Reviews

142 Links & Letters 1, 1994 Reviews demonstrates that they can be argued to sentation. MIT Working Papers in display a number of functional categories Linguistics 1 O. 43-74. which partially paralel those ofsentences. POLLOCK, J.-Y. (1989). Verb Movement, In English, for example, noun phrases can UG, and the Structure ofII? Linguistic be analysed as D(eterminer)Ps when they Inquiry 20.3. 365-424. contain a Det element acting as the head of the whole constituent or as AGRPs Josep Quer when they contain an AGR element that Department de Filologia Catalana normdly surfaces as genitive subject. Tak- Universitat ~~~o~~~~ de Barcelona ing into account the role played by these elements, many idiosyncratic behaviours of substan-tives (mainly, of the head noun and its subject) receive a principled explanation. MARGARET SPEAS, Phrase Structure in The final part 0f this last chapter is Natural Language. J-JDordrechr: devoted to motivate the idea that Comp elements are nominalisers in the sense Kluwer, 1 990. 306 pages. that their projection on the top of sentential clauses allows for the latter to function T~mSf~rmational grammar (Chomsky as arguments. 1965), the Gouernmentand Binding (GB) As has been noted in the previous m~del in particular (Chomsky 198 11, is a paragraphs, Ouhalla's work draws on theo~ of Universal G~~~~ar (UG), the mainstream research in current syntactic set of principies and parameters that are Studies and rnakes a very important conclaimed to be an innate part of the human tribution to the developrnent of some of language faculv Central to GB theory is the ideas put fonvard in the literature. phrase structure (PS), the hierarchical The &eory of parameters and language representation of the structure of the senvariation is one of the central issues at tence; it is to elements in this structure stake in the present stage of linguistic that the transformation move alphais clairned theory and as such Ouhalla's book should to apply deriving from one PS reprebecome a point of reference in the debate. sentation a new PS representation constrained by universal structural and licensing requirements. A central ques- Referentes tion, intuitively stated, is how does PS start, or where does it come from? A se- BAKER, M. C. (1988). Incorporation: A cond question regards the universalit~ of Theory af Grammatical Function the hierarchical representations which Changing. Chicago: Chicago Univerhave been successful in describing confisity Press. gurational (Con) languages like English: BORER, H. (1983). Parametric S'ntax: do n~nconfi~urational (NCon) langua- Case Studies in Semitic and Romance ges like e.g. Warlpiri, which have beeri Languages. Dordercht: Foris. argued not to have a hierarchical PS (Hale FUKUI, N. and M. SPEAS. (1986). Speci- 1983), differ in some fundamental way in fiers and Projections. MIS Working their syntax? NCon languages pose a se- Papers in Linguistics 8. 128- 172. rious challenge to GB theory since in CHOMSKY, N. (1981). Lectures on Gougeneral the evidence in favor of a hierarernment and Binding. Dordrecht: Fochical PS is weak and further, the evidence ris. a ainst such structure is strong. It is this CHOMSKY, N. (1986). Barriers. Camc allenge that Margaret Speas' (1990) bridge (Mass.): The MIT Press. F, Phrase Structure in Natzdral Language CHOMSKY, N. (1988). Some Notes on addresses. The claim Speas (S) defends is Economy of Derivation and Reprethat PS is projected from lexical items Reviews Links & Letters 1, 1994 143 thernselves in a highly restricted rnanner and that the variation found is associated to variation in these lexical iterns frorn language to language; there is no fundamental difference between Con and NCon languages. In what follows 1 briefly surnrnarize each chapter of S's book, highlighting its conclusions and particularly interesting discussion; also, as appropriate 1 discuss its weak points. Overall though, we will see that her conclusions are quite well-justified. Ch. 1 introduces sorne of the basic notions which becorne irnportant later in the book; 1 will sirnply introduce such notions as necessary &d-say nothing further about Ch. 1. In Ch. 2, S rnotivates a view of PS which is different in a fundamental way frorn preceding versions: no "rules" or schernata are required. PS is ~roiected freelv frorn lexical iterns 1 .I thernselves as long as every elernent can be licensed in a certain way. Thus, the structure is licensed only if the elernents in the structure are licensed. Licensing involves three particular "theta" (general lized thernatic ) relations which are realized under sisterhood: (1) discharge, one lexical itern "needs" another as in the case of a verb needing an object; (2) rnerger, two iterns with the sarne type of requirernent "rnerge" into one category which retains this requirernent; and (3) binding, two iterns with the sarne requirement satis& each other (Higginbotharn 1985). An interesting result of this restricted set of licensing requirernents is that D- structure (DS), which is the level of PS associated with the lexical iterns thernselves. can contain onlv elernents which are thernatically related; an elernent which is not associated with the PS tree by any of the above three rnechanisrns is not licensed at DS. This yields a quite restricted view of DS: everything rnust be licensed thernatically. There will be no "extra" ele- 1. Thetalthematic relations are the inherent relations held between a lexical item and the elernents associated with it; e.g. the relation between a predicate and its arguments. rnents allowed at that level. S argues that this is correct by pointing out that certain types of adjuncts (ty ically thernaticall~ unrelated elernents) g ehave syntactically as ifthey are not present at DS. Specifically she argues, following Lebeaux (1988), that such ~hrases are adioined to the PS later in ;he derivation by a "generalized transforrnation (an operation which combines PSs). S's view of PS does not rely on the notion of rnaxirnal pbojection (XP) or rninirnal projection (X ); and she specifically argues against the notion of a defined "interrnediate" projection (X') by showing that no rule of ra refers to such a level. XP an f x are sirnply "' the two "ends" of the PS projection, what she refers to as a Projection Chain, analogously to chains of other syntactic elernents which have a ead and a "foot" ''b (in this case XP and X , respectively). By showin~ that PS can be built based on independently rnotivated lexical requirernents, S rernoves PS "rules" or schernata frorn UG, elirninating redundancy and geatly reducing the power of the PS cornponent. Having spent Ch. 2 arguing for a universal theory of the rnap ing frorn lexical iterns to DS, S faces the P orrnidable challenge posed by the so-called NCon languages, which appear to differ fundarnentally in their PS frorn Con languages like English. Her rnain clairn in Ch. 3 and Ch. 4 is that there is no variation in the principles of projection defended in Ch. 2, which include those rnentioned above as well as several other GB principles: UTAH, that thernatic roles (like agent and therne) are hierarchically arranged and project into the syntax in a specific universal order; the Theta Criterion, which roughly states that every theta position rnust be discharged once; and the Projection Principie (PP), which in SS sirnplification just states that UTAH and the Theta Criterion rnust hold at every level of representation. The result of the PP is that if rnove alpha applies to an elernent, some "trace" of the elernent rnust be left so that its original thernatic relations can 144 Links & Letters 1, 1994 Reviews be recovered; this trace is then subject to other svntactic reauirements. I The specific claim S argues against is that of Hale's (1983) Dual Representation Hypothesis (DRH). The DRH states that universally, every language has two separate syntactic "structures": a lexical structure (LS), where hierarchical lexical and thematic information is represented, and a phrase structure (PS), a "flat" (not hierarchical) representation of the items in a sentence. LS and PS are related by "rules of construal". The DRH further states that in NCon languages the Projection Principie holds only of LS, not PS. Thus, move alpha can apply to PS (vacuously) with-out having to leave traces. First, S shows that the supposed diagnostics of nonconfigurationality do not, in fact, pick out two lists of fundamentally different languages. Many languages show characteristics of both types. Second. she shows that the DRH allows a great amount of redundancy since in NCon languages the mapping from DS to PS. via move aloha. is redundant as I ' it is vacuous, and in Con languages the mapping between LS and PS, via rules of construal. is also redundant. If the redundancy is removed, however, we are left with one type of language with LS and PS (NCon) and a second type with DS and PS (Con); S argues that this model must be rejected because such a fundamental difference between the two types of languages cannot account for the fact (point 1 above) that languages do not divide cleanly into one type or the other. Third, she examines each of the supposed NCon languages in detail and shows three things: (1) each language does in fact display certain asymmetries which could support a hierarchical PS. Although it should be noted that many of these asymmetries involve binding (the theory of the distribution of reflexives, pronouns and "names"); and if binding is not svntactic. as has been ar~ued bv a U number of researchers (e.g. Williams 1987), these arguments are weakened but her main ooint remains valid: (2) the , %, arguments in favor of a "flat" structure can be reinterpreted in her framework to be consistent with a hierarchical PS; and (3) in Ch. 4, S shows in a detailed study that even Navajo, which appears to show independent evidence for the DRH, falls into the framework she has developed throughout the book. Especially interesting in al1 of these cases is the fact that the parameters required to account for the variation observed are al1 consistent with the "Lexical Parameter Hypothesis", first articulated in Borer (1984), that language variation is to be associated with variation in particular lexical items. Thus, the lear~ in~ of oarameters is reduced to the learn- 0 ' ing of lexical items; the principles of projection and other universal (and presumably, innate) syntactic principles are not subject to pararnetrization. The conclusions that this work clearly supports are the following: (1) PS is projected from lexical items and is constrained by universal principles of projection, not PS schemata; (2) variation among languages can be traced back to variation among particular lexical items; and (3) universal principles of grammar, in particular principles ofprojection, are not subject to ,parametrization. S's contribution, then, nicely advances us towards the goal of understanding Universal Grammar, by showing that fundamental principles of grammar need not be parametrized to account for what aoDear on the surface to be fundamental dikferences among the world's languages. BORER, H. (1 984). Parametric Syntax. Dordrecht: Foris. CHOMSKY, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. (1981). Lectures on GovernmentandBinding. Dordrecht: Foris. HALE, K. (1983). Warlbiri and the Grammar of Nonconfigurational Languages. Natural Language and Linpistic Theory 1.5-47. HIGGINBOTHAM, J. (1985). On Semantics. Linguistic Inquiry 16.547-594. Reviews Links & Letters 1, 1994 145 LEBEAUX, D. (1988). Language Acquisi- Natural Language and Linguistic tion and the Fom of the Grammar. Theoíy 5. 151-180. U.Mass, Amherst: Ph.D. diss. SPEAS, M.J. (1990). Phrase Structure in Jeffrey T. Runner Nahra/Language. DOrdreht: Kluwer. University of Massachusetts at Amherst WILLIAMS, E. (1987)- Im~licit A%u- & Universitat Autbnoma de Barcelona ments, Binding Theory and Control.