Referring and non-referring phrases: A study in the use of the gerund and the infinitive By Bent Conrad (review)

682LANGUAGE, VOLUME 62, NUMBER 3 (1986) Referring and non-referring phrases: A study in the use of the gerund and the infinitive. By Bent Conrad. (Publications of the Department of English, University of Copenhagen, 11.) Copenhagen: Akademisk Forlag, 1982. Pp. 188. Reviewed by Arnold M. Zwicky, Ohio State University Numerous works on English grammar have examined the use of the gerund (G: Seeing them is nice; I like seeing them) and the infinitive (I: To see them is nice; I like to see them). As Conrad observes in his first chapter, traditional and pedagogical grammars provide (informal) semantic analyses: some assign different meanings to these non-finite verbal NP's in subject vs. object position, while others assign a single meaning to each, regardless ofposition. When only one of the two NP types is possible, this is assumed to result from an incompatibility between the meaning of the other type and the meaning of the construction in which it occurs. By contrast, structuralist and transformational treatments tend to assume that any distributional restrictions are a matter of (semantically arbitrary) syntactic rule: when both NP types can occur, they are 'free variants, which are only distinguished by relative frequency or stylistic differences' (20). C rejects extreme versions of both approaches, and he maintains that any 'meaning' differences are a matter of pragmatics rather than semantics. He proposes that distinct pragmatic effects are associated with the two types— 'The speaker undertakes two different kinds of commitment by choosing one or the other of the two forms' (91)—but that a number of 'ad hoc rules' (18) operate to select only one form in certain syntactic contexts (e.g. only G as the object ofa preposition, only I as the object ofthe verb expect) and to permit only one interpretation in certain syntactic contexts (e.g. only the G interpretation for either form as the object of bother or manage). He does not mention the 'double -ing' constraint that prohibits It was beginning raining as parallel to It was beginning to rain and It began raining. Cs analysis is restricted to these non-finite forms as subjects (G's in Chap. Ill, l's in Chap. IV) and as objects or 'object-like complements' (Chap. V). Wisely, he makes no attempt to cover the many constructions in which only l's occur, such as adverbials (in order to go), postnominal modifiers (the person to see), or adjective complements with personal subjects (Kim is easyleager to please); C does not propose to assign an interpretation to all l's, but only to those functioning as NP's. His analysis explicitly excludes subject examples with formal it (It is nice {seeing them I to see them}), on the grounds that the it might contribute pragmatic effects of its own (90-91). In addition, by taking gerunds as one of his two types, he also excludes gerundial NP's with expressed (genitive) subjects (his minding the baby) and their infinitival counterparts (for him to mind the baby), though these should be expected to submit to the same analysis as the forms with unexpressed subjects which C does analyse. The crux of Cs proposal is that G's refer uniquely—'the speaker implies that the referent is such that identification is possible (i.e. that there is an individual phenomenon)' (45)—while l's refer non-uniquely, since no such identification is possible. The notion of unique reference is developed in Chap. II, a loosely organized discussion of the literature on reference and specificity that consumes about a third of the book. Unfortunately, this chapter introduces unique reference as a pragmatic property of certain tokens of indefinite NP's, e.g. a fish (for which a test for a uniquely referring NP is that it must introduce a discourse referent, in L. Karttunen's sense); REVIEWS683 and it extends the notion to non-finite verbal NP's (for which, as C notes, the discourse-referent test is inapplicable) with almost no elucidation. In any event, the proposal is intended to predict that a subject G 'always refers to one locatable instance, or several locatable instances, of actions, processes, states &c' (93); a subject I has the negative property that...