Uttering trees (review)

The study of the syntax-phonology interface is emerging as a highly productive area of research. This book demonstrates just that. Two novel proposals about conditions imposed on syntax by phonology are developed in the two main chapters of the book, with far-reaching consequences for future research. Each proposal reveals new aspects of phenomena, unquestionably interesting, and invites further exploration of them. Understandably, there is a direct correlation between the originality of the proposals and the explorative style of working them out. Inevitably one is led to appreciate the gains that each represents, putting aside the many questions they raise.

R’s first proposal is that linearization statements of the type proposed by Kayne’s (1994) linear correspondence axiom (LCA) are subject to a new well-formedness condition called distinctness (there is no commitment to the details of LCA, including directionality of heads). Distinctness is about the content of linearization statements. R suggests that linearization statements cannot contain ‘more than one node of the same kind in an asymmetric c-command relation’ per cycle (141), perhaps because they are not distinct enough to be interpretable as an ordering statement. The question of what nodes are ‘distinct enough’ takes considerable discussion, and may vary from language to language. Linearization of pairs of nodes of the form <α, α> are banned if they have the same label in some languages (although nodes are identified by features rather than by labels; see, for example, the discussion on Polish complementizers in section 2.3.1); in other languages, only DPs that ‘have identical values for case, grammatical gender, and/or animacy’ create problems (6).

Distinctness is only relevant for functional heads. R exempts lexical heads from distinctness on the grounds that the timing of vocabulary insertion is different for each: late insertion for functional heads and early insertion for lexical heads. There is a strongly speculative character to this aspect of the proposal, details are not worked out, and the distinction does not always yield the desired results (p. 210, n. 34).

R is to be commended for the heterogeneous crosslinguistic materials dealt with by distinctness, including aspects of case phenomena, bans against two consecutive verbs, and word-order matters. The one common factor in all of them is that the content of the linearization statement is insufficiently distinct.

R convincingly argues that distinctness is about syntactic structure, not about linear adjacency. He follows Chomsky (2000, 2001) in identifying spell-out domains with strong phases. In addition to transitive VP and CP, R treats PP and KP as strong phases (K is exempt from its phase status at least in Irish; see section 2.4.2.2). R’s use of phases is both extensive and skillful throughout. A neat result is the difference that emerges in contexts of A and non-A movement. NP-traces and wh-traces yield different results in sequences with two string-adjacent verbs in English: *John was seen __ leave (17), derived by NP-movement, and How many prisoners did you see __ leave? (19), derived by wh-movement. The key element in this difference is the phase boundary of the strong VP-phase in the example with wh-movement, absent in the passive example (passives have weak VP-phases). With the addition of a phase boundary, spell-out can linearize the conflicting nodes in two separate cycles.

Four specific syntactic mechanisms serve the principle of distinctness, two concerning structure (adding or removing structure), and two concerning movement operations (creating or suppressing movement).

As we may expect, a distinctness perspective yields entirely new ways of grouping grammatical material. For example, the fact that nouns do not take nouns as complements (the destruction (*of) the city) and certain relativization patterns familiar from English (and Romance) come out as belonging together (*a man who to dance with). In both domains, a KP between two DPs adds a spell-out boundary, allowing each DP to be part of a separate spell-out operation. Differential marking is involved in whatever grammatical domain DPs interact. Undoubtedly, the fact...