Alcohol and speech By Steven B. Chin and David B. Pisoni (review)

204 LANGUAGE, VOLUME 74, NUMBER 1 (1998) The next four chapters discuss focus particle constructions in PPs, DPs and NPs, APs, and VPs and complementCPs. Most oftheevidence in these chapters comes from comparing relevant structures in German and Dutch (SOV languages) with those in English, but the final, perhaps most interesting, chapter extends the theory to complement clauses in Bengali , a strictly head-final language. Ch. 3 (86-121) contains the primary evidence for B's theory. The contrast between EngUsh, which allows preposition stranding, and German, which does not, indicates that overt movement must satisfy the CGH. The fact that focused NPs inside PPs are possible in English but not in German indicates that the CGH also applies at LF. This chapter also deals with some potential problems for B's theory. The most obvious one is that overt P stranding is not allowed in some languages where the CGH appears to be satisfied, notably Romance languages, and SOV languages, like Turkish and Japanese, which use postpositions. B explains the Romance case by adopting Kayne's view that P is not a structural governor in Romance languages. This also predicts correctly that extracting a focus particle adjoined to a DP inside PP should be ungrammatical in Romance languages. In SOV languages overt P stranding is impossible because P is a bound morpheme. LF movement should be grammatical . However, the examples for such movement cited from Turkish and Japanese concern LF movement out ofan adjunct, where the relevance ofcanonical government is not clear. APs (Ch. 5, 157-85) provide further support for B's theory. German has a number ofadjectives which can be preceded or followed by a PP complement. The fact that DP complements must always precede the adjective indicates that German adjectives, like verbs, canonically govern to the left. If the complement precedes A, and the CGH is satisfied, both overt and LF movement out of AP are grammatical, while both are ruled out when the complement follows A. Movement out of a verbal CP complement (Ch. 6, 186-250) also supports B's thesis. He shows that in German focusing a CP complement is possible only when it precedes the verb. Sentences corresponding to the English He said only that he might be late are ungrammatical in German. But it is not clear eitherfor adjectival PP complements or for verbal CP complements that movement out ofa complement following its head is ruled out, because the complement is not canonically governed. B thus develops the theory that CP and AP complements which do not need case move to the right of the head and are reselected there so that they are in a governed but not canonically governed position. Since he appeals to case theory, it would have been simpler to say that case is assigned under canonical government and that complements which do not require case can be generated to the right of V and A. It is only at this point in the book that B deals with another problem for his theory, namely that German allows what looks like long WH-movement out of a CP complement to the right of the verb. He argues that in German long distance wH-dependencies are not the result of movement but result from the insertion of a wh scope marker in the higher clause. WH-movement out of CP complements in Bengali (Ch. 7, 251-309) clearly supports the claim that in SOV languages WH-movement is constrained by canonical government. Bengali is an SOV language with two types offinite CP complements which must eitherprecede orfollow V. wh constituents in general remain in situ. B shows that wh constituents in CP complements to the right of the verb can take scope only within their own clause, while wh constituents in a CP preceding V can take wide scope. This confirms his analysis of German complements and in general provides support for his theory that LF movement is grammatical only if it satisfies the CGH. Extraction from DP and NP (Ch. 4, 122-56) poses a challenge for B. It should be ungrammatical in both languages, since D, N, and V do not canonically govern in the same...