You are on page 1of 213
Gare Sade in Linge: ‘Samuel Jy Keyser general editor 4 dro the Sari Groans 3. Sta eter ‘Spon prettin Coon Grama Ry Saket Tre Sint of te fro Laue Su Kans Spcoh Sut nt Fare Coa Fo tne O Re of ag Grama a Thee Peto, ‘Bouse Peal Preach Spe: The Psonic Richa. Kaye ‘os are Farner Pn Kip DS aor Seman Cogito Ry no ey Ste A Sty of mead i, A Kan 10, Phong Spree: he Reto ener Sel on Src, Babi oS ‘ie Grammatical Be of Ling Referee Lrg Us lei hon Roben € Berk od Amy 8 Westar 12 bare othe Tho of Grammar Henk var Roma Ea Mian 18, Wont a Soene Pry Seat, te Lie eo Paste i 15. The Raptr ef lene, Ee Resa Aloe. Bt Maen tos [anaes on Stes, Montel od Sn Roger rg mage an! Paton of Koowle The Mana Lecire, Noe Glomty 1A Cour G8 Syea:Lecre om adn and gt Ceres, Homa ‘as ant far Urge 18. Seman sans Ra cent 19. von eh Soman Engh & Sid Semi Smanic. Tete Parone 20. Prine on Para n Conger Grammar, Rober Fei eter 121 Randa of Goma Syne Raber Fein 2 Mowe «Connon on Pe Apteanon and Opt, Howard Lasik end Maren Sao 23 Pata Bes, Bary Sebin 24 The Tie rom Ding 20° Eye inguin m Hem of Sybae ‘Brrr Kent a Sere uy Reo ees 24, Grad Phony, Diss Arsene end Dogs te 24, The Map of a CenmonLangog: Jaton. Maths. Teale, a n & ‘he Prog Lng Cre, teh Toman ore Smet Hagricer on Casa, Desi Petey ‘eimai Progra, Nos Ch) ‘The Minimalist Program Noam Chomaky The MIT Press Cambridge, Masaacactts ondoe, England {© 1985 Meche Inia of Tehooey Al hs serve No psf ths bok ey te erode in ay Frm by any eter: or mahaneal mers eons peecpyine ore lore ‘nslorapand ene wll person wt he pth, “Ts tok was sen Tine Ronan by Aico Trae Typing Ld Hoag Keng rd esr oud nth Une Sirf Are. uta of Cong Castopne so Pubison ta ‘Chamsty Noses “The minal program Noam Chay em (Corot sts ings 28) Anson itgrapia eens on oN O2«2090595 (ah pape) SBN O20 1263 (peat pepe) Minster ings). Tie Sere Canetti ere Piseincen ins fea st Gentenis Cage The They of Pes ad Siow Lak 13 hoger Same ote Een of Deaton ‘ma Regecinion TS hope 8 ‘A Mia Progra or Le Tey 17 meee Cours od erste Reteness 35 Indes 4 Inirodiction “The chapters that follow are based in arg parton regular Ise seminars at MI from 1986 though 1994. These have beep conning ‘ow for over 30 year, wth bread participation by ster cy, a ‘ters, from various insintions and disciplines. Io thes ntodtory temaris I wil oullne some ofthe background forthe material that Tetons "This work is motivated by to reste questions (2) ware he gen- fal conditions thatthe human tage fculty should be ented 10 sathf? and @) te what extent ie the Ingunge Fcuy determined by ‘hese conttions, witbout spas seurtre that le beyond then? The frst question in ara has (Wo aspects: what condone ate posed on the langue faclty by vite of (A) ts place wai the array of og ‘ve systems ofthe mindtrain, and (B) general considerations of eon cept suturales tha have some independet aust, nel. sing, economy, symmetry, rowredundstey, athe ike? ‘Question (8) rao pres, but not without Conte aterton Wo these mates ean provide guidelines here, aim rational ei generally ‘ofa as sch considerations canbe cad and rendered Hlausibe we ‘an ask sister a paicular stem saises them in one or another Fo. Question (A. in conta, as an exact ans, though ony pas ‘oF it ean be surmised in the Light of cute undesnnding about lat ‘age and related copie systems. ‘To the exent thatthe answer to queton (2) is positive, language is something ike “perfect system," mctng external eoeatrsins 96 well scan be done, in one of the eaonable ways. The Misra rogram ‘or inguisti theory seks to explore thew pees. ‘Any progress cmard this goal wil daepn a proble fo he Wslogical sciences that 8 tea fr fom nk: how cata ster sch as rm language arse in the mind/bean, oF for thet mtr i the onic orl, unieh one ses ot wo find anything tke te bese popes of human language? That proble: has scmatines ben posed ee a forthe copitve sciences. The ecers ate erpropit, butter lacs ‘s isplaced: they are primanly a pete fo cogy snd the bev se fs, whch as eurenly understood, do nt prowde ony bas for wha Appear to be fait well exablshedconcusons about lreuaye” Much ‘ofthe brosdr interest of the detale and technical sty of fnguage lies ight ere, n my opinion, ‘Tne Minimalist Program shares several underyng fetal assomp- ions with i predesesors tack te the early 1950, though thas have taken somevtat diferent forms as iquiry his proceeded One tet ther is «component of he han midbrain dedicated ¥o language the langue faculty interacting wih other ystems. Though rot se ‘oust correc, this assniption sess eesonably wel-esabiahed, and wil cootius to tae it for grated hex, long withthe further riper {hess thatthe language fc bas a least wo components cogeine ‘tem that stores information, and performance tems that acer tht ‘information and use it in various ways. Iie the copmtive ste tat ‘rims concerns tere, Performance systems are presumably at eas in part langugespeciti, ‘cose comporenis of the laeeunge faculty, But they ave ener sk ‘ned not to be specif wo partir langage: ty do not vary in the 'anner ofthe gm system as Hinge evionvnemt ty, TRs the smpist assumption, and snot known to be ae, though a may ‘ell be, Knowing of to iter ideas will Kept suming anguage ‘ation to he etic f the copie sate "ao bortow fom ear werk the sstmpton thatthe cognitive sys tem tracts wth the performance systems by meaas of eves ing ‘te representation. inthe echo vense of ths noion? A more spec sumption i thatthe ceive syst interacts wih jun two auch enteral” systems: the artaatory-erceptil ayer AP and the ‘onecivalinetional system C-1. Acton. there are two ierface ‘evel, Phone Form (PF) at he AP ntetare and Logeal Fort (LP) the Ct trace. This “double inerface™ propery is one way t ek ‘ress the rational description of language sound with ¢ mesing, Inacable at et ck to Aree, ‘Though comrenly adopt, a les tails, thee assumptions abut the imal wcitecire of the language Tacully and place song ‘ther systems ofthe mindibrain ae nt at ll ebous. Even within te genera ramework, th idea that arteation snd perception involve the same interface representation is comtroveria, and arpealy incre in some fendamental way Problens relating othe © antec at tl more obscure and poorly understood. Tilt keep to thee fay cone entinal assumptions. nly noting here that i they tim out to be ‘ore, even in part, that Would be 3 surprising and hen intersting iscovery “Te lending qustions that guide the Minimalt Progr came into focus asthe peinspler an parameters (PAP) model took Hage shut fiten years ag A look at recent itory ray be help in placing ese question in content, Needs 104. thete rma ae schematic ad aleve, ad beet rom hid arty generative grammar fice two indie problems: 10 finda ‘say to acount forthe phenomena of particu languages (decipive dcquaey") and to explain how Koowedge of thes fats aie In the tind of the speaker hearer ("explanatory wdeyuary”) Though it was scarcely recognized atthe tine, hie rseatch program revived the cow ‘ims ofa rch tation, of which erhape telat jor mpresetatie was Ouo Jesperen* Hpertenreognied that the xrucres of lat guage “come ino exisnee inthe mind of «speaker by abstraction fiom experience with uteranes, ying “notion of ther strcre™| {hat “deine enough to guide hm in Framing stnces of his om oc “Tew expressions that are typically raw wo speaker end heart We ca take these properties of language (0st the priary onl of "puis theory: spell ou clearly this “notion of stucare” and the Procedure by which i yields “ree expressions." and 1 explain how it aie in the mind of the speaker -tbe problems of descriptive nd Plaatory adequacy, respectively. To atin deserptive neque for arcu language L. the tory of Lis grammar) rst characteie {esate stained bythe language facuky, or at eas ome of aspects “To ain explanatory adequacy. theory of langue must characte ‘hei sate of he language fact and sv how tas xpi 0 the state tained Jepsen bed frther that iw oly “with regard to ‘tax that we expect "iat there must Be soeting in common tall aman speech”; there canbe a “univers er peer) grammar” hence ‘perhaps larseaching account of te ntl state of ie langue aly in this domain, though "no one ever dreams of «universal orp gy" That ies too ak eran vonage eet work the modern period thew adtonal concems wove dispiced. in pat by bhavivsetrents in prt by various structurale appa, ‘ich raicalymarowed the domain of inquiry while greatly expanding ‘he tse for some fate inguiry that mish retun t he aio ‘ad surely valldconoers, To aes tern equ attr under standing ofthe fact that language eves fii se of este meu.” in one uric formulation. Advances in the fornia sciences provided thal Uundestanding, making it feasble to deal with the problems construe. ively. Generative grammar can be regarded as Kind of confuses of long-erporten concems of he study of lrguage and mind, and new understanding provided bythe formal iene. “The est forts te approach these problems quickly reveled tha ta- Aitionalpranmatics| and lesa! dee 60 not hep to devrte, et lone explin, the mest elementary fis about ven the bests lan ‘guages. Rather, they provide hints that can be used by the rider who vey har tai knowledge of language, and of particular langues the eral Lope of equry was, substantial measure. spy ignored. Sine the requisite tact Leow is easly aened without ree ‘lon tadienalgramars wad daionarenappenr to have very breed coverage of lnguisic dat. That aniston, ower a We qu) {scour when we uy to spel out whet i taken or granted the nature cf the enguge uly an its sae paniculr cases “Tiss harly a Staton unique to the sty of language. Typical, wher quesicas are mae sharply formulate, i lewaed that ven ce smenily phenomene had exaped notice, ad that itive acount tha scat spl and persuasive are emily inadagute ve are mid ‘hat ape fl to he ground because thats atoral plac, there ‘wl be no srioae science of mechanics The sue i tue i oe st {sfed with taditioal rus for forming quesions, or with the leical otis it the most elaborate dictionaries, none of which come cose to esting imple properties of hese igus objets. -Recoptitin of he unsuspected ices and compesity ofthe pheno ‘ns of langue created aversion atheen the goal of deserpive snd ‘rplnatory adequacy. I wa cst tht to achieve explanatory seus fu theory of the nal sate mst low only ited arom parc. ‘lr Iaeguages must be largely known in advance of exprine. The ‘pens pent in Univers! Grama (UG) ms be highly ese [Experience mist sfc to Bx thm one way oe smother ing a sate of the lamespe fact that dtemines the vaied and complex aay of expen. hit sod and meting’ ad even the most super ot ee hn hare tno eg ter the dat of exer But the gl ef explain nea) Fe cell utero he dtc pert syns we need ‘n pur of doeipieadeqsy in aceasta fore {nt engine Te problem ws cred ty he ge eof omen dicoveed ben steps were mee To ferme eal ‘steno Lng “Ths teson dine the reac program of ety generate rarer attest he dey within tha cence ee. From ely 1s is ema eine ws 0 aap in ey the ome le sens deed for articles ving als {silent in hr praton by thaw UG pce Sey ints don reese vanity of neugespestic pers th Contig to expansion adeacy. They sho tend to eld net fd more tral hors, ping grounder oan erental tals appreach Theres ey ht ths eth ci cul tm outta “ope rier and more cme veraon of UG recs permsile vay. thar comune to he pny cla Explanatory adequacy. Te ral, howe, theo ener hve proven ob mut refracted ey se One usraton cones tefondent peice mth evelapeg empl coverage: Rept, thas foun tha hoa eng Sted en mast be eed by mnredidan os. The Cateye Dee So repr tha the nd cae eugene hes bso ‘ching prin nai. Aes in eure oye a & ological ace “The rt mist inthe PAP mde Com 1h, or oe fsslaton). This cena ra ea fom he Ach a ton of tesa of yoo of tei gtin, far mere so taney cate ona, hh eal ens rl tonal com Ces an approves to theta een vy a te ee and $-Srutre; he Empty Category Pisce: Xbar theory sally; the operation Mowe the spit hypothess and other. All ae minted o subsanialy rized in secre chapters, prey ete. ‘The co ret ia picture of langage that difers considerably fom ‘vents ied peers: Whether hee sep arc onthe it rack ‘ener, couse. ony time il tl Notes . 1, Forse dscsion of tht ia eChmsky 1M. fering Ena 9m Eran ths the eri obese fatal for cone sore 2. Acai, cs, oet Chery 1S 3. The em ory oo nro i tht gs th he egane - ‘hyn pce, wah pc elt a rae Werk oe ps dear nan ngage ceri hie uso! saren wl enn 0 {he the ters, tet hoot any ipl soow prey of eu ye Wheering othe cue of sponge, 41 Fr seme dcuson ee Chak 177, ht 5. mupret teri of coun to dren theyre se Ta oom foals, egestas bythe haga ay 09 {pera st) oars of tet cing ages: omens nee Sos expen of othr gia, post nent emer mel (6 Ths, wit ell “Eaghs” “French” “Spun” oo en oder ‘elon owl in haraateus Spt commen el the Now tnen Congost proxy (0 Corman soc = Beg tish oter Fates tht ear Svs te reared ar poyeras oh age ey Pau te ova easing si har ona ta the pei he Tange Tc) et eject ft ar! wenda san ey ‘umpios retake for gale he yo crams erly Chapter t ‘The Theory of Principles and Parameters ‘vith Howard Lasnik’ 1 Introdecton Principlevand.zarameters (PAP) theory is not 9 precisely articulated theoretical stem, but ase a parcel apprcech fo cas problems ofthe sted of anges, guided ty certain fading ideas at had een taking shape since the origins of modem generative grammar some 40 years ago These iat ery ito a Gtncve approach 10 the topic by abou 1980. nthe years Se, many spe variants have Been eveloped and explored. The epic base ofthese ings bat bo _weaty expended as they have ended o languages of wie vain Iypes and have eugaged «eich broader range of evden concen Jangvnge and its ws, also pentating wo far greater depth, In his survey ‘we will not attempt to delineate the vanity of propos the fave en investigated o oases theremin success and nadequacis Rasher, we will pursue a yarcula path hough the ara a des and ins that have been develope, scmetimes ating oer directions that have been persue, bot without any ate 0 be compehensve: similarly, biiographic references ae fete comprehonsive, usualy indian only few sie of particaar questions. The de of 2 tcl path should be regarded only a an expository deve, sh cf fort oda the kinds of questions tat ae being addresed, some of the thinking tha pues mich earch, and it erie ciation, We “Tis chop, couthred wits Howat Lae, wa ly plied in Sy lax An ura! Honk of Cneguray Revers hey Johan _hect,Ari on Seow. Wollgen Smad. a'Then Van (hn ‘and New York Wal de Goer. 199). le oppess Rv wth Ton oe ‘y permiton ofthe pub. " caper o not man opty that shes prt choces ive en well eta Ssh in eontas to cthers, ony some of which we wl eal ven fo “The study of pererative grammar has bee ei by sever fund mena poblems ach witha waditoral aver. The bane concer f 0 termite and characterize the hnguistie pace of particular ini tah, We ar concerned then. with sae of he nigga. wich 1c undestand to be some arayofcomtive tate and apc pat ‘ular component of te human mind brain. The lena faculty ha 4am ini ste, genial detemied i the nogmsl course of devel ‘pret it uses trough a sees ates in early childhood, teaching 2 rately sable sady sta that undergoes lite subseques change, part from the lercon. To «good fst afprenimation he ital ate sppens to be uniform fort specie, ASAP tratinal fers Yo Speci segs we cl the theory of teste atid i roma a the theory of he iti state Une Grammar (UG). ‘Thetis ao reason to Beleve that tet ste is in cul respects «spiel characters of humans, ith prope thtaphear to be ur, ‘estan the Biologia work. IF tut tht a matter of broader ster, ‘ut one of no get reeranc to determining the properties ad nature ofthis facly ofthe rindi. ‘Two fundamental problems then. are to dstermin, fr each ini ual ay, Jones) the properties of the steady state that Jones language Faculy ati. and the properies of the nial state tha sa common human endowmen. We Gitingish tetneen Jones competence (new le and understanding) std fis performance (what he does with that rowledge ard understanding). The steady state contines Jones's ma ture guste competence ‘A salient propery ofthe seuy ste that it permits nlite we of frit means te borrow Wier von Humboksaphor Apert hole of hee ite mean sa por lnguage taking langue to bea way to speak and understand a tradtonal formulation Jones's ‘competent is consitue by the particu ste of ite means Be has equ ‘The nocon of “infinite we requites further ana. f the gh of| Insights of the fora scienes in the 2th enty, ee singh 40 seme ofthis notion, the st reatingo competence the sco to fe formance. fn the fst sense a language species am infinite range of sen Dole objsts, which we cll senna despre (SD), We may thik Pitino Pareto s ofthe language. then. a8 fitely specif penectve recede une= ton) that enumerates an infinite set of SDs. Each SD, in ur, species ‘he fall ray of phonetic, semantic, sn sybacie prope of parte ‘ie ingle exresion. This sense of “infite we rls to Jone linguistic competence: the perrativ procedure wit sft seope "Tne second sens of iit use™ has to do with Jones's performance a he makes use of his competence (0 express his though to ref to reduce signals, to interpret what he ear, sad 50 on. The langues acl is embedded in performance systems, which acces he generative procedure. I in this broader conte shat quero of eaten an tne of SDs ars, question of arculaion, intonation, tnd the ke: How does Jones say X? What Jones talking about? What ‘oes Sones ake Smut o be sping or intern to convey! And seo. We might think ofthe SD ar providing isructions vo the performance stems that ema Jones to cay out these ection. ‘Wen we sy the Jones asthe anuape Ly we cow mean tat Jones's lanpuage fut i in he sate L, which me iden with» proertve -roeedare embeded in performance sytes, To sings this cor ‘Spt of anguspe rom eters, et us fer to 8 Hogue, here Hs 0 ‘sugs “ier.” “individual” nd “itesiona.” The coop of an ‘age is intra in tha it deals with am nner sate of Jost ‘in, independent of other eles inthe woeld nil tha ‘deals with Jones, ad wit language commits only deviate) as ‘soup of people wth smitar Elanguaes I intensional nthe thai a) sense that the language sw finetionspetied in intension, not ex ‘selon ts eterson isthe st of SDs (vt we might cal the ere ofthe engunge) Teo dstnst Hanguages igh, in princi, have the fame structure, though asa matter of enpinca fact, human lngings ‘nay happen net perm thi option. Tha i tight arn et ha he range of Hanguages permite by UG isso nattow tht the theoretcl ‘ion is snpy ot relied, ka here are no distinct Manghages gen ‘ning the sane set of SDs. This seem in fat, no uly Bat i is ots lope necesity. When we we the term ange Below. we mean angus. “ln the cast work in generative grammar, i was sumed that Jone’ language penerates un SD foreach of the perme phonetic forms for han language 0 beeps! by UG. Thus, Jeness Janguage assign a particulr satus to such expuessions a (1) wher ¢ (ace) indicates the postion a which the quasion word tconsued (2. otis seeing rk hat Mary Bd + (anew: the cr) 4. what do you sonder wheter Mary Bed 1 answer theca) © how do you wonder whether Mary fixed the carr (ensne: with awench) © expression of Seah, Hungtian, fact. some of the mos insti recent work has bron concerned With he difleencs sete by (1c, heh in some sense “deviant,” but assign a fleet stats by Jone’ langage (tions 3% 141) and one might wll enn about the languages of Jones and Wang by sying thet reactions to slerancs of Sah “Anotber sion that appears commonly i the Hsrture He “ormal angus” in the technical sone set f well formed formals in afar tar vatety of formal arithmetic, "2 + 2)= 5" bat aot "2+ ~2)5¢~ (Cast eset a Elanguage where Es to sgget “eternal ad “ek ‘ersona” In the ceory of rma anaes, tbe E-angue is defied by aipulation, hence is unproblematic. But i i «question of emptied fact whether atarllnghgs has any eter 0 this oto thet wheter Jone’ Flineuge generates not ony set of SDs bul alo iatnihod language some subset of the ponte forms of UG. i ‘lating ome burro ao hese of 1). Apt from expiry passes the concer of Eanguageseareely appetite tradion of generate aroma that we are considering ere. As diet fem the notions di ‘sed err, i tat no Known statis i the stay of langue. One ‘nih deine Etanguage in one or another way, but # dows ot seen fo ‘uate how tis is done: ther so keown gp i ings thors, mo xplnatryFuneton that would be filed were rch cones presente ere, wl play o ole or dacaton, nthe ty of formal anguoges, we may distinguish weok generation of Elanguage from stone generation ofthe sircture of the langue (these of SDs). The weak generate capac of x theory of Uanapes ' the st of Flnguspes weakly seer, adits song gerenve poets isthe set of strates strongly genersted. Ia the sty ef natura language, the conceps of stroctire and stor generation ae ceil the umepts of Elanguape snd wotlk gesatiom best rare an et haps not empirically mearingfl ata. Not at f Eanes do ext, ‘hey ate att considerably further resve fom mechani si Beha ond Pacts ” age Thus, the cid is peste with specimens of = Ibvioe in particular ccemstanes aad acquires an Tanguage io some ‘manner to be determined. The angnage sate ofthe inden ‘hs a certain structure (ie, strongly penerates a et of SD). tm oF ‘may not allo weakly generate an Eangege, 8 Tigly abstae cet ‘emote rom meckansms ued behavior. Inthe terms rst outlined, we ean comider soe of the case prot sem of te study of tangs, @) 2 What docs Jones know when he has patel lige? 1. How id Jones acquit this knowles? Mow dees Jones pt this knowlege to use? 4. Hw did these properties ofthe idan coe in the spe? How are these properties realized in merase of the brain? Under 2). we want o account for 8 wide varity offic or example, that Jones knows that a Pinshyrnes with fn Eich expression of (1) basis specifi sae. © IPMay is too elever to expet anyone tea then we dont ‘expect anyone to catch Mary (bt noting eat bout whether Mary expets anyon oat ws) {41a stoo angry o ros the mesg, then iter May 0 angry that she cet run the meting. o he sso angry ha ‘ust rum the meting (empate:the coed sto gr tr he ‘meeting in contr, which meetings Mar o0 amet ra bk ‘only the Former ocndeviand interpretation ©. Mary pred the Rouse row, then i exterior (nak ‘neces isinter) x brown {AF May persuade ilo got college, he Bl caret intend to go twealleg (utile Mary mayo tay not have) The proposed answer to problem (28) would be tht Jones has a sage L yeneratng SDs that expres such fet as 3), Note tht Jones ‘as this owls whether oF not he rear of thee ft Abt Ht fe it may take sme effort to eit such awareness, andi might even te beyond Jone’ capacities. Ths i 4 question that fills within the ‘cade content of performance sytems. “The ansocr to problem (2) sn substan pan in UG, The covrect ‘teary ef th inl ate wl bee enough east forthe atment ‘of spf language onthe basis of be evden aval othe eid, bu oot so ras 1 ext atainabl Inganges We may proceed fo ‘ska wll how environmental factors and mattatons proces ni, 2x wit heist ate ested by US. Treble 20) cals for the development of performence theereh ‘mong than, thors of proton and inkpetation. Put Eenerally the problems are teyond rac: would be unreasonable to pose ie prot of ow ones dies to ay hat he dso how Deiter ‘what be ens in priculr circumstances, Bt highly seized aspects af the problem are simerable to study. A standard emprical hypottet ‘atone component of the min rain ia por, which assent 06 to signal (abstracting from otter crests relevant wo ln Dretaton), The parser presumably incorporates the language and nich ss, and the Rypotheis is that inlepretationinvolss such a sey, embeded i ther Tehas sets been argued that Engst theory mua met the em ‘scl coniion that account For theese sed pity of pring But baring does no, infact, have these properties Parsing maybe low ed “ic, or even imposible, and it may be “in ero” in the sence that the pereetasipned i any) Fito match the SD eseiaed wih the sional: many fair eases ve ben suid. Im ger ts ot the ‘case that language is eadlyasale or “designed for use.” The sebparts tat are wed are ube vl iologkal considerations He! Us to ‘expect no more tno tat. Sitar, etening to prcblem (2), there '8 1 «pron reson 0 exec that the languages permite by UG he ‘esenable-—tht atainuble ue normal cteumaonees. All de we ‘am expect that some of thr may te; te ots wil nt be ound ‘numa seis. If propesas within the PAP approach ae clse to the ‘mar then it wil fll tht neuapes are in ft eat, bu that sm empscl discovery and a fathe surprising one, Problems (ae) appear to be beyond serio ingly forthe time Be ing, along with many sna queionssboutcoeition general, Her apuin one must te wary of many pitts (Lewontin 190), We wil put thse mater ide A grammar for Joes is rif (or othe exe hat) the erp it estes isthe one Jones has. In that case the grammar wil accoun! for sich Fact 5 (3), by providing ensue tat genotes appropriate SDs A true gramiar i ai co weet the condition of deserting ade ‘ocr. A ieory of UG is tue i (oro Be ener thai corety 3 operand Prancrs » sere the ii state of he langage fey. that case wil presi descriptively adequate grammar foreach sine language. A tue {ory OF UG mers th condition of explanatory adequacy. The terme oly is intend to suggest a cerainplausbe pte of explnaton Giver an aay of facts Such as (), we ean give eh socount of thee tne fev! by providing gramme for Jones, ahd we can provide an ek [laoation fr them st a deeper lel by annering problem (2). that by showing how thse facts dene rom UG, piven he "bur cond tine” set by experince, Note that this pater of explanatcn, thoveh standard, makes cernn epic sesumpons shows the actual poset ‘af susion that are by 80 means obviouly tre for example ha the ‘process a i were istantancs. Sack atumptions te inivetty fupponed to the etert that the explanations see, ‘Any stows approach to complex phenomena snvolves innumerable ‘teaizations, and the one just sketch ts na exception We do not ex: et to find “pre istamiaions” ofthe iil sae ofthe langage fac Aly iene of UG). Rater. Jones wil have some jumble ef rats, tau en the peculiar pattern of fs experience. The enplaratry model cuted dea specifies with language acqulion ue the Weaized ‘xndtions of @ homogeneous sprch community. We acute Hat the ‘yslem described by UG i real eompunent of the mind in, pt to ‘Se in the compe ercumstances of ordinary We. The vabty of thie assumption is hardly in question. To eect would be 0 ase ther that comtomogercous(conictng) data are regi for ngage ‘ssouson, oF (2) thatthe mind/ratm does indeed have the system de Serbed by UG, bu i isnot wed in langue sequition. Neither se ‘mption ie remosly pluie. Resting therm, we acep the appreach {Nat ound asa reasonable apreach fo the tuth about herons and & iy prerequisite 10 any serious inguin ite the complex and cote Phenomena word Furthermore, een if a homagencousspech cominunity existe, we would ot expect its linguistic system to Bea “pure ce” Retr, all sorts of acedems of history woul hve contained the tem, a the properties of (ough) Romance versus German origi inthe kx 2m of Esgsh, The proper topic of ngury, then, shoul be a theory of then tate that abatacs from suc asin vil mater, For ‘othing purposes (and nothing more than ht} we may make 4 rough hd tentative distinction betwen the core of x language and its pera where the core consis of what we tently atu to te: pe = chap \nstuntitens of UG and te penphary consists of marked exceptions CGreguaeverts ee). Note hat the periphery wil als exhibit properties ‘oF UG (¢gsblut phenomena), tough es raneparsely A reasonable approach would be to focus atcaion on the core system, puting aside phenomena that esl rom histo cident, ire mitre, perso ‘tiosynorses and the lke. AS in any eter empiielinguy, theory internat considerations enter int the effort to pore this Oued We capet further dstnctions to be necessary (consider, for exarple,the phenomenon of dosnsrtion i Engith ain (le) ot on 8 pe with nega verbs, but not of the peneralty of fronting of question werd). “The preceding remark are argey concepts thous ot without ee pirical comsequtnes, We now pried along paritlar path in the ‘manne nfs eri, assuming fer emia kat cach pnt ‘We assune thatthe Lnguage (he genertne procedure the langue) has two components: a compuaicnal system ands Fexion, The fest _serates the frm of SDs the second caaceies the ea ems that ppeat in them. Many crcl questions aie am to how thee ste ‘ere We wil sume that ove aspeet of an SD i 4 system of ‘enreseration elled D.Smnerare. st which lea items ee laser Dire expresses excl properties in a form accesible othe omputtional tem. We assume rhea dtinetion beeen ecto aod diana! process of morphology, the ater intra tothe tren, the former Javalvng cmputaional operations ofa broader sytactic scope. Tete onputationl operations ight involve werd format or checking Conse for example the pst teme Form wale. Te leicon conta {he oot [ak with its iiosneratc properties of snd ening. and form sped: nd the iesionl fare fre coe va of ich i Inst. One ofthe computational re call t R,esocsts the two by combing them (cher ain [walk] to flee), or convey). We nigh interpret hs deseripve comment 80 way Ore possi) tha: fk is drawn from the texicon as sus then combines it wih (pes). second possiiy tat processes interea o the lencon (re stander) form the ord Bid withthe properties (valk ind Ie} akeady speed. The rte R thee combines the amalgam with [past checking and ensing i inns eta [pst Un thi ese the fescon is more srctred. I contass the element [wake before, along wit rules indicating that any verb ray’ ele ietiesicaly possess iran Poremtr, a such properties a gas purl. and the like. Similar gustons ais shout complex werd (ciustvs, noun corporation structures, comm ound nouns ete) AS these topics are puri with more peson, Shin more closely articulated theories, important an often sl em ical asus arse (Marantz 1988, Fabb 1984, Baker 198%, Di Scialo fd Wiliams ISH, Grimshaw 190), "The SD provides information (o be ire by perfrmance sys tems) sbout he properties of cach linge caprenon, including sound and its meznng. We sssune that he design of langue provides fs eriay of symbote sjstems (lr of mpreenroin Tallin thee fats, incoding the level of Phenetc Form (PF) athe evel of Lege ‘Farm (LE), spciving aspects of sound and meaning, repectvly, iv tofer as thy are Knit determined. Ancthe i the level of D- Sucre, which lates the eormpzational system andthe ion, ‘The kel PF rust say thre bass conditions of adequacy. Te mst be acral inthe roe thal an expresion of ny actual o pot human lngge is representable within itt cnust be en interface, in that its elements ave iterprettion in terms ofthe sensoramotor oy ‘ems Ad it must Be wor, in at Chis interpretation ufos for sil languages, 028 to capture ll and oly the ropes ofthe sys ‘flangeage 0 such ‘Tee sane thes conditions Nol for LF. To capa what the language faculy determines about te meating ofan expresion, it mist be Wn ‘vers in that any theupht expres in » oman language is pte ental in ita inefacein tha hese seresenations have an interne {ation in tems of other stems of he mind ban involved in hough ‘efering, planning. and so on; and nifrs, in jot the ses that the ‘Mhonete system is We wal put aide importans ques concerning the fsature of the LF itrfce does i iveve conceal sem (Icke dot 1983, 1990), a ae theory of meaning «clusal theory of ref ‘rece, ete? The conions are more butte than inthe case ofthe phonetic analogue, because the ystems atthe terface are much es ‘sll uidetood, but there none a wel of evidence Brn eno {allo substantive inquiry According to ths conception, then eich SD contin the interface lees the extra interac eves PF and LF, ad the intr itoce level of D'Siucue. The elements at these levels ae furleranjecd ‘mo festre:phoeoogia, selecina,esegria ands on Ingenta, 2 ayer! ‘ich symbol ofthe represerations ia Feature Sin espects to be fr ther specie, ‘A furter assumption, developed ia the Eten Standard Theory (EST) that these levels are vot elated cy rater, tei elton, are medated by an ntemedate level of Site. Adopting ts ew, ach SD is a Sequence (2 2.80), where and 4 are represen’ at the external incerfae feels PE ane LE. 8 is the etna strfce of ‘computational sjsew ard lexicon, a 6 i envative. The fist three eels mee epiialconitions imposed by tbe performance systems ithe leon, The level ofS Strctre rst late t thee te eee tthe manner specie in UG: we might hak of formally 9 he (peesunatiyunigue) “clon” to eset of condos. the subse eent dacusion we react ours largely fo the els D Stacie, SSirsture, ad LF and the relotions among then (ht sh mare Sense) We are thus concerned pamly wih he dation Ieee D> Structure to LF in o DSinture «—~ Lercon PF = --Ssinsture 1 Lr Sune questions arise ao ow the lations among these els ato be consrua: special, there an inberent "rections," 3 tt the relaions should be consuued as « mapping of one eel ence, ‘ris thre Simply a nonditetionl elation? Te Foeulte this sa el fepircal sue i not siple mater, and empirical evidence 10 As tingih such possi s nox easy fo come By. Bot mtresing (sod coufitng arguments bave bee presented, Discrimination rong tes erative tacos pel if we adope (as we wil) the ‘sandard EST assumption, from the early 1970, tha representations nay ich emp eteorcs (ECS ements (Fate set) ha are pe Fey substantive fom he point of view ofthe computations stem, but that donot happen o be assigned an interpretation by te mapping from Stuur 10 PF though they may have indir phoneti ees, ‘hs the contraction rules of English comert nent 10 te the phenol ‘cal word wana when there no trace terse (who do yew wa see ut rot whe dy wenn se fel (Consky and Lasik 197), Principe ond Parts » ‘We wil temavely proseed on the assumption tha the ations aes ‘in ft, cvectonal DStrctreis app to S Siac, which (nde pendent) mapped to PF and LF. “The caren modern work in petratve grammar borrowed standard teas of tditional grammar, whic recognized (1 that seen as 4 Iherarhy ef phrases (noun phrases, clases, et) ad that ts (oie ead) ene foto certain grammatical relations; ad (0) that sentences talons to various grametcalcomioctions wih systematic relations “mong then, some more “bas” thom others actives mone base thn pastes, declartves more tsi then interrogates, ete), Comeipond- Fags, the cars versions of UG provided wo kinds of re (0 phase race rules geerating SDs tha express the berry of phrase, nd {AD transformational rules thar for grammatical eonsruciene fro absretunderyng. forms, with mote transformations tevebed in fot ination of theless bese constructions (us, only abiatry transfor tation apply 0 fom active declarative (Reel sentence), Dut sme ‘ptonal ones are involved in formation of fsves, evogse, ce) “The pase wractere rules provide x "seometrial™ acount ef games al Faton, understood felatonaly nat, sujet not aye ‘epory tke noun phase or ver, but Undestood ax the rion Sbjetof holding of the pair Ue, ki) im Jon jt, and 30.00 (Chomsky 1951, 1965, 1975). These notions were dein in sich nay that he phe trite rules (1) generate D-Stscures deep sr tors), each a phase marker that represents eazchy ane ratios “Transformations conver thes object mo new psc markers the later EST version. at noted. DStrctre art mapped to $ Stuer by fh rotons an he er af mapped nen 10 PF and “Tie coor to phase stuctre rules nar ako sugested by other ‘esideration. The earkest work concerned on wht ow called ‘erative phonology. and inthis drain “rewcing rules” ofthe frm XY, wiereX ican expression “ewriten” ae in the cote of er ‘ation secms an appropiate device I thse rls are restricted to the foam KAY = XZ¥./A sane symbol end Z nonnhthen we have & ‘stem of oes that can form phase swuctre representations in 2 net Fal wy (comevefee rules, ¥ ae nl) Further movation derived rom she theory of foal sjstens. Grimmatiol transformations 3s erative dvi were spgested by work of Hari (195), wich ed ™ caper {formal riaions among expressions a devi to orm” txt for the analysis of dacoure 'As for UG, the eats verions ase that provided a format for rule syste anid an evaluation mee that sega a “ae to etch {creative procedure of the proper format. The eral empirical eon- ion on UG. then is hat the system provide ony fev highvalud [Hamas onsen withthe kinds of dta available tothe i per haps only ene. FUG is fase ia this sens, the furdamena pele (2b) canbe adresse (Chasey 196). “This approach recorded many achievements but feed a fundamental and resuren problem the lesion between deeipve and explanatory ‘adaquacy. To achiew deseiptive adequacy it seemed noes 0 each the format of permite stems, bein doing 30 we lee the property ‘of feast, so that problem (2b) stil wsesoled, The crf arses 2s t00n awe move fromthe inline hints and exroples of wition franmar to exp generate peoeduts. IW was quickly repre "har the prom is inerent in te hans of re systens tat were be ‘consider. The most plausible approach ro its to try to “factor ou” fvearcing principles that gover sue appscason seca. esiging ‘hem to UG: the actual rues of pranar ca dn be ven i hes it foe, with thee princes ensring hat they wil operate in tach 1 way a8 10 yiid the observed pienomens in their fll complex (Chomsky 1964, Ross 1967) The li that might be eeched i tht rules se eliminated ciel, the upparet cues” being deduced from genera niles of UG. inthe sna hatte interaction of the principles woud eld the phenomena thatthe cus hid been constructed to esrb. To ‘he extent that thie sult can be achieved, the ues postulated for patic- ‘lr languages willbe show ro te epiphenerena ‘Such ess were pursued with g00d deal of ster from the ealy 1940s, teding to the PAY approach which assumed tha the bt ean it fact be steed: the yp fe tht al pincipls ee signe to UG and that language vrtion eerie to crtain opis fo how the ‘anes api Iso, then rule jsems ae eliminable, at east fr the Coreof the tnguage. “Tost, comider agin (Le), repented hee (6, wha do you think that Mary aed 4. shat do you wonder whether Mary Raed ‘© ow do you wonder wheter Mr) fed the ear Pipe and Parent » ‘The gol ito show that the question words or fom the poston of # bya general pind that allows movement ite freely with the op- ‘ions, nterpreiatons, and varying satus determined y the interaction “ofthis price with oer. ‘What she sates of he rls (9 (phrase rr) and (1 (anor rational) under ths cosention? The warsformational rls sil exist, bt only principles of UG. rey pleat ro arbitrary expressions ‘Sich deweer appt o be unavcdsbe inane or another form, whether taken 10 be operations forming derivations or reions ebished on -mpresntons. AS for phrase srcture rust appears tht they ay be completely superfuous. That would not be too surprising. With the “advantage of Binds we can sce sha, wake transformational rele. they were «dubioor device opin with, seapitlting information tat mos be pres incimiobly, inthe Fxcon. Fr example, the fac that perma rakes a noun phrase (NP) ad lesa phrase (CP as om Pemex ay Isl property, equi tht thee be phrase trace fleseling V-NP-CP an istincaion of he phrase XP headed by the verb persuade: and completely general properties require further tat XP nat be VP (er psc, ay say, NP- The apparent cena ‘otphasestrocture rls become cer bythe te 1960s, ith the sepa tion of he lexicon fom the computational sem and the evelopment “Abr theory tion 1.3.2), ‘Theiss canbe sharpened by considering two properties that de- ‘evptivestements about language might tave or lack, Thy may or ‘ay not be language they may oF may nt e construction: rial. The snerens ef traditional gramme wpa both ‘us and consimetion-peicular, and the sme ise of the eux of ‘rly aeneane grammar, Consider the role analyzing VP as V-NP. the rules roming the question words Eiftret ways (ee). Splled ‘ut in fl ett these phrase tutes trnefrmatoel als are ‘esi 10 English and to thse constructions. Tere ae few exceptions tothis pase, ‘The PAP sppeaach sins to race descriptive statements (0 t4o snegores langunge-iarot, and lnguage-paricuar. The lngusse invariant statements a pocinesGnchdg the parameters, each on 3 ‘er with» prise of UG he lengoage-prtilae ones sr specfes- ons of patiular yah of parameters. The notion of consrton, it th tational sen, efeively disappears: 8 pera Uf for de- serpiv tone but hat no thereto tats This there ae no sch Py hope comsiructions as Verb Phrase, or interopaive and rave cause, or amie and raising constructions. Raber, thre ares peel prin lesa teract to form these dsciptive artifacts “The parametric pons avaiable peat 1 be quit Tested, An a smpion ha sents ot uealsuc that tee ise ene compute tinal ster hat forms desvations fom DStruure to LP: at some ot i the dentin (& Strutt, the proces branches to form PF by En independent phonclopial derination (es 0). Options would hed eves 10 two ess: (1) properties of the lescon, oF 2) the point tn the deviation (3) from O-Stractore to LF a which sisures ae ‘mapped o PF (Stuer) (Stowell 1986), In the ctegory (pat fom Soasueanariraries an some lan ovary i the choice of substantive elements we Neve options to how nowuteantive (tinal) elements ae veazed (Borer 1984 Fokui 1986, Speas 1986) ane variations i slbel peepee of heads (6 do verbs pra oF Fllow ter complements) Trai 1984). In the eatery (2) e fn, for example, nage with overt 0x rent of questo plzae (Eas, tan, cc) and languages without ‘vat rmovenent (Chinese, Jape, ee). In thse ms languages tri the question phrse inthe poston that woul! be ocepid by a {race in languages wih over movement, there good evidence th simay movement operations ake place, bat ony inthe mapping rem SSiructure to LF, with no inition ithe psa foe isle: the ‘ranch point a which PF is ford rom S-Sirute precedes tse Op catens in te denvaton @) from D-Strctre to LE (Hang 1982, Lani and Saito 1988, 1992), Sin, we Br Iarauges wih over ‘matifesttion of gmmaical case (Gree German, Japanese te.) and ‘ers with vitally no such marsfotaion (Engh, Chinese, ec). But ‘sn, there ocd reson to bee that the ease ten ate bsicaly ‘le eres ngutialy and tat the erence pany in thei onesie ealzation (he mapping to PF). “The gen expectation fr al consractins, & tha langues wil be sey similar a the DStrcture and LF lvls, 36 the example as ‘dacs. us tha there or parame that act efor of EF representation or the compaatina! procs from §Strcture 10 LF: tie evidence is avila 10 the tanguage Kerner bearing on thee ‘ater and thre would be no way fr vals to be determined with ny ‘alibi. According, any varios atthe LF level must eres of DSwrctue parameter settings. or of variations i the mapping from Ponca Parancure ” DDSirscture to SStcture fo the extent thats proper are dtr ined rom iepscon of PF rornsD-Snustore, nt. rect exit proprts these 100 pear be nie in varie insofar as hey alec ‘De computational system. At tbe PF level. properis ofthe language an be eaily beret ed variation is pombe thin the Bred eeper tere of plone propertie athe insartan pricier of iver pb ‘sin SSirntures dre hot eontrsned by interface conics and ‘ary win the ange peste bythe variation of he strat el ‘he banc point o the PF mapping, and any independent contin tot may bold of SSirscure “The principles tat have ben iniesiated al into wo general este ‘ores: prices that ate applied v0 construct derivations (wansforns opal options and eonions onthe way they operate. amd pings ‘Oat apply to repeentations(iceng conditions). The warsormatona ‘operations are movement (adjunction, subsitition, deletion, aed per bape inoerion: we may think of thee ae ratancr ofthe general oper ‘ion Afect 2, arktrar (Lasmik and Saito 1988). Condon of tealty ‘nd cher constrain the aplication aed fetonng of these operations Lens contons atthe extem intro: eels PF and LF estabish therelation of language o other fis ofthe mind/brai, DStrcre ‘sontons specify the mone in wtich Knel propertes are expesed in grinmatiealsrctures. That tere should be S Suucturecondions les obvious, but it sens that they may ost Ge secion 13.3. "The principles have frtbersrctre, These eer groupings into modes oF tnguig:(intng tory, Bxbeory. Cae theory. ce). Cetin nipingeonepes eer ite many ral medoles: cenitons of fealty. somal” properties defined on phrase markers. and 50 en. There re also certain general idest that appear 1 have wide appcaiiy, ‘upong then. pencils of eensomy stating tha thee ca Be no sper ‘Maou symbots in repesnitions (he principle of Full Interpretation, For supefuoes sips derivations (Chomsky 186b, chapter 2-4 of this book). As these pins are giver an explicit Formlstion, they become empirical rypothess with spc mport and range “The prnope Fis assumed as ¢ matter of cure in phonology if 8 symbol in a represcition has no sensorimotor interpretations, the resentation des not quel asa PF representation. This is what we ‘alld the “imteface condition” The sane condiion sped to LE ‘entails that every element of the epesenttion have & (angae- Infependeu) interpretation There an, for eam eno next, m= cher cx vacuous quantifier at the LF eel. The principe of concmy of dei ‘ston reqeies that competstinal operations must be daven by sone “ondtion om Yepresenatons. as @ “lst resort” 0 overcome failure tomeet such a condition. Interetng wath eter peitipes of UG. sich sconemy principles hve wideranpngeflccts ad may, when Mars Fe properly understood, subsume meh of what appears to be the specie ‘Saracter ef pera penile. “The hfs in fcus over the years ker te tsk of inguin considerably sd yield ferent conceptions of what eonstites 4" sul” i he Study of language. Suppose we have sme eolsion af phenemcos in ‘tcl langung. Inte aly sages of generative grammar the task ‘9 find a rule stem of the preted form frm which tee phen ‘mera (nd intel any eb ould be dened. That harder sk than the ones posed in prepeneraie arama. but not an impossible fone there ae many penta role stems. andi seen posse to ‘evs one thst will more lene work though the problem of expla tory adequacy at nes aie 36 noe ‘But this ahieveren, however die, dees ot count a rel resale itwe adope tne PEP approach ao apo, Rather it rerely ss the pro Jem, The sk i ow to shew how the phenome derived hy the mie system ean be deduced Fem the invariant pines of UG wth pre Tes tn one of the permite ways. Ths i 2 far herder and more challenging task I san inportant fst thatthe problem can now be posed ently 2nd seed in ntereting ways in cme range of ease, ‘vith Fares sot af also interesting insofar as Chey po THE Wy 10 taster solution. The departure from the long ard xh eration of Ur utc inquiry i much Sharer an ore rial has ety generative ‘rapmar, with problens at are qste new and prespects tha appeae roising ‘Otter tradition! problems ao aime a diferent form under 8 PAP ppreach. Questions of rypoley and langage change wil be expec in tems of parameter cc (Lighafoat 1991). The theory of language scgustion wll be concerned with sequin of ecu tes, Ka of porameters dad parhaps ratration of principles (Hyuma 186, Rosner nd Willams 1987, Borer and Wevler 1987; Chen nd Weler 1991, (Crain 1991, Pierce 1992) 1 might tre oot that pares re beslly| ‘unform for all languages: he parsers for Engh snd Jopanes wou (ifr only in that parame are set dieremly (Fong 1991). Otier Preps nd Foams » sues would also rear soe sehihing. ith proach tars ot 10 ‘econ, Much of the mst Fit ini into generative rata the pst years has pursued te workin hypothesis that UG i spleen le ‘et heey, With fondamertal rnp hal havea inti character and broad generality. By doling he notion of eoscition and mov ‘ng omar "rule" stems the PAP apqronch cares this tendency onsideratiy forward. A jelaedassumpsion i that UG is “poms “dan,” the sere thet penonens ure expe by interaction of prin ‘pin one prticlr ay. Discovery That phenomena sre “ovedeter mined” has commonly teen ten to indent w torial deiserey that should be overcome by se of refined princi. The working Jypotneses have proven sucemel wt «guide to ngury.faing 0 He Sscovery ofa vast range of empire! phenomena in widely varied le ‘ges and to Torn of explanation that much excad what could be fonerplted wot many yeas age. Thee are eather arpiing fs. THE ‘iin ideas resemble thse afien adopted in the wady of ioxpaic ‘Phenomena, where success has fen bues spectacular nce the 17h ee "wry But lngoge i «biologal psiem, and biog sytem yp cally ate “mes.” iat, the esl of evolusonay tikerns” and stp by accidental circumstances and by physical condition har eld ‘complex sytems wih vara! Tunetons and eons, Regundency | oc anya pial feature of sh systems. Bt an expected nein ha i Indps to compensate for injury and defer, and to acormodate 10 4 diver of ends an functions. Language ese appears © have theo sted proper a noted ie «far fet th large parts of I ‘age ae "unable, he unl parts appt ofr a hoe ad “unpre pled spc ofthe al anguage Nevers, it Bat Bee & ‘nfl working hypothesis ha mts Baie tire, te lsage ae ‘ty as proper of simplicity sn elegance that ave not characteristic of complex organic stems, just ste infin dg character seers Fiolopeaty rater slted. Posably these concusions are etfs re Aecing a pariulr pattern of nga, the range of completely utr ‘lin an eppareny choc plenomer of anpinge lends ced ‘osauchshepcism, Sl. the propre that has ben made bythe contrary stance anot be onerioked “The PAP approach st sometimes tered Goverment Binding (GB) ‘Theory. The teinloyy i sending. Tro, ealy ots to see 0 Caper tent thinking these tems happened to concentrate onthe sears ff government and of binding (Chemsky 1981s), bu these medules of language stand alongside many eter: Cas theory, beer. ee 50 en. ay turn out tat the concept of poserinet hs a kid of unig role, but there is othing inherent 16 the approach tht requests. Furermore, insofar ax the theories of governent and binding del ‘sith real phenomena they wil appear in See Tom in every apoech to language: this arpraach tat no special ci on hem. Determination ofthe mire of hese and ether sens is «common projet, ot se ‘oft thi parila concepion of th ature of language ad sue. 2 The Leicon ‘A person who has a language hes aces to detailed information abst words of the fanguape. Any theony of language mst refs this Fst {hs any theory of anguage ust inode some sre of erica, the re poston ef al iosycrac) properties of partial lexical ies. These Property include a representation of the phonological form of eich tana speifeaton of te yar enepry. and its seman character ities, OF partial interest inthis dain ar te seman see tion aed thematic proper of lecal heads verbs. nouns, adjectives ad pre o postpontions. These sper the "argument sree” of & fxd ndiating how many arguments the bead Seems and what 5e- mantic oleh eee. For example the eb ie most be spied as ‘signing an agen ec, a theme rl, snd » gelrecpent role. 1a (3) ota bock, an Mary have these respective rematc(& os. (6) on grve a book to Mary ‘Te sssacntion etwen assigned Eles and argurem positon isto 8 lage exem predicate, For example, “agent” i apparently never a= signed to complerient Ard to the extent that de association is p= ‘table ater than ionyerac, need nt (hence, must net) be aed in paniculae ea ents. . ‘This conepion ofthe lericon i aed o that developed in he 1960s (@omeky 1965, bu depart from ti certain aspects. Thee, sob ‘cxegorization and scketonal cottons played cata rte. Te Fr ‘ner condone fr ical Head what phase enegones kes a8 emplerente for example, that Kick takes an NP complement. The ter comiltons speci insase semantic features of he complements) Principles nd Pasar 4" and sujet fn this case the NP complement of ik i [ concrete ‘us noted in section 1. that pluse sacar rales er (rey) rn ant wih subenepoizatin, ence sre Carey) elma, But now ‘ot that subeatporizaton follows alms ent rom ble specif tion, A verb wth ne ole to assign fo a complemen wil Po Be alo take complement. A ver wih obligatory) Ooles toes will ave to occur i = configuration with enough arunents (pws inttng complements) to racivethowe Doles. Further, at Hast in prt see ‘ional restriction wll ako be determine by thematic properties Tore ‘te «particule Grae she inherent semattic etuer ofan argent ‘mst be compatible wih that ole “These tentative conclusions about the ergarization of the lion ese importnt questions about he seqsstion of ene Knowledge. Suppose that subeategoczntion (cseecon) is artic is lcs derived from secant properties (selection) Is reasonable to ak whether this ‘sensequence ofthe acqustion protdur el (Patky 192). Peet (Geveoping ides of Grishav (1979) sogets tha this st beso. He mpares the primitives of eselection (ymactic categories such as NP, (CP, ec) with tow of theory (Ypen,” "pes." “peal” ee.) end farms iat the be, but not the former, met hat we my call the ‘nition of epistemotoicelpriovay. That they can play be pled by the earner Io provide a pretiminay, pings anal of eatonabe sumpe of data ane thas can prove the bass for develop en rom the seit site othe tend sat Ths is an tractive Ee of ‘essing, but. piven our crrent understanding of th is, i ot ‘concsve. Whe i doer seem core thatthe primitives of eseectin oot have episteaolopcal rity, i is not a all ler that those ‘of sseection do have such x wat. Although the notion “gent of an ion” is possibly avaiable tothe child in advance of any syntactic vowed itis les lee tha the @aeorese nation “agent of = tence” i. Tat is, before the eld know anything about the syntan ‘of his or ber language (beyond what is ges by UG), ean the ehild de ermine what portion of a sentence coitus the agent Fete, the ‘idence availabe to the learner ely consis of sentences ater than simpy individual webs in oaton. Bu such entenesexpicy spiny celection ropets: they exit verbs slong with ther complements ‘Thus, the cll simultaneously presented with cadence bearing on ‘eth eskection (given tat semences are present in conte, and ws ‘ming tha the revant coments can be determined) and esteton, 2 aps is reasonable to ase that both aspects of the evidence cotibte to the development ofthe knowledye. Alongside the sta of afi outined by Poetsly, he converse stuation with exaction evidence tfc po- ‘ing infermation about te meanings of verbs (Lasik 1990, Ghat 1950) igh lo obtain. For example, expose to a seneace contig {lua complement to sn onfamiar verb would lea! the leaner 10 Iypotheze ha the verb oe of propositional tide, "Tas serio not neces efit wth Pees’ ini pint out the onsnizaton of leucal eri. The mens by which krowledge is arved st & pt insriably refed inthe form tat the Know shimately takes For example. Grinshaw (181) ages ht he seq. tio ofthe syenaciceatgory of leaker based in par onthe rion “enon uct! realization” (CSI), The CSR of hyscel jects that of un action is Vand 30 0m I the sbsener of eines ‘he cid will assume that a word belongs to ts CSR tat 2, a word refering te am actions verb. AS Geisha indicator, wile Such “semantic botsrapping™ might consti prt ofthe ecqision po ‘eure, the resulting sca) state lexicon bas no such requienent. Lat fies commonly have aouns, ke dracon, fering 0 ato 8 elas verbs ike, tat dot eer te aetcas) "Note that ths consideration indicates tht lea enc contain a test some syntacte information in adton to tbe phoologe! od ‘sea inflormaton that surely ust be preset. Grins arpues that farther syntactic seication s needed as well, ction i addition to saelation. To conser ope example, Grimshaw observes tha sean category "quesion™ can be srutualy realized 85 either ‘dase, asi (6), 028 NP, a (6) Mary ashe vba imei as) (7) Mary asked te ise] The vec ack sentially svete» question, Grimshaw argues thai is alo necessary 16 specify thar it estes dause or NP i order 10 ‘isting from wonder which nly takes a cause (where ® indicates ovine) (8) Mary wondered fat imei wo (9) *Mary wondered the ine} Smee as sugested above it spose tore mest of eaeation 10 rcestion, she question arse whether such redichon mighesonatiow ‘be sill inthis instance as well. Psetsky argues ha ii, Ase wil ser in section 143, NBS must esive aract Case fom a Cate signet ‘wile uss need ot. Hetecorth, we wil captaie the word Casein isechical use J Given tis, Pesky propos tat the difleence be ten ask and wonder ped nat be stated i tere of easton, but rather follows ffem « Case difereoce: ask asens objective Case but ‘onder doce not In this rept, wonder pater with adjective, which ko do mot aespn objective Case. (1) Marys uncrai ae ietis} (1) *Mary uncer the ins} Pscsky presents Farber evidence fr this Case asiging dstinton be ‘ween vets lie wk and thse tke wor In English. generally oly ‘objective Case-ssignng verbs can occur in the passive, Given this (6) and (8) contrat in precy the pated fashion (12) st as asked whan time was (13) Si yas wonder what tie twa ‘As Peseshy notes, descriptive generation pine! au by Grimshaw now follows: among the verbs tat sselect guesions, seme eet ‘se or NP site otberse select ony lave: none case only NP. ‘There are Case-assning Uiferenss among verbs nd hese a relevant to esskction of NP (because ofthe Case rouirenent of NS), but ot kas, This reduction sms quite succesful for 9 wide range of cases, but it is important to ote that formal syntactic pcan in tenes! eres hve not been emily ebininate in favor of semantic ones. Whiter ot Dota verb asigns objective Cae a8 far ai known at preset 4 purely formal propery et deduce from semantics. While much of ‘celaaion follows fom sadetion, there ea rytzctie vsidu, sata, it Pesky is comet. in terms of texaly esynerati Cae properties We will introduce further properties of the lexicon as rguied By the 13 Computational System 12.1 Geneal Properties of Deiticos and Reprseatons ‘The generative procedure hat costes the I langue consis of & lexicon and 4 computational system, In teton 1.2 we oud some 4 Chaper proertie af the lesion. We now ur ote compusational system. Un ‘er the general asimptiors of sceton I. we conse the four levee ‘of eepesntaticn of the EST syste and th relations that fold among then, fects atti oe “harow stan” that the eration lating DStuctre, S Stuur, and LF DiSirucue, LF, and PF are irae eves, which sty the general ‘condition Flin manner wo be made practe Each Inel in a ombobc fysem, consisting of atomic elements (pines) ad objets onstrated ftom them by conctcration and eter eperasons. We ake thee objets to he phrase markers inthe friar sense epeseted conventionally by wes or labled bracketing): Each prime a eature compe, though for orthographic convericnre we will gntrally use conversion sya bole For conretena, take eatepris to be asi (14), for noua seb, jetves and pre-and pesipestions respectively 09 & NeEEN Vy) bVeI-N a) © AM[N AY] 8. PoEN VI ‘The feature [+N] ¢ the wadtionsl sbstantee; the Rute (+Vh sedate, ‘The primes consiuing the termina ining of 4 vase marker sre ‘dawn frm the lscon oes at projected from these hea by opera Sins of the computational system. Elements that pet no farther ae ravine projection: tn inleral nottion, XP the maxi projection| {om he terminal eatery X: tha NP is he mania projection oft ad Nand 0 08 So ston 1.3.2 he io base relations of = phe marker are domination ad fn- ‘eariy 0 te pase mater (15) we Say that B dominates D and E, © ominats F and G, snd A dorirate al ther ealesories (node) Fur thermere, B precedes CF. and GD presses EC, F, ad Gr ad 20 09 A x oN ZN, Cy £ S icp sn Parr % TEX is 4 ead. is “sister” i is complemen: ths, iD and Fare heads, then E and G are hr comriements in (15), We ssue that or ‘eng eclations are determina By a ew parameter stings Thus in English. ighr-vonching anpuae, 0 heads precede thet complements wile i Japanese. lef tranche Language, a hed follow thir comm pHements the errs determiged by one sting ofthe head preter Eyamples below tat absract frm partes Inginges ae uly 13 te interpeted independently of the order given. Domination elatons sr determined by general prices econ 13.2), ‘One findamenal eonzpt that applies thraghout the modes of grammar is camend (Khia 1964, Langacker 1949, Lasnik 1976, [einbert 1976, Stowell 961, Aoun ane Sportiche 1961. We sy that 9 “ccommands Bi «does vot dominate and every y that dominates 2 ominates 8. Ths, in (13) B ecormmands C, F, G:C ecommiands B,D, E.D ecommards E and comers, F ccommands G and convey ‘Where 7s resrcted fo manimalprjeton, we soy tht command ‘A Second fundamental concept government (Chek 1981 1980 Riz 1990), 3 more “ca yey f command to wich we retin seston 141 ‘Given ie language L, cach SD is a seqoence (8,0 hese being ras markers dawn rm the lets PP LF, Struct, and Struc lure, respectively. The creat 8 recs properties ofthe ems sete fom the Fescon as these are interpre by the prices of UG, with the parameters fed for L. The ements @ tnd 3 ae formed by suc ‘ye application of operations of te eampurationsl tem to 8 ey will Ine he properties of 8 8 modified by thee opertions. The PF rque= ‘enition x a ting of phonetic pres ith slic and intonation ‘srocture indicate, derived by a eemputaion from o. We assume tha the primes themsees ae not mes in the course of the derivation from 802, ‘Apical exe entry consists of « phonological six sod other feature, among hem the eategoal features N, Ven s0 0m nd io tne case of Ns, Case and aprezment Fears (Gerson, number, gene) ‘znetorth ectes. In piel. any of hese Fates may be lacking Tn one case of particle tee, the entre pons ti i ck ing. this case the element is am EC mp entegory). Atsong thes Cs we have the elements # of (16), (17 we ase * to indicate seers evan, 7a weaker aie Ps gir (16) John expects [eo hur ise) its commento hurt ones] 07) earned yesterday (he anived yesterday") We refer to te EC of (16) as PRO, an element that can be conrollad by ts accent Ua, i (16) oF eam be aba a itespetaton, 35 in 16). Pnsbly te ater is alo a case of contol by an EC eccurying ‘he same postion as fru in (18 (Epstein 1964) (08) is convenient for us for others to othe har wk) so, PRO's alvays conto. See section 142 ‘The EC of (17s a proserial clement, tenefoth po, Is not pe ‘ted in this position in English he counterpart woald be rama Sn Mins, 2 ml subject lavage. On factors relevant to tig the pe rameters see Riz 1982, 86a. Huang 1984, Borer 1984 Jeg and Safir 1989, This BC ute mich inthe manger ofan ordinary pronoun, Irving reference ine by context or by seme anteedent in wn approp te position. The ruc! Felton of antec pr) pie se, fr thermore ereral tke tow of omeedent, pronoun) ana nthe those ‘of congo. For example, na ell subpet language we find the eau [ent of (14-6), analogous to the pair (18-€} eb taken to Be the anced of pro, (19) a he people thax pro taught amie Soha . ‘pm adiited Jobe the people hat he aught admired John “he adie Sohn “The behavior of pro and eis simi, while PRO can ever appear in thee postions 'A thd ype EC, not denen from the fesicon bat crested in the coun of x Sein, tated in 20, (20) 2. wonder vo Jobin expected fe10hurt inst} dob was expected eto Ha msl] Wie refer to his EC a rave (9, telational neon mace-ofX, whee X Is the moved element seving a the antecedent bd the ce. Thus, Jeb Winds in (200) much 25 ebieds the reine of 35 ehey ids he ‘espa ia (21), ie turn binging the refx. 21) they expecta each otber to hut themes] Principe rd Parmeter ” In 0a) the Wace ofthe NP who. Te trace futons aa variable ‘ound by nha, understood se restricted quantifier: Tor which e, «| person. Here © tira binds hi uns each oer tnd thoeles 4 (21) and il binds hin iv C2), with Bul substiaing fo the ve able of) (@2) fot expect [Bi tout hist] 1 Qu) both «and Aone ncn ws variables bound by the ested quantile 0 tha the LF fon would te ioterpeed “1 wender (or ‘eh cc psc Job expect eto hurt ef. Note that ear ing the erm find ere vo cover he assecition of at antecedent wht ace ‘ule geeraly ncn he case of he (sya) bing ofa variable by aguante ihe elena we aso we heer, at Ln the sense ‘oF quantir arable nding In (208) the ve-b war is composed ofthe lene element be and the Jafletonat elements [pos 3 person. single} Assume now thatthe roces of competion niin the copa to the nfltional eres (ing. Real that there are two interpsasions of this proce (I) rasing of fe to the inflection positon of the setense Yo coberect the combined form [+ eflering. oF (2) ring of [t+ intons] (vas, dra rm telecon wth its fener ated assigned) to the Toficon posion, whee the features rechecked. she way, we have & ‘cond ace in (200) = (23) (23) del was expected, to hart hist] ‘The EC ey isthe ince of fe oF war: ci the ace of Jon, binding msl Ia each ese the tac decpies the poston from which sane ‘dent was moved. For cincretenets, we ado the checking theory So that we ave wasting in 3). Raising of wa othe efetion postion is neesay to check ees onal proper. The see is tre ofthe other ifsc ves fo xa ple, wand (2). whieh spree I person. singunr Ths, 9 ler (thoueh sil oni prt representation would be (24), whee ei he tre of wonder. (24) T wonder who Fon expected to burt ise] “There i reason 1 Believe that in English (24) ean LF rpreseeition while the cousterpare sn other sma anguagss te, Freeh) is an ‘Structure represtation: (23) and is courterpars are § Struct ep rsentations in both kinds of lnguape (Ernonds 1978, Pollock 199) » upte “Thus, Engh auniares reat S-Strotue but mn verb rise only CLF, while the comesponding Pench cements all aise at SStroture English and French would then be identil in leant respects DSirvctore and LF, whl dftring a SStructre, with Eopieh (25) (coresponine othe haul shared DStracure) versus Frese (25) (conesponding tothe Bascal shared LF for). (25) «Job olen ses Mary) by Jean embrase souvent Marie Sean se often Marie leformaly, dhe ace functions througout asf the antecedent were is {Hat postion, receiving and aspring sntactie and sete properties. ‘Thus, es isthe normal poston of the antecedent of a refenve both (2a) and 20). Ande 25). che race isthe ert head of VP, as: ‘png parcla emai vole and granimatial Case to is sonal objet ‘Note that PRO and tae ate quit diferent in ther properties. Thus, ‘an dament tht eorzos PRO isan iadependant argurontn the sense of Section 12, esigned a independet setae ee: bu an element that binds w trace fot Compare (16a) ané (200) peated hee: (26) a, Jobmenpect fe to hur ire) 2. dob wt eapected feto Bur hist] 1 (2) Jon isthe sujet argent of expected xsl ain (22): he EC contrled by Jal as s independent farcton a sbjat of art 1 (2G), in coast Jon has no semantic fle ther tha what i “inher” from its tac, as subject of fr. Since the subject of 6 “espero s asigoed no independent argument rec bea onan tment (an ele) 22 2 (20 theres expected to beam ti tomer Other dierences of interpretation fellow. Compare, For exaene (28s) and 2, (28 a. you fiends oped [to ish the meeting happy) 1 your frends sered [to ish the mec bps] J (8a) your fronts and e ve independent aeumenis,asdened thie serantc roles assets of hope a fish, espetvl therefore, the EC rust be PRO, enti by yur re Bu over as 0 sea ‘nero ot subject whic can agin beam expletive a8 5). rips aod Parton, » (29) ise your reds isha the teeing hp] 2 thee seems [eto te mistake in your argent) ‘Aecortingy, the EC in (258) ust be trace, wih its anedent jour nents tesving es semantic role a8 an argument ata were sn tat postin. We know further that the adstive happy modifies the subject Fis own clase, not that of « higher dause Thos i (30) Mary med meeting ot your rend: the sentence mens that your fends ope that che snospere would be happy when the mesting nds. (20) your ends hoped he meeting would Bish happy} tn (2), the, dappr modes PRO tn () a trace (6) Example (248) ths means that your neo ha crtain wih: hat they would be hap pj as he mctng ens. Bur (280) ha ought the means of (2). wth app) medying your ene Otter dierences of measing also appear, as in (a) and (316) urae 980), (01) & ene ransator ath ws expected / wo be signed 0 the ising diplomats >. cue wanstor each hoped PRO tobe asgned to the wsting Aiplomats 4n Gla) acer ene womr ach or its ae sf 3 postion with Independent argument satus, Therefore, the argent phripe ane one lave och interpreted as if were inthe peton of the tae with te argument satus of objet of apne: the easing that Hwa ‘expected that on tramlator each would bead to the visting it ‘as (ie, enc diplomat would be assigned one Wansaton). In (3D, It ne motor rach aed PRO ae indepenert argues PRO, not one transtrer each, that Winds and is ieerpected a were in that position The sojer one rant ed ish Be with 49 Jnverpetton very mh as tr in the sma consruction (3). (02) one wansatr each hoped that he would be asgned to the visting ‘ilomars Alibouph the argument status ofthe anacent of trace deter- red in the position of the tree, the antecedent may stil have a8 independent semantic roe in other espe, Compare the examples of 3, « haper (63), ‘tse to each eter lat your frends ae happy] 1b your fend see i esch oer tobe Ha © itees hat al your ends have ro et ried {all yor nls sem to ae not se are In (3u) yeu friends can bind he repeal each cer, bat it can in| (630) ths fonconig in its overt poston, et that of its wace: Tn 33) nd (38) the oven postions are revan for determining ops P| {ie tha, only (35) ean mean ta seems that nt all 04 end Fave arived, with no aking see over all, We se, then, that sco property and arpument satis are celeined be dillerent ays fe nteedenaraceconsrucions. Sch fics a these cught (0 fl ut 2s consequences of the theory of ECs and semi ioterpreltion. See section 142, PRO and ice alo difler inthe sytacti dsrbation, Thus. in (34) we sce the properties of contl ith the sntcrten snd PRO funtin- ing as indepndert arguments; bur the preperis of tae, with only foe arpuren. cannot be exhibited inthe analogous struts, 28 (3) irate. (28) a. ob ake wheter [PRO to ese) ‘akin expected that it would be fin [PRO 10 vst London} (68) a, thot was asked whether eo ae] John was expected thar it would be fun (vist London fe fe, ace ad PRO do not overlap im thir dation: the fet sould, sin, al ut ofthe theory of ECS ‘We aho allow fourth ype of EC, one hat has only the eteporiat Sestures 4, £V), projecting the uml wa. They serve ony at fa for movement oe filed in the courte of demotion, Sine thee ements have no senate role, thy will not saisy the econ FT UD Strecre (as we wil sharpen thi blow), and we may tentatively ‘sume tht they andthe srtres projected fro them are ised in the couse of éereation, in a manne prt bythe theory of phrase struts See tien 43 fo farther core Tithe kinds of EC are inde dinint, han we expect ser to die in featsre compostion (Chomsky 1982, Lasik 198). Opin the {estres soul be st those that chugs over: cements AS 3 HSL sppreximation, suppese tha overt NP fal into te eategoves ene Poi ond Peas 4“ eerie, reciproeals). pomaun, aud rexpression atm, the rina ‘Square root of 2, an the expresso that are "quatre in the internat sense of scion 11). We might sue, then, shat we have two two-tlued features, fanpher! and fponominal. with potentially Tour catezeis (06) | +naphor, ~pronasina 1b (anepton + pronominal) [-anapher = pronominal] & [anaphor + pronominal ‘Au expression tat s+ anapor functions eférently ony in intra tion with ts aniecden te reference of a expression that is (+ promo minal may be deterine by an antecedent but cs rer), Retenies and reciprocas thal it category (3) and pronouns nto category 166). The tid category contains clrents tha refer but ae not refer tnvlly dependent: neapeesions, The four Ee discs above would Te she ypotegy of 37. (G9) & Trace of NPis[-+ nophor, —pronemirall 1. Frois[—araphor. +proneninal. © Trace af operator (anal) 5 [~anephor, — pronominal 4 PRO'S +smaphor, 4 prosominal) “Thus, te Of NP is soneferets, ro hes the properties of peo oun aa vanes are “referent im thi they ate pct foe expressions. Cooled PRO alone category (7), here all PRO apparent uncontrolled PRO actly hss a hidden conto (e (18) ‘We would expec, then, that race of NP ro, and vable wont share felevintproperics of overt anphors, pronouns and rexpresons, Spectvy. Soc ements as English en, French am, German man eight ‘paral overt counterports o PRO, sharing the modal interpretation of abirary PRO and is ressicuon to subject postion (Chomsky 1986) “These exputations are meget satisfied, when we aback aay fos ‘ther Fairs. Tha, he stretra lan of eet it terse ‘nsicaly that of an anaphor ots antecedent: bath ea the arsed fn mus command the race, and oer sretral eons mist be ‘s,s asrated io (89) wit the examples hep sighly dierent Yo vod fotos that br the urate rote, 2 Caper (8) 8 Job ur ims fi John war hart “bine housht Hoh sams to be inttigent i. “rahoupht ohn ses that its rain 6 ob dei inst ef eal) 1. ob ak died ft eve er) “Thee properties sharply restrict the options for movement of NPs rising not lowing. bjec-to-subet ut nor conversely, and 50 on (engo 1977. ‘Siolar but aot que idenical conditions hold of PRO. Ths, she ©-Command Condition iatated by C9). 9 eon expcts[PRO of fms) 1. “o's mother expects PRO wo hus hms} Sob expects [PRO to el Mary's voter about Reet] 13%) PRO isin poston to ind raf But the Command Con tion requires tht its antecedent Re Job, not Mary. Sires, varies share relevant properios of expressions, as expected (40) 8. They dhnk od wl eave tomo} iT wender who they thik [wil lene tomorrow bk sitseens Hot tobe itelizes] wonder who i seems 10 be iliget} ‘© hedhiks ohn sittigns) fi Tere who he hiks is iteigent} Si John thinks Peis ineigen) '%. Twonder who [tins fh tigen Go (40a) she name aod the viable appear as Case-marked subjects of fre das, andthe expressions are well Fr, satisfying the Case _markng edition on rexpresios, © which we return diet Bo (08) the nate ad the variable pear es wb of initves lskng Cs, land the expressions ae severely deviant. I 40) isnot feel ound by Jl (we cannot ake oreo Tok, a we may (0) and in the parallel svuctre (4c) he nd the variable 7 are unrelated ‘ferential (we cannot take et bea vasable bound by te opersor sto, which binds «35 we may in ei). Agim, many conditions om ‘overeat fal out a pci aes, rips ana Parte a “These ECs als have ote features of over eapressons, epi, fetes Ts. the eae in (2) as thefts masculine, ga, ence the choice of overt sao. ’An EC lacking the typoloisl Feats of (37) or etre i i serpreabie. bene mperissible at LF by the piste FL. Such an ee ment denied only by its eatepoil Features (NP, Ve) may appear ithe cous of derivation, bat ony a6 poution to be ied o eet ‘wi eliminate. 11 an open qustion whether movement alvays lve «ice, and ‘wher whan it dows, thve ar independent reas for thi. Fer the putpeses of exposition, we tentatively assume that moverent of en ele nena lays leaves trace an, inthe splat cat, frre choi (@.2) where othe head of the ein, the moved elements i ie ‘nace, The chain sun char its ea asthe property X: we ern to ‘elevaol choices oF X. Te ements sujet to iterpetation atthe iter face lel LP are chains (Sometines oucrnenbers). cach ah abst presentation ofthe head of he cin, “The movernent operation henceforth Move 2) ia invariant prince fe of computation, stating tht x catepery can te moval to 4 tree positon. We tate tbe moved enegory and the target to be primes (excel items, EC tarps for movement, or potions fom the mini ‘ements, ith we options her the awed eager a races the target (substitution). of # ago to (adunction. sei (41) (ower relevant, rte tac of Band, te cccurrens mA . Any further consents on moverient wil be derivative fom othe pine ‘Pe including eondiers on representations ‘There ae to nstura interpretations ofthe eaments fred by a jmeton: we might assume tat ech ecumence of fin 2) a catenory ‘io Ms own sight (Lasik and Soo 1992) or that together they farm ingle eaegoey (BB wih the two occurrences of as its cps “ capt (ay 1985, Chomaky 19869), Eira! direnes Follow, a usu. fuer theoreti stucture i arculate “The sepmen-caesory dstintion requis a sharpening of the cow cept of dominance and those drive frm it (command, ee), Let vs ‘ay tha the eategey I. Bi 41) ches X, exces 1 and coma ‘x tand whatever i dominated by these element). We esr domination to inlsion, Ths [fu dinates ony X. We ay that segment oF ‘enlegory 2 corer ii cota nes Ror =P. Defining the cor trand eations as tfor, 2 ecommands 3a (4), sce it wot om ‘ott (oly sontsined) by bot ¥ ince in 9 des not. We eur ‘ove the poperticn of bead and command tothe pestadjunetion tras tre, Thus, iy was the beud of the predjecton category and e- ‘emis 8, then in the pstaduncion sete [Py Bah ¥ semaine the ead and ecorimands 8, Where ne confsion ares we wi feo ‘he postadjuntoneatepoe BB simpy es ‘Sibsutuion x consraned By 8 UG principe of recoverabiliy of dle tion, bch egies tat no iforsation be lest by the operation; hus 2 say sbatiae for only tere & 0 feature confit erwsen then ‘The tact of svn wil hetcore bays be am EC with he sme tego features st the moved category (te srctre preserving hy othe of Emonds 1976. saisr property hokts for adjunction, i oper se Sean 13). ‘Mowe permits mukiple (steces- sive rom he D Structures (43). with the argets of overen inserted (42) Jol seems [to hate bor expected 1 eve] Tender who Hon toveht Bil expected ft ewe) (43) 4 esses to hav ben expected Jor to ewe] . T wonder e Sohn tought eBilexpected a ose Io 2a) we have te esi Job. Fit he ks Ulin) an in 2b) the ein wha 7 ao wih wo irks. The ead ofthe hin re al ho. respective ‘We have so far sued thatthe operation Move @ forms a singe link of chain, Asatvely, we might ase bat te operation isnot Move 2 bar rather Fort Chai, an operon hat fr tefl chains (oF 2) from the Structures 2) 4 ingle tp. With her theo reveal context. the dsnction mayb moe than merely notational (Se ‘apie 3). We tenanively aoe the more conventional Moe inter retain. The operation Move» sais narrow lcaiy conditions. ‘Soppse thatthe postion of he intermedi ee #82) ied. a in (G4) 50 that he ehain rust he formed with a single ok, skipping the ‘odked poston (ocupied by i, wheter here. respective). (44 4 oto sens ht was expected [tere] bb eka he John remember eter Bi ed {© "how dd John ember [rer Bil ied the i “The ehns ohn. (shat, how, A) wae the fealty conditions, and he expressions are deviant, hough stingy diferent ways fat ‘hat demand explanation i ters of properties of UG. Note that nese e) i the BF for ith ths nrpetaion that yi how Con ‘raed inthe postion ofthe taco that deat if how is sons sith remeber ee no devine. The singe PF form has to diet ‘Ss, on sharply devas the ether no cal that each element rust have wifem, anguageedependent interpeciation atthe trace leet LF (he pip FD), Some Cements “re argumens assigned specie senile colts (Coles). sich 8s gent ld goa (Se section 12s over anaphors, PRO, and rexpresions (i= ing varibis) are a arguments. Expeves ea Empl there Hak inne ae axsgred no Gels. Some cements (em Engh, French Iteian gro) may ambiguously serve as apumens or expives. By Fl, ‘xpiciver mst be somehow eave wt LF Geto 1.33), ‘As argument must recive oe fom «bed (@ barking) Am argu ‘ment may also recive a semantic roe (wher be eonierd a ole ‘oF not uheorierea question that we pu ade) by prdeation by an XP (ee Wilms 1980), possibly an epen selene (the reative ‘use OF 69), ith variable postion 0, (45) the job was fee to Mary [who everyone pred ha the best ‘ushieatons] ther Ps (ajc, such a veri pases aga semantic le 19 predict. a ead o another june. AS stested i (eb), move ‘ent of adjunets and arguments hes quite dileens properties (Hoang 1982, Kayne 1984, Las ad Saito 1964 1952, Aout 1965. Rize 1990, inqu 199. postions position to which w Pole sane. ‘The lemenis seeing iteration at LF are chins. Hence cach srgurnet tain (46) ms conan atleast on postion 46) a5.) “ chop Fusthermor, oy he postion aecupied by a) at D'Strcur, mis! te f Gpostion, The etn les in the interpretation of D-Srvture 35 1 ramimatial retzation of kcal properties. Accordingly, Oamarking trust ake place at Structure: an element, moved er no, wil hve at EF cuacty the Oomarking properties (asgsng and ress 0k) that it has a B-Siractue: From the same comiertion, eDOws tat toting am move into Spontion, gaining» Bole that was Dot 8 Siena to iat D-Siratre. Thus,» chon car ave no sore than ene pesto, though any ur of semantic roles maybe ashe in hs pono, fn(4) for example the wall cies 2 tric ole fom both Pot a ed (40) e paid the wal rod “Te theory of Case (extion 1.43) requires hat eer argument have rac Cae (goeubly reaiaed vey im one oF another way, depen ing on spctie morhologcl propre ofthe tnguogs). ese. ana urent cain (46) mest have ne and only one Opsitin (amy) find at lt one poston im wtch Case is asiged (a Case pesto) FFtoing Joseph Aoun, we might sink of the fmion of Case as 10 snake an argument hain vale for Omerking The Last Resort condi tien on movement (ss section 1.1) egues that movement peited tony to satisfy some conn, in prt, to ati ity (hence, FH). Once an element has moved to » Case postion, it ean move no further, al tevant conditions no being satisfied. I fellows the, tha ‘ery argument chain ust be Heide by Case postion and mus ‘inate ina postion ihe Chair Conon). Note that these conclusions held only for arguments othe han PRO, ananomaly to whith we return section 143-On the tans of ins eae by expletives ith regu othe Chin Conon, see section 1.33. ‘We have ofr considers chine that ori from an NP argent postion of D.Stucture These al no the two types usa in (2), peated bee (48). obe sens to have been expecta to evel 1. Twoeder to Joti though (Bil expected oes} tn 68) we have, among others. he arguzent chain en 1) and in (42b) the opernornanable chain ho.) ‘Chairs may she erga frm non-NP pasitions. Oe ese, areaty inenioned, he riovemert of lec eaeyory Ged movement), an Proiglerond Peete © (22), repeated here ilustraing the sing of V to the inflectional postions (49) 9 Jobo was expected ro hur seh 1. Iseonder wo John expsted our its Here we ave the chains (rx. 1) and (son), te ter an LE chain for Engl Hed eovement is also involved information of compound wont i szany langges Suppose we were to form a caumtive verb meaning “Gusto al fem the underying D-Suctar (St) by aehoiing fll 0 owe ——_ i ee eae tothe V i ca “This operation yields the strata) the ace ofl ob ‘See Boker 1988, ere came is the head of @ two-sgment ves cate 00, if we assume a segment hey of adnction. A sezand kindof ean ciginating trom & noo-NP postion arises {Gon ovement of nonrguments (adores pects) a in (52, ($2) a. {to whose benef woud that preposl ber 'b. Mos: carly does he eae to theca ist England, Reever ot? [assur as Mary dont think hat Jn wil eer bee « ‘creer In cach cave the basket! ronarpuet isthe antecedent ofthe trace the ebting, then ae (Uo hase en (ow cars) England) le scsi ax Mary) respcovey. The quesione ele ‘nen in (528) ely whos the est "eased along” beers ho cam ‘ot be extracted from the D-Suucture position (53) pet-pping Ros 196), (69) shat props woud e to who + POSSESSIVE benef “The nasal interpretation reas the D-Strcture Frm; the meaning is “or wich person «that proposal would be ts benef” ‘ence thatthe LF form should inde be constued In someting Re this saaner ee seton 13.3. Case (526) might be ierpreted ilar, ‘ha the interpretation would be Tor what degre x he expec fx the ar fv earflly We tughy, then. argue tht these are not really cass St movement ef adhnet pass es sch, but rather ofthe question ee- rene wh, nt withthe adjunct pease carted along. We might con ‘he further that operator movement i the only kid of moverest 10 ich ade phrases ate subject, uke arguments, which ea form 3 fimeht chins The concuson twippered by the cservtion tha though adjnets can yp appear ss may sentence postions, thy fav not interpreted as i they had moved fom some more deeply em eae poston (sito 1945). Ths, (Si 201 gen the interpretation (oF (58), as it woud be if crf im (84) ad een moved from the DSirvetere postion of wei in (8), (58). carfly Joh tl eet te ear 1 Job old met he the ex ently) “This gets tha (52) might also be regarded a kind of pit ing ith the moved element how carping along the arse pase how eae ily, Se capers 3300 4 ‘Within theory of empty estepres ane chaise ca ete othe vesion of erection of nell relations rise in scion LA AS ota there, such questions ae obscure at Pes, and asne even more ‘ube under the wsumptions of trace theory” Consider again the S- Struciee representations (42) deed from the D-Simtur represents ‘ions (4) (repeated here) (65) x, John sems [to base Boe expect to evel 1. wonder ho Jobe though Billexpete (0 evel] Penal aed Parancters a ($0). eseemsfeto have ben expected oh to ens Tone fe Jon hough Bil epected fh 0 vel We now ask whether (550) are derived from (560-b). respectively, by movement of Jl, hs or water D-Strctar is deived from 8. Suture by algorthn (Spore 1983, Riza ABN, so tet DStraure is infect, derived property of §-Siructure: or wher thee simpy ' pondvctional elation betwen the pared expresions. Tse ar ale tative expressions ofthe retaion between $ Strate andthe icon. [A thee approaces are “aesformational” inthe ubact sense that they consider a eation between « “spaced element” and te potion in which such am element i stndorely interpreted, and in the case of (55) the postion in which i wou be over a §Stoctr inLguages ff he Chinese Japanese variety (e sttion |!) Sach dapacemet re tions ares fundsnenal feature of hua lage, which must be cap ged somehow. Apparent iferences ance slernative frmulions ten disotve. ov iui. to notational questions about how this prop ‘xy eaprese sila questions arse with par to apparent die cs Deticen “nulilee” approaches and “lee alters that tae plobal props of phrase marker in compen sito (Chomsky 195t, Herman 1863, Gazdar 191). 18 the present as the empl di ‘ingisabity ofthe approaches tans om ily thon-itermal eons erations We wil continue to adopt the derivtionsl approach of section TUL We asame that thsi at root,» question of tith and ft, ‘hough soe one ‘To se some ofthe problems that ais, conser he losaycontions fom Move a. A perl eoeiton ists in (4) thatthe target of ‘movement must be the closest possible postin, wih varying eft de ending onthe hind of movement involved. The condition vey st For ead movement, which cannot pass over the eles ecormarding ead (the Hend Movement Conran (HMC), «special cas of tore vert principles: see yecon 14.1). Ths, (5) formation of rom ‘he D-Sinture (ising lo the clase ntl postion, sats he HMC, but raising of rea to tis postin, erosing the possible target sition cocupied by wil ilstes the HMC, king the sharply evant Imerrogatie expreion (570) 7 & John wit cod the book 1b dot ead he hook (tread Join wil the book » apie But the teaity eons expressed in the sephysep computation ‘night ot be dry expressed tthe ovnput ees That servation tray stisy the HI an each step, but the cutpat may appear tn (ate that tbe condition is seated, Consider again the formation of @ “ausative verb meaning “euseto-ful By adining fal 10 case, 28 in (Gih Reza hats verb mut also be ised to the infection position ence, he reuly formed eategory cave must now ase this pos tion, forming the nuctre (8) (bere TP is ense-headed phrase. 15 thea of al, ad is tbe ace of eae fl). oo ee Here me have two chains: (afl) an (fal, 1) Each tp of ean formation sti the sre ecliy condition. But che resuling ein fended by Jal doesnot Inthe S Structure, the cain (fall) wots the FNC because ofthe ietervening bed a possible tart of move- ‘ment hi skipg™ by the chain. The fora soul hus be 28 devin I S70), bat iis wel formad. The lest condos are ssid ey ‘ssi te derivation, but are not sats y the ouput cain. Moa Chtom requied ter nonderivaiond approaches are net ently ‘srigiforvard 132, p-Stasure “The computational sytem fos SDs that express te basic struct Tacs oymtacte phonological and semantic) of the language inthe orm tt phrase marers with tena rnge draw from the exon. We ate SSvuming the sch properes of tara saguige a “placed elements” fre cxpresed hy multiple represtiona eve, cach spleen ‘cps Pacers and wit simple operations suchas Move eating thr. Fach fve ‘captures certain atest specs ofthe fll comple The Ration of ‘he computational system othe lesion expend a he ater ser face lvl of D Structure. DStructure 8 mapped ¢© LF, the interlace ith conesptual and performance systems at sre point (Stricture) Fethaps varying somewhat from langusge to lnpsnge the deivetion| “traches” and an independent mapping (phonoloey) form the PF rep resnation that provides the interlace wth the setsonmotor stone S68) ‘The east atempis to develop generative grammar in the modeen sense postulate a single lve of syntactic representation, fred by rules ofthe frm (9), where A iba Single symbol and X, ¥. Ze stops (and possibly nul) Ss the desensed nits, rd hee fs set of dina terial symbols that ar ten mopped by oer rls to phonetic fore (8) XAYxZv ‘The symbols were assumed to be comple, consisting. of two Kinds of leer categoria und sructoral Categoria mets were NP, Wad 40 on Sieur elements were fntures that cde global proerics of | Die markers for example. NP-VP agierent in the mam ae here 8 ‘ded by the [+ phar} ature asiend to § and “nheried™ by NP {0 VP through application of the ule [+ plea» (NP, + pal) IVP, + plural (Chomsky 1851. Subsequent work “factor the com ety imo two component, reiting the symbols te jst thet cae oil part (phrase stucture rks forming phrase markers) wed wing ‘ransformations! rues to express global properties of expressions (Chomsky 1975, Lexs 1963, Matthews 1968, Kim 1968). A ftr step festrced the recursive part of the generative proce to rules of the form (59) apd seperated the lexicon Iran the somputational sytem ‘(Chomsky 1969) This pronded » wove syst: phate acre ‘ules and Yes insertgn form D Strctize and tanto. Form the derived phrase markers of sufoce srr. thn subjected to pho te imerpretatcn. The Standard Theory assimed urbe that ony D. ‘Structure are subjected to smanti interpretation, a postin haborated in Gonratve Semantics (Lakoff 1971). Tee Extended Standard Thor (ESD proposed that eure stwture determines eral elements of ‘Semnantiinterpreation ackendo 1972, Chomsty 1972) Laer work Jed to the fouclevel conception of EST ctlined eal andthe PEP

You might also like