LECTURETTE #9 (#2 on X-Bar Theory): 'UNIVERSAL BASE', FUNCTIONAL HEADS As developed during the '70's and '80's, X-Bar Theory holds that as a general rule the internal structure of a given maximal projection 'XP' follows a generic pattern. An XP includes a lexical (or Bar-0) head X0 and a variety of 'satellite' constituents. These 'satellites' are rough- ly grouped into two types: complements and specifiers. Typical comple- ments are things like objects for which a verb or preposition is subcate- gorized, i.e. requires and licenses; we'll discuss this issue further when we get to Theta-Theory. Typical specifiers are articles and other determiners, also adverbs like 'very' in adjective phrases (APs). In 'classical' (i.e. ca. 1980) X-Bar Theory complements are assumed to be necessarily maximal projections themselves while specifiers may be either maximal projections (e.g., genitive NPs modifying larger NPs, as in '[[the neighbour's] dog]') or minor lexical (i.e. Bar-0) items (e.g. 'the' in 'the neighbour') or morphological elements performing similar functions (e.g. the suffixes representing definiteness in Scandinavian and Balkan languages). Some work done since the late '80's, notably Steven Abney's MIT disserta- tion The English Noun Phrase in its Sentential Aspect, has tried to resolve this complication in the definition of a possible specifier by regarding all non-heads, including specifiers, as necessarily maximal projections; such work also argues for redefining the things traditionally called 'noun phrases' as 'determiner phrases', i.e., the real head in a phrase like 'the book' would be 'the', 'book' being merely a complement thereof. Not everyone (by a long shot) is convinced by this argumentation, but you will occasionally find some syntacticians talking about 'DPs' instead of 'NPs', which is the main reason i mention this hypothesis here. Getting back to 'classical' X-Bar Theory, the distinction between comple- ments and specifiers is assumed to be critical to the difference between Bar-1 and Bar-2 projections. A maximal, i.e. Bar-2 projection XP is sup- posed to dominate a Bar-1 head X1 and a specifier; the Bar-1 projection X1 dominates the lexical head X0 and any complements. Fuzziness in the distinction between complements and specifiers is one of Peggy Speas' principal grounds for rejecting the Bar-1/Bar-2 distinction, as mentioned earlier. The projection structure described above is assumed to be common to most if not all category types. The fact that different category types take different types of complements (anything can take a PP complement, but in many languages, including English, only verbs and prepositions can take NP complements) is considered due to constraints of Theta and Case Theory which we shall discuss later. The question arises, Do specifiers occur with all category types? Assuming for the moment that Abney's theory doesn't hold, the specifiers of NPs are obvious; cf. the highlighted con- stituents in (1). APs and PPs can have adverbial modifiers as specifi- ers; cf. (2-3). But what about VPs? Like APs and PPs, VPs can have ad- verbial modifiers, as in (4), but unlike the case with APs and PPs, i have yet to see these described as 'specifiers'. (1) a. *a* bird b. *the* book c. *some* tortillas d. *the neighbour's* dog (2) a. *very* large b. *not as* early as expected (3) *far* under the table. (4) a. *suddenly* turned to the door b. *carefully* closed the envelope c. *softly* walked across the patio A hypothesis has been gaining ground during the past 10 years (to the best of my knowledge, the earliest published expression of it was in Lisa Travis' 1984 dissertation), according to which the subject is actually the specifier of VP. This hypothesis is known under three names -- at least, that's as many as i've come across so far. It is known as the 'Internal Subject Hypothesis'. This name is self- explanatory, or nearly so: the claim is that the subject NP is base-gene- rated inside VP and moves from that position to the specifier position of S as a result of Case Theory constraints which we shall get to in the near future. The second name is 'Lexical Clause Hypothesis'. The notion encapsulated in this name is that a full-fledged syntactic clause consists of two parts which at least at DS can be easily distinguished. One is the 'lexi- cal clause', including most if not all of the overt lexical items in the clause. It is dominated by the VP node; its self-evident lexical head is the main verb, with the subject NP in the specifier position and all other arguments of the verb in complement position(s). Above this VP or 'lexical clause' is the 'functional clause', consisting of a series of 'functional heads', which i shall be discussing shortly. The third name, most general of all in its implications, is 'Universal Base Hypothesis'. Under this label, the hypothesis drives to its logical conclusions the above-mentioned notion that all lexical heads, including verbs, project the same X-Bar structure. Every lexical head has, at least potentially, specifier and complement(s) associated with it. As the '80's progressed, this structure was extended also to so-called 'functional categories'. In one form or another, functional categories were always present in the Standard Theory, at least by implication in the background. The Aspects model posited a node 'Aux' that was supposed to dominate auxiliaries. But even in clauses without auxiliary verbs, or in languages that don't use them as lavishly as English does, the Aspects model assumed that something was needed in constituent structure for the base-generation of verbal inflexional morphology (cf. the Aspects model's 'affix-hopping'), why not put it under Aux? By 1980, what in the Aspects model was called Aux was redefined as 'Infl' (short for 'inflexion'; typically pronounced 'in-fl' with a syllabic 'l'). This was a node assumed to be present in all languages, in all clauses (or at least finite ones), dominating what- ever was needed in the way of verbal inflexional morphology. Not neces- sarily the morphemes themselves, but the abstract features that licensed them: information about tense, mood, aspect, as well as the relevant agreement features of the subject NP and if appropriate the object NP as well. ('Affix-hopping', note, is no longer a feature of REST. Given the more explicitly lexicalist assumptions of the current theory, it is assumed that words are inserted from the lexicon into syntactic structure with all their inflexional morphology, if any, already on them. The purpose of a functional head like Infl is no longer to provide the inflexional morphemes but rather to license or justify their presence on the appro- priate lexical items. Nowadays it is assumed that the finite verb is base-generated as head of VP with all its inflexional morphology, subse- quently rising to Infl either at SS as in French or at LF as in English in order for its inflexional morphology to be 'checked' against the ab- stract features there.) Another functional head that began to be talked about fairly early in the history of the framework is Comp. Among the earliest book-length discus- sions known to me in the Chomskyan framework of the syntax of subordinate clauses are Peter Rosenbaum's 1967 The Grammar of English Predicate Com- plement Constructions (MIT Press) and Joan Bresnan's 1972 MIT disserta- tion Theory of Complementation in English. Gradually, mostly during the '70's, the theory developed that any clause might have associated with it a Comp node, which might be a base position for overt complementizers like the highlighted words in (5) (or non-overt complementizers as in (6)) or a landing-site for the fronting of wh-elements, i.e. relative and interrogative pronominals and the phrases dominating them as in (7). (5) a. We know *that* politicians cannot be trusted. b. I would be very much surprised *if* Sam were late for the party. c. *For* Leslie to be stuck in bed with the flu would be disappointing. (6) a. We know 0 politicians cannot be trusted. b. They thought 0 Terry would be back this morning. (7) a. *Whom* did you see near the bridge Sunday? b. First prize in the Science Fair went to the girl *whom* you met yesterday. c. *Whose dog* did you say has been wandering loose? d. They've cut down the tree *under which* we had the big party tent last year. By 1986, in his monograph Barriers, Chomsky explicitly extended to Infl and Comp the X-Bar structure previously defined as general for lexical heads. Thus, both Infl and Comp were defined as syntactic heads licen- sing full projection structures. Infl takes VP as its complement and its specifier-position is the surface position of the subject. (This per- spective doesn't necessarily obviate the validity of the Internal Subject Hypothesis, since as will be discussed under Theta-Theory it is desirable to base-generate the subject NP under the projection of a lexical consti- tuent, i.e. a verb, rather than of a non-lexical functional head like Infl. The subject is assumed to move to the Spec-of-Infl position at SS.) The maximal projection of Infl, 'IP', is equivalent to what until then had been called 'S'. Comp, in turn, takes S = IP as its comple- ment. ('Comp', of course, is short for 'complementizer'. We just have to live with the awkwardness of talking about the 'complement of comple- mentizer', or even 'complement of Comp'.) Comp0 itself is the base posi- tion of complementizers as in (5-6), while its specifier-position is the landing-site of 'wh-movement' as in (7). The maximal projection of Comp is referred to as 'CP'. Whether every clause (IP) is necessarily the complement of a CP or not is a question about which there seems to be some debate, though i have noticed a general willingness to assume the existence of a CP node whenever it's not blatantly absurd. Towards the end of the '80's, especially in work following upon Jean-Yves Pollock's 1989 paper 'Verb Movement, Universal Grammar, and the Structure of IP' (Linguistic Inquiry 20:365-424), the single node-label Infl has come to seem overly simplistic. Nowadays REST researchers are more like- ly to subscribe to a hypothesis defining a distinct functional node for each of the verbal-inflexional feature-categories Tense (Tns), Aspect (Asp), Mood (i haven't actually seen this one yet in any published work, but i don't see why it should be excluded), and Voice (i *have* seen an occasional reference to a node 'Pass(ive)'), as well 'Agr' (pronounced 'ag-r', with syllabic 'r') for agreement features; in fact, one commonly distinguishes between Agr(s) and Agr(o), the first being for subject- agreement features and the second for object-agreement (languages that mark their verbs for agreement with both direct and indirect objects would presumably need two of these). All of these functional heads are inserted into the hierarchical position previously occupied by Infl, so that instead of the structure posited by Chomsky in 1986, represented below in (8a), we have something like the structure in (8b). (8)a. CP b. CP / \ / \ Spec Comp' Spec Comp' / \ / \ Comp0 IP Comp0 Agr(s)P / \ / \ Spec Infl' Spec Agr(s)' / \ / \ Infl0 VP Agr(s) TnsP / \ Spec Tns' / \ Tns0 AspP / \ Spec Asp' / \ Asp0 Agr(o)P / \ Spec Agr(o)' / \ Agr(o) VP The hierarchical order of functional heads under Comp' is not necessarily fixed; there is lively debate in the framework now about whether langua- ges can vary on whether, for example, TnsP is the complement of Agr(s) or vice versa. What i have represented in (8b) is merely one possible structure commonly proposed. As can be seen, according to this hypothe- sis everything above VP at DS is a sequence of functional heads, devoid of any necessary lexical content, and their projections; all the clause's lexical material is presumably base-generated under VP. (I'm not even going to talk about the notion of 'NegP', a putative node (with full pro- jection structure of its own) somewhere between VP and Agr(s), which is supposed to be the base position of clausal negators, which are typically lexical. We're also ignoring the fact that Comp0 typically dominates le- xical material, as in (5).) Hence the above-mentioned dichotomy between the 'lexical clause', i.e., VP, and the 'functional clause', i.e., all the other stuff. This hypothesis is sometimes referred to as the 'Exploded Infl Hypothe- sis', and Peggy Speas in referring to it has at least once in my hearing spoken of the 'Functional Big Bang'. Typical versions of this hypothesis assume that the node labelled 'Tns' is necessarily associated with the clause's tense features, the node labelled 'Asp' with its aspectual features, while anything licensing agreement morphology is necessarily associated with one or another of the Agr nodes. But Elly van Gelderen has recently been arguing (cf. especial- ly her 1993 book The Rise of Functional Categories (Linguistik aktuell #9), Amsterdam: Benjamins) that this is not necessarily true. She argues that even though Dutch obviously has tense marking it shows no syntactic evidence of a Tns node, or indeed any functional node other than Comp. She presents similar arguments for Old English, claiming that Tns arose in English as a distinct functional node only in the late Middle Ages, and that English does not now and never has had any kind of Agr node, even when it was young and had vigorous agreement morphology. I think she's right. I'm also aware of arguments calling into question any at- tempt to reflect morphological features in syntactic constituent struc- ture (cf., just for one example, Joseph & Smirniotopoulos 1993 'The Morphosyntax of the Modern Greek Verb as Morphology and not Syntax' Linguistic Inquiry 24:388-398). And in fact i'm working on a major research project that challenges the whole REST notion of 'functional heads'. 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! ***