^1. Austin, J. L., Sense and Sensibilia (reconstructed from the manuscript notes by G. J. Warnock) (Oxford, 1962).Google Scholar

^2. Some of the formulations of this paper owe a great deal to Harrison's, BernardMeaning and Structure: An Essay in the Philosophy of Language (Harper & Row, New York, 1972).Google Scholar Part of the intent of this paper, although I cannot yet make it adequately explicit, is to try to show how Harrison's linguistic devices may come to be modified through argument.

^3. First edition (Macmillan, London, 1940). My reference is to the paperback edition of 1964.

^4. It strikes me as fantastic, and, were it not so much part of an unexplored post-Wittgensteinian orthodoxy, one would have to say it had been ‘insinuated’.

^5. In particular, I am not trying to adhere to Austin's terminology.

^6. This is itself a very important point. I think it is frequently overlooked that dictionary entries are of different sorts. In the case of substantives such as ‘sheep’, it looks as if we are given lists of features which, say, something must have to be a sheep; and, further, it looks as if this unstructured list gives a paraphrase for ‘sheep’. Then for each item of our very large vocabulary of substantives, we would have to learn such a list:

but my claim is that with the entries for ‘sheep’, ‘goat’, and ‘ox’, we have an implicitly structured list:

In the first case, however, it is not at all clear why—logically—the list of criteria is finite; and, in fact, it is very difficult to see how such a list will pick out just those things which are sheep and nothing else: marginal cases will always force us to add to, and amend, the list. In the second case, the criteria are taxonomic criteria. Because the list is structured, the criterion of being gregarious needs only to distinguish among ruminants; and the process of classifying is finite. More simply, in the first case, the list of criteria is matched with one word, where, in the second, each criterial choice-point has at least two exits. Thus the second has something like generative power, where the first does not.

These remarks, and the whole emphasis on criteria as properly taxonomic criteria, owe very much to Bernard Harrison, although I am not sure whether he would agree with certain of my claims (op. cit., esp. sections 3.7 and 3.8).

^7. Austin also wants to exclude, or take as excluded, the nameless, sceptical horrors of exploding goldfinches, etc.: he can only succeed in this by not letting them be specific ways in which a thing can fail to be a real S. Otherwise, we would seem to succumb to the sceptic.

^8. Cf. also: ‘… real pearls, real ducks, real cream, real watches, real novels, and the rest—all those uses of “real” which Ayer overlooks entirely’ (p. 83).

^9. Talk of souls here is a version of the substance talk discussed in section 6. The monstrousness of intermediate cases and the popularity of doctrines of human exceptionalism (the latest of which is Chomsky's) do not show the naturalness of the categories, but reflect our moral order and its sanctions. Cf. Douglas, Mary, Purity and Danger (Routledge, London, 1966, and Penguin, 1970).CrossRefGoogle Scholar

^10. I am deliberately shying away from giving some account of the logical relations between ‘real’ and ‘really’. They impress me as complex and perhaps largely irrelevant to my argument.