Page 260 - Encyclopedia of Philosophy of Language

Reference
like 'tiger' or 'sloop.' These terms have received less
explicit attention in the philosophical literature, but
there has been a substantial theory of word meaning
developed in linguistics and related disciplines, a
theory known as 'componential analysis'. The idea is
simple, and is also common in philosophy: these terms
acquire their meaning from definitions. 'Tiger' means
'large, carnivorous, striped Asian feline'; 'Sloop'
means 'a one-masted cutter with a fixed bowsprit.' metaphysical intuitions. Kripke points out that
Once more, this idea is naturally seen both as a theory of meaning and as a theory of reference; the term refers to those entities which fit the definition. As a general view of the semantics of natural kind or arti- fact terms, this story has problems. Putnam has poin- ted out that tigers are carnivores is very different from the genuinely definitional triangles are three-sided, for it is both contingent and empirically corrigible. Fodor (1981) has pointed out that it is in fact extremely difficult to construct extensionally correct definitions in which the defining terms are more conceptually basic than the defined term.
There is a deeper problem for componential analy- sis. Definitions do not yield an explanation of the relationship between language and the world. Rather, in Devitt's felicitous phrase, they 'pass the semantic buck'; they explain the meaning, and hence the refer- ence, of the term defined, only insofar as some further adequate explanation can be given of the semantics of the defining terms. 'Passing the buck' is not pointless if it is passed to terms that are in some way special; terms whose semantic properties can be explained, not just presupposed. Logical empiricists thought that observation terms were special, but the official 'anti- psychologism' of that movement, the ban on appeal to psychological process, prevented them from giving any explanation of the semantics of observation terms. Many contemporary semantic theories are far from being anti-psychologistic. But the clear failure to reduce theoretical language to observation language, and the difficulty of reducing quite ordinary terms to a reasonably small set of more basic ones, threatens our ability to profitably 'pass the buck.'
3.3 Singular Terms
Kripke has constructed a similar argument about names. He has argued that names cannot be abbrevi- ated descriptions because a name has the same bearer, or none, in every possible world. In his terminology, names are referentially 'rigid.' But the reference of a description, or even a cluster of them, can vary from possible world to possible world. Hence there can be no semantic equivalence between names and descrip- tions. In some respects then, Kripke supports John Stuart Mill's view of names, namely, that while names, like other terms, might have connotations, those con- notations do not determine the reference of the name; 'Kripke' does not refer to Kripke in virtue of any connotation of 'Kripke.' Proponents of 'direct ref-
'descriptivist' theories of names imply that we can use a name to designate its bearer only if we can identify that bearer, if not in a police lineup then at least in knowing some unique characteristic of the bearer. He goes on to illustrate the implausibility of this view. Human language (unlike many animal systems of communication) is not stimulus bound. We can speak of the elsewhere and elsewhen. Given our cognitive limitations, and the stimulus freedom of language, referring to Einstein had better not closely depend on knowing about Einstein. Putnam has made much the same kind of point about general terms. But the most critical point against the description theory of names is not that it forges too close a link between referring and knowing, for no doubt suitably cunning amend- ments to the theory might be able to finesse that prob- lem. Rather, it is a buck-passing theory.
3.4 Causal Chains
So what kind of theory might avoid buck-passing? Since the early 1970s or so, causal hypotheses about reference have been explored (see Field 1972; Devitt 1981; Devitt and Sterelny 1987). The central idea is that the referential relation is constituted by a causal chain between a term token, and its reference. Let us follow Putnam in imagining how this might go for natural kind terms. Joseph Banks disembarks at Bot- any Bay, and is confronted by the first kangaroo any- one in his speech community has seen. There is causal commerce between him and the roo, as a result of which Banks says 'We will call these animals "kanga- roos".' There is a causal chain between the term token and its bearer in virtue of which the term refers to kangaroos. Others hear him talk of kangaroos, or read his letters about them, and acquire a causal con- nection to kangaroos indirectly, via Banks's linguistic behavior. They 'borrow' their reference from his.
Naturally, this sketch is oversimplified. The intro- duction of a natural kind term involves not just an ostensive but also a structural component. A term is introduced into the language by perceptual contact with samples of the kind. But the extension of the term goes beyond these samples, concerning all those objects having the same nature as the samples. In a first approximation, the nature of kinds is determined by internal structure. So the term 'kangaroo' applies to all animals having the same internal structure as that of the samples; the term 'plutonium' to that stuff having the same internal structure of the samples pro-
238
erence' theories of names have taken his arguments to support the stronger thesis that there is nothing to the semantics of names but their denotation; more about this in Sect. 4.
Kripke's provocative metaphysical views have aroused much debate, but Devitt has pointed out that his arguments 'from ignorance and error' are clearer and more convincing, for they do not rest on corrigible