Prolegomena to inferential discourse processing By Roger G. van de Velde (review)

BOOK NOTICES 733 nicative function of the presence or absence of subject pronouns. E demonstrates that they are used to force an unexpected or 'non-favored' interpretation of a message—i.e. for emphasis, to avoid ambiguity, and as indicators of subject switch. E struggles to conflate these functions, but he should probably have concluded that they all serve to maintain face through observing the cooperative maxims of quantity and manner. J. Kaplan, 'VP anaphorchoice in discourses' (47-59), is tortuously presented, though not uninteresting . K says that very few or no instances of Eng. VP deletion or do it/so are found in environments where so do or too may occur (viz. where only subjects are compared). Too differs from so do in being much less frequent and in normally having a pronominal subject (most often /); it tends to be contrastive. R. Tomlin's excellent 'Interaction ofsubject, theme, and agent' (61-80) argues that the thematic NP will be clause subject; but if no NP is more thematic than the others, the agent will become subject. (A thematic NP conveys information significant to the development of the topic.) T proves his point from play-by-play descriptions of ice hockey. To his credit, he emphasizes the need to keep the encoding devices (forms) distinct from the semantic phenomena encoded; and he also recognizes the need for something more concrete than the analyst's intuition in deciding what is thematic. (He puts his faith in cognitive psychology, but I would not.) Finally, R. Longacre, 'Discourse peak as zone of turbulence' (83-98), argues that discourses have a possibly recursive structure: prepeak , peak, and post-peak. An action peak is the point ofmaximum tension and confrontation in a story; a second action peak (if there is one) is a decisive event that makes the resolution of the story possible. There may also be a 'thematic or didactic' peak. In different languages, peaks are marked by (a) repetition and paraphrase ; (b) shift of clause type, tense, aspect, or person; (c) rhetorical question, dialog, or direct speech—where these have not been used before; (d) varying clause, sentence, or paragraph length; (e) sound symbolism; (f) packing the action line by increasing ratio of verbs to non-verbs; (g) special markers; or (h) reduction of conjunctions. Beyond the sentence is a useful little book. One wonders why it took five years to get into print. [Keith Allan, Monash University.] Prolegomena to inferential discourse processing. By Roger G. van de Velde. (Pragmatics & beyond, 5:2.) Amsterdam & Philadelphia, Benjamins : 1984. Pp. viii, 100. $20.00. 'No realistic account of discourse interpretation can be given when the cognitive and communicative function of human natural language is disregarded ... considering the fact that verbal , coverbal, and nonverbal actions are constitutive parts to communication, it is essential to take the inferential ingredients of communicative actions into account' (pp. 1, 2). The ingredients which V discusses are syntactic processing , semantic processing, and action processing—with the demonstrably valid assumption that these 'operate in interactive and integrative ways' (79). V presents a theory of language use, not a theory of language (or a component subtheory thereof); but about a fourth of the text is taken up with aspects of cognitive or action processing which, although not irrelevant to understanding language use, seem to me to lie outside the linguist's discipline . The boundaries of linguistics are a matter of opinion, but the discussion of apraxia in§4.4.2 is surely way outside. Generally, though, Vs book makes good sense, and backs up his claims with satisfactory argumentation and some empirical evidence. He maintains the distinction made by M. A. K. Halliday & R. Hasan between a formal category cohesion (anaphora etc.) and the semantic property coherence, and he identifies several types of the latter (81): language-bound, cognition -bound, reference-bound, action-bound, communication-bound, and personality-bound (some of these go beyond linguistics, and therefore beyond semantics). V demonstrates the pre-eminence of coherence; thus the recognigion of cohesion may depend on the coherent construction of what I would call a world from the text. Again, a text can be coherent even though its cohesion is, by normal standards, deficient —as in...