1Episode 5b. Agree and movement5.3-5.4CAS LX 522Syntax IThe Big Picture Now that we’ve gotten some idea of how thesystem works, let’s back up a bit to remindourselves a bit about why we’re doing whatwe’re doing. People have (unconscious) knowledge of thegrammar of their native language (at least). Theycan judge whether sentences are good examplesof the language or not. Two questions: What is that we know? How is it that we came to know what we know?History In trying to model what we know (since it isn’tconscious knowledge) some of the first attemptslooked like this (Chomsky 1957): Phrase Structure RulesS → NP (Aux) VP VP → V (NP) (PP)NP → (Det) (Adj+) N PP → P NPAux → (Tns) (Modal) (Perf) (Prog)N → Pat, lunch, … P → at, in, to, …Tns → Past, Present Modal → can, should, …Perf → have -en Prog → be -ing An S can be rewritten as an NP, optionally an Aux, and aVP. An NP can be rewritten as, optionally a determiner,optionally one or more adjectives, and a noun. … What we know is that an S has an NP, a VP, and sometimesan Aux between them, and that NPs can have a determiner,some number of adjectives, and a noun.History Phrase Structure RulesS → NP (Aux) VPVP → V (NP) (PP)NP → (Det) (Adj+) NPP → P NPAux → (Tns) (Modal) (Perf) (Prog)N → Pat, lunch, …P → at, in, to, …Tns → Past, PresentModal → can, should, …Perf → have -enProg → be -ing In this way, many sentencescan be derived, starting from S. The tree-style structure is away to record the history ofthe derivation from S to thewords in the sentence. We model our knowledge ofEnglish as a machine that(ideally, when it’s finished)will generate all of thesentences of English and noothers.NPVVPSeatlunchNPNPatNAuxModalmightPat might have been eating lunch If you can sayPat ateyou can sayPat had eatenorPat was eatingorPat had been eating It looks like the verb canbe past or present alone,but with have it takes onan -en (past participle)form, and with be it takeson an -ing (presentparticiple) form. The first verb or auxiliarytakes on tense forms.AffixHopping So, Chomsky proposed:Aux → (Tns) (Modal) (Perf) (Prog)Tns → Past, PresentModal → can, should, …Perf → have -enProg → be -ingPast → -ed Yielding something likethis: If you build a sentencethis way, things aren’tin the right order, butthere’s a simpletransformation that canbe done to the structureto get it right. Empirically, tense,perfect have, andprogressive be eachcontrol the form of theverbal element to theirright.NPVVPSeat lunchNPNPatNAuxTnsPastPerfhave -enProgbe -ing-ed2AffixHopping So, Chomsky proposed:Aux → (Tns) (Modal) (Perf) (Prog)Tns → Past, PresentModal → can, should, …Perf → have -enProg → be -ingPast → -ed Yielding something likethis: Affix HoppingSD: afx verbSC: verb+afx The affixes all “hop to theright” and attach to thefollowing word. An ancestor to the kinds ofmovement rules we’ve beenexploring, and thisphenomenon specifically isclosely related to the Agreeoperation we’ll be talkingabout.NPVVPSeat+ing lunchNPNPatNAuxTnsPastPerfhave+edProgbe+enHistory continues Through the 60s there weregood people working hard,figuring out what kinds ofphrase structure rules andtransformations are neededfor a comprehensivedescription on English. As things developed, twothings became clear: A lot of the PSRs look prettysimilar. There’s no way a kid acquiringlanguage can be learning theserules. Chomsky (1970)proposed that thereactually is only alimited set of phrasestructure rule types. For any categories X,Y, Z, W, there areonly rules like:XP → YP X′X′ → X′ WPX′ → X ZPX-bar theory If drawn out as a tree,you may recognize thekind of structures thisproposal entails. Theseare structures based onthe “X-bar schema”. XP → YP X′X′ → X′ WPX′ → X ZP YP being the “specifier”,WP being an “adjunct”, ZPbeing the “complement”.Adjuncts were consideredto have a slightly differentconfiguration then.WPZPXX′YP X′XPGB Around 1981, the viewshifted from thinking ofthe system as constructingall and only structureswith PSRs andtransformations to a viewin which structures andtransformations couldapply freely, but thegrammatical structureswere those that satisfiedconstraints on (variousstages of) therepresentation. First, a “deep structure” (DS)tree is built, however you likebut Selectional restrictions must besatisfied θ-roles must be assigned Etc. Then, adjustments are made toget the “surface structure” (SS) Things more or less like AffixHopping, or moving V to v, ormoving the subject to SpecTP. Further constraints are verifiedhere: Is there a subject in SpecTP?Etc. Finally, the result is assigned apronunciation (PF), and,possibly after some furtheradjustments, an interpretation(LF).Which brings us to 1993 The most recent change inviewpoint was to thesystem we’re workingwith now (arising fromthe Minimalist Programfor Linguistic Theory). The constraints thatapplied to the structuresin GB were getting to berather esoteric andnumerous, to the extentthat it seemed we weremissing generalizations. The goal of MPLT was to“start over” in a sense, to try tomake the constraints followfrom some more naturalassumptions that we wouldneed to make anyway. This new view has thecomputational systemworking at a very basic level,forcing structures to obey theconstraints of GB by enforcingthem locally as we assemblethe structure from the bottomup.Features and technology The use of features to drivethe system (uninterpretablefeatures force Merge,because if they are notchecked, the resultingstructure will be itselfuninterpretable) is a way toencode the notion that lexicalitems need other lexicalitems. What the system is designedto do is assemblegrammatical structureswhere possible, given a set oflexical items to start with. A comment about thetechnology here: The operations of Merge,Adjoin, Agree, and featurechecking, the idea thatfeatures can beinterpretable or not (or, aswe will see, strong or weak)are all formalizations of anunderlying system, used sothat we can describe thesystem precisely enough tounderstand its predictionsabout our languageknowledge.3Features and the moon We can think of thisinitially as the same kind ofmodel as this: The Earth and the Moondon’t compute this. But ifwe write it