LECTURETTE #11: SOME CONTROVERSIAL HYPOTHESES FROM RICHARD KAYNE I spoke in my last lecturette of how work in X-Bar Theory has progressed towards the goal of the complete elimination of language- and category- particular PS rules. Before proceeding to discussing Theta Theory, Case Theory, etc., i'm going to briefly discuss a couple of theoretical propo- sals by Richard Kayne, both of which develop this tendency to an extreme with which i disagree. I'm mentioning them mainly so that you will be acquainted with them, since they are much discussed, and also much taken for granted, amongst REST researchers. The first is the 'Binary Bran- ching' hypothesis, presented most accessibly in Kayne's 1984 book Connec- tedness and Binary Branching (Studies in Generative Grammar #16; Dordrecht: Foris). The other is the hypothesis presented in his book 'The Antisymmetry of Syntax' (Linguistic theory monograph #25; Cambridge, MA: MIT Press). These hypotheses wold obviate the need for any parameter- setting in the generation of DS; just like LF, DS would be exactly alike in all languages. So one can understand why Kayne and other theorists are attracted to them. In both cases Kayne is to be given credit for bold attempts explicitly to restrict the power of the theory by limiting the range of structures it can generate, but in both cases i believe des- criptive adequacy is lost. The 'Binary Branching' hypothesis holds that any given non-terminal node may dominate at most two daughters. This is obviously in some sense a default; we tend to assume that a phrasal node dominates precisely two daughters, the head and either a specifier or a complement, unless we have good reason to think otherwise. Obviously, a non-terminal node may have only one daughter; both of the NP nodes in (1) directly dominate single lexical items. (1) [I](NP) like [peas](NP) The question arises as to whether it is ever possible, or appropriate, to describe a phrasal node as immediately dominating more than two nodes. With regard to the NP in (2a), the question is whether the two adjectives 'big' and 'blue' are independent modifiers of 'notebook' as in (3a) or both members (perhaps conjoined?) of a single modifying AP as in (3b). Semantically, it would seem that they are distinct, since the notebook's size and colour are presumably independent qualities. The French equiva- lent, (2b), definitely seems to favour the former analysis, since the two adjectives are on opposite sides of the noun (a fact which, i think, also militates against an 'Abneyan' analysis that would treat the modifying ad- jectives as the single specifier of the NP). (2) a. The big blue notebook on the table b. Le grand cahier bleu sur la table (3) a. NP/DP / \ / \ / \ Det NP/N1 | / | \ the / | \ AP AP \ | | N1 big blue / \ / \ / \ N PP | / \ notebook / \ /___________\ on the table b. NP/DP / \ Det \ | NP/N1 the / \ / \ / \ AP N1 / \ / \ big blue / \ / \ N PP | / \ notebook / \ /____________\ on the table If the analysis in (3a), in which the two adjectives are independent modifiers, is correct, than they presumably should be sisters not only of each other but at least also of the N0 or N1 that is the head of their mother (and possibly also the PP, which might eliminate the X-Bar-theore- tical distinction between them and the PP 'complement'). The problem also arises in the case of verbs with multiple complements. For instance, what is the internal structure of the VPs in the sentences in (4)? It is a priori plausible that in each, the two bracketed argu- ments are simultaneously sisters to each other and to the verb as in (5), in which case the VP has not just two but three daughters. (4) a. Sam [gave [a ball](NP) [to the baby](PP) ](VP). b. Sam [gave [the baby](NP) [a ball](NP) ](VP). c. Sam [told [Terry](NP) [that there were no tickets left for next Saturday's concert](CP) ](VP). (5) V1 / | \ / | \ / | \ / | \ V NP PP/NP/CP Attempts have been made to get around this problem by iteration. You will notice that in (3) there is at least the possibility of two N1 nodes, one dominating the other. Likewise, it is possible to analyse the sentences in (4) as involving iterated V1's, one dominating the lexical verb and the NP that is its nearest argument and the other dominating it and the remaining argument, as in (6). (6) V1 / \ / \ / \ V1 PP/NP/CP / \ / \ V NP The main problem that i have with this hypothesis and the above-described attempts to get around its descriptive problems, besides its insistance on multiplying entities beyond (apparent) necessity, is that i have dif- ficulty convincing myself that any one type of complement (or specifier, as the case may be) is *necessarily* closer to the lexical head in terms of constituent structure than another. As already noted, Speas argues that the difference between specifiers and complements is inadequate to justify a distinction between projection-levels. On the other hand, she argues that indirect objects are cross-linguistically more likely to be base-generated closer to the verb in constituent structure than direct ob- jects. I have difficulty buying both arguments together. My experience with languages which have rich morphological case-marking and -agreement systems is that it is often difficult, given a clause in which the verb has both an accusative and a dative object, to say which is likely to be closer to it. Such languages seem to me often to be much more amenable to an analysis wherein the verb and all its complements are severally sis- ters to each other. The 'Antisymmetry' hypothesis boils down to an assertion that, at the most abstract level, all languages exhibit the same ordering relations amongst heads, specifiers, and complements that English does on the sur- face, i.e. all languages at DS exhibit the order specifier-head-comple- ment, all languages are underlying SVO, however you want to put it. The basic justification for this hypothesis in Kayne's paper is that it makes it possible to describe dominance and precedence relations in terms of the same linear-ordering properties: transitivity, totality, and anti- symmetry. (Kayne also argues that all movement must be to the left, be- cause rightward movement, were it allowed, would undermine this equiva- lence.) I personally fail to see why this equivalence is desirable. And the hypothesis necessitates what are in my opinion outrageously complica- ted analyses of a variety of linguistic phenomena that are amenable to much simpler analyses if the hypothesis is rejected. For one thing, while i do not claim real fluency in any 'verb-final' lan- guage like German, Japanese, Sanskrit, or Hindi, i have over the years acquired enough facility with these languages to i have some general in- tuitions about them that seem to be fairly reliable. And according to these intuitions, these languages really are verb-final, that is, the verbs are base-generated at the ends of their clauses, preceded by both subjects and objects. Extra processing (i.e., presumably, movement- transformations) seems to be required to get the kind of verb-medial order that is typical of English or French. It seems to me that if Kayne's 'Antisymmetry' hypothesis is correct than languages like these should be a lot harder to acquire than languages that are superficially verb-medial. And i know of no reason to believe that this is in fact the case. I am also strongly inclined to reject the 'Antisymmetry' hypothesis on the grounds that in order to account for observed data it requires a lot more movement than alternative hypotheses about DS. Especially given PPA's explicit preference for MINIMAL application of Move-Alpha, the amount of movement required under the 'Antisymmetry' hypothesis and its corollary prohibition against rightward movement in order to account for a large variety of observed data strikes me as prohibitive. I am ac- quainted with several explicit arguments in favour of rightward movement in a variety of languages; i've presented some myself, including argu- ments to the effect that trying to account for the same data by leftward movement alone, with or without the antecedent assumption of basic SVO order, makes the languages in question seem unusually difficult to mas- ter. There's no question in my mind that such analyses are logically possible. There's also no question in my mind that they are of zero empirical value. Best, Steven --------------------- Dr. Steven Schaufele 712 West Washington Urbana, IL 61801 217-344-8240 fcosws@prairienet.org **** O syntagmata linguarum liberemini humanarum! *** *** Nihil vestris privari nisi obicibus potestis! ***