Indexicals as Token Reflexives
1. This talk is a defense of the "token reflexive" view of indexicality first propounded by Reichenbach, and supported nowadays by Perry and Crimmins, against the alternative view defended mainly by Kaplan, and also by Braun and Richard.
2. Let us present first the two views. The token reflexive view takes into account three semantic properties. There is first a linguistic rule associated with an indexical type, say, 'he', which goes like this: "for any token t of 'he', the referent of t is the male who has been made salient by a pointing gesture or any other means for manifesting a demonstration at the occasion of the production of t"; analogously, the linguistic rule associated with the type 'I' is this: "for any token t of 'I', the referent of t is the speaker who has produced t". As these rules make clear, reference (truth conditional import) is only ascribed to tokens. In addition, there is also, associated with any particular token, a token reflexive mode of presentation: an individuative property, known to any competent user who correctly identifies an utterance including a token of an indexical; this property is obtained from the linguistic rule associated with the type, by instantiating the quantifier with the relevant token. According to this view, the properly referring expressions are tokens and not types.
3. The token reflexive proposal shares with Kaplan's its taking indexicals to be "directly referential". This just means that the possible world truth conditions of an utterance of 'I am hungry' are determined, relative to every relevant possible world, by the referent of I. (Henceforth, a bold face indexical refers to a contextually relevant token of that type.) The same applies to utterances where the sentence is inside a context governed by an alethic (or "metaphysical") modal expression. That is to say, indexicals are "rigid designators". Given that descriptions, including descriptions like 'the producer of I', are not rigid designators, this means that the description expressing the token reflexive sense of an indexical is not synonymous with it; in Kripke's expression, the description merely "fixes de referent" of the indexical. This is compatible with knowledge of the description being required for proper linguistic understanding of an indexical; it is also compatible with neo-Fregean views.
4. The difference between the token reflexive view and Kaplan's lies in that the latter refuses to ascribe reference (or "content", in his terms) to concrete entities. He only considers two semantic properties for indexicals, a character assigned to types, and content; he ascribes the latter to entities which are abstract, like types: what he calls "occurrences", or "types-in-context". The bearers of the fundamental semantic properties in Kaplan's theory are thus individuated by the type they instantiate and the context where they occur. This is mainly intended to deal with the objections to the token reflexive account to be consider presently.
5. My argument for the token reflexive account has three parts. Firstly, I reply to the objections by Kaplan and others which are supposed to be fatal for it. Secondly, I show that the token reflexive view makes the right semantic prediction regarding "true demonstratives", as opposed to "pure indexicals". Kaplanian accounts can only handle true demonstratives properly at the cost of running into the same (alleged) difficulties they attribute to the token reflexive view. Thirdly, I claim that the token reflexive view provides a most needed Fregean complement for theories of intensionality which seem to be on the right track, like the nowadays popular "two-dimensional" explanations of the distinction between epistemic and alethic modalities advanced by Kaplan and Stalnaker.
6. (i) The main objection to the token reflexive account is that it is incompatible with a "logic of indexicality". (There are related criticisms, which we can deal with at discussion: the of lack of an expression consisting of 'I exist now' conjoined with itself a trillion times at this very instant, to be counted as true; the problem posed by 'I am not uttering now'.) Consider any logically valid inference, like: p, therefore p. The validity of this inference seems to rely on the assumption that the very same expression is first premise and then conclusion of the argument; this is how it is guaranteed in a purely formal way that the argument is truth-preserving. In the token reflexive theory, however, the fact that premise and conclusion instantiate the same type will not guarantee this; for they will be different tokens, therefore different expressions, possibly with different semantic values. In Kaplan's theory, we can assume that premise and conclusion are produced in the same context. This will guarantee that they are the same expression, as required. Consider, however, an argument like: he is hungry; therefore, the formerly mentioned person is hungry. It is perhaps not formally valid, but it is certainly analytically valid; and its analytical validity is not foreclosed by the token reflexive view, which by itself does not say anything about anaphora. This is enough, for it is the most that a "logic of indexicality" can aspire to obtain. 'I am here now' is Kaplan's paradigm validity of such a logic, and his explanation of its validity hardly allows for their "formal" character.)
7. (ii) True demonstratives are indexicals which require a demonstration ('he', 'you', 'this'). Pure indexicals allegedly do not need it ('I', 'now', 'here'). In fact, the latter sometimes require it, and the former sometimes can pass without it (unless, of course, demonstrations are taken to be intentions of the speaker, in which case they can also be presumed to accompany pure indexicals always). This partially confirms the token reflexive view, which does not make any difference of principle between indexicals in the two groups. This view also fares much better than Kaplan's original account of true demonstratives, which involved assigning a different character to all instances of one and the same indexical type occurring in different syntactic positions in the same utterance or fragment of discourse. This is required to make the required prediction that the two instances of 'this tree' can be taken to have different references in utterances of 'this tree is older than this tree', or of 'this tree is a chesnut, therefore this tree is a chesnut'. Because the context is not supposed to change in these cases, if the demonstratives were supposed to have the same character (to instantiate the same linguistic type), they should be assigned the same referent. Maybe they refer to the same entity, but this is of course not linguistically guaranteed. This is why Kaplan took them to be, essentially, different expressions. But this is incorrect, because it misses (invidiously, relative to the treatment of pure indexicals) the semantic commonalities relating different instances of true demonstratives. Now, there are different ways of capturing those commonalities. Perhaps the most natural is to assume that every demonstration is a contextual element, which, by occurring, triggers a shift in context. Under this assumption, there is no difficulty in assigning a common character to every instance of a true demonstrative. The proposal, however, is not compatible with Kaplan's criticisms of the token reflexive approach, for it shares all its (alleged) difficulties. They may just be notational variants of each other.
8. (iii) We will only consider the "two-dimensional" account of the distinction between epistemic and alethic modalities, although the point can be extended to cover indirect discourse in general. Under the assumption in (3), the condition primarily asserted by an utterance u of 'I am here now' is the same one that could be asserted by an utterance of 'Manuel Garcìa-Carpintero is in Genoa on Genuary, 21, 1999. It seems absurd, however, to say the same about the truth-conditional import of the very same indexicals in an utterance of 'I know a priori that I am here know'; for this utterance would make a true claim, while an utterance of 'I know a priori that Manuel Garcìa-Carpintero is in Genoa on Genuary, 21, 1999, seems to make a straightforwardly false claim. Stalnaker's resorts to what he calls "diagonal propositions" to account for the difference. Stalnaker's characterization of a diagonal proposition is this: "this is the proposition that is true at i for any i if and only if what is expressed in the utterance at i is true at i." (Stalnaker 1978, 280). There are no restrictions, in this general characterization, on the facts that, at the possible circumstance i, determine "what is expressed in the utterance". Thus, at its most general level, the diagonal proposition signified by u is this: that u says something true. This proposition selects every possible world which includes an instantiation of u that says something true, no matter what it says. In some of them, for instance, u says that snow is white, and snow is white. This diagonal proposition is of little use, for the purposes for which Stalnaker deploys those theoretical entities; we are not told anything relevant about the mental life of a subject, when he is described as holding (a priori or otherwise) a belief with such a general content. For those purposes, we are only interested in much more constrained diagonal propositions. We want to consider only possible worlds where the relevant linguistic facts are very much as they are in actuality. We want u to be syntactically parsed in all of them the same way it is parsed in English, we want the resulting word-types to keep their conventional meanings in English, and so on and so forth. In fact, the diagonal proposition we really need is the one that could be perspicuously stated so: that the time at which now is produced is 3 pm. This is of course the proposition determined by the token reflexive account. Kaplanian theories lack these explanatory virtues; for in such accounts content is ascribed to expressions-in-context. In the most natural understanding, those entities are individuated by the entire context in which they are actually produced. Again, a further elucidation of Kaplanian views might be envisaged, sharing the virtues of the token reflexive approach; but, until any such proposal is carefully elaborated, we are entitled to think that it would just be a notational variant of the view here advanced.
A Presuppositional Account of Reference-Fixing
1. Setting up the problem:
- Singular terms include indexicals, proper names and (nonliteral) referential uses of descriptions; most of the talk will focus on indexicals, next day proper names.
- The core of Fregean views on the semantics of singular terms: a correct semantical account cannot associate with them only an extralinguistic individual, its referent. It must in addition associate with them a property having the necessary features to count as a Fregean mode of presentation (predicative; individuative; known in a privileged way, on the basis of linguistic knowledge). If t is one such singular term, let us use D(t) to express the descriptive material constituting its Fregean sense.
- The alternative view of direct reference theorists (DR) is not that there are no such properties, but only that it is not for a semantic theory to incorporate them; they rather belong in a "metasemantical" (Kaplan 1989a) or "foundational" (Stalnaker 1997) account of language.
- The subsidiary reasons for the Fregean core are that senses of singular terms characterize their truth-conditional contribution to identity statements, singular existentials and indirect discourse. The main reason is that they are already needed to properly characterize the propositional content of ordinary statements.
- Loar's (1976) example: "Suppose that Smith and Jones are unaware that the man being interviewed on television is someone they see on the train every morning and about whom, in the latter role, they have just been talking. Smith says 'He is a stockbroker', intending to refer to the man on television; Jones takes Smith to be referring to the man on the train. Now Jones, as it happens, has correctly identified Smith's referent, since the man on television is the man on the train; but he has failed to understand Smith's utterance. It would seem that, as Frege held, some 'manner of presentation' of the referent is, even on referential uses, essential to what is being communicated." Heck's (1995) gloss of Loar's example: the conventional communicative point of some speech acts is the transmission of knowledge, i.e., justified belief; acknowledging senses in a semantic account is required because they are essential to account for the preservation of justification in sucessful communication.
- The overall goal of the project of which the proposal in this talk is a part is to defend the Fregean core, without excrecences added by other Fregeans. Excrecences: (i) Expressions are types, repeatable entities; here, tokens. (ii) Context does not intervene in ascriptions of sense; here context intervenes, compatibly with the core (token-reflexive senses). (iii) Internalism; here, object-dependence of senses. (iv) Synonymy of t and D(t); here, D(t) merely "fixes the referent" of t (to account for the modal differences between singular terms and literal attributive uses of descriptions unveiled by DR). It will be assumed that sentence-like expressions signify ("refer to") states of affairs, specifying possible-world truth-conditions, the contribution to which of singular terms is an object (= Russellian propositions of DR),
- Kripke expresses the problem which this poses for the Fregean: "Some of the attractiveness of the theory is lost if isn't supposed to give the meaning of the name: for some of the solutions of problems that I've just mentioned will not be right, or at least won't clearly be right, if the description doesn't give the meaning of the name [Š]. So the analysis of singular existence statements mentioned above will have to be given up, unless it is established by some special argument independent of a general theory of the meaning of names; and the same applies to identity statements" (Kripke 1980, 33).
2. The presuppositional account:
- Distinction between presupposing and asserting (Soames, 1989): 'it was Victor who made a cake' presupposes someone made a cake; but not 'Victor made a cake', which asserts the same. At a pretheoretical level, presuppositions are identified by two criteria. First, they are "inherited" in contexts where what is asserted is not: if the sentences carrying them are negated (or are constituents of some conditionals, conjunctions, and other complex utterances), the resulting utterances still take the same for granted, but they do not usually assert the same state of affairs any more: compare 'it was not Victor who made a cake', 'it might have been Victor who made a cake' and 'if it was Victor who made a cake, everything is in order'. Second, they can be "cancelled"; this occurs, for instance, when the presupposition is explicitly asserted or clearly entailed in certain contexts: the utterance of 'it was Victor who made it' in 'if someone made a cake, it was Victor who made it' does not presuppose any longer that someone made a cake.
- Stalnaker's (1973, 1978) account: Linguistic exchanges take place with respect to a "conversational context" of states of affairs which speakers and their audience assume, believe that the others assume, believe that the others believe that they assume, and so on. (To assert is defined by Stalnaker as to restrict this conversational background further, eliminating some possible worlds in it.) We can then say that an utterance u presupposes a state of affairs p if it is reasonable to infer from u that the speaker takes p to belong to the conversational context; a sentence carries a certain presupposition if its utterances normally presuppose it.
- Literal attributive uses of descriptions carry presuppositions of existence and uniqueness (Frege-Strawson); with respect to contexts which do not include as part of the conversational background the claim that there is a unique F, u(the F) does not assert a state of affairs and is neither true nor false. Still, the contribution of a description to the signified state of affairs is property-like, not object-like. Given that the presupposition is satisfied, an assertion of u(the F) restricts the conversational set to worlds in each of which a unique F, possibly different in different worlds, satisfies u(x). The proposal, therefore, is not that utterances of u(t) presuppose that there is a unique D(t). That would make t and the D(t) synonymous.
- What is it to genuinely refer? First approximation, Evans' Russell's Principle: To grasp the singular state of affairs expressed by an utterance of 'I', 'he' and other indexicals, it is not sufficient to think of the intended referent on the basis of the purely linguistic mode of presentation given by linguistic knowledge alone, that he or she is the producer of the expression, etc.; it is necessary to put together with it, for identificatory purposes, some other mode of presentation individuative by itself. Think of Perry's (1993) example: "Ellsworth goes to Hawaii and sends me a postcard. Unfortunately, it gets a bit wet before I receive it. The postmark, return address, and signature are all illegible. The message stays dry: 'I am having a good time now'." But Russell's Principle is not generally true (Boër and Lycan, 1986, on knowing who; the politician-journalist exchange).
- The alternative Principle of Reference: Utterances including 'I', 'he' and other indexicals express genuine singular propositions in that: (i) an intended referent is presented relative to a purely linguistic mode of presentation given by semantics alone (that he or she is the producer of the expression, etc.), on the assumption that (ii) for identificatory purposes, this linguistic mode of presentation will be contextually supplemented with other senses, or at least could be so supplemented with senses adquired later, so that (iii) this actual or potential supplementation is relevant to evaluating the correctness (truth or falsity, etc) of the utterance.
- Grice's dossier metaphor: "Let us say that X has a dossier for a definite description D if there is a set of definite descriptions which include D, all the members of which X supposes Š to be satisfied by one and the same item Š the speaker intends the audience to think (via the recognition that he is so intended) (a) that the speaker has a dossier for the definite description D which he has used, and (b) that the speaker has selected D from this dossier at least partly in the hope that the hearer has a dossier for D which 'overlaps' the speaker's dossier for D (that is, shares a substantial, or in some way specially favored, subset with the speaker's dossier)" (Grice 1969, 141-2). An utterance u(t) involves, I will say, a presupposition of acquaintance: the speaker represents himself as acquainted with a particular object which is not only uniquely D(t), but also a repository for additional modes of presentation, actually available in the context or potentially obtainable. They collectively individuate an entity semantically correlated with t on which the truth or falsity of u(t) depends; it is the truth-conditional contribution of t in u(t). The presupposition associated with t encapsulated: there is a unique D(t), and it is the referent of t.
- This proposal agrees with the facts summarized about presuppositions. Consider an utterance of 'that car is running into a bus', and let t be the uttered token of 'that car'. The presupposition in that case, according to the proposal, is the proposition that there is a certain contextually relevant car intended by the speaker when he produced the instance t of 'that car' to which t refers. It seems indeed to be taken for granted in an utterance of 'that car is running into a bus', and would still be taken for granted in utterances of 'that car is not running into a bus', 'that car might be running into a bus', etc. The presuppositions meets the first intuitive criterion for presuppositions. It also meets the second: the candidate presupposition is cancelled, for instance, in utterances of 'If this is not virtual reality or a hallucination, that car is running into a bus'.
- Presuppositions of acquaintance are always linguistically constrained; they may include contextual extralinguistic elements, though. (Russell's Principle would require them to include always contextual extralinguistic elements.) Loar's example again: in that scenario, the proposition that there is a unique male most salient when the token t of 'he' is produced, and t refers to him is a conventional presupposition of Smith's utterance. By contextually determining the determinable most salient when t occurs with male on the TV screen with such-and-such visual aspect we obtain the proposition that there is a unique male on the TV screen with such-and-such visual aspect when the token t of 'he' is produced, and t refers to him. Communication requires grasping the state of affairs signified by the speaker by sharing with him the relevant presuppositions, in particular presuppositions of acquaintance. (Relaxation: this is not necessarily a matter of having the same dossier, but only a matter of there being sufficient overlaping.)
3. How the presuppositional account helps with the original problem: - First and foremost, it explains why descriptions associated with singular terms merely "fix their reference" (i.e., why the contribution to truth conditions of singular terms is the object itself, as according to the modal intuitions unveiled by DR), while at the same time ascribing to those descriptions a properly semantic role, as required by the core Fregean views and the arguments supporting them. (This can be shown to meet the criticisms of previous Fregean accounts of reference-fixing in Soames 1998.) - Subsidiarily, it accounts for the widely shared Fregean intuitions that the descriptive material associated with singular terms becomes a part of the signified state of affairs in some linguistic environments: identity statements, singular existentials, indirect discourse. As Kripke envisaged, what is required is "some special argument independent of a general theory of the meaning of names". This special argument is an account of why presuppositions are inherited in some linguistic environments, and why they are cancelled in some others.
- Without attempting to derive them here from such a general account, indicative outlines can be given for each of the three cases. In identity statements and singular existentials, cancellation is required to make the intended conversational contribution by the speaker a sensible one, given the linguistic meaning of the identity and existence predicates. (Compares favourably with DR accounts; for instance, Stalnaker 1978 for identity, Salmon 1998 for singular existentials.) In indirect discourse, the general idea (Carnap, Quine, Sellars, Davidson) is that words inside indirect discourse structures are "quasi-quoted": they are used with the purpose of modelling some semantic properties of the reported "speech" by means of some of their own semantic properties. (Pietroski 1996 provides a particular form of these accounts which I like.) Given that presuppositions of acquaintance are such semantic properties, it is only natural that, in some contexts at least, they become an essential part of the modelling. Terms playing that function cannot be substituted by others just because they have the same referents.
The Mill-Frege Theory of Proper Names
1. Setting up the problem:
- The already announced project was to defend the core of Fregean views on the semantics of singular terms: a correct semantical account cannot associate with them only an extralinguistic individual, its referent. It must in addition associate with them a property having the necessary features to count as a Fregean mode of presentation (predicative; individuative; known in a privileged way, on the basis of linguistic knowledge). If t is one such singular term, we will use D(t) to express the descriptive material constituting its Fregean sense.
- Today we confront the main problem, proper names. Here Millianism seems to most people the correct view: "According to Mill, a proper name is, so to speak, simply a name. It simply refers to its bearer, and has no other linguistic function. In particular, unlike a definite description, a name does not describe its bearer as possessing any special identifying properties" (Kripke 1979, 239-240).
- The alternative view of direct reference theorists (DR) is not that there are no such properties, but only that it is not for a semantic theory to incorporate them; they rather belong in a "metasemantical" (Kaplan 1989a) or "foundational" (Stalnaker 1997) account of language. As Salmon puts it: "Millianism does not entail that a proper name has no features that might be deemed, in a certain sense, intensional or connotive. Unquestionably, some names evoke descriptive concepts in the mind of a user. Some may even have particular concepts conventionally attached [Š]. It does not follow that this connotive aspect of a name belongs to semantics, let alone that it affects the propositions semantically expressed by sentences containing the name" (Salmon 1998, 311).
- The subsidiary reasons for the Fregean core are that senses of singular terms characterize their truth-conditional contribution to identity statements, singular existentials and indirect discourse. The main reason is that they are already needed to properly characterize the propositional content of ordinary statements. Last time we used an example of Loar's involving indexicals to show this, and as a test for the token-reflexive proposal then advanced. This time we will be using an example by Heck for the same purposes: "Suppose Eric Blair were to become amnesiac and check himself into a hospital. The doctor, Tony, deciding that she needs to have some name by which to call him, dubs him "George Orwell". And suppose further that Alex says not intending to refer to Tony's patient "George Orwell wrote 1984" and that Tony forms, in reaction to Alex's assertion, the belief she would express to other members of her staff as "George Orwell wrote 1984". This belief is true: Tony's new patient happens to be Eric Blair, that is, "the other" George Orwell. But surely it would not count as knowledge, even if Alex knows that George Orwell wrote 1984: it would not even count as justified" (Heck 1995, 95). The conventional communicative point of some speech acts is the transmission of knowledge, i.e., justified belief; acknowledging senses in a semantic account is required because they are essential as a measure of the preservation of justification in sucessful communication.
- A reminder: the overall goal of the project is to defend the Fregean core, without excrescences added by other Fregeans. Excrescences: (i) Expressions are types, repeatable entities; here, tokens (G-C 1998). (ii) Context does not intervene in ascriptions of sense; here context intervenes, compatibly with the core (token-reflexive senses). (iii) Internalism; here, object-dependence of senses. (iv) Synonymy of t and the D(t); here, the D(t) merely "fixes the referent" of t (to account for the modal differences between singular terms and literal attributive uses of descriptions unveiled by DR). It will be assumed that sentence-like expressions signify ("refer to") states of affairs, specifying possible-world truth-conditions, the contribution to which of singular terms is an object (= Russellian propositions of DR).
- Three sorts of arguments have been proposed in favour of Millianism. Firstly, a semantic argument designed to show that speakers use instances of N to refer to o even if they lack knowledge of any description the D(N) uniquely identifying o, or even if the description the D(N) they associate with N in fact identify a different object. Secondly, an epistemic argument, arguing that while utterances of the D(N) is D(N), if something is uniquely so are knowable a priori, utterances of N is D(N), if something is uniquely so are not knowable a priori. Finally, a modal argument, pointing out that the (alethic) modal properties of utterances including N are different from those of sentences with the D(N) substituting for it.
- The modal argument can be disposed of along the lines indicated in the previous talk (the relevant description merely "fixes the referent", which is why the name behaves rigidly in modal contexts). This is possible only in so far as we can provide descriptions relative to which the semantic and epistemic arguments can be overcome; that is to say, descriptions provided by linguistic conventions for the terms, or at least "guided" or "constrained" by linguistic conventions, like the following for indexicals:
'I': For any token I of 'I', I refers to x iff x is a unique producer of I
'He': For any token he of 'he', he refers to to x iff x is a unique most salient (in virtue of an act of demonstration by the speaker, or otherwise) male when he is produced
'That': For any token that of 'that', that refers to x iff x is a unique most salient (in virtue of an act of demonstration by the speaker, or otherwise) entity in contextually specified class F when that is produced
- Kripke expresses the problem which this poses for the would-be Fregean: "[Š] what [Š] conventional 'senses' [Š] can plausibly be supposed to exist for 'Cicero' and 'Tully' [Š]? Are not these just two names (in English) for the same man? Is there any special conventional, community-wide 'connotation' in the one lacking in the other?" (Kripke 1979, 244). The challenge for the Fregean is to provide descriptive material which can be plausibly taken to be semantically associated with proper names, over and above their referents.
2. The Mill-Frege proposal:
- Once the problem is set up in that way, it is obvious that it only admits a solution of the "metalinguistic" variety, according to which the conventional sense of a proper name N is, roughly, being called 'N'; previous writers who have defended such approach include Loar (1976), Bach (1987), Katz (1994) and Recanati (1993). The Mill-Frege account, however, is much more influenced by suggestions by Kripke (1980) and Evans (1982).
- Kripke's own alternative picture of reference involves a certain protracted event or process, a chain of communication leading to the use of a name from an "initial baptism" in which the referent is authoritatively referred to (by other means than by using proper names, we might assume to avoid circularities) and conferred the name. As different people have noticed, this provides a (token-reflexive) relational descriptive property of the used name-token, which determines the referent. However, as Kripke (1980, 88fn.)) points out, this view would only count as Fregean in a Pickwickian sense. For, no matter how tacitly, senses have to be known by competent users; and there is no serious sense of knowledge relative to which it can be said that ordinary competent users of proper names know the Kripkean process.
- The Mill-Frege theory does not rely on such trivializing maneuvers. Instead, it takes its inspiration from Kripke's "initial baptisms". The suggestion is that they illustrate a much more widespread specific variety of speech-acts, which I will call acts of naming. They are specific speech acts, to be distinguished from the ordinary speech acts which mainly constitute natural language assertions, requests, promises and so on. Acts of naming are explanations of the meaning of a lexical unit, whose conventional purpose is that which definitions are supposed to attain: typically, to introduce someone who is learning the language to the practice, or to clarify one's own usage. The lexical units whose meanings are explained by acts of naming are what were traditionally considered categorematic expressions, as opposed to syncategorematic ones (like logical particles and syntactic features signifying logical category, force-indicators and so on, which are "contextually" explicated without acts of naming).
- Competently performing acts of naming involving common nouns ('water', 'tiger', 'AIDS', 'square', 'red') requires specific extra-linguistic knowledge identifying (sufficiently, for the contextually relevant purposes) the kinds or properties in question; at least, it requires the ability to defer to experts with that knowledge. In these cases, the conventional reliance in linguistic acts of naming on the first-order identificatory traits gives rise to a second-order, emergent or subsidiary metalinguistic identicatory trait of the kind or property: it is the kind selected by speakers in such-and-such community in acts of naming involving tokens of 'water', say.
- This fact is important to handle Twin-Earth cases, but here it is relevant for a different reason. Proper names are, we may say, a conventional device to solve this problem: to refer to entities which cannot be referred to by means of indexicals, or referentially used descriptions. In putting the issue this way, we assume that we already know what it is to refer by means of those devices, but not that we know what it is to refer by means of proper names. The fact we have highlighted concerning acts of naming involving common nouns, adjectives, etc, suggest a way of solving the problem: namely, to device procedures making sure that acts of naming involving tokens of a distinctive type pick out the individual we want to bring into the discourse. This is were tagging procedures (the seal parable) enter the picture. Notice that, to competently use these devices, no "first-order" knowledge of the referent is needed; only the knowledge that there is a proper procedure to secure the relevant acts of naming.
- Acts of naming, like other speech acts, are actual concrete events or potential ones. They are purposeful events, governed by specific linguistic conventions which I will call appellative practices. As is generally the case with purposeful activities, attempted acts of naming can be in particular cases dysfunctional, merely putative acts of naming: although the participants in the act justifiably assume that they serve their intended purpose, they do not. We individuate the practices by describing their "felicity conditions"; that is to say, the conditions under which they would be performed successfully, without assuming that they are always thus performed. Successful instances of one and the same appellative practice have three features. First, instances of N in the acts instantiating the practice are relevantly of the same type. Second, if there exist concrete acts of naming instantiating the practice (appellative practices may remain uninstantiated in some cases), they conform to a common causal history: some act (modelled on Kripke's "initial baptism"), causally independent of every other act in the practice, constitutes the causal origin of the practice; the rest are causally dependent on it. Third, all the acts in the practice have a common purpose. The primary conventional purpose of acts of naming is, as I have said, to contribute to "fixing" (to put it in Kripke's terms) the semantic value, the referent or truth-conditional import of a given class of expressions in ordinary speech acts, and to exhibit the sense of the expressions in doing so.
- The main difference distinguishing the appellative practices governing proper names from others is that the former are purely nominal: only the second-order, metalinguistic aspect is linguistically invoked to determine the referent of the expression in ordinary speech acts. Linguistic rule for proper names:
PN: For any token n of a proper name of type N, n refers to x iff x is a unique individual picked out in the acts of naming involving N which instantiate the appellative practice (of contextually indicated type S) on which n relies.
The Mill-Frege theory contends that to knowledgeably produce or to fully understand a proper name, the following is sufficient: (i) identification of the token and of the type it instantiates; (ii) general knowledge of the nature of appellative practices, both ordinary and purely nominal; (iii) adequate sortal knowledge of the referent, and (iv) knowledge sufficient to separate the practice relied upon from others involving the same type. Items (i)-(ii) count as features of specifically linguistic knowledge; items (iii)-(iv) are to be obtained from context on the basis of general knowledge of the world, guided by knowledge of the linguistic convention PN for proper names. None of the items under (ii)-(iv) involves individuative modes of presentation, nor do they amount in conjunction to one. Because the knowledge mentioned in them constitutes a sufficiently stable disposition across discourses, proper names are anaphoric devices, like most other expressions in the language aside from indexicals. Nevertheless, proper names share with indexicals the token-reflexivity mentioned in (i). To understand a proper name it is not required a sufficiently reliable capacity to recognize the referent, or to preserve memories originated in a perceptual encounter with the referent, or to be in a position to produce an individuating description not involving the name. The only required cognitive identification is the identification of a particular appellative practice. For that, items (i)-(iv) provide a sufficiently reliable basis, even though, of course, not an infalible one.
- Gloss of Heck's example: To fully understand Alex's utterance, Tony should grasp Alex's reference-fixing mode of presentation. But she does not; she applies her linguistic knowledge of proper names by assuming that Alex's utterance of 'George Orwell' relies on the appellative practice she herself has introduced about the amnesiac Eric Blair, while Alex's utterance does not purport to rely on that practice. Alex is relying on the common practice which we associate with George Orwell, the writer. This is different from the one initiated by Tony according to our second criterion of identity for appellative practices (section 2), because it is causally independent from it. In this case it is thus the appeal to contextually available general information (to select a particular practice among those associated with 'George Orwell') which is liable for the misunderstanding. The sense of 'George Orwell', as uttered by Alex, is (let us put it this way) the writer called 'George Orwell'; as understood by Tony, it is the patient called 'George Orwell'. Alex has good reasons to take for granted that something like writer is contextually available to select the specific appellative practice on which his use of 'George Orwell' relies, over and above that name type; we may assume that Tony has good reason to think likewise about something like patient. (Perhaps the doctors at the hospital have the custom of using names of famous writers to dub their patients, Alex is ignorant that they have an Orwell, while Tony justifiedly thinks he should know.)
- Kripke's puzzle: Kripke (1979) offers the best reply to criticisms of Millianism based on substitutivity problems, by presenting "a paradox about names in belief contexts" derived from two principles "apparently so obvious that their use in these [anti-Millian] arguments is ordinarily tacit" (op. cit., p. 253). A form of the paradox can in fact be derived from just one of them, a "disquotational principle": "If a normal English speaker, on reflection, sincerely assents to 'p,' then he believes that p" (op. cit., pp. 248-9). This principle is formulated for expression-types; Kripke warns us that it would not be intuitively valid if the sentence replacing 'p' included "indexical or pronominal devices." It can be shown, however, that we do not invoke any such principle when attributing beliefs involving proper names. Consider a hospital like the one in Heck's story, whose literate doctors customarily dub their new patients, for restricted hospital use, with names of writers they like. They just introduced 'George Orwell' to dub a new amnesiac patient who checked himself into the hospital. In Alex' presence, Tony (who, like Alex and the other doctors, is well aware of the dubbing) is reading a newspaper in which the famous writer George Orwell is reported to have suffered a heart attack; showing the newspaper to him thus making clear her intention to refer thereby to George Orwell the famous writer she asserts 'George Orwell has suffered a heart attack'. Afterwards Alex is discussing with other doctors the condition of the amnesiac patient whom they dubbed 'George Orwell'; the doctors lack any information about why the amnesiac patient checked into the hospital, and they are still wondering what his condition might be. Would it be correct for Alex to report to the other doctors in that context Tony's previous assertion, by uttering 'Tony believes that George Orwell has suffered a heart attack'? Of course not; and it would still be incorrect even if, unbeknownst to the doctors, the new patient happens to be Eric Blair, that is to say, "the other" George Orwell, the author of 1984. Examples like this show that, far from being obvious, Kripke's disquotational principle is not one we use in our ordinary practices of attributing beliefs. Tony would certainly assent to the sentence-type 'George Orwell has suffered a heart attack' (in some contexts); but ordinary speakers would not consider this sufficient to attribute to her, in some other contexts, a belief worded with those very same expression-types.
- Circularity worries: aggravated by the function of the third criterion of individuation for appellative practices, intended to deal with 'Madagascar'-like cases, cases in which the semantic referent of a name intuitively changes even if, according to the first two criteria, it remains inside a practice in which previous tokens have a different referent. To explain 'purpose' in the third criterion, think again of the grain of truth in Evans' Russell's Principle and Perry's "Ellsworth" example. It entails that in any case of reference we can distinguish two sets of referential intentions, "ultimate" ones (providing the speaker's referent) and linguistic "ancillary" ones (providing the "semantic referent"). In some cases, speaker's and semantic referents come apart. In 'Madagascar'-like cases, the tension in insisting that they come apart is excessive, and the natural thing to say is that the semantic reference of the term has changed; in our framework, that a new practice has in fact begun. But this makes clear that this is one more point at which the concept of reference is used in the analysis.
- Reply to the circularity objection. We do not explain the reference of a token of N as whatever N refers to. We explain the reference of a token of N in ordinary speech acts in terms of (i) other speech acts, acts of naming (in which the name-type is not essentially used, but mentioned, while the referential use of other expressions there, indexicals or descriptions, can be independently explained); and (ii) the role such a token plays in ordinary speech acts, role which, although as the preceding discussion shows is also to be elucidated partly in terms of the concept of reference, does not involve either the concept of reference of a proper name. What is indeed the case is that the present account of proper names does not help reduce intentional notions to non-intentional ones. Similarly, our explications of the token-reflexive senses of indexicals used in fact indexicals (the bold-face letters referring to tokens), which make clear that a reductive account of indexical reference is not promoted by them. I do not think we should be worried by any of this, however. The Mill-Frege theory presupposes a Gricean conception of language. It assumes only that the intentional notions involved in language-use can be illuminatingly explained relative to the intentional notions of folk-psychological explanations. It is, I believe, a mistake to interpret this kind of theoretical enterprise as devoted to produce conceptual or a posteriori reductions of intentional properties to non-intentional ones, or of social intentional notions to purely psychological ones. The real objection would be that we have not explained the nature of reference by means of proper names, not that we have not explained it refraining from using intentional notions. - Comparisons with other metalinguistic views can be provided if desired. The main difference with the view of Evans lies once again in his acceptance of Russell's Principle. This is what explains his deferential treatment of proper names (his distinction of producers and consummers), which I think entirely inappropriate to understand proper names, and made only plausible by inadequately focussing on the specific case of names of persons (and in their most ordinary uses for that matter, not the ones involving official registers and so on).