LECTURETTE #8: INTRO TO X-BAR THEORY (This is the first of what's shaping up to be three lecturettes covering X-Bar Theory in REST. Get all three ...) Precursors of what is nowadays known as 'X-Bar Theory' go way back; in- deed, unless i am much mistaken, if you know where to look you can find hints of it all the way back in the Classical grammarians, Thrax and Varro and their like. But for our purposes X-Bar Theory enters the scene with Chomsky's 1970 paper 'Remarks on Nominalization' (published in Roderick A. Jacobs & Peter S. Rosenbaum, eds., Readings in English Trans- formational Grammar (Waltham, MA: Ginn), pp. 184-221). This is a major paper and accomplished a lot of things, not all of which are directly relevant to our present concern. But the foundational insight of X-Bar Theory, presented in that paper, is that node labels such as NP, VP, etc. are not atomic. Rather, they are sets of feature specifications, some of which are common to a variety of node labels. Thus, a lexical noun like 'book' and a complex NP like 'the emerald-green book that Terry bought yesterday at Brentano's' share a common 'nominality'; like all nominals they are inherently referential, they can occupy similar syntactic posi- tions, etc. Likewise a lexical verb like 'speak' shares certain 'verbal' qualities with whole VPs like 'speak the speech trippingly upon the tongue'. On the other hand, lexical nouns and verbs like 'book' and 'speak' have certain characteristics in common that they do not share with NPs and VPs (for one thing, the fact that they're *lexical* items, freely insertable into a wide variety of syntactic contexts), while NPs and VPs likewise share certain characteristics not shared with lexical items (e.g., the quality of being relatively saturated semantically, a notion we shall return to in the future). Thus the node-labels NP and VP are analyzed into two parts, a CATEGORY TYPE (N or V) and a PROJECTION LEVEL. Projection level is usually spoken of as Bar level and given a numerical value. Lexical items are said typically to have a Bar level of 0 (I say 'typically' in order to exclude from consideration idiomatic expressions that are lexicalized although, as in the case of 'kick the bucket', they may be syntactically full VPs). Higher Bar levels are ty- pically indicated either by one or more horizontal lines over the catego- ry-label or by 'prime' marks after it: N', V''. The former is in my opi- nion much to be preferred, since when properly done it is easier to read; horizontal bars stand out more than little prime marks (which one of my professors once disdainfully referred to as 'chicken scratches'), but it's harder to find a type-face and word-processor that can handle such bars properly (note that i'm not even trying in this forum). A third convention, especially within the GPSG framework, is to write after the category-label a numeral corresponding to the number of Bars: N0, V2. I think this is a good idea, though it doesn't seem to have caught on much in transformational grammar; nevertheless, i'm going to try to use it consistently in this forum, when it's necessary to refer to specific Bar level. The number of distinct Bar levels defined has fluctuated in post-1970 generative grammar. In his 1970 paper Chomsky didn't worry too much about this issue but thought maybe 2 levels above the lexical might be sufficient. Ray Jackendoff in his 1977 monograph on the subject (X-Bar Syntax: a Study of Phrase Structure (Linguistic Inquiry monograph #2), MIT Press) argued for 3. The developers of the GPSG framework explicitly define 2 as the universal maximum, and this is commonly assumed in many frameworks today, though Peggy Speas has said, repeatedly and in print (cf. e.g. her 1990 book Phrase Structure in Natural Language (Studies in Natural Language and Linguistic Theory #21), Dordrecht: Reidel), that she sees no good reason to define a Bar level higher than 1. (In my own research i have suggested that free word-order languages like the Australian and early Indo-European languages may not generate Bar levels greater than 1, but i am not as yet completely convinced by Speas' arguments with regard to all languages.) In any case, the Bar level of nodes like NP and VP is identified with the maximal value, whatever it is (which means that if you assume a maximum of 2 you never have to actually notate N or V with two bars or two primes or whatever, since you can write 'NP' or 'VP' instead to mean the same thing) and they are defined as 'maximal projections'. (Negative Bar levels are invoked by some frameworks, such as Autolexical Syntax, that incorporate morphological structure into their analyses.) I'll have more to say about projections and Bar levels in the next lectu- rette, but first i need to say a little more about category types. The theory introduced in 1970 does not treat category types like 'noun' or 'verb' as atomic any more than 'NP' or 'VP'. Rather, it defines all ma- jor lexical classes as conjunctions of 2 binary features, 'nominal' and 'verbal'. In this theory, the word 'noun' (whether referring to a lexi- cal noun -- an N0 -- or an NP) is an abbreviation for the feature complex [+N, -V]; 'verb' is similarly an abbreviation for [-N, +V]. For reasons that might be worth discussing some other time, adjectives are regarded as [+N, +V]. This leaves one logical possibility open: [-N, -V]. This feature complex is typically identified with the category 'preposition', while 'adverbs' are regarded as a sub-class of adjectives. Emonds, how- ever, has argued ('Evidence that Indirect Object Movement is a Structure- Preserving Rule', published 1973 in M. Gross, M. Halle, & M. P. Schutzen- berger, eds., The Formal Analysis of Natural Languages (The Hague: Mou- ton), pp. 73-87), and i believe rightly, that adverbs are properly regar- ded as [-N, -V], prepositions being merely 'transitive adverbs', i.e., adverbs that take objects. (This is, of course, assuming that the things traditionally referred to as 'prepositional phrases' (PPs) really are headed by prepositions. In some of my own work i have suggested that (at least in some languages) some 'prepositions' collocating with NPs bearing certain cases, rather than 'governing' those cases are actually no more than modifiers serving to clarify the semantics of said case-markings (a given case in a given language often having a wide range of possible semantic connotations), in which case (pun intended) the phrase in question should perhaps be regar- ded as really an NP with an adverbial modifier.) Jackendoff, in his 1977 book, argues for a different, more explicitly syntactic set of distinguishing features. In his view, nouns and verbs are alike in taking subjects -- this was, in fact, one of the prinicpal insights discussed in 'Remarks on Nominalization' -- while verbs and ad- verbs are alike in (at least some of them) taking objects (adverbs taking objects being, of course, called 'prepositions' or 'postpositions' or 'adpositions', depending on the language). I think there's a lot to be said for Jackendoff's view, and as far as i know it has never been argued against, but neither has it been generally adopted. In some of my own work i have very occasionally suggested 'enriching' the standard organi- zation of distinctive features '+/- N' and '+/- V' with Jackendoff's '+/- Subj' (his '+/- Obj' feature being completely predictable from the value of the feature 'N'), which among other things might make it possible to fit the 'functional' categories like Infl and Comp (to be discussed in the next lecturette) into the system. 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! ***