Page 245 - Encyclopedia of Philosophy of Language

nowhere discusses what could be the sense of an index- ical. As an alternative, Perry suggests treating senses as roles, and thoughts as information expressed; so that (1) expresses different thoughts on different occasions, although its role remains constant. We can thus no longer equate thoughts with senses as Frege wanted to.
2. RussellandReichenbach:CanIndexicalsbe Eliminated?
In the wake of logical empiricism, several philosophers have attempted to reduce the importance of indexical expressions in natural language, or even to eliminate them altogether by translating indexical sentences into supposedly equivalent nonindexical or 'eternal' ones. In An Inquiry into Meaning and Truth (1940) Bertrand Russell sets out to reduce all indexicals or, as he calls them, 'egocentric particulars' (defined as terms whose denotation is relative to the speaker) to nonindexical terms plus the one indexical this; so 7 means 'the biography to whom this belongs,' now means 'the time of this,' etc. But this leaves the problem of their status: what is the constant meaning that they have? Unlike names (with which they appear to share a lack of descriptive content), they refer to different objects on different occasions without being ambiguous; unlike descriptions, they always apply to one thing at a time only, and not to everything that is ever a 'this'; and unlike general concepts, any instances they may have are an instance only at that particular moment. To get rid of this embarrassment, Russell tries to elim- inate this single remaining indexical element by defin- ing it in terms of so-called 'minimal causal chains' between a nonverbal stimulus and a verbal response. But if an expression like this is really to apply to something we directly experience, it cannot refer to an object in the outer world, but only to our own percepts. In other words, its designation is a sense datum (1940:114). However, as sense data are private phenomena, this analysis makes communication between two individuals, or even within one individual at different points in time, impossible.
In his Elements of Symbolic Logic (1947), Hans Reichenbach treats indexicals in a manner similar to Russell's; but with him, they do not refer to sense data but to the act of speaking itself. He calls them 'token- reflexive,' because they refer to the particular token of their type used on the occasion of utterance. There- fore, he holds, they can also be defined in terms of the phrase 'this token,' so e.g., / means the same as 'the person who utters this token,' now the same as 'the time at which this token is uttered,' etc.
Reichenbach states that the symbol 'this token' is not really a phrase but an operator; it is not a term with a fixed meaning, as its different tokens will not be equisignificant to one another. He therefore calls it a 'pseudo-phrase,' an operator the meaning of which can only be formulated in the metalanguage; in other
words, all token-reflexive words are pseudo-words that can be eliminated. Thus, like Russell, Reichen- bach first reduces all indexicals to one kind, which he then attempts to eliminate from the object language. We will seewhether this treatment isfeasible in the discussion of Kaplan below (Sect. 3.2).
Reichenbach's application of these ideas to one par- ticular class of indexicals, viz. the tenses of verbs, in which he postulated the need for distinguishing several different contextual factors, has exerted a considerable influence on many linguistic theories of tense, and on Hans Kamp's theory of double indexing discussed below (Sect. 3.1).
3. Model-theoretic Semantics
For a variety of reasons, in natural language semantics verb tenses have received the most attention of all kinds of indexicality, especially through A.N. Prior's work in tense logic. Prior rejected Reichenbach's treat- ment, which he thought would lead to an undesirable proliferation of reference points for the more complex tenses; in his own approach, these could be easily treated by the repeated application of a small number of tense-logical operators. Richard Montague applied these ideas to natural language semantics.
3.1 Montagovian Pragmatics; Index Theory
Richard Montague defined pragmatics as the study of indexicals, and proposed it should study the notion of truth, not only in a model or under an interpretation as in semantics, but also with respect to a context of use. Sentences of a pragmatic language should be interpreted with respect to an index or point of refer- ence, instead of merely at a world-time pair. An index i = <s, w, t> is a complex of all relevant aspects of the context: the speaker s, the world w, the time t, etc. An index is proper if s exists at t in w. Logical validity can now be defined as truth on every proper index in every structure. This gives a unified account of meaning for indexicals and nonindexicals alike by extending the semantic notion of world-time to that of context of use.
The standard treatment involved a single index for the time of utterance; however, Kamp (1971) showed that a double indexing was required. A proper account of sentences that combine temporal indexicals with tense-logical operators must distinguish the context of utterance (i.e., the time of speaking) from the cir- cumstance of evaluation (the time determined by the tense operators). In sentences without both indexicals and tense operators, context and circumstance coincide, but in (2) they do not:
One day you will be grateful for what I do now. (2)
In purely tense-logical terms, (2) would be true at time to if you are grateful for what I do now is true at some later time t|, i.e., if the hearer is grateful at t, for what
1
the speaker does at t . But obviously, now should refer
Indexicals
223