indexicals ‘I’ and ‘you’. Recall the identical twins case discussed earlier in which the twins were playing hide-and-seek. Suppose we now change the example so that Ernie and Bert are fraternal twins raised under qualitatively disparate conditions. The case becomes far less intuitively satisfying. Imagine that while Bert was hiding from Ernie, he encountered their father, who told Bert to tell Ernie he had better clean his room on the double. Eventually, Ernie and Bert meet, and say to one another, ‘I finally found you!’ Ernie playfully emphasizes the ‘I’ and ‘you’ in his utterance, since what is cognitively signified to him is that now Bert is ‘it’ and he himself can go hide. When Bert tokens those same words, however, he is annoyed at how long the search lasted, and thus emphasizes the ‘finally’. What is signified to him is that now he must tell Ernie that Dad’s looking for him, and he’s in big trouble. So it seems that as soon as their general intentions vary, the cognitive significance of expressions varies, even if they have the same linguistic character.[ 8 ]
One might respond that all this shows is that we need a finer-grained notion of character to explain the variance in cognitive significance.[ 9 ] On this view, however, the cognitive significance of expressions is no longer explained in the manner of Frege and Kaplan, namely, as differing just in case someone might affirm a sentence involving one expression while denying a corresponding sentence involving another. For whenever there is a difference in total cognitive state, the ‘cognitive equivalence’ of uttered sentences cannot be used as a predictor of behavior. Specifically, someone might affirm one and deny another ‘cognitively equivalent’ sentence if his or her total cognitive state varies.
On this non-Fregean criterion the supposed cognitive equivalence (or nonequivalence) of utterances seems to become a matter to be investigated by cognitive psychologists, and the like. While such entanglement with detailed psychological structure is something Fregeans might have hoped to avoid,[ 10 ] such a finer-grained conception of cognitive significance is not obviously out of line with Kaplan’s approach. But I think the holistic approach to narrow meaning envisioned by this seemingly weak and accommodating version of CST is more problematic than it may at first appear.
My worry is that such an approach amounts either to giving up the idea that cognitive significance can be accounted for in terms of linguistic meaning — character — at all, or to individuating characters so finely that we lose our ability to generalize about them at all. The first alternative constitutes explicitly giving up CST. To avoid this, one might be driven to accept the second, on which any differences in cognitive role must be equated with differences in character. But allowing all manner of cognitive idiosyncrasies to show up in character is remarkably like having no theory of character at all. While character was originally advertised as shared linguistic meaning, what’s being sold here is a view on which there is scarcely any constancy in character across two uses of the same word by the same person, much less across my use and your use. If we accept it that every instance of ‘I’, ‘you’, and ‘that’ has a different character due to differences in total cognitive state, we have to accept both that character is no longer shared linguistic meaning, and that we have no philosophical theory about the meanings of such expressions here.[ 11 ]
Prima facie, if an account of demonstratives — much less, of pure indexicals — allows only perfectly identical twins to actually share character, then something is wrong with the account. Relatively cognitively similar agents — like, for example, you and I — present when an assertion is made using a demonstrative had better be able to entertain the same character. A theory that fails to allow this must fail to capture the sameness of the generalizations true of the thoughts and actions of two such agents. Likewise for relations between thoughts and other thoughts. Such a theory fails to capture agents under the intentional characterization.
However, it might be thought that Kaplan’s more recent work side-steps many of the criticisms levied here.
2.1. Directing intentions
In ‘Afterthoughts,’ Kaplan (1989) expresses his new view that it is not a demonstration, but an utterer’s directing intention, which completes a demonstrative. He regards ‘the demonstration as a mere externalization of this inner intention …. of no semantic significance’ (582). Indeed, if Ernie were blind, and paralysis prevented him from making any pointing gestures whatsoever, he could still use the demonstrative ‘that’ with perfect competence. Although we may not always know to which of two things he refers, that doesn’t prevent him from successfully referring. When he hears footsteps stop at his doorway he may ask, ‘Who is that? Is that Bert?,’ referring to whoever produced the footsteps — all the while remaining perfectly still, apart from the vocalizations. Kaplan’s new view is that this inner intention is the factor that differentiates the prepropositional meaning and cognitive significance of one occurrence of a demonstrative from another (588). He thinks each occurrence of ‘that’ in ‘That1 is that2’ is accompanied by a separate inner intention, which accounts for how each may differ in cognitive significance: The inner-ness of these different intentions seems to automatically explain how (inner) psychological states could differ in cognitive value. This account of how reference is determined in terms of directing intentions may seem to alleviate some of the problems which arose in the discussion of context dependence above, pertaining, for instance, to the vase. I said that when Bert points to the vase and says, ‘That is beautiful,’ something other than the utterance and demonstration must account for whether he referred to its color or its shape. It might be said now that that something is the directing intention: ‘The directing intention is simply focused on the color, not the shape.’
We now turn to the role of the directing intention in the semantics of demonstratives, and beyond. Then we will examine the directing intention account of demonstrative reference and see how it fares in comparison with the demonstration account. I hope to show that it is no better.
Kaplan distinguishes the way a directing intention aims at its object, which is like perceptual focus, from the way in which Donnellan says a speaker may have an object in mind, which need not involve perceptual focus (583).[ 12 ] But Kaplan thinks that often, referential uses of descriptions involve tacit perceptual demonstratives. Rather than ambiguously asking, ‘Who is that?,’ in a room full of people, Ernie substitutes, ‘Who’s the guy with the martini?’ On this analysis, the ‘that’ and its directing intention are there, albeit tacitly.[ 13 ] (Still, Donnellan’s ‘having an object in mind’ need not involve a demonstrative. A tacit demonstrative’s directing intention may pick out an object which is not the one the speaker has in mind, as when I am perceptually focused on the line on a road map labeled I-90, and I say, ‘We’ll drive on that, over to …’ — but I have in mind I-90 itself.)
Kaplan’s analysis of Donnellan’s referential use indicates the great extent to which he thinks his semantics of demonstratives can help explain the connection of language with the world. ‘The theory of direct reference, with its prepropositional semantics, seems especially open to such off-the-record elements in language’ as are implied by the tacit-demonstrative analysis of Donnellan’s referential use (1989: 584). Of course Kaplan thinks that our perceptual relations to the world play a role in the account of how demonstrative reference works. But he has a tendency toward the stronger thesis that the prepropositional semantics of demonstratives, specifically, of directing intentions, plays a role in the account of how language in general connects up with the world. However, in burdening the inner directing intention with such vast work, he has transferred the onus of justification of his theory to the to-be-filled-in-later account of the directing intention.[ 14 ] The problem is that it’s looking like the prepropositional goings on aren’t semantical in any straightforward sense, and don’t admit of systematic generalization by a philosophical theory.
Kaplan himself has reservations about directing intentions, since he rightly fears they will not secure a semantic account of cognitive significance. This is because he has doubts about whether directing intentions are identifiable with character, and thus about finding an aspect of linguistic meaning which will account for the cognitive significance of language. According to the ‘Demonstratives’ account, a given demonstration, since it is a type, is most tenably definable in terms of a narrow psychological state type, and thus might have demonstrated a different demonstratum. If we grant all this, it is plausible to associate characters with demonstrations, although we have seen reasons for doubts about doing so. But it is even more doubtful whether a directing intention of ‘Afterthoughts’ might have determined a different object as the referent of a demonstrative. I think this attempt to separate directing intentions from particular contexts will fail — much in the spirit of my corresponding claim about demonstrations. As Kaplan realizes, if they are not thus separable, the difference in cognitive significance between two coreferential occurrences of ‘that’ will not be explicable in terms of an aspect of linguistic meaning which is transplantable into other contexts, such as character. This would not leave us resources with which to defend any version of CST I can think of. Thus, we would not have a general semantical theory which would enable us to explain the phenomenon of intensionality pointed out by Frege. How bad would that be? Not very bad at all.
3. Toward a final exorcism
The alternative picture I have in mind is motley indeed: When demonstrative reference works smoothly, what’s cognitively signified by an instance of ‘that’ depends at least on the case-relative details about the linguistic, visual, auditory, social, inferential, and other pragmatic skills in virtue of which the relation between the agent and the demonstratum obtains, plus the actual demonstrated thing out in the agent’s environment, plus the linguistic context in which ‘that’ is embedded. An agent’s link to a demonstratum is multiply realizable since the particular pragmatic, perceptual, or other skills which implement it may fluctuate. It is also context-dependent since the efficacy of such skills presupposes some stabilities in linguistic, social, and physical environments. So the agent’s locally supervening states can only be understood by paying attention to the way in which they contribute to securing reference — i.e. by turning our gaze toward the referents in the world. The Kaplanesque conception of directing intentions as features of locally supervenient psychological states won’t suffice to explain the systematic ways in which we behave with respect to demonstrated objects. To do justice to this intentional systematicity, our explanations here need to make specific reference to such objects themselves.
To help see more clearly the true nature of the directing intention, imagine a fairly common type of case: Movement. Suppose Bert and Ernie are sitting in the back of their father’s car on a ride, when Bert sees an unrecognizable roadkill just as the car is passing it. As he sees it, he asks, ‘What’s that?’ And as he slowly speaks, his finger, which at first points out the side window, moves continuously until he’s pointing out the back window, compensating for the car’s movement. This example brings out the usually under-emphasized (or unnoticed) degree to which the relation between speaker and referent is dynamic. The details about the cognitive significance of ‘that’ here fall out of the account of the concrete way Ernie’s skills link him to the roadkill — for instance, the way in which he exercised his ability to maneuver his body appropriately so as to maintain focus. The semantics here, on the other hand, ought to idealize away from these psychological and other details. Ernie’s link with the object — however secured — is what fixes it as the referent.
The point of the roadkill case is in part to highlight the extent to which demonstrative reference depends intimately on agents’ sets of extra-linguistic skills and the state of the demonstratum in the world. Further, given this, if it is right that the mediating processes linking Ernie to the demonstratum are multiply realizable and context dependent, then the details about those processes — for instance, about Ernie’s psychological structure, about which skills he utilizes at which moments — don’t figure systematically into the semantics of demonstratives. Thus, to get off the ground, this semantics will have to idealize away from the details of that structure to a level of organization at which the systematicity does emerge. Consequently, it idealizes away from an account which could be pressed into simultaneously serving as an account of cognitive significance. On this picture, from the standpoint of semantics, all that is important is that Ernie is reliably keyed to the demonstratum — that he is able to sustain the relation between it and himself under various factual and counterfactual changes in, for example, relative location. The details of how this is done are off the record from the standpoint of any semantics.
Kaplan engages in idealization, too — but he over-idealizes in precisely the wrong direction.[ 15 ] He abstracts away from skilled agents in suggesting that we can characterize demonstrative situations such as Ernie’s in terms of relations between occurrences of complete demonstratives and contexts — with no reference to speakers. In ‘Demonstratives,’ Kaplan (1977) said, ‘it does not seem to me to be essential to a demonstration that it be mounted by any agent at all’ (525): All that matters is that it is associated with a demonstrative occurrence in a context. Then in ‘Afterthoughts’ it is assumed that even if we abstract away from the way in which agents interact with their environments, demonstrata can be determined as functions of narrow psychological states and contexts. Both of these ways of idealizing take it for granted that demonstrative reference doesn’t essentially require robustly skillful agents (i.e. agents that use a multiplicity of ways — even strategies — to secure reference). However, contra Kaplan, due to all of the extra-linguistic facts about speakers like Ernie which are essential to their ability to refer, these situations ought to be characterized in terms of the relation between an agent — not just an utterance or even a narrowly specifiable psychological state — and some thing in the environment. Just as I’ve argued that you can’t factor the world out, you can’t factor the agent out either.
So ‘Afterthoughts’ does not constitute a reversal, but just a sophistication. In essentials, the newer theory is not too different from the solipsistic view of demonstrations I criticized earlier, on which demonstrations are like mental pictures definable in terms of narrow psychological states. While demonstrations were represented in LD by nonlogical demonstration constants, here Kaplan has in effect simply replaced them with nonlogical directing intention constants. As he says, ‘We will need to be able to formulate sentences of the formal language in which different intentions are associated with different syntactic occurrences of a demonstrative …’ (587). This is why the criticisms wielded against demonstrations still apply against intentions. For LD to work the way he wants it to, he will need directing intentions which are separable from their contexts, so they can be included in LD as nonlogical constants. I am arguing that they are not so separable. What makes a demonstration the one it is depends on what the actual demonstratum is, and the way in which certain skills tie an agent to that demonstratum. The tendency to think demonstrations are separable from their contexts is an artifact of an implicit commitment to a methodologically solipsistic view of meaning.
Someone might object that the roadkill case is explicable in a Kaplanesque manner which allows for the identification (or at least correlation) of directing intentions with some kind of linguistic meanings: ‘There are many directing intentions — one for each successive perspective on the object — each indexed to distinct utterances of ‘that’, only one of which is voiced, and the rest of which are left unsaid, tacit.'[ 16 ] If this is right, there may very often be unvoiced occurrences of demonstratives — even more often than implied in the discussion of tacit occurrences in connection with Donnellan above. As far as I can see, to defend CST, Kaplan needs something like this to work in the end.[ 17 ] In particular, it has to be feasible to view directing intentions as constructions on the basis of agents’ locally definable focal reactions to patterns of stimuli on perceptual apparati. Only such ‘directing intentions’ could be analyzed as features of narrow psychological states, and thus thought to be separable from contexts.
But, first, the idea that an agent’s link to a demonstratum is continually re-established rather than constantly maintained strikes me as unintuitive. Second, in light of the above-mentioned problems with culling a directing intention out of a narrow psychological state — much less a whole series of such — I think it best to avoid the Kaplanesque way of dealing with this. And happily, there is an alternative. The interactionist picture of demonstrative reference I’ve been defending has the virtue of allowing us to take the intention to be singular in virtue of the constancy of the object intended. Due to the dynamism of agents’ interactivity with the world, the details of intentional-relation implementations vary; but thanks to reliable skills, this flux doesn’t interrupt the robust constancy of the intended object. In our specification of this intentional relation we rise above the miry miscellanea of implementation, making reference to coarser-grained relational properties featured at a higher level of organization. In particular, what matters here is that an intention is of a particular object; as such the intention is only specifiable by reference to a wide psychological state in which that object enters as a constituent.[ 18 ]
If this alternative story about demonstrative reference is right, it’s wrong-headed to think that we can get away from the notion of the skilled agent, and that by factoring the world out of our account we can distill to a portable, context-independent directing intention. Chunks of the world help constitute the intention in question; and without such versatile agents such relations could not be counted on to obtain.
The problems encountered above are spawned directly from CST — particularly, the Fregean conception of semantics as being centrally concerned to offer an account of the cognitive significance of expressions. Frege said that since ”a = a’ and ‘a = b’ are sentences of obviously different cognitive significance,’ the semantic accounts of the expressions ‘a’ and ‘b’ must differ. It should be clear by now that I think a more promising direction can be found if that requirement on semantical theories is self-consciously and radically rejected — in which case, accounting for the cognitive significance of language becomes the job of the psychologist of knowledge, and not that of the semanticist. The anti-Fregean revolution advanced by the likes of Kaplan, Perry, and others, must cut loose both FIT and CST.
The cognitive significance thesis is just one of an host of demons of narrow content which belong outside the gate of the city of wisdom. But we shall have to exorcise qualia and the rest on some other day.
Manuscript submitted February 12, 1996
Final version received April 10, 1997
Ackerman, F.: 1989, ‘Content, Character, and Nondescriptive Meaning’, in Almog, et al.: 1989, pp. 1-21.
Almog, J., J. Perry, and H. Wettstein (eds.): 1989, Themes from Kaplan, Oxford University Press, New York.
Burge, T.: 1979, ‘Individualism and the Mental’, in P.A. French, T.E. Uehling, and H.K. Wettstein (eds.), Midwest Studies in Philosophy, vol. 4, University of Minnesota Press, Minneapolis, pp. 73-121.
Burge, T.: 1982, ‘Other bodies’, in A. Woodfield (ed.), Thought and Object: Essays on Intentionality, Oxford University Press, New York, pp. 97-120.
Donnellan, K.: 1966, ‘Reference and Definite Descriptions’, Philosophical Review 75, 281-304.
Fodor, J.: 1987, Psychosemantics, MIT Press, Cambridge, Massachusetts.
Frege, G.: 1892a, ‘On Sense and Nominatum’, from H. Feigl and W. Sellars (eds.): 1949, Readings in Philosophical Analysis, Appleton-Century-Crofts, New York, pp. 85-102. (Trans. H. Feigl)
Frege, G.: 1892b, ‘On Concept and Object’, reprinted in P. Geach and M. Black (eds.): 1980, Translations from the Philosophical Writings of Gottlob Frege, 3rd ed., Rownman & Littlefield, Totowa, New Jersey.
Kaplan, D.: 1968, ‘Quantifying In’, Synthese 19, 178-214. (My pagination corresponds to the reprint in A.P. Martinich (ed.): 1990, Philosophy of Language, 2nd ed., Oxford University Press, New York, pp. 370-391.)
Kaplan, D.: 1977, ‘Demonstratives’, in Almog, et al.: 1989, pp. 481-563.
Kaplan, D.: 1989, ‘Afterthoughts’, in Almog, et al.: 1989, pp. 565-614.
Kripke, S.: 1977, ‘Speaker’s Reference and Semantic Reference’, in P.A. French, T.E. Uehling, and H.K. Wettstein (eds.), Contemporary Perspectives in the Philosophy of Language, University of Minnesota Press, Minneapolis, pp. 6-27.
Lewis, D.: 1983, ‘Attitudes De Dicto and De Se‘, Philosophical Papers, vol. 1, Oxford University Press, New York, pp. 133-159.
McClamrock, R.: 1995, Existential Cognition, University of Chicago Press, Chicago.
Perry, J.: 1977, ‘Frege on Demonstratives’, Philosophical Review 86, 474-497.
Stalnaker, R.: 1978, ‘Assertion’, in P. Cole (ed.), Syntax and Semantics 9: Pragmatics, Academic Press, New York, pp. 315-332.
Stalnaker, R.: 1989, ‘On What’s in the Head’, in J. Tomberlin (ed.), Philosophical Perspectives 3: Philosophy of Mind and Action Theory, pp. 287-316.
Taschek, W.W.: 1987, ‘Content, character, and cognitive significance’, Philosophical Studies 52, 161-189.
Wettstein, H.: 1986, ‘Has Semantics Rested on a Mistake?’, Journal of Philosophy 83, 185-209.
Wittgenstein, L.: 1953, Philosophical Investigations, Macmillan, New York.
[ 2 ] Kaplan’s (1968) vivid names likewise depended on there being narrowly individuable content of mental states: ‘The notion of a vivid name is to go to the purely internal aspects of individuation …. it depends only on … current mental state, and ignores all links … with the actual world’ (383).
[ 3 ] The same type of problem arises in the case of the pure indexical, ‘here’. One can easily imagine the following scenario occurring on ‘Star Trek: The Next Generation.’ A guest is being given a tour of the ship, when he and the tour guide step onto a holodeck. The guide says, ‘Here …’ — the scenery changes completely — ‘is here.’ The guest is in disbelief. Someone might counterassert that this simply shows that in this case at least, ‘here’ is not used as a pure indexical. But I won’t press the matter.
[ 4 ] As we shall see below, Kaplan thinks the referential use of a description in fact often involves a tacit demonstrative. So the theory of demonstratives ought to shed light on such referential uses, rather than vice versa.
[ 5 ] For another discussion of important differences between demonstrations and descriptions, and other arguments that demonstrative reference is more distinct than Kaplan allows (i.e. not explicable by analogy with descriptive reference to objects), see Taschek (1987). I want to thank the referee for bringing this valuable paper to my attention.
[ 6 ] It will not do to assume that for each situation there will be a non-ad-hoc set of such assignments — e.g. that there will be some narrow state which is a function from certain features of context to content. I think such sets can only be constructed on an ad hoc basis, and are of no use in semantics or psychology: They do not explain the putative function in question.
This is why I think, for instance, that Stalnaker’s (1978) diagonal propositions — which are functions from contexts and context-worlds onto truth values — don’t yield the sort of narrow content needed here. (In light of Stalnaker (1989), it seems he doesn’t think so either.) Likewise, Lewis’ (1983) attitudes de se — which take objects of attitudes to be properties (rather than propositions), and attitudes themselves to involve the self-ascription of properties — merely presuppose that behavior-influencing attitudes can be characterized in terms of what’s in the head (143).
I actually think this presupposition is false. I’m not denying there’s a function from everything about the context and everything about the speaker that gets you to content. As will become clear below, I simply doubt that the facts about the speaker’s psychological state play a systematic role. Hence, I’m claiming it’s not an interesting function since it’s so highly idiosyncratic to the particular details of particular cases (cf. Stalnaker 1989). I elaborate on this in the final section of the paper.
[ 7 ] Interestingly, Kaplan (1977) acknowledges the possibility that demonstrations are nonrepeatable nonseparable features of contexts. But he inclines to think that since on such a view any given demonstration couldn’t have picked out a different demonstratum, the view can’t do justice to the epistemological situation on which Frege aimed to shed light (528 f.). (See the sub-section ‘Demonstrations’ of the section ‘Cleaning house’ above.)
[ 8 ] One might suggest that the characters of these expressions in fact differ due to differences in, for instance, prosodic information. But for reasons I’ll presently unfold I think attempts which slice characters finely enough to respect such variations fail. So the logical-form type identity of the expressions at hand is sufficient to make for their character type identity here.
[ 9 ] According to Wettstein (1986) John Perry suggested this in conversation (195). On this view, the contributions of the characters of expressions to total cognitive states are the same, although the states still differ due to the contributions of many other factors.
[ 10 ] Frege (1892b) insisted that his notion of sense was meant to have no part in explaining idiosyncratic features of cognitive states, but only the something that different expressions share in common (196).
[ 12 ] But the latter will involve some kind(s) of causal link between the speaker and the referent, whether by virtue of previous perceptual encounters, or causal chains which allow the speaker to refer a la the linguistic division of labor, et cetera.
[ 13 ] Interestingly, Kaplan says, ‘this classical case of the referential use of a description can be seen as an attributive use of a tacit perceptual demonstrative’ (584), since Ernie means ‘the man with the martini’ to refer to whoever the tacit demonstrative’s directing intention is picking out.
[ 16 ] Indeed, Kaplan paved the way for this sort of response. He says that in the sentence, ‘You, you, you, and you can leave,’ each demonstrative has its own directing intention (586), and that since ‘there is simultaneous perception of all addressees, I think it correct to say that [there] are several distinct, simultaneous, directing intentions, indexed to distinct intended utterances of the demonstrative ‘you’ (which are then voiced one at a time)’ (587). The speaker may of course choose not to voice every ‘intended utterance,’ so it need not be the case that for each intention there is an utterance. Applying this account to the roadkill case, we could say Ernie has in effect uttered a small fraction of a very long sentence, something like: ‘That, that, that, that, that … What is that?‘ (Where the occurrences of ‘that’ are indexed to directing intentions which succeed one another in stream-like fashion, each of which determines the roadkill to be the referent in its respective context.)
[ 17 ] A referee for this Journal suggested that the example might be dealt with by temporally extending the context of the demonstrative — so we need not attribute to Ernie a different inner intention for each variation in the relation between himself and the roadkill. I think this suggestion is just right, but not for a follower of Kaplan. Such a theorist would have to make a corresponding temporal extension of the character, which presupposes that we have a solipsistic in-principle way of carving out which set of narrow features of cognitive states count as part of the character. I’m not willing to presuppose this, since if I’m right, what counts as a set is context dependent.
[ 18 ] The growth of empirical evidence favoring object-centered approaches to perceptual psychology over solipsistically oriented approaches only reinforces the view I’ve been presenting. See McClamrock (1995), chapter 9, on the philosophical implications of the evidence for non-individualistic theories in perceptual psychology.