Word meaning and belief By S. G. Pulman (review)

998 LANGUAGE, VOLUME 60, NUMBER 4 (1984) cations, the book suffers from the lack ofa clear thesis. Some chapters seem mislabeled, and the editing throughout is annoyingly poor: there are a number ofsubject-verb agreement errors, and awkward sentences like 'The application of functional categories such as Jakobson's could be applied ...' (177). 'Linguistic' is mis-spelled in the back-cover blurb. [Barbara J. Koch, Indiana University, Fort Wayne.] Word meaning and belief. By S. G. Pulman. London: Croom Helm; Norwood, NJ: Ablex, 1983. Pp. 179.£13.95; $22.50. The starting point ofthis book consists oftwo claims made by Quine: (1) No distinction between analytic and synthetic sentences can be drawn; and (2) translation is indeterminate. In Quine's hypothetical language, gavagi could as well be translated by 'rabbit stages' or 'collection ofrabbit parts' as by 'rabbit' . Contra Quine, P argues that a semantic metatheory can be an empirical theory; hence there are some things about which we can be correct (or incorrect). With respect to the issue of analyticity, P argues that some statements are analytic, e.g. Circles are round or Bachelors are unmarried, because we cannot imagine anything not round being a circle. However, statements like Cats are animals are not analytic, for we can imagine cats being robots. P accepts Kripke's analysis that it is metaphysically necessary for cats to be animals, given that the world is the way it is. According to P, we can provide necessary conditions in our definitions, but we are not likely to find sufficient conditions: our definitions will be conditionals, but not biconditionals . With respect to the nature ofdefinitions, P rejects semantic primes, and advocates using a natural language as the metalanguage; thus John is a bachelor means that John is a physical object, living, human adult (ifthat is what bachelor means). The definition looks similar to that of Katz, but the status of the terms is different. They are simply English words, not semantic markers. P's line ofattack against Quine's other thesis, the indeterminacy of translation, is to build on psychology to show that all human beings conceive of at least some things, e.g. physical objects , in the same way. P provides an account of nameables—'things which have proper names, and things which can be referred to by a single non-collective countable noun' (60). Four principles of nameability are proposed: (Nl) Individuals which are (relatively) homogenous , continuous, and bounded are nameable . (p. 63) (N2) An individual which is a proper part of an Nl nameable is itself nameable if and only if it has a characteristic function, appearance, or behavior. (64). (N3) A sum of Nl nameables is itself nameable if they are (relatively) spatio-temporally contiguous, or the product of human (or animal ?) agency, or jointly fulfil a function not served by any of them (67). (N4) A sum ofN2 nameables is nameable only if the individuals making up the sum are adjacent to each other (69). Nl is obvious. N2 allows us to have a word for the wing of an airplane, but not for the left half of a wing. N3 accounts for names of stellar constellations and widely separated and dispersed geographical and political entities, e.g. Oceania or the Commonwealth. N4 rules out a word that would designate only the four legs of a dog. Unfortunately, P does not go into the question of which kinds of nameables are (or could be) given proper names. After we have principles for nameability, the next task is to discover the principles of categorization . Similarity—principally of appearance , behavior, and function—is the basis for categorization, but it 'can be overruled by more specific beliefs about the internal structure of the member' (73). P summarizes the work of Berlin on folk-botanical taxonomies, and of Rosch on prototypes and basic-level categories. Experiments conducted by P, using Rosch's experimental methods, are reported; but P uses verbs, whereas Rosch concentrates on nouns. P is surprised to find that subjects do indeed judge some verbs as significantly more prototypical of their category than other co-hyponyms : murder is more prototypical of kill than commit suicide, and survey is more prototypical of...