1. It is true, for example, of the views expressed in Grice (1957, 1969), Austin (1962), Lewis (1969), Strawson (1964), Schiffer (1972), or Bach and Harnish (1980). [BACK]

2. One could offer a theory of mind that employed another notion of "representation" (e.g., pictures, maps), but it would be a distinctly separate theory from CTM. It would have different strengths and weaknesses (e.g., it would not yield a syntactically based account of productivity and systematicity) and would require a separate analysis. Even more fundamentally, however, it is not clear at the outset that there can be a general account of representation, because it is not clear that there is one property called "being a representation" that is common to symbols, pictures, maps, schematic diagrams, flow charts, and the other things to which the word 'representation' is applied. Any attempt to supply a "general continue

account of representation" would have to wait for a careful analysis of several specific kinds of "representation" (i.e., analysis of the uses of the word 'representation' as it is applied to apparently distinct paradigm examples) before one could decide whether there was some feature they had in common, or whether they were called by the same name in virtue of family resemblance, or simply homonymously. This would be a worthy investigation, but stands outside the scope of this book. It is worth noting, however, that only one basic kind of general account has historically been offered. On this account, put forward by writers as diverse as Thomas Reid, Edmund Husserl, A. J. Ayer, and Daniel Dennett, R is a representation of X just in case some P uses R to stand for X . One might note that this "general" account carries the same dangers of interpretive regress as the semiotic account presented in chapter 4. [BACK]

3. This criticism was raised most forcefully by Rob Cummins, who read a draft of the manuscript for this book. It was also raised by one anonymous referee of an article developing the same view. [BACK]

4. Richard De Witt makes a similar point in his "Vagueness, Semantics, and the Language of Thought" (1993). [BACK]

5. One might additionally observe that there is no such thing as "the pattern" associated with, say, the letter rho. As Hofstadter (1985) has argued, there are infinitely many patterns that can count as rhos. Moreover, what can count as a rho in situ is highly context-dependent, so it will not do simply to take the whole set of patterns that can ever count as rho and treat that set as constitutive of rhohood. This is arguably even more true with phonemes than with graphemes. (Opera goers are quite familiar with this: on the high notes, all of the vowels tend to gravitate towards [a].) [BACK]

6. It was Rob Cummins who initially made me see this point during his NEH Summer Seminar on Mental Representation in 1991. Rob was kind enough to show me that the point was already pretty clearly implicit in my analysis of symbols and syntax. But, however clear the implication might have been, it had been entirely lost upon me until pointed out. To the best of my knowledge, neither he nor anyone else has really explored the point in print. But as far as I know, the original insight was his and not mine. [BACK]

7. Indeed, Davidson takes this view to the logical conclusion that differences in usage between speakers amount to a difference in language, since a language is determined by a unique mapping from expressions to interpretations. Thus it is idiolects (at particular times) that are languages in Davidson's sense. There is no such "language" corresponding to the public language English, since there are many variations upon English in individual idiolects. [BACK]

8. The word 'model' here is, of course, ambiguous. In the terminology of settheoretic modeling, it is the interpretation of the set-theoretic construction—i.e., the mathematical domain—that is called a "model." Here I am using terminology in precisely the opposite way, taking the set-theoretic entity to be a "model" in the sense that one speaks of "models" in the sciences. Thanks to Sanford Shieh for alerting me to possible confusions on this point. [BACK]

9. It is the problems with semantically closed languages that lead Tarski to another important conclusion: namely, that the T-equivalences for a language L and the truth theory T(L) may not be articulated in L (else L would be seman- soft

tically closed, hence inconsistent, hence unsusceptible to a truth definition), but must instead be articulated in a metalanguage M , which contains L as a proper subpart, but also is "essentially richer" in that it contains variables of a higher logical type (Tarski 1956a: 55). To this Tarski adds the following crucial point, articulated not so much as a logical necessity as a desideratum:

It is desirable for the metalanguage not to contain any undefined terms except such as are involved explicitly or implicitly in the remarks above, i.e., terms of the object language; terms referring to the form of the expressions of the object language, and used in building names for the expressions; and terms of logic. In particular, we desire semantic terms (referring to the object language) to be introduced into the metalanguage only by definition . For, if this postulate is satisfied, the definition of truth, or any other semantic concept, will fulfill what we intuitively expect from every definition; that is, it will explain the meaning of the term being defined in terms whose meaning appears to be completely clear and unequivocal. (ibid., 54-55)