Eli Hirsch

This is a strikingly original, rich, and trenchant study. Its point of departure is the notion of discrimination, which is shown to illuminate a range of topics in metaphysics and epistemology, including subjectivity, observationality, sorites paradoxes, and identity criteria. A central problem involves the phenomenal character of experience. We are intuitively tempted to say that character is subjective in the sense that distinct characters must be discriminable. This seems to imply that matching experiences--that is, experiences which are indiscriminable in character--must have the same character. We find familiar cases, however, in which the experiences x and y match, and the experiences y and z match, but x and z do not match. Williamson defends the subjectivity of character by arguing that a pair of characters may be indiscriminable as presented by one pair of experiences but discriminable as presented by another pair of experiences. Hence the indiscriminability of characters is indeed sufficient for their identity, which makes characters subjective, but the matching of experiences is not sufficient for the experiences' having the same character.

The central question in this book is why it seems reasonable for the words of our language to divide up the world in ordinary ways rather than other imaginable ways. Hirsch calls this the division problem. His book aims to bring this problem into sharp focus, to distinguish it from various related problems, and to consider the best prospects for solving it. In exploring various possible responses to the division problem, Hirsch examines series of "division principles" which purport to express rational constraints on how our words ought to classify and individuate. The ensuing discussion deals with a wide range of metaphysical and epistemological topics, including projectibility and similarity, alternative analyses of natural properties and things, the inscrutability of reference, and the relevance of such pragmatic notions as salience and economy. The final chapters of the book develop what Hirsch contends is the most promising response to the division problem: a theory in which constraints on classification and individuation are seen to derive from the necessary structure of "fine-grained" propositions and the necessary dependence of some concepts on others.

The Evans-Salmon position on vague identity has deservedly elicited a large response in the literature. I think it is in fact among the most provocative metaphysical ideas to appear in recent years. I will try to show in this paper, however, that the position is vulnerable to a fundamental criticism that seems to have been virtually ignored in the many discussions of it. I take the Evans-Salmon position to consist of the following two theses: Thesis I. There cannot be objects x and y such that it is indeterminate whether x is (identical with) y. Thesis II. The only way for an identity sentence to be indeterminate in truth-value is if one of the expressions flanking the identity symbol is referentially ambiguous.] The argument for Thesis I is essentially as follows. We are assuming that the sense of identity under discussion satisfies the standard formal logic of identity including Leibniz's Law. Suppose, now, that it is indeterminate whether x is y. Since it is determinate that x is x, x differs from y with respect to the property of being determinately x, from which it follows by Leibniz's Law that x is not y. Since the supposition that it is indeterminate whether x is y leads to the conclusion that x is not y, this supposition is incoherent.

I would expect many readers of my book to want to agree with either Mark Heller or Alan Sidelle. The very idea of “rational constraints on lexicons” will immediately suggest to many people that either the constraints are of a purely pragmatic nature or there really are no such constraints. I can take some cold comfort in the fact that many philosophers will join me in rejecting, and many others will join me in rejecting, but since I have nothing to offer in place of these positions—except mystification—I’m afraid that few will join me in rejecting both of them.

We can describe languages in which no words refer to objects. Such languages may contain sentences equivalent to any sentences of English, and hence may allow for as much objectivity as English does. It is wrong to try to deal with such languages by claiming that there are more objects than those accepted by common sense ontology. The correct move is rather to acknowledge a sense in which the concept of an object might have been different. A consequence of this position is that we cannot have a general semantics applicable to every describable language in which words are referentially connected to objects. The point here is not that reference may be inscrutable, but that different concepts of ‘referring to an object’ may be required for different languages.