Composing questions by Hadas Kotek (review)

The immensely rich crosslinguistic empirical landscape of wh-questions has motivated many theoretical innovations in syntax/semantics throughout the history of generative linguistics. Hadas Kotek's recent book Composing questions presents the latest examples of such innovation. Based on new data gathered from both off-line and on-line judgments and careful examination of existing analyses, K offers an elegant analysis of the complex interplay of the structure and interpretation of simplex and multiple-wh-questions across multiple languages.

The central question K tackles is how wh-phrases are semantically interpreted within a formally explicit theory of the syntax-semantics interface. Since the earliest era of formal semantics, two competing analyses of the semantic contribution of wh-phrases have been put forth in the literature. According to Hamblin (1973), wh-phrases denote a set of alternatives that undergo insitu semantic composition—point-wise functional application—projecting a set of propositions as the denotation of the whole sentence. By contrast, Karttunen (1977) suggests that wh-phrases denote existential quantifiers that will give rise to the proposition-set denotation when they are scoped above the question operator residing in the complementizer position. The Hamblin-style analysis has been further developed within alternative semantics (Rooth 1985, 1992, Kratzer & Shimoyama 2002, Beck 2006, Beck & Kim 2006, Shimoyama 2006; cf. inquisitive semantics, Ciardelli et al. 2019), while the Karttunen-style analysis has been a basis for many subsequent influential analyses of questions (e.g. Engdahl 1986, Chierchia 1993, Dayal 1996, 2016).

An important tension between these two lines of analysis concerns the status of phonologically in-situ wh-phrases in the logical form (LF) of multiple-wh-questions like 1.

(1) Which student read which book?

According to the Karttunen-style analysis, a phonologically in-situ wh-phrase like which book in 1 must be scoped above the question operator for the sentence to receive an appropriate interpretation. Many researchers have used the device of covert movement at LF to represent this wide scope for the in-situ operator, an approach that is supported by a range of crosslinguistic evidence (Huang 1982, Rudin 1988, Richards 1997, 2001, Bošković 2002, É. Kiss 2002). In contrast, according to the Hamblin-style analysis, there is no semantic motivation for the wh-phrases to undergo movement, since they can be interpreted in situ. Thus, if there is any movement of a wh-phrase overtly or covertly, it must be for entirely syntactic reasons. It is within this tension that the current book is situated.

In the book, K proposes a hybrid view, where the grammar is equipped with both the mechanism for in-situ semantic composition of wh-phrases à la Hamblin and the mechanism for the scoping for wh-phrases via covert movement à la Karttunen. In this sense, K follows in the foot-steps of Pesetsky (2000), who argued for a similar hybrid picture, but goes substantially beyond Pesetsky in two respects: she offers a formally explicit theory of compositional semantics that integrates insights from both the Karttunen-style and the Hamblin-style theory, and she further provides precise conditions under which a wh-phrase undergoes in-situ composition and covert movement. The proposal enables three particularly interesting theoretical contributions: (i) a compositional semantic analysis of the pair-list interpretation of multiple-wh-questions (Ch. 3), (ii) a new framework for the analysis of crosslinguistic variation in the distribution of pair-list readings (Ch. 4), and (iii) a hybrid analysis of English in-situ wh-phrases in which they may undergo a covert movement followed by an in-situ composition (Chs. 7–8). Below, I briefly review each of these contributions in turn.

Multiple-wh-questions, such as 1 above, are known to have both a single-pair and a pairlist reading (Pope 1976, Bolinger 1978, Wachowicz 1974, Comorovski 1989, Dayal 1996). The single-pair reading of 1 is one that can be fully resolved by an answer like 2a, while the pair-list reading of 1 requires an answer like 2b for it...