Events, phrases, and questions (review)

In this book, Robert Truswell takes a novel perspective on the long-standing issue that extraction is more restricted out of adjuncts than out of weak islands, yet is not entirely impossible (in contrast to strong islands), by taking into account not just narrow syntax but also the role adjuncts play in event structure. He proposes that adjuncts are weak islands, with extraction being further restricted by the single event grouping condition (SEGC).

1. 1. SINGLE EVENT GROUPING CONDITION: An instance of WH-movement is legitimate only if the minimal constituent containing the head and the foot of the chain can be construed as describing a single event grouping.

(157, ex. 64)

An event grouping is defined as a set of (core or extended) events, the subevents of which necessarily spatiotemporally overlap and where at most one (maximal) event is agentive. Given that purely semantic constraints should not restrict the distribution of syntactic operations, the SEGC is ultimately embedded within Gibson's (2000) dependency locality theory, according to which the processing of filler-gap dependencies, measured in discourse referents (including events), is costly due to the requirement to keep a filler in memory while processing other linguistic material, and to integrate fillers structurally when a gap site is encountered.

The book consists of two parts, as well as an introductory and a concluding chapter. Chs. 2-4 introduce a model of event structure that is largely independent of syntax. Chs. 5-8 relate this model to the empirical findings in the domain of extraction out of adjuncts. Ch. 1 provides an overview of the history of syntactic treatments of such extraction, arriving at the currently unresolved debate about adjuncts constituting weak or strong islands. Observing that acceptability of such extraction is a gradient phenomenon and largely based on semantic considerations, T concludes that the restrictions cannot be purely syntactic in nature, unlike commonly assumed. He addresses four empirical puzzles for strictly syntactic approaches, to be accounted for in Part 2, as summarized in Ch. 9.

• • RESTRICTED EXTRACTION PUZZLE: Some classes of adjuncts freely allow for extraction, while others do not.

• • RESTRICTED ANSWERS PUZZLE: The nature of the assumed answer to a question formed by extraction from an adjunct can render extraction acceptable (see 2 below; cf. exx. 46 and 47, p. 31).

• • INTERPRETIVE PUZZLE: Extraction out of bare present participial adjuncts (BPPAs) is possible only with telic matrix VPs; the adjunct specifies a causal relation with accomplishment matrix VPs (3a below), but a strictly temporal relation with achievements (3b, in contrast to nonachievements in 3c) (cf. exx. 50 and 54, p. 32).

• • UNLIKELY ANTILOCALITY PUZZLE: Extraction can improve if more material intervenes (3a vs. 4 below; cf. ex. 56, p. 33), whereas under current locality theories, more material intervening, if anything, blocks extraction.

1. 2.

   1. A. Which book did John design his garden after reading?

   2. B. An introduction to landscape gardening./#Finnegans Wake.

2. 3.

   1. a. ?*What did John drive Mary crazy trying to fix?

   2. b. ?*What did John die whistling?

   3. c. ?*What did John paint this picture eating?

3. 4. ??What did John drive Mary crazy fixing?

Part 1, 'The structure of events', investigates the conditions under which single (Ch. 2) and multiple VPs (Ch. 3) describe a single event. T argues for a maximal CORE EVENT template composed of a process and a culmination, which are necessarily contingently related by direct causation (similar to Moens & Steedman 1988), resulting in accomplishments or achievements (differentiated not at the level of event structure, but with respect to agentivity). States are argued to lack a process and a culmination, activities to consist of processes, and points to consist of culminations (e.g. notice). Based on empirical and experimental data, multiple VPs are argued to be able to describe extended events, which, in addition to causation, can involve another contingent relation between an agent and two events—enablement.

Building on Higginbotham's (1985) proposal that the two λ-abstracted event variables of a VP and an adverbial phrase are...