PU cea SRMOS TAA r TTR ea PEM ASD aTTELECD LEY CET IP VEN COB aU Lal NLL Parameters to Minimalism a aA tao WY aieeh ES Second edition Jamal Ouhalla Pee Sed) ta CA UL eh Reet BUNT Edward Arnold (Publishers) LimitedSR APES ES SAS PRE Introducing Transformational Grammar: From Principles and Parameters to Minimalism FERAL RIPE INS RAEI Second edition Jamal Ouhalla %& HE Sot SpE ST LH Et Rie - etCR) ES 155 FS SCBLPESE : 04 - 2001 - 2864 Pel feh8i28 El (CIP) SG PARE ROR Sib: MY SB BS fe Ha 9 ASE) BR BL (Ouhalla J. F BEBE. — JR SNE BE SAI HNL, 2001.9 ISBN 7 — 5600 - 2515-3 Le 1. OK OF We —/ERURE - BI -HR XV. Ho4 *F BAR AR 3 fir COP Bd BF (2001) 085279 S © 1999 Jamal Ouhalla All rights reserved. No part of this publiearion mey be reproduced, stored or transmitted by any means without the prior permission of the publishers, This edition of Introducing Transformational Grammer: From Principles and Parameters to Minimalisin is published by arrangement with Edward Arnold (Publishers) Limited. It is for sale in the mainland territory of the People’s Republic of China only. ARH ks Be ABO - BRAG th ALAR AD Pe Be SS HH AA A Rt MAS: ARUSHA GAR Jamal Oulvlla THE FR x oe « ERR: PROD ARR IT: SMBS STR Ah ARE ALTTTPR = =PILH 19 S (100089) hep: /Avww.flerp.com.en ARP Ae + 650% 980 1/16 » 33,25 2 2001 12 ASB LHe 2001 242 AME 1 KERRY 2 £5000 Ht ISBN 7 ~ 5600 2515 - 3/G+ 1189 243.9036 SVE FA SET PI AL TARREMLO: HBS Hh TREE D2 SBI HLTA: (010)68917519 *MHEDBREHSRH *SOPSRH BRE5K GQ 8435 BMBF AH ERERE = OE RR BEE (ARR SBA) Mee R-S PML RES sey RA BR BR AUN HAR AQ RH HK BA (REE YE) TKS F HL LRH LHP Lay FHS £2489 2 KR BAR RE The HR f BRM Nes RAL PR MIG MRE ER Ree RR RR F R SPH SER PTE BAK WES War MPR EN RRA RR MRA AR REA MbR aa ARS HEM RAR RB Raw FM mM BR & Fw Ha & F 4 2 €RS Hak Hen Roe RPA MR RB KR WM BRLHea SS CARBONE G25 AR St XE) Ht 54 AY 2000 4 9 Att CLE HARALD , HAD 5000 HK OER 10 A Gr EP 6000 B., HEE ADIL AP Me EFT AE A ESE ER ER HRRTREBAS WM, AMA PS A A A ABT, BEE EM APACE ALN ER BOBS SP AEC EB ESR. SE ARUERA , SMALE MY AC PED SE BY TBA eM READE AGH CHMHRAR DAA EAS SHR AINE RS ST SA — TEE FED HOC FED ET AL So BE LE, OME RL BR Ce HE tO E it 58 A. (LEB OMAARARE:— 4 58 PRX ASAR Te REA? PLD Ht mY 26 PH BR EY 33 hare a A AF UMBAS ARSE BAO BALERS, ETE FOR LR DYER SWAT EIR IN RT OR FER AR A $6 AL TRIE NTT MEM AS Ee ROA EA AEP OER HE = TA a RE Oe EL, HH SCE LHR ODE FS UR gS eS aR EA OAH BH EABA EH HRES GS ER POEM AARD ES RAERLR HUBER TEE. ACUEDE— TAR TT RHE RA. 2 OR BB a Se SFE AME RS 5 RA SCURVY HE. SS, SMALE MY BE, PTB GEE BEY A a LT AD AG EH HH AL BT KRE. Shi BE 5 FE Us HL BASES BB 2001 4 8HPreface by Halliday Foreign Language Teaching & Research Press is to be congratulated on its initiative in making these publications in linguistics available to foreign language teachers and postgraduate students of linguistics in China. The books are a representative selection of up-to-date writings on the most important branches of linguistic studies, by scholars who are recognized as leading authorities in their fields. The availability of such a broad range of materials in linguistics will greatly help individual teechers and students to build up their own knowledge and understanding of the subject. At the same time, it will also contrihute to the development of linguistics as a discipline in Chinese universities and colleges, helping to overcome the divisions into “English linguistics”, “Chinese linguistics” and so on which hinder the progress of linguistics as a unified science. The series is to be highly commended for what it offers to all those wanting to gain insight into the nature of language, whether from a theoretical point of view or in application to their professional activities as language teachers. It is heing launched at a time when there are increasing opportunities in China for pursving linguistic studies, and | am confident that it will succeed in meeting these new requirements. M.A.K. Halliday Emeritus Professor University of Sydney FoERK? KEK, SRREARAR, PRHZARRR, BRT AW RA AR: FL AB. AH AST: AEPERENERERAST; FRAH OS LR ES J, BEFRLEPRL EES T. RAPUMAEL, FI BEER ONESENEFZERA, RT BMERPE WA SE ARBRE, ARH - RBA, PALEAUARE FRAME, TARAS, EMRE, ANRELAEARH RHE RRAB ROM CYREAB EPS RAB SSX Do REPKE H BAERS, CHE TERESA BEF OSPR, KHPARMSSMAKSRRARS FRLEREGRA. PRRPAMR HESS oe AH, REREBSSROAERSRRRARE— HK. RMNRAK, MET AABRY SARK, RRPAAB AT PE-REOAA MER, EERLR, BH- BPRS PROB RAGAM, ARLE -ASEL ES REARS RHEE, SESS. RUNRBEAXER RAAF PADRE RE DR WR: RUHPERAABRYRA METH SE BF FRAO ARAE A RRA MH KAE, BAB, F TRAD, BRE SRK a Fey eM MEM A a TT fe BNEZ, WREREH, TEPER MH, REPRE HARREARA, BAREMY A, ULRRNH RS, TARP ARHAR, ADRES RGAR SAME? RM, REARH-FSAWAWE O Wy OSA SUMNER REE) edt Sa ATE. FIOKH, RERKEPARSPRAEA, RUAREE-BEW th Hie RAE AH, A Bb Be ie Be 4 AL BEZHRARRAARN, PRALARZRULAMBE. EEPHAAH, HABLERSESARAK, RBBER He, REATHERERWAT, CAAREH, TERER UAT MHRER. BERK AHEMRR, BARRE HAGA, ERE RH, MEME SOFRRAAB, KR WMHARESFHOAHH. EBELCAH, ANH -MHMH, RLRaD ill, MAK, STEM RARAB AAR RS, Hel dH Z Chomsky; WAHAB, BPARHST LM, RA Be, MEDARRABAPEA BASHA, RAARTHA ~ tbe RABE, RHSKMRERUKMAREM. A LHERRER PRU. CHES. RAR AARP ERRE AME, SHUG. EOHRARME REE, SAT ORS, AL, RAR EM ERR E, TERRA RR AR, eae AMA, LERBR A, ANEAH SE, RERNEMHA, BR ARKH, REPFRETART. ATRH, RNRSSHEH Ge, BOR OR A. AK, BER MLPRAWAR. EMAAR PREM, RMEHFTE I WR, —-AG GRBM ANRS RH RAAA BAT, RNARR-RLE WARK RPO UMA. BES SMBS*ASTIR BRST RARER GOKYATES Bil, MERE PATER R AER RMKAT ES © KE ALPR EAA RUGRH TSA WR EE, A-HH, ABT HAs LAR, Rte we RSPR A SE, ARHAPEH, CREP MBE e Se FitHOPES, MEHRRATERACL AA. BT MMI PL, ARE PR-APRAN BRERA IT -HELERHE AR. PUPP ARR L, BRECK SM A- PRE HA, RHERAAEG-R, -RRFRKS, PRES, EERAXH, ANRAREZEHAN RAB. BF PASEERRBA, AMEBSSRABE; MTR GREK, RRP BAMARPRE RR: ARERRAAE th, BURA HAR, BER RRARASA PEMRNALHSB HE, E-REESON TE, A MTPERMNWERER, AMAL EE HE EAB. RNEGEKE HA -AHH, AAR BHR, BLUE Sb HK LER SFL Be one aod Fi2Preface by Chomsky Tt is ebout half a century since the study of language undertook a rather new course, while renewing some traditional concerns that had long heen neglected. The central change was « shift of attention from behavior and the produets of behavior (texts, corpora, ete.) to the internal mechanisms that enter into behavior. This was part of a general shift of perspective in psychology towards what became known as “cognitive science,” and was in fact a significant factor in contributing to this development. With this departure from prevailing structuralist and behaviorist approaches, the object of inquiry becomes a property of individual persons, my granddaughters for example. We ask what special properties they have that underlie an chvious hut nonetheless remarkable fact. Exposed to a world of “buzzing, booming confusion” (in William James's classic phrase), cach instantly identified some intricate subpart of it as linguistic, and reflexively, without awareness or instruction (which would be useless in any event), performed analytic operations that led to knowledge of some specific linguistic system, in one case, a variety of what is called informally “English,” in another a variety of “Spanish.” It could just as casily been one of the Chinese Ianguages, or an aboriginal language of Australia, or some other human language. Exposed to the same environment, their pet cats (or chimpanzees, etc.) would not even take the first step of identifying the relevant category of phenomena, just as humans do not identify what a bee perceives as the woggle dance that communicates the distance and orientation of a source of honey. All organisms have special subsystems that lead them to deal with their environment in specific ways. Some of these subsystems are called “mental” or “cognitive,” informal designations that need not be made precise, just as there is no need to determine exectly where chemistry ends and biology begins. The development of cognitive systems, like others, is influenced hy the environment, hut the general course is gencticaily determined. Changes of nutrition, for example, can have a dramatic effect on development, but will not change a humen embryo to a hee or a moust, and the same holds for cognitive development. The evidence is strong that among the human cognitive systems is a “faculty of language” (FL), to borrow a traditional term: some subsystem of (mostly) the brain. The evidence is also overwhelming that apart {rom severe pathology, FL is close to uniform for humans: it is a genuine species property. The “initial state” of FL is determined by the common human genetic endowment. Exposed to experience, FL passes through a series of states, normally reaching a relatively stable state at about puberty, after which changes are peripheral: growth of vocahulary, primarily. F13As fer as we know, every aspect of language — sound, structure, meanings of words and more complex expressions — is narrowly restricted by the properties of the initial states these same restrictions underlie and account for the extraordinary richness and flexibility of the systems that emerge. It is a virtual truism that scope and limits are intimately related. The biological endowment that allows an embryo to become a mouse, with only the most meager environmental “information,” prevents it from becoming a fly or a monkey. The same must be true of human higher mental faculties, assuming that humans are part of the biologicel world, not angels. ‘We can think of the states attained by FL, including the stable states, as “languages”: in more technical terminology, we may call them “internalized languages” (T-languages). Having an I-language, a person is equipped to engage in the “creative use of language” that has traditionally been considered a primary indication of possession of mind; by Descartes and his followers, to cite the most famous ease. The person can produce new expressions over an unhounded range, expressions that are appropriate to circumstances and situations but not caused hy them, and that evoke thoughts in others that they might have expressed in similar ways. The nature of these abilities remains as obscure and puzzling to us as it was to the Cartesians, but with the shift of perspective to “internalist linguistics,” a great deal has been learned about the cognitive structures and operations that enter into these remarkahle capacities Though the observation does not bear directly on the study of humen language, it is nevertheless of interest that FL appears to be biologically isolated in critical respects, hence a species property in a stronger sense than just being a common human possession. To mention only the most obvious respect, an I-language is a system of discrete infinity, a generetive process that yields an unbounded range of expressions, each with a definite sound and meaning. Systems of discrete infinity are rare in the biological world and unknown in non-human communication systems. When we look beyond the most elementary properties of human language, its apparently unique features become even more pronounced. In fundamental respects human language does not fall within the standard typologies of animal communication systems. and there is little reason to speculate that it evolved from them, or even that it should be regarded as having the “primary function” of communication (a rather obscure notion at best). Language can surely be used for communicetion, as ean anything people do, hut it is not unreasonable to adopt the traditional view that lenguage is primarily an instrument for expression of thought, to uthers or to oneself; statistically speaking. use of language is overwhelmingly internal, as can easily be determined by introspection. Viewed in the intcrnalist perspective, the study of language is part of biology, taking its place alongside the study of the visual system, the “dance faculty” and navigational capacities of bees, the circulatory and digestive FI4systems, and other properties of organisms. Such systems can be studied at various levels. In the case of cognitive systems, these are sometimes called the “psychological” and “physiological” levels — again, terms of convenicnce only. A bee scientist may try to determine and characterize the computations carried out by the bee's nervous system when it transmits or receives information about a distant flower, or when it finds its way back to the nest: that is the level of “psychological” analysis, in conventional terminology. Or one may try to find the neural basis for these computational capacities, 2 topic about which very little is known even for the simplest organisms: the level of “physiological” analysis. These are mutually supportive enterprises. What is ieerned at the “psychological fevel” commonly provides guidelines for the inquiry into neural mechanisms; end reciprocally, insights into neural mechanisms can inform the psychological inquiries that seek to reveal the properties of the organism in different terms In a similar way, the study of chernical reactions and properties, and of the structured entities postulated to account for thern, provided guidelines for fundamental physics, and helped prepare the way for the eventual unification of the disciplines. 75 years ago, Bertrand Russell, who knew the sciences well, observed that “chemical laws cannot at present be reduced to physical laws.” His statement was correct, but as it tumed out, misleading; they could not be reduced to physical laws in principle, as physics was then understood. Unification did come about a few years later, but only after the quantum theoretic revolution had provided a radically changed physics that could be unified with a virtually unchanged chemistry. That is by no means an unusual episode in the history of science. We have no idea what the cutcome may be of today’s efforts to unify the psychological and physiological levels of scientific inquiry into cognitive capacities of organisms, human language included. Jt is useful to bear in mind some important lessons of the recent unification of chemistry and physics, remembering that this is core herd science, dealing with the simplest and most elementary strucrures of the world, not studies at tbe outer reaches of understanding that deal with entities of extraordinary complexity. Prior to unification, it was common for leading scientists to regard the principles and postulated entities of chemistry as mere calculating devices, useful for predicting phenomena but lacking some taysterious property called “physical reality.” A century ago, atoms and molecules were regarded the same way by distinguished scientisis. People believe in the molecular theory of gases only because they are familiar with the game of billiards, Poincare observed mockingly. Ludwig Boltzmann died in despair 2 century ago, feeling unable to convince his fellow-physicists of the physical reality of the atomic theory of which he was one of the founders. It is now understood that all of this was &ross error. Boltzmann's atoms, Kekule’s structured organic molecules, and other pastulated entities were real in the only sense of the term we know: they had a crucial place in the best Fisexplanations of phenomena that the human mind could contrive. The lessons carry over to the study of cognitive capacities and structures; theories of insect navigation, or perception of rigid objects in motion, or Llanguage, and so on. One seeks the best explanations, looking forward to eventual unification with accounts that are formulated in different terms, but without foreknowledge of the form such unification might take, or even if it is a goal that can be achieved by human intelligence — after ali, a specific biological system, not a universal instrument. Within this “biolinguistic” perspective, the core problem is the study of particnlar Llenguages, including the initial state from which they derive. A thesis that might be entertained is that this inquiry is privileged in that it is presupposed, if only tacitly, in every other approach to language: sociolinguistic, comparative, literary, ete. That seems reasonable, in fact almost inescapable; and a close examination of actual work will show, I think, that the thesis is adopted even when that is vociferously denied. At the very least it seems hard to deny a weaker thesis: that the study of linguistic capacities of persons should find a fundamental place in any serious investigation of other aspecis of language and its use and functions. Just as human biology is a core part of anthropology, history, the arts, and in fact any aspect of human life, so the biolinguistic approach belongs to the social sciences and humanities as well as human biology. Again adapting traditional terms to a new context, the theory of an I- language L is sometimes called its “grammar,” and the theory of the initial state S-0 of FL is called “universal grammar”(UG). The general study is often called “generative grammar” hecause a grammar is concerned with the ways in which L generates an infinite array of expressions. The experience relevant to the transition from SO 10 L is called “primary linguistic data” (PLD). A grammar G of the I-language 1. is said to satisfy the condition of “descriptive adequacy” to the extent that it is a true theory of L. UG is said to sstisfy the condition of “explanatory adequacy” to the extent thet it is a true theory of the initial state. The terminology was chosen to bring out the fact that UG can provide a deeper explanation of linguistic phenomena than G. G offers an account of the phenomena by describing the gencrative procedure that yields them; UG seeks to show how this generative procedure, hence the phenomena it yields, derive from PLD. We may think of $0 as a mapping of PLD to L, and of UG asa theory of this operation; this idealized picture is sometimes said to constitute “the logical problem of language acquisition.” The study of language use investigates how the resources of language are employed to express thought, to talk about the world, to communicate information, to establish social relations, and so on. In principle, this study might seek to investigate the “creative aspect of language use,” but as noted, that topic seems shrouded in mystery, like much of the rest of the nature of action. The Diolinguistic turn of the 1950s resurrected many traditional FI6questions, but was able to approach them in new ways, with the help of intellectual tools that had not previously been available: in particular. # clear understanding of the nature of recursive processes, generative procedures that can characterize an infinity of objects Cin this case, expressions of 1.) with finite means (the mechanisms of I). As soon as the inguiry was scriously undertaken, it was discovered that traditional gremmars and dictionaries, no matter how rich and detailed, did not address central questions about linguistic expressions. They basically provide “hints” that can be used by someone equipped with FL and some of its states, bur leave the nature of these systems unexamined. Very quickly, vast ranges of new phenomena were discovered, along with new problems, und sometimes at least partial answers. It was recognized very soon that there is @ serious tension between the search for descriptive and for explanatory adequacy. The former appears to lead to very intricate Tule systems, varying among languages and among constructions of a particular language. But this cannot be correct, since cach language is attained with a common FL on the basis of PLD providing little information about these rules end constructions. The dilemma led to efforts to discover general properties of rule systems that can be extracted from particular grammars and attributed to UG, leaving a residue simple enough to be attainable on the basis of PLD. About 25 years ago, these efforts converged in the so-called “principles and parameters” (P&P) approach, which was a radical break {rom treditional ways of looking at language. The P&P approach dispenses with the rules and constructions that constituted the framework for traditional grammar, and were taken over, pretty much, in carly generative grammer. The relative clauses of Hungarian and verb phrases of Japanese exist, but as taxonomic artifacta, rather like “terrestrial memmal” or “creature that flies.” The rules for forming them are decomposed into prineiples of UG that apply to « wide variety of traditional constructions. A particular language L is determined by fixing the values of 2 finite number of “parameters” of S-0: Do heads of phrases precede or follow their complements? Can certain categories be null (lacking phonetic realization)? Etc. The parameters must be simple enough for values to be set on the basis of restricted and easily obtained data Language acquisition is the process of fixing these values. The parameters can be thought of as “atoms” of language, to borrow Mark Beker’s metaphor Each human language is an atrangement of these atoms, determined by assigning values to the paramciers. The fixed principics are availahle for constructing expressions however the atoms are arranged in a particular L- language. A major goal of research, then, is to discover something like a “periodic table” that will explain why only a very small fraction of imaginable linguistic systems appear to be instantiated, and attainable in the normal way. Note that the P&P approach is a program, nota specific theory: it is a Fl?framework for theory, which can be developed in various ways. Lt has proven to be a highly productive program. leading to an explosion of research into languages of a very broad typological range, and in far greater depth than before. A rich variety of previously-unknown phenomena have been unearthed, along with many new insights and provocative new problems. The program has also led to new and far-reaching studies of Jenguage acquisition and other ateas of research. It is doubtful that there has ever been a4 period when so much has been learned about human language. Certainly the relevant fields look quite different than they did nat very long ago. ‘The P&P approach, as noted, suggested a promising: way to resolve the tension between the search for descriptive and explanatory adequacy; at least in principle, to some extent in practice. It becume possible, really for the first time, to sec at least the contours of what might be a genuine theory of language that might jointly satisfy the conditions of descriptive anil explanatory adequacy. That makes it possible to entertain seriously further questions that arise within the biolinguistic approach, questions that had been raised much earlier in reflections on generative grammar, but left to the side: questions about how 16 proceed heyond explanatory adequacy. {i has long been understood that natural selection operates within 2 “channel” of possibilities established hy natural law, and that the nature of an organism cannot truly be understood without an account of how the laws of salure enter into determining ils structures, form. and proverties. Classic studies of these questions were undertaken by D'Arey Thompson and Alan Turing, who believed thai these should ultimately become the ventral topics of the theory of evolution and of the development of organisms (raorphogenesis). Similar questions arise in the study of cognitive systems, in particular FL. To the extent that they can be answered, we will have advanced beyond explanatory adequacy Inquiry imto these topics has come to be called “the minimalist program." The study of UG seeks to determine what are the properties of languages its principies and parameters, if the P&P approach is on che right track. The minimalist. progfam asks why language is based on these properties, not others. Specifically, we may seek to determine to what extent the properties of language can be derived from general properties of complex organisms and from the conditions that FL must satisfy to be usable etall: the “interface conditions” imposed by the systems with which FL interacts. Reformutating the traditional observation that language is a system of form and meaning, we observe tha: FL must at least satisfy interface conditions imposed by the sensorimotor systems (SM) and systems of thought end action. sometimes called “conceptual-intentional” (CI) aystems. We can think of an Llenguage, to first approxima ion, as a system that links SM and CE by generating expressions that arc “legible” by these systems, which exist independently of language. Since the states of FL. are computational systems, the general properties that particularly eoncern us are FIsthose of efficient computation. A very strong minimalist thesis would hold that FL is an optimal solution to the problem of linking SM and Cl, in some natural sense of optimal computation. Like the P&P approach that provides its natural setting, the minimalist program formulates questions, for which answers are to he sought —~ among them, the likely discovery that the questions were wrongly formulated and must be reconsidered. The program resemhles earlier efforts to find the best theories of FL and its states, but poses questions of a different order, hard and intriguing ones: Could it be that FL and its states are themselves optimal, in some interesting sense? That would be an interesting and highly suggestive discovery, if true. In the past few years there has heen extensive study of these topics from many different points of view, with some promising results, I think, end also many new problems and apparent paredoxes. Insofar as the program succeeds, it will provide further evidence for the Galilean thesis that has inspired the modern sciences; the thesis that “nature is perfect,” and that the task of the scientist is te demonstrate this, whether studying the laws of motion, or the structure of snowflakes, or the form and gtowth of a flower, or the mest complex system known to us, the human brain. The past half century of the study of language hes heen rich and rewarding, and the prospects for moving forward seem exciting, not only within linguistics narrowly conceived but also in new directions, even including the long-standing hopes for unification of linguistics and the brain sciences, a tantalizing prospect, perhaps now at the horizon. Noarn Chomsky Institute Professor at MIT F19MRE WWHAA-PREERAESFER EMAL RKAR AUR, REXAW BARA ED EERE EHH Bib yk ew WAH. ROKAMEEREFR REDE, H AZTHRAE, ROFRRRI RP ERS HM, BE RETHAAK, ERROR T BRR ME, AHKATRORR HSH, ELTIMRETEMAB. & ARRAS THER, BTR PLR RAH RS AANA, WER ARR RE TBD, NREMPAS ST SAWRRPRR, RNA a ey Hh, AHR EM RAE EAP A, REBRWHE, AM Ae, REF ELAS AR - MRS CSKRRE) LA BRED PRTE A ERMAR, LHAPHRMERERB CPA ARM, MRE BARTRE EH - +E ORR BK, ETLGESS UES, OHEP SRR MAK APRABMHE RED R, CNARAKMAAREMS Y AREDHRARTHA, HHEREFRUR, AWBTER MRT KUO RRA, SRA SHER, GBP A BR AAEM IREMAUAY REA. CHUA ERE RANGES, REM EHF EF fe PEGA HBLPAE, R-RARR RSH PERS SO HAG aH BM CSR BB ES ARR FRE), HM S4 Aw 20007 9 AMHR, RARA Rio UE RMEDA HMI, RA pk Bed CRB) BARAK, AAAS Rae CRE) Maw Ree HR, BA A ode SB, RS Re aR ARTA, HMM TARE, BOR, AMEE. Ue Bee. RHR, BRAKE. PREKRADEAR, F20HRI OT RERAW SH, MRE TAMER RARR, RENEE, (KE) F-KAFE HORA BALK, FE PRBREER SD: POUT ROBBER, Bie BR, BAR, HAR. FRR. BAT. RRM EAR ER ANEH-EEEKEL, MM AAR CARE), & Ri CEBTA), BH CRA), BAAM CRM HE) SE. ROME TSO EAS MRD HSE, HGREMEERERRE-+EBREWA KR. RHRB, HIRW ERT ETRE EMAAR AS ROSEN BE, EHUD RARER. PR CREE, SRARELURWBA, VROPRR OW REL CASPER E RR AE AAU. WHESPRDAREREAPRRABEA EH HES, PRAWE TRE ES, RE CER MD vf AVE Foy PEAR Ra. WEA, tee ee, BMAD, ERRAAH HE, HOhEWSAS TE “RRRM LUE. RRR KZY OD”, RERMD BRD BNLS, MHS, RMT A. PERM EA “RR BRT, BRARP AR MAAS, xb Sh HEE SH, MAPLES LARA -MERw, HeRED AE BRE MAPCR- BREWS, CEM AK, TR, AMNZEN REA, SRPAPLLRAE DHE. RR WEAM THERE, ALA PBEM APOE RT MOMS, THRE LSHRA, BARES, Bee RHE. APRAERANARPT RE, AUB RERA HRA, HAV WR ERBE WE KM REWER, RPP, Bk COED HURM KES, Agee EHRY, AMAR (OE) MARA, REBRALARH, CRED RAR, WPM, RAM, APPRRE, FE BHA, LAR GOY SORA RER RSH, CZE REE F21HERR. GSHWRBERCEY, HEEPERE. — BHETAR, EHOR-K, PLAERARE, #RARRH H, BLA PREAH, AERA A-ARM. ATR HALeKRT MLSS, PARMAR.) BERRA S. RAT. RA-KAFRARHHEHRTTR, HAH AGHBPAARKREEANH MER, PRRIFD, HBABA KLE R REM MCLE) HH, SPHSAR PREM. we, HR” WR RE EER Sha, CEDAHS FH, BREACRSAAHR AMT &, REARRART EA - ROR, RET MARTE KH AEEE. BAY, PRADEHEKAARR, RERER BERETS HRT, MARRPEH-PERRHRR ARBORS, AMR HER RRR RR RR SNe TAREE) RTL RS, HAAR REG BRAHE: RATCHARREBA, RY RAB A XK RK. RUERE-PMESHR, HHRARREEWNER RA. HAASE Lee EERARAR, CHE EEA RABRKRERHA WEA. MALARGHBESE RK A ROR, COR) MRT RAW RS RR A Re PSPS BIER STR. BR WEA SIE F22HERAT, ONE AR ERUATA AEM IB IK. IA Chomsky 1957 46 1 epee) Mi Wa, AT SSR ( Cake) HB), STR CREAR) ESTE APOE RAM), GBR (HSI a) AMP CHM) SSN 40 SEHR, BS TE CREAMY, PHRMA SR BERRRAEE, RNRA, PD RE TASK EA Se te RIA, ABP SS h, GAD ERA RARER RE, LRADR Mm EDERAL ERR. KERMA RR MS AUR LATER. BA, PERALTA PUIG SEI EET TAA 2 R—, PRE RIPE AMER IE SU BMA, SREY URCEM EAE, UA-TILB EAT RNR, EIT A A BRERA, PRATER BMA RTER ABR ALP, HRA MMS DL” AHERN, KAT ADA. TE Us UA BAY Chosmky 69 “a EL” (Mentalism) at “APR EX” (Nativiom), UH -PLBEE PARA BARA Ee TE, A 2818 SOLAN ARES —, EIATWAWHAEE TMM RER AS H-PRP RR, REPRE ESREMEMT. HES ARM AA RK, CETRATMAMERRARR. KI” T- SMB ARH ED SOREN “SR”, AREAL BR PRAY OE. TARAS BE BD TA RE Mt 50 FALE! 60 EAR ER ay “UA de” (Cognitive Revolution) ABT T AR. AUR MGA ICI EAT aR SE A ES cy SR, HARA ALTE PUIG HIB AOE RR, PE RE GAMER ARE, WAMU Fi SORE TA ABT, le AFRFERE A TEA TERE LM AMAR 2 WHE RE EE Oe LASER. FELT A Os fh MR A DAS Be EAB EL 3 SHR, HEARNE N ERAS, ROMP HS HA ee F23RATA S, SRAM TARA DEA, MiP ERT FEA ATR PUEDE EEE, HRT, PRE BRKT TRO MAAN ERR, Bl, RRERAKT FATE SAT APE RAD, TORRE BRR RRR RT A BR ATARATET ASSIA FMB. SB, FERRE RI Te Ae IBA FM Da. EK A RA RUE ACCUEIL Ie, RE SHARSIC MILB AR ARM, CHRCERLRERRE MC, PTE HY PR SESE A A ee Sa EAR GERREM PM BRAK LG, IRAE RE HAE FEMS ARES BL Re ERE RAAT EL (EE TARA RE S CER, PA EMBER BS, SOL, FRR EK ALAR IA AO TCE A hid, PP ETEMARBAW AM. MAM OS “HAR” ERA BAEK FRESE IE A VR, Ae ake eR TE & RAR SEW WEE a, AT SARA RASH "MRR". BABA AAG RNR, PREAH R Ee FEHR ROMA ARES, GEES He -MSLM®, He ANE RMA TAA, iL PE RIE THRE. “78” (Government and Binding) 23% BE ABR BRE AR ROR, SHS IS SB oy AS SAB So Sh A OR SO op — aE PYAR IE FEBLTG . Chomsley( 1992 ) 48 tH “I 24 i", Hab ae ET Sha Chomsley 99°18) 2576" 15 98 EE“ 4938 LES “RB” a BS, BRAG" AM ER ET ER, BOS AIO FREAD BIZ BR BRK. By FE T “at G8 DM” (Subjacency Principle), “23 70 0 OW” (the Empty Category Principle) § 2% 2 Hi i MRAM 20 EL EEL, ARGS LOE. LORE PART OX RR, Ss T PAE BUA «P 115“ BRIE SUH IR” (Generalized Transformation) 414] TREE, AS AGT” (inflectional morphology) 4 & THESE =" (derivational morphology), FLA — AY ORF BUM" Ceconomy principle) #& BT Ra FR 7 5 ER) FS OLB ak Me AF I EEMTA, 24ARTES — BT A Aah DE Ste SET IRAE RB TLE WA: (—-) ATRIA, AMEE SA RAR SSR MAME HED OE EMS, KT AA A ] & WL Chomsky (1980) ff] Rules and Representations. (New York: Columbia University Press), 2 T 1b 1 6K FAT AP A linguistic knowledge) 3X -AIB, BIRR ATARZ, (1 Chomsky RH Pik a AR AB AHADLRALRRAN, HEBCWE: BAMA RS BD POSER, FAW SOOM TARE EI Se A Re E PES AL RAS. ER A TB PA Be SE MRA EMZBA (BAPRA, 2001). (=) AMIDA RH AB BPUMRAN-- TORR, MARA EUR RAAB ES. (=) WEE A RIZE ATE Fy od BLE ETA S Initial State) , 3% BEAR SALA RIM G PERE, ARSE TT. A REI OS ER te, SE HEED AE ARB ES RN PP. A AS (Steady Stare), FE ART (RR ALE FY ADA J) (0) He SP MR (ARR AS” BE AGL 1” FE TTB Oe EAH SBE A o PTA SAR Ab,” Fa EE 8S APG Td OE BL FIA MARIA 223, WRIA AES AR. MIT AS NE GOK, Hey AAS RIE LA, WADA RULE ES AB SR SP DRE RT BEAR A EAR ARE "FE fh “AURA HW” (Phrase Structure), “i#) HE” (Lexicon) . “S15 wi me” (Phrasal Categories) #128478 HF” CLerminal Categories}. HEY BBI— ERAN ER PER, ROL W AIS BOSOM, GR RA, MURS SR PETG. BLE eB ay ie ME ae a FOR NLR FERC ME T BASSE (PS rales) ZEA op 2 OA AS TOME FE, SX ARTI” (Xbar Theory) AUDEN RU" (Projection Principle), SORT aA Be EAA, ON Ee Pe Raw) BAMA Ch 6.2 416.3). ABE EAI" A BRGY” (complement) #1 “REE IR GP” (adjunct) 19 Bl, “S549 4 ARE” (Structure Preserving Principle, 586.5) th f&—) 4 WY Fa Ro, EE A a Ah 2 PEN HEB CAN ah BE aR) ETB BR PK BO HH “XRF” AN BA DA” eA RL GY, EB a FE fg HSB, MRA G NE ROR SME, 6.6 The F25RO“ vP-448" CIR RG* VP-SbFC44 HY" CV P-shell), SASTIL SER FEHR AR HR FE SRT EA EAN OKAYS. RPAH RM WERE B.D 1s RK WG BR EE — 7c ASA” (Binary Branching), MAAS ALBWRDLRGAAD ARM AEAD ROAM, PERS PABA. RENTRHBRRALM SH. HE, (PR HHA FEE 4} 87K John put the book on the shelf 3 FRA4 META HH ER T TLR, TR pur 28" = TCA A RAIS | ARIE a vP W”, SRL GAT REGEN SAWKM PBA BHA, “ETB” (O-Theory) B*BMC" Ti PRE MIC (BTW). RERAUPILT: (2) ETL ANA ID AS MR, RR RH. el Sri) | VM, (AAA ah ABW i, MBE PAY seem BAKE, FA HE AUD THAAD WA LWA, HTLAZWRS TAMAR, ARRAS ARS AMAR T, RAAA RRR LN SSAA RA FER LX. LAM, Te rained heavily. 414 it ILA TE, AMRIT, I like it POS it BATE, WRU Te. PBR RR” Cocelect) EW RITE, HET URI AH (G-roles). — i TER ALTE AFR Ua LE — 4 8 Fe Pd dS TB BY OR. TH CR ARRA— +> OMG, Fn —+ 0 fie RR iO. RE EP SET" ORE”. ROR HE RE ER, CS ah) GEE MOTERL, BARE 9-98 ROARS) BEI OEE Bd BB. AE “RIE. A URE EO MUP ER: BAT RDI Ch CAI 1 oi OL, PR A ee ee RO, SY MET 0 E MB RA OARIC ALE” (8 marked position), maw MAE G OIC HE” bP RA BLAIS Oe, BS, OS FOR TUR Oe. UREA", BPO MEA 2, PNG IC Fi C8" FP i BN 90 TATA OH A |“ Shae A ee JAAS EL V aL SHIRA Spec of IP BE Spec of VP_L. LANIER: PRO BAP RAB CHE, 1A MIMD io, (2) FRSC HOE i AE) 9“ wh AL” BEB BH FA) OD EY OH Spee of CR” SSPE PIR AERY BL PGE UE” (trace), MAIER EA OR SR AGE — MiB ABE" wh ST “ARERR” (scope). ISBT HY“ wh. SA” MCB” 2 MRO REE (operator-variable) KK. ik PRE ARIS OK IF HE GE A 9" — BERL” (the first order predi- F26cate calculus), WTRSHART RAM HERKAMR, SARE BH RULFt (raise) FI IPR VP ORE, REM TR TBE" MLR Hh. RAP BARBS £—TUBG TRA” (head), APH OTERO” (tail) BS“ EE IGE” (chain), REAR OAR HE LSE. APE Hie” RBS AT 8 “#BWBIC” (Case Theory) a“ R216" 4+ ASb—-TREMBC OLS 8 FW). RNB, “RBI RPE TERY” (lexical NP) (O24 aS} AR SEIS, BRST Ze OE we oe A LT BE A ER BYRD” Bese. ST AO TDSC EES He a AB SWB HE IK. PRO MRICS AERO RA BER AER”. “HE BEG” Se Je BSB el ALTE AR Ya A TE A 9 AE RAILT ABE EHRAE, (—) KF" a AEB” (Visibility Hypothesis, JL 8.1.3) Pit TG EAR A Oia BT PH”, RA A ET J FERS LE bo BAP PLR SEER TERR” (Case Filter) MARL, HE AR His SUER” ASRS, wh STH Bh Aa nd TB ah Pa ABBE, PACES PERE, MENA MRA BRIE AE, BEAD EE IE — PB. PE A RTD MAK, FARE ARAB TR EY Ceo, ( FREER: “RVG Ade PF EO dea hE, ma BLE SH He LP RO RUR 1 (=) “HRP BEAL” Cc-command) . “SK RRL SEA" (m-command) , Al “Hs” (government) RRRERAKER PRN RE Re WRAR RAN TEAR ER, PRR HE ORR HO 7p ia AN A BY RAL RAT OTE. Mut, WAN AE RI OE RE FEMI. HP BAL WL F a oe oe OE ie. BRK ae FES AE LP — TERRIA AR MRR. UTR, “Re BL” ERE ALAA, RE A A EE A BE TRE SSE RHR A BAMEAIE” (Minimality Con- dition, JL191 WA, (3) HF *DpP-BUe” CDP Hypothesis). ARE 1S MIRE AY Sal TEAL ICTF NP, i NP HU Poti) (head) EN, Ape BBY T — ok TAB CHL 201-202 9), PA bet T — “DP”. HEAR AB, Bi HERERO Ravi Ade N, TEED. (se 8A DP AHL 203 HO), Ma F27(AES. DR IP PBS 1, REA A AE DP aha FY IP RAR, UAE RA A I GS aT. 2a ERA) LUSCHE SE DP, FT Lge SOP Se — Le LAE CED “fleb RD” Cexceptional Case Marking). DP-# 3th PT EM wR. “Zoi” (the Binding Theory) #1“ #24122" (the Control Theory) “MAO HAT AT Ee CLO), He Re EE MAA. (1) AASEE (anapbors) (SJE ARI, DPB); (2) FR fl (pronouns) HYALE(CIs (3) HERTE— KS (expressions) (EA THERE Xai AE tat) (4) PRO; (5) BFE HE—4 fH (parasitic gap) BRE ay es A + Be — (269 Canaphoric J M1 pronominal 3x BARRE. 1 Hi + anaphorie, — pronominal], ( — anaphoric, + pronominal}, [ + anaphoric, + pronominal] fi[ — anaphoric, — pronominal] OAD, 43) SY RUSEIA” “PU”. PRO WER FET PERE “AURIE" HST RPT AL. IE EA HEAD OTR Tt iB THLE TE USGS SPE ae (Governing Caregory) AY X, FEAL HE” (SUBJECT) Rh HE Bee SRNR REE TR (ML 289-290), FI“20 RR" MRO RIE Hel AAR” CNIC, 223 WH), AUPE EGBA” (SSC, 235 FL), ORR 48" (Crossover, JL 241 TC) ALA AUR OTST 8 TE UE 47 SCT tat RUA AMR. “PSHE” (Control Theory) 4 PRO A) DAH PEALE. PRO “ARG ay RETR GC + anaphoric AY RHE, FMA A + pronomi- nal AYRE. FAG" + anaphoric Ht ARTE, DALI HE BA“ 29 ea” BOE aE PRO Weve" HP RAR” , ALPE BA + pronominal] 4958 PE, RAR CES ae AP Al edt (BEAR BLA)”, SEAR BL LT —— ED EDT Hae be A RG, RO eR, I Fe, AAP — 8 A FE, BUTT A RAT ge. BH PRO WYER”, DALE ts BL AN A “He PE PRP HER, BUA BP PRO Theorem 34 HABE “Bat” HAUSE Bee AR Oe BT OE HH ty ak Rs RE Ra BF, WAN “FSI” (raising verd) ARN BORA IK, “pat OD APARNA, CASA FORMA EM. 28AY ABA ABAR BOAR, “whl” A “wha” ATK, fe “IE” PRT MUS AHR A A PPA ” He TLL ALM — AEM, JE “Boo” KA RH. AAR. B10 WHEW RE “Bao” NAL. PLAS AO PPA GE IAL A SE LE, OTA HREM. APSO “IE”. ERE ee AR HES IE BH RE IB i OY TAR AK. AWHIG WRT HE AE (1) SPR IG” (the Bounding Theory) # “at SBM” (the Suhjacency Principle), (2) “4398 MRM" (the Empty Category Principle), (3) “i8BHI8” (the Barriers), (4) “isi DAE” (Head Movement Constraint). JL + Beh & Bl AURA A “ARH EE” BR AAHDAG RENI” — CIsland Conditions 2% Island El fect), “ERA PMR” (Subject-object asymmetry), “HMRZE” (in termediate traces} Al “{LH#BUH" (superiority effeccs), BLM B) Rizzi 2A “MERRIE” (the Relativized Minimality) #1 Anou (1985, 1986) BEMGAY “St—S YL MEBLIE” (the Generalized Binding Theory). “FERRE” AL “ASS RUI” FORLGAT, SC BEAP AI. ES PL ER AURIRIG RECREATE”, ea SERRE AGG" A “MEN”, CARAT AE A CoE ERR HAMM ARA, EPA “ERAS. i BRC” #H Rizzi AY “ERP ATHE” LAB Anoun iy “#R--2R 8" TEAR EE ate FERIA MER NDE AY. (2 SRR” PR ABE e “4A EPA” (proper govern. ment), BET AER — TBS. HATES 264-270 RL OD SA? WALE, Satake “SMe”. EAD REL 267 IR. RSARS ee FRMT, — TR o AY RAR PR % Bs (1) BB, Both; CS BAR” BL ANAS AE FRE RATE SPLOT AE “and only i", SE Lae F “it and only it”, KB FE", F29K (2) 0 Hi PHF Bo HER, APTA “RM” MRR, HES” WHR. AHBINK, ROAM A OS" RIN “NRE”. BATT ARR —FLRAS BTA RICA RTE: Bye TE IE HPN AMARE, REM ARESRRMALR LLM. £ RFA EAE BAK AS ARREARS, AEE iC Mt OR AE, BARUVASH RAM MEH. HR, Rh ER KE T-TEE HAM “He aR". Mt, ta (5 267) —FHEM: AAA (1) a MiB BER; A (2) a RSPB B A (3) HoMBZMBAAE, a TOPE Be ES, RETARAZMR “S" WRK, HRPM. BRKT MELVRAl “mash” MTR” a. MA RE WW Hh Fatt, “WAKA” Jb Chomsky (1986) #E{eMS—AZ Oy GEL BY BPRUE HOS. URETHRA, A 10.4.1 PERT ILREAR, E Oh. ART, FR RB CU TBR”) BORE EE MOY BS CN RH"), (ARS EA, A Ke Be Med. OF EE” te a LO RL, A “HS RRB BCR L A HE, A, 10.3.3 Pe RA Bo FEAT AY “TE” RD I EF" (O-govern) HELA (OL 271). F30284 “HBR” AT RO, MP RST — Pie ER A” PR iC” | -marking) CHL 285 11). BTM RRR BH. BRA Csuper-raising) Atta aH Mt — SBR PUL iB ah RA” WIRE, Rial HS te $e", RMT RE ROAD + SD ARTE AR PE” RL 287 SO, ERAT A “ASS” RAT “a a ee at SVE RAR” CBN 86 FRA iv), it) “URE ACE” “OH Re BT RE” Sys AE 84 #085 Pea # VK WATERMAN CREE, DHE, RBH, MESES, RRR. EH TUS DERE RBA ES PUTIN TRIN USA RAM, APRA, ae WARS SAKE, REMIT HK, ~- ea. BARABOO HARES LETRA, RBS, ERMC RA AICAS EASE SHE 0 BIBS: BAW BH “EQiIG” RAE F AMIE SRM Re. KR SS Hise MA PRU", et mA BRON Se GEAR), SRA RAB RARE. bb in Bi: CD) TRAN a AL oP af CER DER) BB, (2) FB AN BRAS CAA Oe “a BRI" FER” HBR, 3) THRASH wh BBE EM EEE MHS LH 11). “2si£70" (null arguments) WEAREAT (licensing) WA AT AREAS, 3k Be “SIE IL” HE pro M “28H” (null operator). WAR T RE PAT Real ss “ER” MSR, WE E+ pro, — pro] A, AR PORT Ra “WI” Ccontext-oriented) SBE “SEVER A" BS (R24) HSK, “ahialig—" (Verb Second) UAE BRM IY. HR F3La, TA “BSI” PULP CEE “MHRA ATE”) HARMS bHS, TH VSORSERE HAGA” RAHEK (UB 13-7). Baker AM “BhéiG” (incorporation theory) TAOe “Riker” Ay as BE ALEOA “AR RA, AP OMA. PAS AS. ERB RS REE. HORNA A “AR ah ile Ae” (eausutivization) (WG 14 HH). WAG SS BALES is i AE EA. (PIR PE DT MeO PA GL DE TOA IS oS PEAS TAL PAIL, Uy HE RE te LAP AMEN MEE RE ROA, HR RRA RE POMGAL, MES RAM RR, BRT ASR. BM AEE aR FT OE oh AS A TE I OS ae ARAN AAR AR oP a Se 5 aT AMO “8B — AE” (che second position) bs BEWARE RR HE SSID bs AIR aL PAB AES MELE TR Agr, MA RBM EMER BL ONS 1S FT). “AGG” TARE RAR, OE” HET OAT ANE” WE RP OR IS PN eA. SALT A LU Agr #1 Neg BETH REME FR AS OFAC CUR 16H). ERE BLAM bo Wk" Csurong) Be “BI” (weak) MAMA, HSE “A257” BAER CORSE IT). AVE” LPB) “GwiE” 2 RAT RAMEE. HMI” FO Ds, Ss ERA, 2 “AD Ae” AE PF OM LE RT RIE AT. AVRO EST WRAP RRR. TERRE, “Me” AAT Re at ey “RR HERE” (Generalized Transformation), 31a) To -7E “TRE” He EE Oy BR EK OR —— 5" (Merge), ~ HARE BH 0” FPG RAE “ETRE AE (the Minimai Link Condi- tion), RATA AMS: “MRO” Ble eae” cee Sy" Fl RTA SNHEIL, COLB 18 4). “BEEIO” (the Cheeking Theo- ty) “RAE” BO REAR CUB 19 Hr), BT ABTA MAUS, AT-M ALTLAT . CBE REALE” (che Linear Cor- respondence Axiom. Sil 454-459 FL) 2-H RAW Aa MID, Y F32AERA, MRAP HIER. At 1999 RUE, 2 “RM AIG” HER PHRERBAR ER RR. HMTRRRAR, LETS PE scm: Chomsky, 1999. Derivation by phase. MIT Occasional Papers in Linguis- tics 18, Cambridge, Mass: MITWPL. Epstein, Samuel David, and Norbert Hornsicin. 1999. Introduction. In Working minimalism , eds. Samucl David Epstein and Norbert Horn- stein, ix-xviii. Cambridge, Mass; The MIT Press. Lasnik, Howard. 1999. Minimalist anatysis. Oxtord Blackwell THA, 2001, “MRA ANM TE PORES BEM RICE TY HIF’, ONES SMBS) 2001 4, 6, 2.5 Ff, F33Acknowledgements J owe a debt of gratitude to many individuals who have contributed directly or indirectly to this textbook. Most of these individuals are students who attended my classes at University College London, Queen Mary and Westfield College (QMW- London University), and, for one quarter, at the University of California at Los Angeles (UCLA). Unfortunately, the list is too long to reproduce here. Iam also indebred to Bettina Knipschild, Lluis Cabre and my daughter Sima for their help and patience. F341. Contents Preface by Halliday ERR Preface by Chomsky TRAE aE Acknowledgements Language and Linguistic Theory PART I. Phrase Structure, Lexicon and Transformations anon . Phrase Structure . Lexicon |. Transformations I: Phrasal Categories ‘Transformations II; Terminal Categories PART H. Principles and Parameters 6. 7. 8. 9 10. X-Bar Theory @-Theory Case Theory Binding Theory and Gontrot Movement Theory PART ill. Language Variation 1. mR 13. 14, AS. 16. Head-Complement Order, Bounding Nodes, and Wh-in-sieu Null Subjects and Objects Verb Second, VSO and NSO Incorporation Phenomena Clities and Cliticisation Llowering versus V-raising F10 F13 F20 F23 F34 7 7 43 61 92 109 109 183 257 27 311 326 361 384Contents PART IV. Minimalism 17. 18. 19, 20. Minimalist Program Copy Theory of Movement, and Binding Checking Theory and Language Variation Bare Phrase Structure and Antisymmetry Bibliography Subject Index Index of Languages and Phenomena SERAI F8 403 403 416 433 447 477 487 4891 Language and Linguistic Theory Contents 1.1 Language and mind 1.1.1 Knowledge of language 1.1.2 Language and other faculties 1.1.3 Grammar and Universal Grammar 1.2 Language and the linguist 1.2.1 The cask of the linguist 1,2,2 Some necessary idealisations 1.2.3 Speakers’ judgements 1.3 A brief historical overview 1.4 Abou: this book Sources and further reading 1.1 Language and mind 11.1 Knowledge of language Linguistics is usually defined as the discipline which concerns itself with the study of language, although what language is taken to be may differ from onc school to another. In the Generative tradition, language is understood to refer to the know- lecige that native speakers have which, together with other faculties of the mind, enables them to communicate, express their thoughts and perform various other funcuons. Accordingly, the task of the linguist is to characterise in one form ar another the knowledge thar native speakers have of their language. Let us cake Chris to be a native speaker of English and review certain ‘things’ that Chris knows by virtue of being a native speaker of English. Obviously, Chris knows the words of the English language, che way they are pronounced, what they refer to, among other things. The word pin, for example, is pronounced in a particular way, and refers to a particular entity m the world, Words are basically sound symbols which bear an arbitrary relationship to what- ever they refer to. As a matter of fact, Chris's knowledge of che words of the English language involves more subtle information, For example, the word pin is made up of discrete sound units, so that the initial sound ‘p/ can be roplaced with the minimally different sound /b/ to obtain the different word bin. Chris also knows that the word unhappy, for example, is made up of two smaller meaningful units, the negative prefix w- with the meaning ‘nor’ and the adjective happy. Put together, they convey che meaning ‘not happy’, The negative prefix can2 Language and {| inguistic Theory be attached to various other adjectives such as kind and sympathetic to derive the complex negative adjectives wakind with the meaning ‘not kind’, and unsymtpa- thetic with the mcaning ‘not sympathetic’. Interestingly, Chris knows that the negative prefix wn- 36 different in meaning from the homophonous prefix found with complex verbs such as unpack, unbutton and unzip. The complex verb unpack, for example, does not mean ‘not pack’, but means roughly ‘reverse the action of packing’. Somehaw Chris knows thar the negative prefix can only attach to adjectives. When Chris comes across a verb which includes the prefix uns, Chris interprees the prefix to mean ‘reverse the action” and not to mean ‘not’. In addition to knowledge of words, Chris knows that words can be arranged together to form meaningful sentences, and that different arrangements (word orders) give rise to different meanings. Consider the examples in (5): The the likes boy girl. The boy likes the girl. The girl tikes the boy. The gitl is liked by the boy. Boy girl likes the the. yeene If the words in these sentences ate arranged as in (1) and (5) they do not make a meaningful sentence, bur if arranged as 1m (2), (3) and (4) they make meanmgful sentences, (2) means that the boy experiences a certain feeling towards ihe girl, and that this feeling is not necessarily shared by the girl. {3} has the (opposite meaning that the girl experiences a certain feeling towards the boy and chat this feeling is nut necessarily shared by the oy Finally; (4) has a meaning which is similar to that of (2) rather than that of (3) even though it has a word order which is similar to that of (3) rather than that of (2). As with knowledge of words, Chris's knowledge of sentences involves some quite subsle information. Consider now the paits of sentences in {6a8cb] éa. The girl asked the boy to leave. 6b. The girl asked the boy to be allowed to leave. (6a) can have the meaning whereby the boy is supposed to leave (‘the girl ordered the boy to leave’) ot the (more subtle) meaning whereby the girl is supposed to leave (‘the girl asked permission from the boy to leave’). However, if zo leave is replaced with to be allowed to leave, as in (6b), only the meaning whereby the vir! is supposed to leave is accessible. Likewise, (7a) has the meaning whereby the hoy is supposed to do the talking to somebody else. However, if the girt is dropped from the sentence, as in {7h}, she bay changes fom being the individual who is supposed to do the talking [to somebody else) to becoming the individual who is supposed to be talked to (by somebody else) 7a, The boy is too stubborn to tatk to the girl, 7. The boy is too stubborn to talk to.Language and Linguistic Theory 3 In addition to what has been mentioned, Chris knows that sentences can convey different messages in different contexts. Consider the brief dialogue in (8a8cb) from Sperber and Wilson (1986): 8a. The boy: Do you want some coffee? 8b. The gicl: Coffee keeps me awake. The girb’s answet to the boy’s offer can mean “yes, please’ (acceptance) or ‘no, thank you’ (decline), depending on the ger/’s intentions and plans, and the ume in which the dialogue takes place. If the girl intends vo stay up late in the night, (Bb) is likely to convey an acceptance of the offer. However, if the girl intends to have a good night's sleep co be able to wake up carly the nexe morning, (8b) 1s likely to convey a decline of the offer. Presumably, these two possibilities presuppose thar the dialogue in (Ba6cb) rakes place some time in the evening, If ic 15 assumed to take place eacly in the morning a different message might be deduced from the girl's answer. The ability to snfer the right message from the the girl's answer depends on knowledge of concexrual information relating to the time in which the dialogue takes place, the intentions and plans of the girl, as well as encyclopaedic knowledge relating to the fact that coffee contains a substance which can cause one 10 be awake, There is a sense in which this kind of knowledge is nor of the same order as the formal knowiedge of how to derive complex words from simpler units and how to arrange words together to form meaningful senccnces, For example, it is possible that a native speaker of English will fail to infer the right message from the girl's answer in (8b) (misunderstanding) for lack of the right background information. However, it 1s unlikely that the native spcaker will fail to realise thar keeps consists of the verh keep and the third person singular marker -s. We can make a distinction between Chris’s knowledge of the English language and Chris's abiliry to use this knowledge properly in differcnt situations. The former includes knowledge of words and the rules which govern pronunciation, word formation, sentence formation, among other aspects of language. The latter, on the other hand, includes knowledge of language as defined, in addition to knowledge relating ro people’s beliefs, the rules chat govern social behaviour, encyclopaedic knowledge, as well as cules of inference which enable people to interprer ucterances in relation to a given context. We now have narrowed down she expression ‘knowlaige of language’ to mean knowledge of the rules which govern pronunciation, word formation and sentence formation, in addition ro knowledge of words. Ik is this specific definition of (knowledge of) language we are Interested in here and which we will assume in the rest of this book. 1.1.2 Language and other faculties The Chris we have been assuming so far could be a normal person with averaze intelligence. Now, Jane may be more skilful than Chris at knowing how to manipulate language to persuade, give good speeches, write detective stories, and so on. However, there is no sense at all in which Jane could be said to be more4 Language and Linguistic Theory skilful than Chris at knowing, for example, chat un (Za) above the boy is the individual who is supposed to do the talking to somebody else, whereas in (7b) the boy is the individual who is supposed to be talked te by somebady else. This kind of knowledge is common to all normal speakers of English irrespective of their other abilities and skills. Now take Chris to be an adult with the mental age of a young child. Chris may have a few problems grasping the rules which govern social behaviour, and may also have difficulties with problem-solving casks which are otherwise not supposed to be taxing for people at a similar bialogical age. However, it is quite possible that Chris knows, much as (ckilful) Jane does, that sentence (4) above has a meaning which is similar to that of (2) rather than that of (3), even though (4) has a word order which is similar to that of (3) rather than that of (2). Smith and Tsimpli {1991) have reported the case of a ‘29-year-old man ... whose non-verbal 1Q averages between 60 and 70, who is institutionalised because he is unable to look after himself? (pp. 316-17). However, this man has a normal mastery of his native language English, and, more spectacularly, ‘when given a passage writzen in any of some 15 of 16 languages simply translates it into English at about the speed one would normally read aloud a piece written in English’ (p. 317). Other cases have been reported in the literacure of people of varying ages who display a sharp discrepancy between their general cognitive abilities, including communicative skills, and their linguistic abilities or knowledge of their language (see e.g. Curtiss 1981, Yamada 1990), Finally, take Chris to be somebody who, as a result of an accident or a stroke, has received physical damage to certain arcas of the brain, and consequently is suffer- ing fom whar is clinically known as Agrammatisin in Broca’s Aphasia. Cheig may not have lost the ability to pick up hints from contexts to interpret sentences (in face, this patticular ability may become substantially cnbanced), and may not have ost the ability to solve taxing problems, and so on. However, it is quite possible that Chris may have lost the basic ability to interpret senrence (4) as having a meaning similar to that of (2) rather than chat of (3). Numerous studies have reported cases of aphasic patients who have difficulties interpreting sentences of the type in (4), though not necessarily sentences of the type in (2) and (3) (Caramarza and Zurif 1976, Geodzinsky 1990). Interestingly, the difficulties mostly arise in relation to an identifiable sub-class of these sentences. Compare the sentences in (9a8¢b): 9a. The boy is kicked by the gitl 9b. The door is kicked by the girl. {9a} is called a ‘reversible passive’, where either of the two individuals involved can in principle perform the action described by che verh (the kicking). The Positions of the boy and the girl can be reversed to detive the equally plausible sentence The girl is kicked by the boy (9b), however, is a ‘non-reversible passive’, where only the girt can perform the action described by the verb. Reversing tie positions of the door and the girl results in the infelicitous sentences The girl is Ricked by the door. The patients reported in the studies cited are likely toLanguage and Linguistic Theory misinterpret (9a) as meaning ‘the boy kicks che girl’, but are nat likely to mis- interpret (9b} as meaning ‘the door kicks the girl’, Presumably, this is because their knowledge of the world and their ability to reason (intact) enables them to exclude the infelicrtous interpretation af (9b) whereby ‘the door kicks the girl. ‘The discussion so far should lead to the conclusion that knowledge of language is probably independent of che other facultics of the mind. It is independent of intelligence, can remain intact when other faculties are impaired, and can itself be impaired when other faculties are intact. The human mind is said to have a modular structure, where each faculty has an autonomous existence from the others, although the ability of humans to use their language normally involves an interaction between all of these autonomous modules, Because knowledge of language forms an autonamous module, it should be possibie to study it separately from tbe other faculties of the mind, We will come hack to this point later on in this chapter. 1.1.3 Grammar and Universal Grammar We have defined knowledge of language as knowledge of words and knowledge of rules which govern pronunciation, word formation and sentence formation. Let us now rry to classify this knowledge into identifiably distinct categories. Knowledge of words can be characterised in terms of an open-ended mental dictionary, technically called the lexicon. Like a commercial dictionary, the lexicon can be considered to consist of lexical entries for words, where each lexical entry specifics various types of information necessary for the proper usc of the word. Knowledge of rules can be characterised in cerms of the notion ‘grammar’, We can say that Chris knows the grammar of the English language to mean that Chris knows che rules of the English language The term ‘grammar’ is understood here to refer to the rules which govern pronunciation, word formation and sentence formation. The sub-component of grammar which includes the rules which govern pronunciation is called phonology The sub-component which includes the rules which govern word formation is called morphology. Finaily, the sub-component which includes the rules which govern sentence formation is called syntax. These terms are also uscd by linguists to refer to the sub-disciplines of linguistics which deal with each set of rules (more on this larer on). Here, chese terms are understood to refer to aspects of Chris’s knowledge of English and therefore are components of Chris's mind. Accordingly, Chris’s knowledge of English consists of an amalgam of more specialised cypes of know- ledge together wich the English lexicon Restricting our actention to grammar, let us now ask the following question: How did Chris come to have chis intricare and highly specialised system of rules which we call ‘che grammar of English’? It ts unlikely thar Chris was taught this knowledge, Cheis may have been mstructed at an early age to say brought instead of bringed and mice instead of mouses, and at a later age t iy co speak “proper English’ instead of ‘teenage gibberish’. However, it is highly unlikely that Chris was taught that if one6 Language and Linguistic Theory substitutes to be allowed to leave for to leave in (6), the meaning whereby the boy is supposed to leave disappears. This type of knowledge is subconscious, in the sense that although native speakers possess it and use it, they do not have direct access to it and therefore cannot teach it. It is also highly unlikely, in fact practically impossible, thac Chris came to know English by memorising all the sentences that exist in the English language. This is because the number of such sentences is infinite. An important property of human language is that a substantial number of sentences produced by speakers are novel, uttered for the first time. Human language is said to be creative, insofar as there is no limit to the number of novel sentences thar can be produced by native speakers. The creative aspect of human language provides the strongest evidence that know- ledge of language is essentially knowledge of rules, a computational system which makes ir possible to generate an infinite number of sentences from a fine number of rules together with the lexicon. A somewhat more plausible answer to the question raised above is that Chris came to know English by observing others speak it, deriving, che rules from their speech, and then internalising those cules, all at a subconscious level. Granting this, one cannot help the feeling thar there is something miraculous about this achievement. The examples discussed above give only a glimpse of the highly complex nature of human language, a fact which 1s all too clear to linguists, though not necessarily to people who do not undertake the task of analysing Janguages mm scarch of rules and generalisations. Yet, Chris managed to master English at an age when certain much simpler rasks are beyond the reach of children. Asa matter of fact, the complexity of buman languages is such that Icarning them from scratch is beyond the reach of any living organism which does nor have some kind of special predisposition, an innate ability of some sort. Let us see whar the nature of this predisposition or innate ability is likely to be. We have seen that knowledge of language is independent of the other faculties of the mind. Therefore, it is onlikely shat chis predisposition is common to all faculties of the human mind/brain, Uf it were, we would nor expect the kind of selective impairment of faculties reported to exist. Rather, the predisposition in question must be a specialised one, specific to language. It 1s plausible to reason that to determine the nature of the predisposition in question we must look at the nature of language itself. The idea behind this reasoning is thar the nature of the predisposition to develop language must somehow be reflected in the properties of language itself. We have seen that language hasically consists of cules of various types which in combinanon with the lexicon make it possible for native speakers to produce an infinite number of sentences. On this basis, we can conclude, as a working hypothesis, thar the innate predisposition to master language basically consists of a set of rules, i.e. a grammar. Once we accept this conclusion, Chris's achievement in mastering the English language becomes amenable to 2 more rational explanation. Chris approached the task of mastering English already equipped with a rich system of rules. The learning process amounts to the comparatively more manageable task of learning certain aspects of the English lexicon and certain cules specific to the EnglishLanguage and Linguistic Theory 7 language. Much of what remains of what we have been calling Chris's ‘knowledge cof English’ was there right from the beginning, and could not possibly be said to have been learned in the way one learns how to drive or play chess, for example. The grammar which characterises the innate predisposition to learn language is called Universal Grammar (UG), where the term ‘universal’ is understood in terms of biological necessity, This is 10 say char UG is the set of rules thas all humans possess by virtue of having certain common genetic features which distinguish them feom other organisms, Consequently, UG rules ase to be found in English, Berber, and indeed any human language, and form part of the knowledge that native speakers have of their own language. Thus, Chris's knowledge of English consists of the cules of UG, certain rules specific to the English Janguage, in addition to the English lexicon, Likewise, Idir’s knowledge of Berber consists of the cules of UG, certain rules specific to the Berber language, in addivion co the Berber lexicon. 1.2 Language and the linguist 1.2.4 The task of the linguist Assuming that there are good seasons ro study human language, not least the prospect of learning something about the distinctive properties of the human mind, the task of the linguist can be described as an attempt to characterise in formal terms the knowledge that humans have of their language. The task is essentially one of reconstruction, in the sense that the linguist tries to reconstruct, via the Process of analysing data, knowledge that exists in the mind of native speakers, In other words, the task of the linguist is to formulare a theory, sumetimes called a model, of language, insofar as theories of natural phenomena in gencral are attempts at reconstructing the mechanisms underlying those phenumena. Needless to say, such theories and models are meant as approximations of reality supported by the evidence available, rather than exact replicas. It is quize possible that the theories (approximations of knowledge of language) produced by the linguist may at some stage be wide of the mark. However, this isan the nature of scientific inquiry in general. The process of building a theory of language (or any other natural phenomenon, for that matter} consists of attempts to accommodate as many dara as possible, including new data. In view of this, it is also possible thar a revised version nught reintroduce an idea that was previously rejected on the gtound thar there was not enough evidence to support it at the time. For exampie, new evidence might support an idea for which there was litle evidence ac an carlicr stage in investigation, thereby justifying its resurrection. We have characterised knowledge of language as involving, in addition co rhe lexicon, knowledge of universal and language-specific rales, ic. knowledge of a grammar. It follows that a given theory of Language 1s itself a grammar, snsofes as it incorporates the rutes which govern the various aspects of language. It is for chis reason that the term ‘grammar’ is sometimes said to be used hy (some) linguists8 Language and Linguistec Theory with ‘systematic ambiguity’. It 1s used to refer to the knowledge that native speakers have as a component of their mind/brain, as well as to the theory constructed by the linguist as an approximation of chat knowledge. The grammar constructed by the linguist should be able co distinguish the language-specific rales from the cules of UG, so that it is possible to have theories of particular languages and a theory of language or universal grammar. Obviously, sentences of particular languages do not come wearing on their slecves the rules involved in generating them. These cules can only be attived at via rigorous investigation. Moreover, once these rules have been identified they are not likely to be wearing labels which classify chem as either universal or language-specific. The cask of classifying a gwen rule as a rule of UG must take the form of a hypothesis to be tested against data from the same language, and ultimately against data from other languages. A grammar which correctly describes a native speaker’s knowledge of her!lis language (ie. a theory of a particular language) is said to mect the condition of descriptive adequacy. On the other hand, a grammar which correctly describes UG (ie. a theory of UG), where UG is understoad to be the set of rules which define human Languages, is said to meet the condition of explanatory adequacy. The goal of achieving explanatory adequacy is very much tied to the question of how native speakers acquire sheir language. Recall that we came to the conclusion that Chas must have been born with a predisposition to learn language on the basis of the fact that the knowledge Chris has is highly complex. A theory which describes accurately the predisposition in question (i.e. UG), and explains how the know- ledge thar Chris currently has follows from this predisposition can be said to meet the condition of explanatory adequacy. A distant goal is to construct a theory of language use, thar is a theory which will characterise how knowledge of language interacts with other components of the human cognitive system in performing various functions. This theory entatls individual theories which characterise the properties of the various interacting systems, including a theory of language. Because very little is known about the properties of some of these systems, the prospect of arriving at a theory of language use seems less realistic, for the rime being, 1.2.2 Some necessary idealisations We have seen that the ability co use language properly, for communication and other purposes, is the result of an interaction becween different faculties of the mind, including the language faculty. Consequently, the speech produced by native speakers is likely to contain a jumble of infurmation not all of which reflects their knowledge of language. Now, a linguist who is interested in studying native speakers’ knowledge of their language obviously has a dilemma. The dilemma is basically to ensure char conclusions reached on the basis of an analysis of native speakers’ speech reflects accurately and solely knowiedge of their language. In view of this, the linguist has to take certain steps to avoid drawing misguided conclusions on the basis of tainted speech. One such step, mentioned above, is toLanguage and Linguistic Theory 9 make a distinction between knowledge of language, that is knowledge of rules, and the use of language in particular situations which involves ather faculties of the mind, This is the distinction between what Chomsky (Aspects of the Theory of Syntax, 1965) calls competence (knowledge of language) and performance (the actual use of language in concrete situations). Having made the distinction, the linguist can then be in a position to work out which properties of speech should be attributed to competence and which should be attributed to performance, by no means a straightforward task. Chomsky explains that speech is usually affected by performance factars such as false starts, hesitation, memory lapses, and so on, all of which are extraneous to language itsclf, and therefore should be purged from the data to he analysed. Tn other words, the linguist should deal with an idealised form of speech hypo- thetically produced by an ‘ideal speaker-listener’, not affected by the performance factors mentioned. This is a necessary idealisation, designed to isolate for investi- gation only those aspects of speech which reflect the properties of language. Another idealisation relates to the notion thomogencous speech community’, that is a community where there are no individual or other types of variation. This idealisation is also a necessary procedural step designed to cnsurc isolation of properties which are common to all speakers of a given language. As a matter of fact, this abstraction away from individual or larger differences is routinely assumed in everyday life situations when people use the expression ‘the English language’, for example, There is no ‘pure instantiation’ of the English language in the outside world. Rather, English exists in the form of a collection of dialects (American English, Australian English, British English, Indian English, ro mention just a few of the larger categories) which are collectively referred to as the English language. Just as English, ot any other language for that matter, has no ‘pure instantia- tion’, UG also has no ‘pure instantiation’. UG is instantiated as a component of larger systems of rules which include language-specific cules. In other words, UG is instantiated as part of Enghsh, Berbcr, Japanese, exc. Thus, to be able to isolate the properties of UG when analysing data from individual languages (the only roure), the linguist has to abstract away from the properties of those individual languages, Like the previous idealisations, this one 1s also a vecessary procedural move, designed to isolate for investigation a specific component of language. Ic is interesting to note that the data the linguist initially bas to deal with, which as we have seen is generally ‘degraded’ in nature, is precisely the kind of data that the child is faced with when learning a language. Yer, an the basis of this "poor evidence’ the child succeeds in developing a rich and highly complex system of knowledge, This is one of the major arguments, usvally known as the “poverty of stimulus’ argument, for the innateness hypothesis, Only if humans are assumed to be genetically predisposed in terms of a rich system of knowledge can this other- wise impoxsible achievement be rationally explained. Presumably, in learning a language the child undertakes a series of idealisations, abstractions away from the non-pertinent properties of speech similar in principle to the idealisations che linguist has to undertake. The difference, of course, is that in doing so the child10 Langnage and Lingsustic Theory is guided by UG (the innate predisposition), whereas for the lingusst UG forms part of the object of investigation. 1.2.3 Speakers’ judgements Since the speech produced in normal conversations tends ro be distorted by extraneous factors, it may include utzerances which, under different circumstances, would be rejected by the same speaker, Dats collected by observation of recording (sometimes called a corpus) will not be a reliable basis on which to draw conclu- sions about language, especially with respect co the subtle aspects of language. Thus, in addition to the nevessary abstractions which need to be undertaken, rhe linguist has to cesort to other means to ensure the database is reliable. One such other means is called native speakers” judgements, that is the intuitions native speakers have about their language. Consider the sentences in (0a—<) and (1a ¢): 1a. J think (chat) Jobn fixed the car with a crowbar. 10b. What do you think John fixed with a crowbar? 40c. How do you think John fixed tbe car? Jia, | wonder whether John fixed the car with a crowbar. ib, ?Whar do you wonder whether John fixed with a crowbar? lle. "Hew do you wonder whether John fixed the car? Native speakers of English are likely to judge the questions in (105) and (10c) as ‘good’ or acceptable. They are also likely to judge (11b) as being slightly deviant The notion ‘slightly deviant’ is conventionally indicated by one or more question marks at the beginning of che sentence, (11b) is slightly deviant in that it is not as “good! as (10b}, but at the same time not as ‘had’ as (11c), with fox! understood to modify the verb fix. (Ile) is not acceptable, a property which is conventionally marked with an asterisk. These subtle judgements surely reveal crucial information about mative speakers’ knowledge of English, and therefore must be taken inco consideration by che lingust. Although these ‘grammaticality judgements’ may in certain cases be affected by certain irrelevant Factors, there is a sense im which they are a reliable source of data. It is possible co think of ‘grammaticality judgements’ as tbe result of ‘litule experiments’, whereby native speakers subject che sentences they are presented with to the test of whether they are generated by their mental grammar. A sentence that is judged as grammatical is a sencence chat 1s generated by the native speaker’s mental ycammar, and a sentence that is judged as ungrammatical is a sentence that is not generated by the native speaker's mental grammar, Now, since che grammar constructed by the linguist (che lingwist’s grammar) is intended to be a theory of the mental grammar, we should expect the lingust’s grammar not to generate a sentence that is not generated by the mental grammar. ‘The expression ‘grammaticality judgement’ is also systematically ambiguous, ansofar as it relates to the linguist’s grammar and ca che mental grammar, If theLanguage and Linguistic Theory LL expected parallelism between the linguist’s grammar and the mental geammar does not hold with respect to a given sentence, the linguist will have to revise the model grammar to accommodate that particular sentence. A linguist who is working on a language of which she/he is a native speaker can rely on her/his own judgements of sentences. In this case, the linguist is said to engage in the process of gathering data by ‘introspection’. Although in principle introspection should be sufficient, it is sometimes useful and instructive to compare ‘one's own judgements to those of other native speakers of the same language. However, by and large this would only be necessary in situations involving so- called borderline cases as opposed to clear-cut cases. For example, it is possible that some native speakers of English would find (11b) ‘good’ on a par with its counterpart in (0b). However, this is less likely to be the case with {[1c). (11b) is a borderline case, but (11c] is a clear-cut case of an unacceptable sentence. Borderline cases and clear-cut cases generally tend to clicit the same type of reaction across speakers: hesitarion and certainty. To the extent that this is generally true for certain sentences, it implies a partern, and therefore a piece of data which should be taken into consideration by the linguist. This is co say that we should expect an adequate theory of language (che lmguist’s grammar) co be able to disongnish between borderline cases and clear-cut cases. In view of this, what might otherwise look like ‘conflicts of judgements’ or “disagreements on data’ are in actual fact no more than a reflection of the borderline stars of the scatences in question assigned to them by the (mental) grammar. Needless co say that clear-cut cascs are not expected to, and usually do not, give rise to “confliers of judgements? or ‘disagreements on data’, 1.3 A brief historical overview Transformational Grammar is a version of a lacger set of different versions of Generative Grammar. Generative Grammar developed in the 1950s in che context of what came to be known as tthe cognitive revolution’, which marked a shift t0 focusing on the mental processes underlying human behaviour from a more concern with human behaviour for its own sake. As far as language 1s concerned, it marked a shift from a concern with the mechanies of certain limited aspects of language (mostly, morphophonemies) to a concern wich the mental processes underlying a broader range of the properties of language. This change led to the articulation of certain ideas about the mental processes underlying language, some of which have been mentioned in the previous sections. Here we will limit ourselves to a brief and broad description of the evolution of some of the major ideas which have influenced the development of Transformational Grammar. Inevitably, some of the specialised terminology will not be transparent wo the uninitiated reader, but, hopefully, will become sa in the course of reading this book. Initially, grammar was considered to consist of a set of Phrase Structure (PS) roles which generate Phrase Markers called Deep Structures (DS), and a set of transformational rules which perform various types of operations on these Phrase12 Language and Linguistic Theory Markers to derive appropriately modified Phrase Markers called Surface Structures ($8). PS rules are ‘rewrite’ rules of basically two types. The ‘conrext-free" type of the form X — Y, and the ‘context-sensitive’ type of the form X > WYZ, where W and Z represent the context. The former generate phrasal categories such NB, VP, S. . . etc. and the latter introduce lexical stems into appropriate concexts in Phrase Markers. Transformarions were Largely construction-specific, so that there was a transformation for passives, a transformation for yes-no questions, and so on. UG was considered to contam a kind of blueprine which prescribes the types of possible rule systems, and an evaluation metric which restricts the cange of possible gtammnars to che ones (ideally, one) compatible with the data available to the child, Ata later stage, it became clear that there was a conflict betwecn the desire to provide a description of forcher phenomena, chat is the desire to achieve descriptive adequacy, which resulted in the proliferation of rule systems, and the need to constrain this proliferation, that is the desire to achieve explanatory adecuacy, The reaction to this conflict was basically to derive general principles with broad scope from existing ones and attribute them to UG. These principles would then serve as conditions on representations, the application of rules of thei output, and perform a restricted range of operations. As components of UG, these principles also serve to define the notion ‘possible human language’ The developing theory went though successive stages with distinctive properties called the Standard Theory, the Extended Theory, Government and Binding Theary, the Principles and Parameters Theory and the Minimalist Program (or Minimalism). Each of these stages represented an improvement on the previous stage, where improvement is driven by the desire co achieve explanatory adequacy. As the theory was developed, its empirical range was widened considerably to include a fairly broad range of diverse languages. This ted to the sharpening of some of the existing ideas, but most prominently to the formulation of clearer ideas about the principles responsible for language variation. It turned our that some of the major aspeces of language variation can be accounted for in terms of simple and well-defined sets of options, technically called parameters, which are largely deter- mined by the lexical properties of a specific class of categorics called functional or inflectional categories. The comparative work carried out within this framework has been largely successful in identifying common underlying properties of super. ficially different languages, 1.4 About this book Although this book is intended as an introduction to Transformational Grammar, it also incorporates as a major objective an attempt to explain some of the fundamental shifts of perspective which have shaped its development up to Minimalism. These inclade the shift from a theory based on catcgory-specific and construction-specific rules to one which is based on general rules with broader empirical range; the shift from a theory with a minimal (or no} internal structure to one which is highly structured; and the shife from a theory of grammar whoseLanguage and Linguistic Theory 13 main focus was to provide an in-depth descnption of the grammars of 4 few individual languages co one which combines the cask of describing the grammars of a broader range of individual languages with an attempt to account for language variation in terms of parameters which define and set the limits on linguistic variation. Obviously, the task of presenting and illustrating these major shifts of perspec- tive inevitably involves a process of selection of issues, data, hypotheses, opinions, bibliographical sources and so on, Combining this task with that of providing an introduction to the format mechanisms af the theory sometimes inevitably invalves a process of adapting ideas from previous frameworks in ways that affect mostly the format of their presentation, but sometimes also their concent. For example, the cask of justifying the eventual reduction of most transformations in earlier stages to a single operation has meant that che definitions of the relevant individual trans- formations in earlier chapters must emphasise their common aspects. The book is divided into four parts. Part I includes a chapter on phrase structure (Chapter 2), a chapter on the lexicon (Chapter 3), a chapter on transformarions which affect phrasal categories (Chapter 4) and a chapter on transformations which affect terminal categories (Chapter 5), Besides the task of introducing the narure of the rules involved in each of these components and their function, the presentation is also intended to give che reader an idca about how the theory looked in its earlier stages Fart I] outlines the stage of the theory known as the Principles and Paramerers framework, with each chapter dealing with one of the modules of the framework. Chapter 6 deals with X-bar theory, Chapter 7 with 6-cheory, Chapter 8 with Case theory, Chapter 9 with Binding theory and Control, and Chapter 10 with Move- mient theory. Here again, besides the task of explaining the major concepts of the Principles and Parameters framework, the presentation attempts to explain why these concepts came to be held and how they fit into the general aim of achieving descriptive and explanatory adequacy. Up to this stage in the book all data discussed are from English. This is done to make the point that it is possible to comstruct a sufficiently abstrsec theory on the basis of an in-depth study of one language to be able to put forward hypotheses relating to the universal nature of somc of its rules, Obviously, the hypotheses need evencually to be tested agains: data from other languages. This is done in Part Il of the book, which deals exclusively with language variation across a fairly broad range of languages. It discusses some of the major parameters of variation dis- cussed in the literature, Chapter 11 deals with variation relating to the order of heads in relation to cheir complement, variation relating to bounding nodes and Sabjacency, and variation relating is whecher the wh-phrase 1s moved or left in-situ in simpie wh-questions. Chapter 12 deals with the phenomenor: of null subjects and objects. Chapter 13 deals with the Verb Second phenomenon, VSO languages and NSO languages. Chapter 14 deals with Incorporation phenomena. Chapter 15 deals with clitics and dliticisation. Chaprer 16 deals with variation relating to whether the verb is moved overtly of not. Part IV outlines the broad aspects of the latest stage in the development of the14 Language and Linguistic Theory theory called the Minimalist Program (or Minimalism). This new framework is suill in its very early stages and therefore the presentation of it 1s generally sketchy, although there are attempts to explore some of its major implications. Chapter 17 presents the general organisation of grammar assumed by the Minimalist frame- work, as well as some of its technical concepts. Chapter 18 deals with the Copy theory of movement and its implications for Binding theory. Chaprer £9 deals with Checking theory and how it accounts for certain aspects of language variation, Finally, Chapter 20 presents cbe theory of phrase structure known as Bare Phrase Scructure and compares it with an alternative theory known as Antisymmetey. The chapters in Parts I, Il and Ill are each followed by a set of exercises. Some of the exercises are designed to test ideas discussed in the chapter. Other exercises invite the reader to think about possible problems and how they can be solved. Some of these exercises anticipate phenomena thar are dealt with in subsequent chapters. The chapters in Parts [and TT each include a section ar the end called ‘Sources and furcher reading’. This section includes the bibliographical sources on which the discussion in the chapter is based, as well as related references. The reader is strongly urged to consult the original sources. The chapters in Parts HT and IV adopt the style found in published material of including references in the text instead of in a separate section. Sources and further reading Discussions of the philosophical and methodological foundanions of linguistic theory in general and Generative Grammar in particular can be found in Chomsky (1957, 1965, 1966, 1968, 1975b, 1980a, 1986a, 1987a, 1987h, 1988, 1991a, 199tb, 1995} and Chomsky et al. (1982). Smith and Wilson (4979) and Newmeyer (1983) include a summary of the core ideas, discussed and explained ar a Farely accessible level. An equally accessible discussion of che underpinnings of lingustie theory and related issues can be found in Smith (1989). Chomsky (1955/1975a}, based on an unpublished text, includes an introduction which provides valuable information sclating to the carlier scayes of the develop- ment of (Transformational) Generative Grammar. A more up-to-date account of the development of the theory and reflections about its Cuture shape ean be found in Chomsky (1987a, 1987b, 1995). Newmeyer (1980) is an excellent historical account of the major debates which have shaped the evolution of Generative Grammar, and which have led ro the development of other distinct, but cquaily important, versions. A general discussion of the shift from a rule-based grammar to a principle-based grammar can be found in Chomsky (1986a). Chomsky (1981) is a detailed and highly technical account of the attempts to replace rules with general principles, and explain crosslinguistic variation in terms of parameters. An interesting and insightful evaluation of the principle-based model and its relationship to eatlier models can be found in Newmeyer (1991).Language and Linguistic Theory Discussions of the issue of modularity can be found in most of Chomsky's references cited above, as well as in che references cired in the main text of this chapter, in particular Curtiss (1977, 1981, 1982, 1988), Yamada (1990) and Smith and Tsimpli (1991}. A general discussion of the issue of modularity can also be found in Newmeyer (1983) ard Smith (1989). For a philosophically oriented discussion, see Fodor (1983).Part I Phrase Structure, Lexicon and Transformations 2 Phrase Structure Contents LL Preliminary remarks 2.2 Constituencies and hierarchies 2.3 Phrase Structure rules 24 Aux and Tense 2.5 Verb Phrase 2.6 Noun Phrase 2.7 Adjectives and Adverbs 2.8 Co-ordinate structures 2.9 Grammatical functions and relations 2.10 Summary Exercises Sources and further reading 2.1 Preliminary remarks In Chapter 1, we characterised the native speaker’s knowledge of Enghsh as consisting of ruies in combination with the lexicon (a mental dictionary}. The lexicon consists of lexical entries for words which specify various types of informa- tion necessary for che proper use of the word, A typical lexical entry will minimally include i) information relating to the pronunciation of the word [a phonetic representation); ii) information relating to the meaning of the word (a representa~ tion of meaning); and iii) information relating to whether the word is a verb, a noun , . . etc, fa categorial representation). As we proceed, information included in cach of the representations mentioned will be made more precise and further information will be added. So far, we bave been using the term ‘word’ without defining i. As a matter of fact, it is hard, if not impossible, to find an accurate definition for this term. For18 Phrase Structure, Lexicon and Transformations example, unhappy and unburton consist of ewo meaningful units each (Chaprer 1), and yet they would normally be referred to as words. Likewise, kicked in (1) would normally be referred to as a word and yet ic consists of che verb kick and the past tense marker -ed: 1. John kicked the white balt. ‘The category Tense specifies the time of the event described by the sentence (past, present or future} in relation to the time when the sentence itself is uttered. For example, the event of ‘kicking’ described in (1) takes place in the past in relation to the time when the sentence is uttered. It is nat clear whether -ed is a word of the same order as kick, since, unlike kick, cannot ‘stand alone’. Ar the same time, the Tense marker -ed, like kick, is a meaningful unit in its own right, and contributes to the overall meaning of the sentence. For the reasons explained, among many others, we will avoid the term ‘word? here. Instead, we will use the less controversia! term ‘lexical item’. We will assume that kick and -ed each constitutes a lexical item with its own entry. As we proceed in thus book, a distinction will emerge between lexical categories such as verbs and inflectional caregories such as Tense. Semences are often said to have a structure. How do we know that this is the case? Our answer to this question will, for the moment, be that we simply do cot know. We are using this claim as a hypothesis which we intend to verify against data, If it turns out that the data support the hypothesis, that is if we can show chat the best way to explain certain properties of sentences is by assuming thar they have a structure, our hypothesis will he valid. On che other hand, if it taras out that che data do not support the hypothesis, that is if we fail to show thar properties of sentences can be explained in terms of a structure, we will either have to modify our hypothesis or give it up altogether, This is one of the main tasks of this chapter. 2.2 Constituencies and hierarchies First, we need to clarify the expression ‘sentences have a structure’. In the present context, we will understand this expression to mean that the lexical items which make up a given sentence are hierarchically ordered with respect to each ocher such that some arc at a higher or lower level than other items, or at the same level of hierarchy, To illustrace with an abstract diagram, consider (2): 2 A B & _—— oO mPhrase Structure 19 ‘There is an obvious sense in which A is higher than B and C, and B and C ate higher than D and E. There is an equally obvious sense in which B is ac the same level of hicrarchy as C, and D at the same level of hierarchy as E, Moreover, D and E are directly linked to C, while B and C are directly linked to A. Using more formal terminology, D and E are constituents [i.e. members) of C, and B and C are constituents of A. D and E, however, are not constituents of B for the simple reason they are not linked to B. It is these kinds of relations which we expect to find among lexical irems which make up us sentences when we say that ‘sentences have a structure’. With this in mind, let us now try to work out the stracture of the simple sentence 3): 3a. This boy can solve the problem. 3b. s Det N Aux v Det N re Ns toy can Soke tha proton Asa first step, we can assign each item in the sentence to its categorial class using the following abbreviations: N for nouns, Aux for modal auxiliaries, V for verbs, and Det for Determiners. These abbreviations are known as categorial labels. The categories in (3) cogether form the larger and more compiex category Sentence, abbreviated as $. This preliminary information is represented in (3b) The eriangle in (3b) and other diagrams below is used to indicate that no specific claim is intended or made (yet) concerning the internal structure of the relevant category. On other occasions, the triangle will be used to include complex eate- gories which ate not directly relevant to the point being discussed. A priori, there are a number of ways the categories in (3} can be hierarchically related to each other and, ultimately, to S. (4) and (5) represent ewo arbiteardy chosen posstbiltties, where the question mark stands for an unspecified categorial label: 4. 8 yo SS, eo I Det N AUK v this boy = can solve §=— the problem20. Phrase Structure, Lexicon and Transformations et ee v Det this boy can ssov@=— the problem (4) incorporates the claim that the determiner this, the noun séadent, the auxiliary modal cam and the verb solve together form a constituent which excludes the determiner the and the noun problem. (5), on the other hand, incorporates the different claim that the verb solve and the determiner the together form a con- stituene which excludes the other categories in the sentence. The correct structure of (3) is an empirical issue which can only be resolved on the hasis of a proper investigation of relevant data. Our strategy will be to select strings of categories and see whether their members behave as a unit/block with respect to the phenomena listed in (6) known as constituency tests/criteria. If a string of categories can, in one block, move, delete, be co-ordinated with a similar string of categories, or be replaced with an appropriate pro-form, we will conchide that they form a single constituent: 6. i} displacement ji) deletion tit) co-ordination and iv} replacement with a pro-form. The string of categories we will select first is solve the problem, which we will examine in relation ro the tests listed in (6), As shown in (7-10), the selected string, of categories can be displaced in onc block (7), deleted in one block (8), co. ordinated with « similar string of categories (9}, and replaced with the pro-form $0 (10), The latter is a pro-form in the sense thax it can stand for the string solve the problem. In (7) and (8), the symbol [--] marks the position where the displaced and deleted string is ‘understood’: 7. Displacement: This boy is determined to solve the prablem and [solve she problem] he will [Hl]. 8, Deletion: Jobn cannot solve the problem, but this boy can [~J. 9. Co-ordination: This boy will [solve the problem] and [win the prize] 10, Replacement with a pro-form: This boy can solve the problem and [so] car che others. On the basis of the data in (7-10) we can conclude that the categories included in the scring solve the problem together form a single constituent. In other words, we ean think of the facts in (7-10) as empirical evidence in favour of postulating aPhrase Structure 21 structure for sentence (3) where the verb solve, the determiner rhe and the noun protilem together form a single constituent which excludes the other categories in the sentence. This conclusion is represented in diagram (11), where the isolated constituent is labelled Verb Phrase (VP). Pending the evidence, the remaining categories in the sentence, i.e. the determiner this, the noun student and the modal will, axe linked to $ by a triangle. Likewise, pending the discussion of the internal structure of VP, its constituents are linked to VP by a triangle: 11. So aa Se Det N Aux wR Da I oof | this boy cen soe =the ~— problem The conclusion we have reached on che basis of the data m (7-10) already excludes both (4) and (5) as possible structures for (3). This is because neither of these stcuctures has the verb solve, the determiner the and the noun problem included under a single constituent. fn other words, {4) and (5) make the wrong predictions with respect to the constituency tests discussed above. Because the categories in question are not included in a single constituent, they are not predicted to cluster together as a unit/block with respect 10 the phenomena of displacement, deletion, co-ordination, and replacement with a pro-form, contrary ro what we saw above. Thus, structures (4) and (5) and che hypotheses they incorporate can be dismissed on the ground that they are empirically inadequate. Following the same procedure, fet us now select anather string of categories to determine the structure of the rest of the sentence. This time we will select the string this boy which consists of the determiner she and the noun boy The data in (la8cb) show thar they form a constitucar: 12a, [This boy} and {that girl] can solve the problem. 12b. [He} will solve the problem. In (12a) the determiner and the noun are co-ordinated with a similar string of categories, and in (12b) they are together replaced wich the pro-form (pronoun) he. Incorporating this conclusion into diagram (11}, we obtain the more articulated structure shown in diagram (13). The larger category which includes the deter- miner and the noun is labelled Noun Phrase (NP):22 Phrase Structure, Lexicon and Transformations * ae NP aux Ve “ N v 7 ; A J can as the problem (13) also incorporates the implicit claim that Aux alone forms an autonomous constituent of S. Since Aux is the only category left, this claim seems to be wartanted, if not mescapable, To make sure that there is an empirical basis for the claim, we can rest it in terms of the constituency criteria. If che conclusion is correct, we expect Aux ro be able, for example, to be displaced and to be cox ordinated with another Aux clement independently of the caregories which make up NP and indeed also VP. In {14a} cam 1s displaced to the front of the sentence across the NP this boy, In (14b) the modal auxiliary is co-ordinated with another modal auxiliary, namely will: 14a, {Can} this boy {—] solve the problem? 14b. This boy [can] and [will] solve the problem. Qu; remaining task with respect to (13) is to work out the internal structure of VP. ‘That is, we iced to determine whether each of the categories included under VP is an autonomous constituent of VP or whether som of thera cluster togerher in the form of a constituent of VP. As a macter of fact, this task has alteady been carried out. Like their counterparts thes and boy, the determiner the and the noun problem are likely to form an NP consticuent too. Generalisations actoss categories of this type are expected on the grounds that categories of the same class rend 10 cluster together in the form of larger constituents. As it will transpire in this book, generalisations of this type form the backbone of linguistic research. On this basis alone, we can conclude that the determiner and the noun inside VP form an NP constituent of VP The consequence of this conclusion is that V alone forms an autonomous consti tuent of VP. These conclusions are incorporated in diagram (15): 1S. NE, Aux Det N Vv UN Det this boy can sove thePhrase Structure 23 Asin the previous station, to make sure that there is an empirical basis for the conclusion thar the and problem form a single consticuent, we can test itm terms of the constituency criteria. (16a) shows that the determiner and the noun inside VP can be replaced with a pro-form, (6b) shows that the ewo cacegories can be co- ordinared with a simular string of categories. Finally, (J6c! shows that they can be displaced cogether to the beginning of the sentence: 16a. This boy can solve [it]. 16b, This boy can solve [this problem] and [thar puszle] 1c. [This problem), I believe the boy can soive [—| (15) is a Fully articulated structure of sentence (3). The numbers associated with NPs have nto theoretical status. They are merely convenient devices to distinguish one NP from the other in the ensuing discussion, Diagram (15) is basically a graphic way of representing the conclusions we have reached concerning the con- stituens structure of sentence (3). The advantage of {L5) as that it shows clearly jin visaal terms) that ulzimatels, to $. For example, Vand NP, arc not ar the same level of nerarchy as NP, Aux and VP. This is because Vand NP, are constituents of a constituent of 5, whereas NP; Aux and VP are all comsatuents of S. Vand NP» are said t0 be immediate constitucars of VP. NP,, Aux and VP are immedsate constituents of 5. Dei and N are atmmediare constituents of NP. The hierarchical relations berween categories are expressed in terms of the relation of dominance. A category 1s said to immediately dominate sts immediate constituents, Thus, $ immediacely dominates NP), Aus and VP. VP immediately dominates Vand NP,. NP {in both occurtences! immediately dominates Det and N, Note, however, that although $ does not immediately dominate Vand NP2, because the latcer are not its immediate constituents, it docs dominate chem, The distine- tion here is berwcen immediate dominance and (mere: dominance. A category is said to immediately dominate its immediate constiments, and to {merely} dominate the constituents of its constituents, The notion ‘constituent of a constituent’ is crucial in determining dominance relations. Because V and NP, are not constitu: ents of a constituent of NP,, the latter docs not duminate them. (15) is called s erce diagram, as with family tree diagrams. As a matter of Fact, terms used xo refer to family zelations, in particular mother, daughter and sister, are also used to refer to relations between categories in a tree diagram, Immediate constituents of a category are daughters of that category. Hor exemple, Det and N are daughters of NP. V and NPy are daughters of VP. Obviously if Det and N are daughters of NP, then NP is their mother. 1f V and NP» are daughters of VP, chen VP is their mother. Also, if Det and N are daughters of the sarne mother, then Det and N are sisters, and of V and NP, are daughters of the same mother, then Vand NP) arc sisters, and so on ‘Tree diagrams such as (15) are not the only means to represent sentence struc: ture, Another equally frequently used means is known as labelled brackets. (17) sepsesents exactly the same information ay (15), using different notation ategories are indeed! hierarchically related «@ each other and,24 Phrase Structure, Lexicon and Transformations 17. és ENP, (Dee this! In boyl] Laux can] [yp Fy solve] [Np, [Der thel (Nn problem]]}] In tree diagrams, categories are represented ir terms of labelled nodes, and related ro each other in terms of branches, In labelled brackers, categories are represented in terms of labelled brackets, and related to cach other in terms of inclusion and exclusion relations. In (17), the bracker Jabelled VB, for example, includes the brackers labelled V and NP2, meaning that V and NP are constituents of VP. At the same time, the bracket labelled VP excludes the brackets labelled NP, and Aux, meaning that the fatter are not constituents of VP. As expected, re bracket labelled S includes all the other brackers, reflecting the fact that all categories in the sentence are either constituents of S or constituents of constituenss of S. We will usc both means of representing structure, with the choice being entirely a matter of convenience. 2.3 Phrase Structure rules The claims abour the structure of the sentence represented in diagrams (15) and (17) can be formulated in terms of ‘rewrite cules’ such as the ones in (18} (tead “9? as ‘rewrite as? ar, less formally, ‘goes to’). 18a. S$ — NP Aux VP 18b, NP -» Der N 18. VP» V NP {182-c) are Phrase Structure (PS) rules in the sense that they incorporate ciaims (specified to the right of the arrow) abour the constituent structures of phrases ispecified ta the left of the arrow), PS rules are said to generate structures, where gencrate is understood to mean ‘make explicit’, (18a), for example, generates the structure of $ by making explicit the information thae § consists of NP, Aux and VP. (18b), on the other band, generates the structure of NP by making explicir the information that NP consists of Det and N. Finally, (8c) generates the structure of VP by making explicit the information that VP consists of Vand NP. Tree diagrams and labelled brackets are (visual) devices of representing claims about constituent structures meorporated in PS rules. ‘The PS rules (18a~-c) were based on sentence (3) in the previous section, repro- duced in {19}. However, their generative capacity goes well beyond (19), ro include all possible sentences in the language with similar strings, (20a-d) are a few examples of such sentences, They al] resemble (19) in chat they include the same patterns of constituency for each category: 19a, This boy can solve the problem. 19. [s [Np this boy] [Any can] [yp [y solve] [Np the problem)j] 2Wa. The police will arrest the thief. 20b. This man can deive that car.Phrase Structure 25 20c. ‘The President will chair the meeting. 20d. The Parliament can impeach the President. To generate a specific sentence of the set of sentences generated by rules (18a-<), another set of rules which generate specific lexical items can be added. (19), for example, 1s fully generated by the set of rules im (21): 2la. $ > NP Aux VP 2b. NP — Det NP 21c. VP —> V NP 21d. Aux > can Qe. Dec — the, this 21£ N > student, problem Qty. Vi salve Rules (21a-c) generare phrasal categories onc constituent of which is a terminal node. Terminal nodes are nodes that do not branch and rhat immediately dominate the lexical item. For example, the phrasal category VP has the terminal node V as one of tts constituents, and NP has the terminal node N as onc of its constituents. § is called the root mode, Rules (21d-g}, on the other hand, generate terminal nodes by introducing corresponding lexical items in the sentence. The structures gener- ated by both sets of rules arc called phrase markers. (15) is the phrase marker of sentence (19), represented in che form of a tree diagram. Obviously, there is also an equally large number of possible sentences which PS tules (21a~c) cannot generate, mainly because thcy have constituency patterns which differ from the ones in (19) and (20a-d), In the rest of this chapter, we will try to accommodate as many types of sentences as possible, essentially by enriching the system of PS rules developed so far, We will discuss each of the PS rules in (21a-c) separately and modify itn such a way that it can accommodate a broader range of constituency patterns for the categoty it generates, and there- fore a broader range of sentences, Our first targct will be the PS rule chat generates Aux, 2.4 Aux and Tense {22a&b] and many similar sentences differ from the ones discussed so far in chat they apparently lack an Aux: 22a, The boy kicked the ball. 2b, The boy saw the girl. {22a&b) are problematic fot rule (21a) for the simple reason that the rule states that Aux is an obligatory constituent of S, To accommodate sentences such as (22a& b), as well as generate scmences with Aux, we need co modify rule (2La). Aa obvious way of achieving the desired result is simply te make the occurrence ef Aux optional. Notationally, this can be done by including Aun between26 Phrase Structure, Lexicon and Transformations parentheses roughly as follows: S > NP (Aux) VP. The lateer states that 5 consists of NP, an oprional Aux and VP. This version of the rule now generares sentences with an Aux ay well as sentences without an AUX. However, there is an alternative way of accommodating sentences such as (22a&h) which, when examined carefully, turns out to he more adequate than the solution which makes the occurrence of Aux optional. Examples {23a-d) show shat Tense, the category which specifies the time of the event in relation to the time of the utterance, #s a constituent of Aux rather than of Vs 23a. The boy [will] kick the ball. 236. The boy [docsn’t/didn’t] like the party 23e. The girl (didn’t) like the party, but the boy did. 23d. [Go to the party] | wonder whether the boy will [—] In (23a), Tense shows up as the Modal iilf, while the main verb does not carty any Tense information. In (23b), Tense also shows up on the auxiliary do. In (23cl, deletion of the verb and its complement (i.e, deletion of VP} docs not affect Tense. Finally, in (23d). displacement of VP does nor affect Tense, Now, given that sentences invariably have a Tense category, and given thar Tense is a constituent of Aux, it follows that ail sentences have an Aux category. Sentences differ only in that some of them have a Modal auxiliary under the Aux node, in addhtion ta Tense, as in (19), while ochers have only Tense under the Aux node, as in (22a8&b). Thas, the optional occurrence of Aux implicd by (22a&b) only reflects the optional accurrence af a Medal in addition to Tense. In view of this, our initial solution, which made the occurrence of Aux optional, is inadequate. This is hecause it way based on che assumption that Tense is not a constituent of Aux, contrary to what examples (23a-d) indicate. The solution which is consistent with the facts illustrated in (23a-d) is one which maintains Aux as an obligatory constituent of $ on the grounds that it 1s che node under which Tense is located. The new version of the PS rule which generates Aux now looks as in (24b) where Tense is an obligatory constituent and the Modal an optional constituent. (24a) is the PS rule which generates §. This rule remains as we stated it originally, with Aux and obligatory constituent: 2a. 8 > NP Aux VP 24b. Aux > Tense (Modal) In view of (24b), in particular the idea chat Tense 1s a constituent of Aux, we need to explain how it is that Tense shows up on the verb in sentences such as (22a&b}), This task is carried out in Chapter 5. For the moment, note that Tense clements such as the past tense marker -ed are morphologically dependent mot- phemes or bound morphemes. They cannot stand alone, and need to attach to a verbal category such as a Modal or a verb, It is fur rhis reason that Tense appears attached to the verb in sentences such as (22a&b). Obviously, Tense attaches to the verh only when Aux does not include a Modal category: In sentences where Aux includes a Modal category, ¢.g. This boy will solve the problent, Tensc appears on che Modal category and the main verb appears unmarked for Tense.Phrase Structure 27 The category Neg{ation) also belongs under Aux, irrespective of whether it has the full form nor or the contracted form n't, This is shown in (25a), where the Modal and Neg are co-ordinated with a similar string of categories, and in (25b), where the Modal and Neg are displaced rogether to the beginning of the sentence 25a. This politician [cannor] and [will not] solve the problem: 25h, [Can't] this politician [—| solve the problem? The tule which generates Aux now looks as in (26a). Tense is an obligatory constituent present in all sentences, and Modal and Neg are optional. (26b) is the sttucture of Aux implied by cule (26a): 26a. Aux — Tense (Modal) (Neg) 26b. Aux Tense (Modal) (Negi 25 Verb Phrase _ The PS rule which generates VP looks as in (28). The cule was established on the basis of the sentence reproduced in (27a), and states that VP consists of Vand NP: 27a. This boy can solve the problem. 27b. this boy can [VP [V solve] [NP the problem]] 28, VP > V NP Sentences (29} and (30) appear to be inconsistent with (28). (29) and (30) seem co have a VP which consists of the verb only: 29a. The doy cried. 29b. the boy [yp fy cried] 30a. The girl smiled. 306. the girl Lyp [y smiled] To accommodate sentences such as (29) and (30), the occurrence of NP in the rule which genezates VP can be made optional by including it between parentheses. (Bla) is the revised version of the VP rule. (31h&c} are the owo WP structures generated by (314). (31b} corresponds to sentences of the cype illustrated in (27), and (3c) co sentences of the type illustrated in (29) and (30): 31a, VP > V (NP) Hb. ye Ble. vp v NP. v | oo | solve the probiem, criod/srruted28 Phrase Structure, Lexicon and Transformations The form that VP takes is largely determined by the type of verb it includes. If the verb is of the type that takes an NP complement (or object), such as solve in (27),VP has the form seen in (31b). On che other hand, if the verb is of the type that does not take an NP complement, such as cry in (29) and smile in (30), VP has the form seen in Gc). Verbs which sake an NP complement or object are called eransitive verbs and the ones which do not take an NP complement are calted intransitive verbs Sometimes, the terms ‘transiewe” and ‘intransitive’ are also used to describe VPs. The VP in (31b} is transitive and the onc in (31c) is intransitive. Just as there are verbs which take NP as a complement, there are verbs which take other categories as a complement. For example, the verbs hint in (32) and knock in (33) take a complement which consists of a preposition and an NP. ‘The preposition and its NP complement are said to form a Preposirional Phrase (PP), so that the verbs in (32) and (33) are said to take a PP as complement. Ps are generated by the PS rule seen in (34a), and have che structure shown in. (34b): 32a, The teacher hinted at the solution. 32b, the teacher [yp [y hinted] {pp at the solution]) 33a, The girl knocked on the door 33b, the girl [yp [y knocked] [pp on the door]] 34a, PP > PNP 34b. a PN at the solution on the doar ‘To generate sentences of the type illustrated in (32) and (33), the VP rule {28) needs to be revised to allow far the possibility chat VP can consist of Vand PR addition to the two other possibilitics discussed. This can be done as in (35a). The Po rule in (35a) includes the information that VP can consist of Vonly, Vand NP, or Vand PP The first two options are shown in (31b) and (318¢c), and the third option required for sentences (32} and (33) is shown in (35b): NP) 35a. VP > V { } PP 35b, ye Bp. at the solution on the doorPhrase Structure 29 It is clear that the rule which generates VP becomes increasingly more complex as revisions are introduced ro accommodate new types of VP. Although the complexity of a given cule only reflects the complexity of the various patterns of constituency it is intended ro deal with, we will avoid formulating complex rules here. Instead, we will restrict ourselves co the version of the rule which is relevant to the example being discussed. Hopefully, the point that a given rule can be revised (made more complex) to accommodate a broader range of sentences is already clear. Examples (36) and (37) imply the version of the VP rule in (38a) and the structure shown in (38b). The verb send in (36) and the verb give in (37} belong to a group of verbs which take two complements, an NP and a PP. Therefore, the VP which includes these verbs consists of V, NP and PP: 36a. The boy sen a letter to the girl. 36b. the boy [yp [vy sent] [np a letter] [pp to the girl]] 37a, The gicl gave a present to the box 37b. the giel [yp Ly gave] [Np a present] {pp to the boy]] 38a. VP > V NP PP 38b. send alter to tho girl ave apresent to the boy Examples (35) and (40) imply the rule in (41a) and the structure shown in (41). The verbs say and rhink take S as a complement, implying that their VP has the form shown in (41b). That the complements in (39) and (40), included under $ in (41b), are of the category $ 1s shown by the fact that each one of them can stand alone as a complete sentence: He would send a letter to the girl and She would give a present 10 the boy: 39a, The boy said he would send a letter to tbe girl. 39b, the boy [yp said [5 he would send a letter to the giel]] 40a. The girl thought she would give a present to the box 40b. the girl [yp thought (5 she would give a present to the doy] 4la. VPaVvs 4b. yp sald ha would send a letter to the girl thought she would give @ present to the boy30. Phrase Structure, Lexicon and Transformations Now, compare (39) and (40) co their counterparts in (42) and (43), which differ only in chat they include the extra item that, called Complementiser (Comp). Comp has the function of introducing the S-complement of the verb. Far reasons which will become clear later on, Comp is not a constituent of $, bur a constituent of another super category called §" {read S-bar). 5! is generated by rule (44a) and has the struceure shown in (44b): 42a, The boy said that he would send a letter to the gil. 42h. the boy said [thar [he would send a letter co the girl)) 43a, The girl thought chat she would give a present to the box 43b. the girl thought [that [she wanld give a present ro the boy) 44a. S' > Comp S 44b. UN Comp § that he would send a letter to the git that she would give & present to tho boy According, co tule (44a), $ 1s also a constituent of S’ represented in i44b) as a sister to Comp. The VP in (42) and (43) therefare has the stracture shown on (45b),, generated by the rale in (45 45a. VP > VS" 45b. vP v S. said that the would send a letter to the gi thought that she would give a present 10 the doy The (internal structure of the) $’-complemenc is generated by rule (44a), in combination with the rules which generate § and its major constituents. An S'-complement is sometimes called a clausal complement, or, alternatively, a sentential complement. The term ‘clause’ is sometimes used synonymously with ‘sentence’, so that (39), (40), (42) and (43) are said to be bt-clausal, meaning that they consist of two clauses. The larger clause (5} which includes che complement clause is called the main or matrix clause. It is also often referred to as the root clause. The complement clause is called the subordinate or embedded clause. As things stand, it seems as chough the complement of the verb in (39) and (40) is a different category from the complement of the same verbs in (42) and (43). In the former, che complement is $ and in the latter it is $°. This would require a different version of the PS rule which generates VP far each set of examples. However, this may turn out ro be an unnecessary complication. That is, it may curn out that the complement is the same category in bath sets of exampies, and that the presencePhrase Structure 31 and absence of shat does not necessarily result in a different category, One could adopt the hypothesis that in both sets of examples the complement is S* and that (42) and (43) differ only in that Comp 1s e(mpty), i.e. not filled with the comple- mentiser that. This hypothesis is shown in (46): said e he wourd sand a letter to the git thought e she would give a present to the boy The advantage of the hypothesis represensed in (46) is that only one version of the VP rule is needed for both types of example, namely (45a). Version (41a) can be dispensed with altogether, ar least as far as the examples discussed are concerned. The hypothesis underlying (46) is destrable on grounds of economy, ie. it allows us to dispense with one rule. However, like all hypotheses, it must be shown ro be empirically correct, thae is, we must be able to show thar clauses have a Comp node even when they apparently do not include a complementiser. Unfortunately, we will nor be able to discuss the relevant evidence till Chaprer $ ‘The PS cule which generates VP (45a) is a guod example of a recursiye rule. Recursive rules are thought to be responsible far the abilicy of native spoakers to Produce potentially indefinitely long sentences, eg. Mary thinks (that) Jane believes (that) John claims (that} Bill heard... (chai) the witness will not testify. The limits on such sentences are extraneous to language itself, such as memory. attention span and so on. ‘This is another respect in which human languages are said to be ‘creative’. By reincroducing $' on the right of the arrow, rule (45a) makes it possible to reintroduce VP into the derivation which then makes it possible to reintroduce $' into che derivation and so on. 2.6 Noun Phrase ‘The PS rule which generates NP established above on the basis of example (47) is reproduced in (48}: 47a, This boy can solve the problem. 47. [Np {Der this] [N hoy] can solve Pyp [Der the] IW problem}] 48. NP Det N Just as the constituent structure of VP depends on the type of verb it includes, the constituent structure of NP also depends on the type of noun ir inchades. (49) includes the noun cancellation which takes a PP complement as well as. a Det. This means thar the NP which includes this noun consists of Det, N and PP. (50)32 Phrase Structure, Lexicon and Transformations includes the noun claim which takes an S’-complement as well as a Det, so that the NP which includes it consists of the categories Det, N and $’: 49a. he cancellation of the party annoyed the boys. 49b, [np the cancellation [pp of the party] annoyed the boys 50a. The gir] resents the claim thac she likes the boy. SOb. the girl resents [Wp the claim [g* that she likes the boy]} The first NP has the consticuent structure shown in (51b}, generaed hy the verston of the NP rule in (S1a). The second NP has the constituent structure shown, in (52b), generated by the version of the NP rale in (52a}. The more general rale which generates both types of NP would therefore have to include the information that PP and S’ are optional constituents of NP, although we will nor try ta spell it out here: Sa. NP > Det N PP Sib. NP an bP the canoeliation ot the party 52a. NP > Det NS’ 52b, NP “yoo the claim that she likes the boy Examples (33) and (54) suggest that the occurrence of the category Det Preceding the noun also needs to be made optional. (53) and (54) include NPs which consist of Det and NP, namely the film and the box and NPs which consist of the noun only, namely Mary and bananas. Both types of NP are generated by rule (35a). (5b) is the structuze of NPs which consist of the noun only. Whether a determiner can occur in an NP depends on the noun, although determiners are nor complements of the noun. In English, nouns such as Mary i.e. names, do not tolerate a determiner, whereas nouns such as boy and cancellation require a deverminer: 53a, Mary likes the film. 53b. [Np Mary) likes [yp the film| 54a, The boy likes bananas, 34b. Inp the boy] likes [yyp bananas}Phrase Structure 33 SSa. NP -» {Det} N SSb. NP N Mary bananas Nor only determiners can precede N but full NPs as well. ‘This is shown in (56) and (57). The relevant NPs in these examples have the structure shown in (585) generated by rule (58a): 56a, John’s cancellation of the party annoyed the boys. 56b, [NP [NP John’s] cancellation of the party] annoyed the boys S7a, The girl resents the boy's behaviour. 57b. the giel resents [yp [Np the boy's] behaviour} Séa. NP» NPN... S8b, aw NP PP) Johnis cancellation of the party the boy's behaviour ‘The pre-nomuna! NPs in (56) and (S7} are in complementary disteibution (cannot co-occur) with determiners, as shown in (59} and (60). The rule which generaces NP would thercfore have to take this restriction into consideration, ‘This can he done by including Det and NP in curly brackets as in (61), a notation which excludes co-occutrence. The formulation of the rule in (61) guarantees that NPs of the type illustrated in ($9) and (60) are excluded: 59a, *John’s the cancellation of the party annoyed the boys. 59b. [yp INP John’s! [per the] cancellation of the party]... 60a. *The girl resents the boy's the behaviour 606... . Lp Ip the boy’s] (Dee the] behaviour] NP. N... 61. NP -» Det 2.7 Adjectives and Adverbs Adjectives (A) and adjectival phrases (AP) typically occur in rhe positions indicated in (62) and (63);34 Phrase Structure, Lexicon and Transformations 62a, The boy is tall. 62b. the boy fyp is [ap tall} e3a. The tall boy likes the girl. 63b. (yp the [ap call] boy] likes the girl In (62}, AP is a constituent of VP together with the verb és. In this particular context, AP is said to have a predicative function for reasons chat wil! be explained below in this chapter. In (63), AP is a constituent of NP together with the N it modifics and Det. In this context, AP is said to have an ateributive function, in the sense that it specifies an attribute of the N it modifies. As is the case with verbs and nouns. some adjectives take complements while orhers do not. The adjective tall, for example, does not take a complement. In contrast, the adjective suspicious takes a PP complement, as shown in (65). The appearance of the complement, however, is optional. This is shown in (64) which lacks a complement for the adjective: 64a. The suspicious girl is Mary. 64b, the [ap suspicious] girl is Mary 65a. The girl is suspicious of the tall box 65b. the girl is [ap suspicious [pp of the tall boy]} ‘The PS rule which generates AP must therefore be as in (66a), where PP is an optional constituent, {666} and (66c) are the ¢wo structures of AP it smplies 66a. AP» A (PP) 66, y 66¢, oN | a PP ta | PN suspicious suspicious. ofthe boy Having established the rule which generates AB, we now turn to the rules which generate the phrasal categories which contain AP in (62-65). The rule which generates VP in (62) and (65) has the form shown in {67a}, which implies the structure in (67b). On the other hand, the rule which generates the NP in (62} and {63) has the form shown in (68a), which implies the structure in (686): 67a, VP > VAP 67b, ve is tall suspicious (of the boy)Phrase Structure 35 68a, NP + Det APN 68h. NP. De AP cy | I | the tall boy the suspicious gi Turning now so adverbs (ADV), they typically occur in at least three maior positions depending on the category they modify. In (69a), the adverb quickly appears between the subject and the main verb, In (695), it appears in the sentence final position. The latter is also the position occupied by the adverb evidently in (70b}. (70a) shows that an advecb can also appear in the sentence initial posicion: 69a. John quickly fixed the car. 69, John fixed the car quickly. 7a, Fvidendy, John fixed the car 70b. John fixed the car, evidently The adverb quickly in (69487) refers to che manner in which ‘John fixed the ca’. Quickly belongs to a group of adverbs which inciudes cleverly, clumsily, deftly... erc. called manner adverbs. Because manner adyerbs modify the verb (and its complements), they are structurally represented as constituents of VP. For this reason, manner adverbs are sometimes also called WP-adverbs. The claim that manner adverbs are constituents of VP is confirmed by the fact that they can be displaced along with the constituents of VP as shown in (71): Tia. Fix the car quickly, 1 wonder whether John will. 7b [yp Fix the car quickly] 1 wonder whether John will j—] 723, VP -» (ADV) V (NP) (ADV) 72b, ye Pe ve | Po £N LA | quickly fed the cer fix the car quickly Manner adverbs are therefore generated by the VP rule (72a) which implies structures (72b) and (72c}, depending on the linear position of the adverb in VP. (72b) is the structure of the VP in (69a), where ADV immediately precedes the verb. (72e) is the structure of che VP in (69b), where ADV is in the final position. The adverb evidently in (70a8cb) refers to a situation which suggests thar “John indeed fixed the car’, In this case, the advecb is said to modify the whole sentence rather than a specific constiment of the sentence. For this reason, evidently is classified as a S{entence}-adverb, together with other adverbs such as presumably, ironically, probably... etc. As constituents of $, S-adverbs are generated by the36 Phrase Seructure, Lexicon and Transformations rule which expands $ shown in (73a). (73b} and (73c) are the structures it generates (73b) is the structure of (70a), where the S-adverb is in the initial position. (73¢] ss the structure of (70b}, where the S-adverb is in the Bnal position: 73a. § > (ADV) NP Aux VP (ADV) 73b, s ADY NP Aux WP Po} | A evidently John Tense fix the car 73. 8 uw NP Ax ve ‘ADv bf ZN 4 John Tense fixthe car _evidentiy The adverbs discussed so far have in common the bound morpheme -ly, and hence the fact that they are sometimes called /y-adverbs. However, there are adverbs which do not have the suffix -ly such as hard and fast in (74) and (75). Like quickly, these adverbs are also manner adverbs in the sense that they descrihe the manner in which the event denoted by the verb is carried out. As such, they are constituents of VP generated by the PS rule (72a): 74a, John hit the nail hard. 74b. John {yp hit the nail hard] 7Sa. Mary ran fast. 75b. Mary [yp ran fast] Note, finally, that in addition to the adverbs discussed, linguists sometimes talk about categories which have an ‘adverbial function’. The most prominent examples of these adverbial functions relate to place and time. The bracketed constituents in (76} and (77) are said to have an ‘adverbial function’, although as categoties they are PP (76a&b) and S’ (77a8cb): 76a. John fixed the car [pp in the garage]. 76b. John fixed the car [pp in the morning]. 77a. Jobn fixed the car [g- where Bill had left it]. 77». John fixed the car [gr when Bill was still steeping). Whether categories with the adverbial functions of place and time are constitu ents of S or VP is a difficult question which we are not going to address here. This does nor prevent the reader from applying suitable constituency tests to find our for herselffhimsel!.Phrase Structure 37 28 Co-ordinate structures Co-ordination is the device whereby two categories are joined together using and to form a complex category of the same type as the two co-ordinated categories. For example, the two NPs the boy and the girl in (78a) are co-ordinaced ta form a complex NP with the co-ordinate structure shown in (78c) 78a. The boy and che girl solved the problem. 78b. [xp (Np the boy} and [Np the girl}} solved the problem 7B. NP. eo Oe | | | | te xy the ae Any category can in principle be co-ordinated with another category of the satne type to produce a co-ordinated category also of the same type, Numerous examples of co-ordination were cited in Section 2.2 ahove involving NP, VP and Aux. (79), for example, involves co-ordination of two PPs: 79a. Mary knocked on the door and on the window. 79h. Mary knocked [pp [pp on the door} and [pp on the window] 79. Pe a pe and Pp a a | ZN | ZN on the door on the window (80) invoives co-ordination of two VPs: 80a. Mary opened the door and closed the window. 80b. Joka [yp [yp opened the door! and [yp closed the window]}. (81} invoives co-ordination of two adverbial phrases: 81a. John resigned quickly and quitely 81b. John resigned [any [apy quickly] and | apy quitely)]. Since all phrase structures are generated by PS rules, and since any category can be co-ordinated with another category of the same type, there must be a co- ordination rule for each category, However, instcad of providing the relevant PS38 Phrase Structure, Lexicon and Transformations rule for each category, we can use the general and abstract rule in (82) where X stands for any category: 82 XP — XP and XP Note that (82) must be interpreted in such a way that X is the same type of category on both sides of the arrow. This is because co-ordination can only invalve categories of tbe same type. An NP can only be co-ordinated with another NP, a VP can only be co-ordinated with another VB, and so on. Thar ewo categories of different types cannot be co-ordinated is shown in (83) which involves co- ordination of an NP and a PP: 83a. “Mary plays the piano and at home. 83b, *Maty plays [[Np the piano] and |p at home|] 2.9 Grammatical functions and relations The terms ‘subject’ and “object” are said co refer co the grammatical functions of categories. They are not catcgorial labels, and therefore should not be confused with categorial labels sucb as NP, VP. V.. etc. In (84), the boy is an NP category whieb has che grammatical function ‘subject’, and the ball is another NP which has the grammatical function ‘object’: 84a. The boy will kick the ball. 84b. s ee NP Au Ve sw “ig Det N v oN | | | an) the boy will kick the: ‘ball Moreover, the grammatical functions ‘subject’ and ‘object’ are ctructurally based. The subject is the ‘NP-daughter-of-S’, and the objece the ‘NP-daughrer- of- VP", In (84) the boy is the subject by virtue of the fact that it is the ‘NP-daughter of S', and the bail the object by virtue of being ‘the NP-daughter of VP’. Grammatical functions such as ‘subject’ and ‘object’ are also said to be rela- tional, and the expression grammatical relations is often used synonymously with gcammarical functions, The NP the boy in (84) is the subject of the sentence in the sense that it is a subject in celation to the senrence $. On the other hand, the NP the ball is che object of the verb, that is, it is an object in relation to the verb or the VP that includes it. The relational nature of grammatical functions can be captured in terms of the formalism in (85):Phrase Structure 39 Sa. Subject-of-S [NP, S} 85b. Objcct-of V: [NP VP] > (VP, S] BSc. Predicate-af- ‘Predicate’ is another grammatical function which is also relational in nature. The predicate of a sentence is usually the string of categories which contains the verb and its complements, that is, che string of categories included under the VP node, For example, the predicate in (84) is kick ihe ball. Its for this reason that the AP in the seatence The boy is tall discussed earlier was said to be predicative. Structural relations which determine grammatical functions are encoded in PS rules, For example, the structural relation ‘NP-daughter-of-S” which underlies the gtammatical function ‘subject’ is encoded in che PS rule which expands S ($ > NP Aux VP), On the other hand, the structural relation *NP-daughter-of-VP* which underlies the grammatical function ‘object’ is encoded in the PS cule which expands VP (VP — V NP). Thus, ia addition to specifying the internal structures of phrasal categories, PS rules also specify the grammatical functions of categories. Finally, it is important to bear in mund that lincarbased nanons such as ‘precede’ and ‘follow’ do not play a role in determining grammatical functions. Thus, ‘subject’ cannot be defined as the NP which precedes the verb, snd ‘abject cannot be defined as the NP which follows the verb. Although the two lincar notions may apply to the sentences discussed so far, there are numerous other types of sentences to which they do not apply as will become dear in subseuuent chapters, The fact that the subject precedes the verh and the abject follows the verb in certain types of English sentences is a by-product of the geometrical aspects of the structures generated by PS rules. As we proceed, it will become clear that linear-based notions such as ‘precede’ and ‘follow’ do not play a role in grammar. Graramatical relations and rules are structurally based. 2.10 Summary Tn this chapter we have postulated a set of rewrite rules called PS rules which generate (i.e. make explicit) the structure of the sentence. A subset of these rules gencrate phrasal categories such as $, VP, NB, and another subset generate specific lexical items. The latter rewrite 4 zermmal node as a corresponding lexical item. Tnitially, the rules which generate phrasal categories were based on a single sentence. Then, it was shown thar these rules could be revised in appropriate ways vo generate phrasal categories with different constituency patterns, thereby accommodating a larger number of sentences. Each revision represents a small step rowards the goal of developing a model of grammar which generates all and only grammatical sentences. We came actoss a situation where the rule which gencrates NP had to he revised in such a way as to exchide unwanted NPs where a prenominal NP co-occurs with a Des clement. Thus, the revisions introduced are motivated not only by the need to accommodate new grammatical sentences, but also by the need to cxelude ungrammatical senrences. We have also seen that grammatical fancrions40. Phrase Structure, Lexicon and Transformations such as ‘subject’ and ‘object’ are essentially structural rather than linear notions. Linear relations have no place in the geammar of human languages. Exercises Exercise 2.1 With reference to the diagram provided: } Fiftin all the missing nodes. ii) List all phrasal nodes, iii) List all terminal nodes. iv) Identify the root node. Y} List all constituents in the diagram. s . aN, wl, a iP N = v = N | the black cat immediately chased the rat Exercise 2.2 Provide the information requested in (/-vi) in relation to the abstract tree diagram provided: i) List all the sisters of NP2. ii) List all the daughters of S‘, ii) List ali the categories of which S is the mother, iv) List all the categories that $ dominates. v) List all the categories that S’ immediately dominates. ¥}) List all the categories that are dominated by VP. vii} List all the categories at are immediately dominated by NP. ee a a comp eo Dé N v NP, BPPhrase Structure 41 Exercise 2.3 Identify the categories described in (Ly) in the sentence provided: i) The subject of the sentence il) The object of the verb iil) Tne predicate of the sentence {} The root clause y) The embedded clause John believes sincerely that Bifl knows the answer. Exercise 2.4 Assign a structure to each of the sentences in {i-v) using both tree diagrams and labelled brackets: i) Bill thinks that John is clumsy, ii) She cleverly avoided the question iii) Bill donated money to the charity. iv) The tall building is new, obviously, Vv) The claim that John is incompetent is absurd. Exercise 2.5 Sentences (} and {li] are said to be structurally ambiguous. Each sentence has two different meanings, each of which corresponds to a different structure. For each sentence, provide the two structures underlying its two different meanings: i) John saw a man with binocuiars. li) The boy called the girl from London Exercise 2.6 Co-ordination is often said to involve only constituents, that is one or a string of categories exhaustively dominated by tho same node. Explain if the sentences in {) and (ii) are consistent with this idea about co-ordination: i) Mary sent a book to Bill and a postcard to Jane. ii) John repaired and Bill washed the new car. Exercise 2.7 The sentences in (-iv) were used above as evidence in relation to constituency tests. Expiain if these sentences can be generated by the PS rules discussed in this chapter or require a new type of rules. if they require a new type of rules, explain what form these rules would take: i) This problem, the boy can salve. ii) Can the boy solve the problem?42° Phrase Structure, Lexicon and Transformations iii) Solve the problem, | wonder whether the boy will. iv) | wonder which problem the boy solved. Sources and further reading Chomsky (1957, 1965) include some of the early discussions of the general proper- ties of phrase structure and the ruies underlying it. The category AUX and its status in the phease structure of the sentence has attracted a considerable amount of actention in the literature. For an overview of the major issues and views see Akmajian e¢ af, (1979), Steele (1981) and Heny and Richards (1983). Jackendoff (1972) is the original source for most studies of the distribution of adverbs and their classification. Chomsky (1965) includes one of the earlier discussions of grammatical functions and relations in Transformational Grammar.3 Lexicon Contents 3.1 Introduction 3.2 Subcategorisation restrictions 3.3 Selectional restrictions 3.4 Separating the lexicon from synrax 3.5 Lexical derivation 3.6 Categories as Feature complexes 3.7 Summary Exercises Sources and further reading 3.1 Introduction Although the system of rules discussed in Chaprer 2 generates a large number of grammatical sentences, it also generates a large number of ungeammarical sentences, In this section, we will discuss two major respects in which the system is said to overgenerate, and chen seek possible ways of constraining it so that at Jeast certain types of ungcammatical sentences are excluded. One respect in which the system overgenerates concerns ungrammatical sentences such as (1) and (2), compared to (3): la. "The boy relied. ib. the boy [yp relied] 2a. * The boy relied the gurl 2b. the boy | yp relied {yp the girl}] 3a, The boy relied on the gisi. 3b. the boy [yp relied [pp on che girl]} (1) is excluded because the complement required by the verb is missing. (2) is excluded because the complement of the verb, though present, is not of the type required by the verb rely. The latter requires a PP complement as shown in (3} (U) and (2), though excluded, are both generated hy the system of rules discussed in Chapter 2, as shown in (4) and (5): 4a. $ > NP Aux VP 45. VP = VND) PP)44 Phrase Structure, Lexicon and Transformations 4c. NP > Der N 4dV rely de. Det > the 4fN boy, girl Sa. [§ INP [Der the] (N boy] [Aux Tense) Ivp fy relied|]] Sb. Is INP [Der the} [x boy! Laux Tense] [yp fy relied] Ip Lp the] Ly girl ill) We continue ta ignore the question of how Tense atraches to the verb until Chapter 5. (Sa) is the structure of (1), generated by the set of rules in (4). (5b} is the steucture of (2) generated by the same set of rules. Note that the options of having or not having a complement of the type NP are both allowed by rule (4b). The other respect in which the system can be said to overgenecate concerns sentences such as (6), which are grammatically sound, bur have an odd meaning. The notacion t is used here to indicate oddity of meaning, understood to be a different notion from ‘ungrammaticality’ (more on this poinr later): 6a. !The boy frightens sincerity 6b. !Sincerity kicked the box The oddity of (Ga8b) arises from the fact thar their meaning is inconsistent with our expectations in the real world. Given our knowledge of the world, we do nat expect abstract concepts such as ‘sincerity’ to be frightened or to perform the act of *kicking’. Obviously, (6a&b) would be interpretable in an imaginary world (of the type found in children's hooks, for example) but this does not alter the fact that they are odd in the context of the real world. (Ga&b) are generated by the set of rules in @a-f), and have the sceuctures shown in (8) and (3): 7a. S > NP Aux VP 7b VP > V CNP) (PP) 7 NP + Det N 74. V > frighten, kick Fe. Det > the 7£N -¥ box sincerity, girl Ba. !The boy frightens sincerity 8b. [s INP [Der the] Ln boy]) Laux Tense] fyp [y frightens] [yp sincerity]]] 9a. !Sincerity kicked the boy. Sb Is Up EN sincerity] [aux Tense] [vp Ly kicked] [Np fer the} [x boy]? ]} The reason (1) and (2) are excluded and (6a&b) are odd has ta do with the properties of individual lexical items, in particular the verb. () and (2) are excluded hecause chey are inconsistent with the fact that the verb rely requires a complement of the type PR On the other hand, (6a8cb) are odd because they are inconsistent with the fact that the verb frighten cannot be associated with an abstract object and hick cannot be associated with an abstracr subject. This suggests that the revisions required to exelude the examples in question are not likely to affect tbe PS rules which generate phrasal categories. Recall rhar cheseLexicon 48 rules were revised in the previous chaprer with the purpose of enabling them to cater for different types of verbs, Rather, the revisions required are likely to affect the cules which generate individual lexical items, chat is the rules which rewrite a terminal symbol as an appropriate lexical item. 3.2 Subcategorisation restrictions Verbs arc said to subcategorise into various sub-groups, depending on whether they require a complement, and if they do, what type of complement they require. The verb kick, for example, belongs to the sub-group of transitive verbs which require an NP complement. The verb cry on the other hand, belongs to the sub-group of verbs which do not require an NP complement. Other sub-groups include verbs which require a PP complement, e.g. rely and knock, verbs which require both an NPand a PP complement, e.g. put and give, verbs which require an S'-complement, e.g. think and believe . . . etc. ‘The subcategorisation propesties of verbs can be formally represented in terms of frames such as in (10), called subcategorisation frames: 10a. kick: [Vi—NP] 1Ob. cry: f 10c. rely: lod. put: [ 10e. chink: [ Subcategorisation frames specify the categorial class of the lexical item (the verb in (10)), and the environment in which it can occur. For example, (1(a| specifies the information that Aick ix a verb, and chat it requires a complement of the type NP. This information implies that kick can only be inserted under a V node in a VP structure where V has an ND sister. Given that subcategorisation frames specify [idiosyncratic) information relating to the properties of individual lexical items, they are associated with lexical items in their lexical entries, Thus, information concerning the subcategorisation propertics of lexical items, which is necessary for their proper use, is an addstional type of information to be added to the other types of information included in lexical entries and briefly discussed in the previous chapter. Obviously, similar subcategorisation frames cxist for other categories: nouns, adjectives and prepositions. Verbs are used here merely for illustration. Subcategorisation frames can form the basis on which a general Subcategorisa- tion rule can be set up which would make eewriting a terminal symbol as a specie lexical item sensitive to the subcategorisation properties of the lexical item. More preciscly, this rule would make it possible to rewrite 4 terminal symbol as a lexical item in association with its subcatcgorisation frame. The consequence of this rule is that a given lexical item can only be associated with a phrasal structure which is consistent with its Subcategorisation requirements. The rule in question can be formulated as in (11):46 Phrase Structure, Lexicon and Transformations » —NP] —t Mev > vi | — PP] — NP PP] —s (11) specifies the various environments in which a given verb, represented by the variable symbol ¥, can be introduced. Which frame is chosen, obviously, depends on the subcacegorisation properties of the verb which substitutes for the variable, If the verb is rely (ic. if ¥ = rely), for example, the chosen subeategorisation frame would be —PP|, and if the verb is think the chosen frame would be —$"]. and so on. With (11) incorporated into our system of rules, we guarantee that sentences such as (1) and (2), where the verb is associated wich an inappropriate subcategor- isation frame, are excluded. Because subcategorisation frames are chosen: on the basis of the Subcategorisation properties of the verb, we ensure that verbs are paired with appropriate Frames from the set specified in (11). To illustrate how the revised system works, example (12) is generared by the set of rules in (13a-g): 12. The boy relied on the girl. aS -» NP Aux VP 13b. VP > -V (ND) (PP) (S’) 130 NP > DetN 43d. V0 => rely/—PP] Le. P > on/—NP] 136 Det -» the 13g N > boy, girl Rule (13d) relating to the verb and role (13e} relating to the preposition are instantiations of two of the options specified in the more general rule (11). Because (13d) specifies the frame of the verb it generates, we ensure that the verb is associated with this frame, thereby excluding sentences such as (1) and (2). (11), of which (13d) and (13e) are instantiations, is a ‘context-sensitive” rule in the sense thar ir specifies the context in which a given category can occur. In contrast, rules (13a-c) are ‘context-free’ since they do not specify contexts. They merely list the constituents that a given phrasal category ean include. 3.3 Selectional restrictions Let us now turn to (6a8th) and see how they too can be excluded. Since the problem they pose involves contextual information as well (ue. which verds can be paired with which nours), the solution to the problem they present is likely 1 involve ‘context-sensitive’ rules as well. As pointed out above, the odd nature of (6a&b), reproduced in {14a8ch), seemsLexicon 47 ro have to de with the fact that the verb freghten is mappropriately paired with an abscract or inanimate abject and the verb kick is inappropriately paired with an abstract or inanimate subject. The solution to the problem posed by these examples will, therefore, consist of ensuring thar the verbs frighten and kick, among others, are not associated with abstract nouns im certain positions. More generally, the system should include a mechanism which ensures that only nouns with appropriate properties are associated with given verds in a given context: 14a. tThe boy frightened sincerity 14b. ‘Sincerity kicked the boy. Features such as [+/—absteacr], [+/—animate], among others, are inherent and idiosyncratic properties of nouns. Therefore, like subcategorisation properties. they are specified in the lexical entries of nouns. Thus, the lexical entey of sincerity, fot example, includes the feature [+abstract], among other features, and the lexical encry of Boy includes the feature [+animate|, among others. Lexical informanon of this type can be used to set up a ‘rewrite’ rule of the type in (15), which specifies the contexts in which a given verb can occur. (1Si) refers to the subject position and (15ii) to the object position (i) (4/~ abstract] Aux — is. [V} > YJ (ii) —~ [+/- animate] (15) is called a Selectional rule, since it specifies certain selectional restrictions associated with verbs, and presumably other categories ay well. Obviously, the version of the cule which would account for all possible combinations is much more complex than (15). The latter 1s simply an illustration of the form the rule can possibly take. With respect to the verb frighten, che corcesponding selectional rule would look. ac least in pars, roughly as in (£6). The verb frighten can take either a non-abstract subject, as in Tho girl frightened the boy, or an abstract subject, as in Smneerity frightens the boy. However, it can only take an animace object, as in The gitl frightened the boy. Consequently, (14a) is excluded, since it involves a non-animate object (i.c. sincerity), which is incompatible with the selectional restrictions of the verb frighten: {+/— abstract] Aux — 16. Vo —» frighten’ | — [tanimate] The rule corresponding to the verb kick would look roughly as in (17). The verb kick can only take an animate subject as in The girl kicked the boy, and can only take a non-abstract objcet as in The girl kicked the boyiehair. (14b) is thercfore excluded on the ground that it involves an abstract (or non-ammate) subject (ie. sincerity] which is incompatible with the selectional restrictions of the verb kick:48 Phrase Structure, Lexicon and Transformations [+animate] Aux — 17. Vo kick! abstract] Like Subcategorisation cules, the Sclectional rules (16) and (17) are ‘context- sensitive’. They specify the environment in which a given verb can occur, where the environment is the subject and object positions. One might wonder whether the problems posed by che add examples (24a&b) are of the same order as the ones posed by ungrammatical examples such as (1) and (2). In other words, ie 1s not clear whether selectional rvsteictions should be dealt with in terms of the same mechanisms (rules) which deal with subcategorisation requirements. Recati that while violations of subcategorisation requirements affect the grammatical status of the sentence, violations of sclectional restrictions do not necessarily affect the grammatical status of the sentence. Rather, they aflect the interpretation of the sentence in relation to a given world. In view of this, one could argue thar selectional restrictions involve an aspect of language (meaning and interpretation) which is different from the one involved in subcategorisation. More precisely, selectional restrictions involve a different component of the gtammar which exists over and above che component which deals with the gram- matical properties of sentences. Ic seems that we have te make a distinction between two different components of the grammar. One component comprises PS and subcategorisation rules and has the task of generating all and only grammatical sentences. We will call this component the syntactic component. The other component comprises a sct of rules, whatever their nature, which assign an interpretation co sentences generated hy the syntactic component in telation to a possible world. For the moment, we will call the component in question the semantic component, although the term ‘semantic’ may be misleading, The relationship berween the twa components is an input-output relationship, insofar as sentences generated by the syntactic com- ponent serve as mput to the semantic component which then assign, them an interpretation, This relationship partly explains the remark made carlier that although sentences such a3 (14a&b) are semantically add, they ate spnracucally sound. Such sentences are grammatically sound insofar as they are generated hy the syntactic componenz. In this chapter we will not have much to say ahoue the rules of the semantic component. Our concern is with the syatacttc properties of sentences, that is the properties which can be accounted for in terms of syntactic rules. In subsequent chapters we will learn more about the semantic component. One of the interesting conclusions which will emerge is that some of rhe mechanisms involved in assigning an interpretation to sentences are indeed syntactic in nature, and, cherefore, should be dealt wich in terms of the cules of syntax, The broad picture of grammar we have so far is une where it consists of two components. One ts the syntactic component which consists of BS rules and Subcategorisation rules and has the function of generating sentences (by making their structure explicit). The other is a semantic component which assigns anLexicon 49 interprezation to sentences generated by the syntactic component. Ie is interesting to see where the lexicon fits tn this broad picture. We turn to this question in the nest section. 3.4 Separating the lexicon from syntax Most of the revisions undertaken so far have been motivated by the twin need to accommodate further grammatical examples and exclude as many ungeammatical ones as possible. Each revision introduced results in strengthening the descriptive power of the grammar, and therefore represents step towards the goal of achiev- ing descriptive adequacy (Chapter 1}, However, our success in accommodating certain types of new sentence reyuired the introduction of new rule systems. For example, to exclude sentences where the subcategorisation properties af lexical items are not properly reflecced, we had to introduce context-sensitive Sub- caregorisation tules to be added to the existing set of context-free PS rules. Given thar the need to achieve explanatory adequacy depends on restricting the prolifera tion of rule systems (Chapter 1), it seems as though some of the steps we have caken cowards the goal of achieving descriptive adequacy have resulted in corresponding steps away from the goal of achieving explanarory adequacy: We appear to have 2 dilemma. One possibie way out of this dilemma is co atttibure some of che functions we have been artriburing co the syntactic component to other components. For example, we have seen that the function of cnsunng that the selectional restrictions of lexi 1 itetns are properly reflected can be attributed to the mantic component which assigas an interpretation to sentences generated by the syntactic component. The consequence is that the context-sensitive rules required to accomplish cus function can be climinated from the syncactic componenz. Although this move has nor resulted in the total elimination of context-sensitive rules from she syntactic component, it represents a step towards the goal of restricting the proliferation of the rule systerns it includes, In view of this, it 18 desirable ro elimmate the remain- ang set of cuntext-sensitive rules, i.e. Subcaregorisation rules, from the syntactic component, Let us see how this can be done. Given their nature as rewrite cules, Subcategotisation rules have the carious effect of equating the rewriting of phrasal categories as individual constituents with the rew:iting of terminal symbols as lexical items in association with their suhcategorisation frames. In other words, Subcategorisation rules ace syntactic in format, insofar as they are rewrite rules, but they differ from PS rules in that they make reference to lexical information. This amplics 2 view whereby the syntactic component and the lexicon are somehow merged, which docs not necessarily have to be the case. It is plausible to think of rhe lexicon as being independent of the syntactic componens, insofar as it includes information which, thongh relevant to the syntactic representation of lexical itcins, 1s essentially fexical in nature. What 1s needed, then, is a general rule that will serve as a link between the lexicun and the syntactic component, that is a rule which will have rhe function of inserting lexical50 Phrase Structure, Lexicon and Transformations tens under appropriate nodes in phrase markers generated by the PS rules of the syntactic component. Let us call the rule in question the Lexical Inser in (18): m Rule (LIR), and define it as 18. Lexical Insertion Rule Insert lexical item X under verminal nede Y, where Y corresponds to the categorial properties of X, and YP corresponds to the subcategorisation properties of X. LIR performs che operation of inserting lexical items under terminal nodes subject to two conditions. First, the terminal nde must march the categorial class of the lexical item. This will ensure that verbs are inserted under V nodes, nouns under N nodes, and so on. The second candition is chat the phrase containing the terminal node, ie. the VP of V, the NP of N. . . etc., must mazch the subcaregor isation properties of the lexical item. This means that if the lexical item is a verb which subcategorises for an NP, the VP containing V must include an NP, and if the lexical item is a verb which does not subcategorise for a NP complement, the VP containing V must not include an NP and so on. Consequently, ungrammarical sentences where lexical items are associated with inappropriate subcategorisation frames are excluded. Ie shouid be clear chat LIR performs the functions that were previously performed by Subcasegorisation rules, so that the latter can be dispensed with altogether. The syntactic component can now be thought of as consisting of one rule syscem, namely the context-free PS rules. fe might be argued that in view of the fact char LIR is differenc in nature from PS rules, we have simply replaced one rule system with another. LIR differs from PS rules in thac ic performs an operation, as pointed above, unlike PS cules. We will see in Chapter 4 that such tules are needed in the grammar for quite independent reasons. Viewed in this wider context, the move to replace Subcategorisation cules with the single LIR does not, strictly speaking, amount to replacing ane rule system with another, but to harnessing an independently needed rule system to perform additional tasks. We now consider the lexicon as being separate from the syntactic component. The postulation of separate components should, in principle, be justifiable. Preferably on the grounds that they have properties which distinguish them from other components. For example, the ausonomy of the semantic component from the synractic component is justifiable on the grounds that it includes rales of interpretation which are different in nature from the rules of syntax, Thus, we should expect the autonomy of the lexicon to be also justifiable on similar grounds. That is, we should expect the lexicon to include rules which are different in nature from the rules of syntax. We turn to this issue in the next section.Lexicon 51 3.5 Lexical derivation ur view of the lexicon so far is thar it is an unordered list of lexical entries, with each entry specifying a range of information necessary for the proper use of the lexical item. Part of this information relates to che categorial property of the item, whether it 1s a verb, a noun... . etc. Another part relates to its subcategorisation properties, whether it takes a complement or not and if it does what kind of complement it is, LIR makes crucial reference to both types of information in performing the operation of lexical insertion to ensure thar lexical items are paired with appropriate contexts. With this in mind, consider che derivation of complex categones roughly shown in (19a- 19a. translate tion — translation (Verh-to-Noun) 19h, sheepish +ish — sheepish (Noun-to-Adjective) 19c, colony +-ise —> colonise (Noun-to-Verb) (19a-c) show thar the process of forming complex categories from simpler lexical items can result in a change of the categorial class of the lexical iter which serves a the base for the derivation. Thus, the affixation of the suffix ion to a verb base results mn the derivation of a now (19a), the affixation of -ish to a noun base results tn the derivation of an adjective (196), and the affixation of ise to a noun hase results in the derivation of a verh (19¢). The derwacion of complex categories not only affects categorial properties of lexical items, but also their subcategorisation properties. As shown in (20a8cb}, the Presence of am NP complement is obligatory with the verb translate (20a), but apparently only optional with the noun translation derived from it (20b), The two categories, though derivationally related, have different subcategorisation proper ties which can be stated as in (21a&b) 20a. Mary translated *(the book). 2b. The translation (of the book) was awful. 2la. translate: [V3 —~ NP] 2th. translation: [Nj — (PP)] Now, given thar the rules of derivation affect the categorial class and the subcatcgorisation properties of lexical items, they must apply at a stage priot to theic insertion into phrase markers. This is because, as pointed out above, LIR makes crucial reference to the categorial and subcategorisation properties of lexical items to ensure that they are inserted under appropriace nodes located in appropriate contexts. Presumably, the stage at which these rules apply is the lexicon, so that the lexicon is not merely a list of lexical entries, It also includes rules of derivation. Assuming that the syntactic component does not include rules of derivation which affect the categorial and the subcategorisation properties of lexical items, the autonomy of the lexicon is justifiable on the grounds that it52 Phrase Structure, Lextcon and Transformations includes rules which are distinct in nature from the rules of syntax. More on this point later on. It is an mteresting question whecher derivationally related categories such as translate and translation have separate lexical entrics which specify their categorial and subcategorisation properties, among other things. The alternative view would be that there is only one lexical enzry for the base form, namely the verb translate, and that the complex noun translation is derived from it by a productive rule. The issue boils down to size, at least initially. The option of having lexical entries for decived forms as well as the base form implies an enormous lexicon. On the other hand, the option of having a lexical entry only for the base form and deriving the other forms from it by independently stazed rules implies 2 comparatively smaller lexicon. However, there is more to this issue than just size. If it turns out thac the process of deriving complex categories is governed by simple rules which apply Actoss categories in 2 consistent manner, these rules must form an important component of human languages. Simply lsting complex categories in the lexicon will amount to a failure to identify these rules. [n view of this, the best course to follow is to try first co see whether the derivation of complex categories is subject 10 general rules, Ir turns out thar the derivation of complex categories is indeed subject ro general rules, Judging from (19a~c) and (20a&b), the rules in question apparently have the effect of changing the categorial as well as subcategorisation properties of lexical items, For the moment, we will ignore the change in the subcategorisation proper- ties and concentrate only on the change in the cacegorial properties. The change in subcategorisation properties is discussed in more detail in Chapter 6. The suffixes listed in (19a~c) and others like rhem are consistent in more than one way. For example, when added to a base form they invariably result in the derivation of the same complex category irrespective of the base form. For example, when che suffix -1on is added to a verb, it invariably results in the decivation of a noun whatever the verb chasen. Likewise, when the suffix -ise is added to a noun, it invariably results in the derivation of a verb whatever the noun chosen. This consistent propercy of the suffixes can be explained by assuming chat, like free morphemes such as table and translate, they too are specified for categorial information. Accordingly, -ion is specified as N; -ish is specified as A, and -ise as V. A more precise representation of the complex categories in (Jac) would be as shown in (22a—c) and (23a-c). The representations in (22e-c) make use of labelled brackets and the ones in (23a~c) of tree diagraras: 22a, [py [y translate] [yy -ion]] 22b. [4 [x sheep] [a -ish]} 22c. [y [1x colony] Ty -isel] 23a. transiateLexicon 53 De. ¥ colony vise The next step 1s to explicate the process chat we have been describing as change of category. Looking at che representations in (22) and (23), it is consistently the case that the category of the derived complex is the same as the category of the suffix. In other words, it is the category of the suffix that consistently predomi- nates. For this reason, among others, the suffix is called the head of the complex category. The head of a complex category is basically the morpheme which imposes ts own properties on the derived complex form. In English, the head of a complex category is the rightmost morpheme. This can be seen more clearly in complex categories wich more than one sufféx with different catcgorial properties, eg. the adverb sheepishly The latter has the structure shown in (24): 24a. Til sheep] 14 -ish]] [apy —ly]] 24b ADV aw oN sheep -ish ty Sheepishly is a complex category of the type ADV by virtue of che fact thae its rightmost morpheme is an ADV. Sheepish is a complex category of the type A by virtue of the fact that its rightmost morpheme is an A. This property of English complex categories can be stated in terms of the rule in (25) called the Right Hand Head Rule (RHR) (adapted from Williams 1987): 25. Righthand Head Rule The head of a complex word is its rightmost morpheme. It turns out that, strictly speaking, there is no change i the category of the base form when a suffix ts added ro it. As shown in (22a-c) and (23a-c}, the base form Preserves its categorial identity The (false) impression that it undergoes a change in category is due to the RHR (25) which results m the suffix imposing. its categorial identity on the derived complex (by virtue of being the head of the derived complex) The conclusion that affixes have a categerial property can be understood t6 imply that they have lexical entries on a par with free standing lexical irems, This is already indicated by the fact thar complex categories have 2 compositional mean- ing determined by both the base and the affix, suggesting thar affixes have a sort of54 Phrase Structure, Lexicon and Transformations representation of meaning, In Chapter |, it was pointed our char che complex adjective unhappy has the compositional meaning ‘not happy’ and the complex verb nnpack has the compositional meaning ‘reverse che action of packing’. There are reasons to believe that affixes have subcategorisation propertics as well. Affixes tend co be highly selective as to the categories they can attach to, The negative prefix wn- attaches only to adjectives. Likewise, the suffix -ion usually attaches 10 verbs, whereas -hood can attach to nouns, as in boyhood, or to adjectives, as in falsehood. These idiosyncratic properties of alfixes must be stared in their lexical entries, presumably in the form of subcategurisation frames. Using the suffixes -ion and -hood for illustranon, their subcarcgorisation frames cat be represented as in (26). The laccer state the categorial identity of the suffix and the type of categories ic can attach to 26a. ion: [N; [V] —} EN] 26b, “hood: [Nz | \a [Al Subcategorisation by affixes targets lexical categories such as V, N... ete.. tather than phrasal categories such as NB PP... etc. Attaching an affx wo a base results in che cteation of a complex terminal category rather than a phrase. As such, it differs from subcategorisation by lexical categories of the type discussed earlier which targets phrasal categories and results in the creation of a phrasal category. To distinguish between the two types of subcategorisation we will refer cv subcaregorisation of lexical categories by affixes as morphological subcategorisa- tion, and co the subcategorisation of phrasal (or syntactic) categories by lexical categories as syntactic subcategorisation. The former is relevant to the rules which form complex terminal categories applying va the leaicon, and the latter 1 eules of syntax. Not all bound morphemes necessarily result in che derivation of a category different from that of the base. There are bound morphemes which consistendly never have this effect. These include, among others, the plural marker -s, the past tense marker -ed, and the third person singular present tense marker -s. The plural marker attaches to nouns, eg. book, to derive complex plural nouns, e.g, books. The past tense tarker and the third person singular present tense mazker attach to verbs, @.g, hick, to form complex verbs, c.g. kicked, kicks. These morphemes are called inflectional morphemes, to. distinguish them from the derivational morphemes discussed above Although inflectional morphemes differ from deniyazional morphemes in the way indicated, they too are subject te the RHR. The plural character of che noun buvks is due to the fact that the number feature of the suffix is plural. Likewise, the pase tense property of the verb kicked is due to the fact that the tense feature of the Tense suffix is past. This is shown more clearly in (27a&biLexicon 55 2a. Niplurel} a7. ‘slplural) | book kick The question of what exactly distinguishes inflectional morphemes from deriva- tional morphemes has generated much discussion. Tie difference mentioned above, namely that derivational affixes result in the derivation of a category distinct from the base. is not absolute. For example, neither the negative prefix nor the reversa- tive prefix results, in the derivation of a category distinet from the base. The former attaches to adjectives to derive adjectives, and the latter attaches to verbs to derive verbs. This follows from the RHR, since prefixes arc not che rightmost morphemes in the complex categories which include them. However, one would be hard-pressed to call the prefixes in question ‘inflectional morphemes’. Another perhaps more useful way of distingnishing between the two types of morpheme is in terms of whether they arc relevant to syntax of not, In Chapter 2 we discussed evidence that shows that Tense morphemes are generated under Aux separately from the verb. In Chapter $ we will see that they are joined with the verb in terms of a rule of syntax. In subsequent chapters we will sce that some cules of syntax make reference to whether a nvun, and therefore the NP that includes it, is singular or plural, among other properties associated with inflectional morphemes. This picture is consistent with che view that inflecional morphemes are the morphemes that are relevant to syntax. There is no evidence that rules of syntax make reference to affises of the type listed in (199-0). According to the scenario just outlined, it seems that the rules involved in the derivation of complex words are nat restricted to the lexicon, They can also operare in syntax, This suggests that these rules may be pact of a separate component, although one thar is not in « strice feeding relationship with the other components. Its rules can apply either in the lexicon or in syntax. We will call the component in question the morphological component. 3.6 Categories as feature complexes In the previous section, we classified the affix -fom as. a noun and the affix -ise as a verb. This may appear strange, sitice it implies that we arc putting these affixes on the same level as more familiar instances of nouns such as table and verbs such as Kuck, The use of categorial terms in relation to affixes reflects @ crucial assumption underlying these terms, namely that they are understood merely as labels for categorial features. They have no implication whatsuever as to whether the labelled item refers to (or names) an entity or an individual or denotes an event of a state. Thus, table, for example, 16 a noun not because it refers to for names) an enticy, but because it is specified for the categorial feature [+N]. Likewise, kick is a verb not because it denotes an action (or event}, but because it is specified for the categorial36 Phrase Structure, lexicon and Transformations feature [+V}. In this context, it is nor steange to classify affixes ss nouns or verbs, meaning they are specified for rhe caregonial features [FN] and [+V}. To be more precise, categorial labels are understoud as “bundies’ of categorial features. The idea 1s bortowed from feature-based phonology where segments are understood as ‘bundles’ of distinctive features. For example, the vocalic sound /i/ is a bundle’ of the features [-thigh, —low, —back, round), and /u! a ‘bundle’ of the features [+high, —low, +back, +round]. This decomposition of segments into features allows for the possibility of capturing certain common propertics between segments which are otherwise different, For example, although /i/ differs from /i/ in corms of the values of the features [+/—back] and {+/—round), itis similar to fa/ in terms of the values of the features [+/—high] and {+/—low]. The sounds /i# and ‘uf are said to form a natural class with respect co their common features. One of she reasons underlying this view is that some rules of phonology make ceference to specific features of segments, sa thar all segments which have a given feature refetred to by a giver tule will be subjcet co that rule, although the segments may differ with respect to other features. The view thar syntactic categories are also ‘bundles of features? has been motivated on similar grounds. We cannot go into the details at this early stage, as we have not yet discussed the kind of rules that have been used co justify the assumption, Here, we will simply give the feature matrix of each categorial class, which we will assume in the rest of the book. Nouns have the categorial feature complex seen in (28a), verbs have the categorial featuce complex scen in (28b), adjectives have the categorial feacure complex seen in (28c) and prepositions have the categorial fearure complex seen in (28di 283, N: |4N, —V] 28b, Vi {-N, +¥] 28. A: [+N, +V/ 28d. P: [-N, -V} The classification in (283d) assumes that the nominal and verbal features are the only primitive caregorial feacures, so that even categorics such as A and P are “bundles” of nominal and verbai features, Ns and As form « natural class in relation to the feature [+N], and Vs and Ps form a nazural class an celation to the fearure [-N], Presumably, categories which form a nacural class are expected to pastern cogether with respect to certain syntactic phenomena. The idea that categories arc essentially ‘bundles of features’ plays an umportant fole in the theory outlined in this book. This will become clear in the subsequent chapters, where additional features are intcoduced that play a crucial role in the syntactic representation of categories. hus, it 1s important ro get used ro thinking Of categories as “bundles? of features. For the moment, note that the idea applics not only to categorial features but other feacures as well. For example, we have been assuming that the Tense category 1s a bundle of the features [4+/— Tense! with 'mplied feature oppositions such as [4+/—past]. Likewise, rhe plural inflectional morpheme os nothing more than the feature (+plural], and so on, Categories as bundles of features are said to be spelled aut by the rules of