/
complementary complementary

complementary - PDF document

faustina-dinatale
faustina-dinatale . @faustina-dinatale
Follow
417 views
Uploaded On 2015-09-22

complementary - PPT Presentation

paradigms Paul Kiparsky Stanford University Paradigms that combine oneword and forms bution have large discussions of McCloskey and Poser Andrews 1990 Such paradigms lexicalist words sentences o ID: 136986

paradigms Paul Kiparsky Stanford University Paradigms that combine (one-word) and forms bution have large discussions of (McCloskey and Poser Andrews 1990). Such paradigms lexicalist words sentences o

Share:

Link:

Embed:

Download Presentation from below link

Download Pdf The PPT/PDF document "complementary" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

paradigms Paul Kiparsky Stanford University Paradigms that combine (one-word) and forms complementary bution have large discussions of (McCloskey and Poser Andrews 1990). Such paradigms lexicalist words sentences organized of grammar . They grist Distributed , a revels every of of morphology But they have even Paradigm Function Morphologists to their morphology . I shall argue, a lexicalist treatment, which the that is a device the output of the generative than directly its derivations 1996). I blocking in show in 2 how it the com- verb passive the missing inectional This Latin verb recently treated from perspective Distributed Paradigm Function Morphology . I compare to treatments argue that is two the synthetic formations yet allows their respective be and covers get 1 three morphological theories 1.1 Blocking lexicalist morphology the blocking the that is not a competing word-formation but competing expressions This a consequence any word-formation, both theories Paul' s W underlich that, offers a straightforward account the constitution of paradigms. view a grammar of two components, a generative a lter . generative — which includes syntax, lexicon, in the sense — expressions the interpretations. lter of a blocking mechanism s expressions meanings set. 1 W makes a of together dene a theory that Morphology . additional will not be 1 is thus quite different that suggested contains language-specic constraints. 1 present I any morphology both lexicalist treats a expressions “lexicalist The blocking mechanism operates competition expres- whose meaning compatible with a given input meaning of it mean- I assume competition holds only with meaning paradigmatically expressed the means. (For example, worse com- petes , but wine compete grape juice I take to be un- controversial that categories a paradigmatic others that a given category may paradigmatic one is paradigmatic German but in English). And I take to be unsolved the Pending a solution of this paradigmatic of a simply be stipulated. Compatibility will be or subsumption. Thus, blocking be- tween outputs express either the input meaning some it. This is done by two CONOMY A void complexity . XPRESS IV ES S : meaning. terms, CONOMY a markedness which equal, simplest expression 2 XPRESS IV ES S a faithfulness constraint, which equal, all the input meaning be expressed the output expression. equal' clause not the but comes ranking. 3 CONOMY XPRESS I VE NE S S gives types 4 expressive expressions, simplest is optimal. expressions, most expressive optimal. expressive unmarked expressions, these constraints make other constraints apply , variation”. XPRESS IV ES S CONOMY conict, decides. they ranked, there variation: gives a different . Cases (b) blocking: semantic blocking ing, respectively . Cases (d) two kinds of variation. An example blocking is the worse rivals, badder , more bad bad most the competing expressions be the grammatical by the blocking system. Though indeed have mean worse — that must lexicalist assumptions, worse in the lexicon, with their respective the morphology , *more bad is the syntax. 2 W e shall assume that complexity measured the morphemes, but other reasonable metrics would give the same results the cases considered here. 3 Koontz-Garboden a stochastic blocking also uses conicting markedness and faithfulness constraints. 4 Kiparsky more details justication. 2 XPRESS IV ES S CONOMY explain the distribution of the ex- follows. Consider rst forms more bad. far they mean same thing worse they compete it. The compositional forms the synonymous simple they violate CONOMY is synonyms tie on XPRES SIVENESS the competition them resolved CONOMY ? Inasmuch `worse' comparison some other, bad worse compete for meaning `worse'. 5 However, express semantic the comparative, a of XPRES SIVENESS not worse worse the is (b): being monomorphemic), they tie on CONOMY competition them resolved XPRESS I VE NE S S. forms ever they have special meanings which worse various reasons) does have. W orse have bad's meaning mean' be the lexical the comparative bad in particular be . “external” comparatives, more bad than unlucky `more than unlucky', be presumably reasons. For competing expressions, blocking. example blocking mechanism the paradigms Paradigms, this view , listed, or they emerge blocking competition expressions. W e that a F intrinsically PARADIGMA T a morpheme F `default' morpheme), that a paradigm COMPLETE a default morpheme every paper, I only discuss competition among forms, that words phrases. , role of blocking goes . Within the lexicon, blocking organizes into A morpheme is itself a micro-paradigm the case) possibly a competing contexts However, extended blocking will role in follows. T o summarize,  a expressions, rules which  results XPRESS IV E NE S S CONOMY  organizes expressions PARADIGMS Paradigm Function Morphology A very different view of blocking paradigms Stump Stump regards a rules, treats paradigms primitives of the ory . For is effected a version the familiar “Elsewhere” says special rules 5 Since the converse does hold, they not the Therefore, worse the only candidate for this the expressions considered here). 3 P ¯ . ini's two or more the same relative expression X a well-formed set  morphosyntactic narrowest takes over the inection X . (Stump The notion of itself dened terms a categories. PARADIGM a lexeme L a CELLS the Y ;  � form Y the lexeme L a complete  morphosyntactic for (Stump the Paradigm Function Morphology . For , two paradigmaticity , lled, that is lled in virtue of P ¯ . s the stipulation that relevant rules belong to the same block. A difculty the fact leaves underspecied forms paradigms. T o say that every a lexeme' s have a morphosyntactic for L effect that inectional categories optional. Y have “defective” categories one or more inectional categories. An example the injunctive in Sanskrit, a (albeit nite) verb (Kiparsky Kiparsky MS). 6 inectionally expressions compete more expressive ones — syntax of a has consequences for treatment paradigms a mixture of monomorphemic complex forms, and paradigms a mixture of synthetic forms. is from Stump' s treatment the English comparative. for blocking of worse introduces blocked a worse DEG:compar ( � 0 ;  � [1.0]DEG:compar,BAD ( � worse, � But rules monomorphemic portmanteau words, such a questionable is done in a way grounds Kiparsky 1982). PFM, only for worse output of a seems be the blocking effect the initial assumption is a As far I can nothing in the prevents listing worse a lexical , would not *badder. effect, amounts to stipulating the blocking effect, than deriving the . Paradigms that forms have unhappy for excludes it must mixed paradigms the morphology . But comparative the morphology sit the fact more (unlike , adjectival adverbial parenthetical expressions, by Poser Is successful, more one is more impressive, or expensive. more expensive) 7 6 Similar have for the imperfective classical Arabic. 4 It a more — — undertaking. More just syntactically say , many Latin below) the tenses any verb. is not possible to periphrastic formations words. Distributed Morphology distinguishing Distributed , developed by Marantz, Noyer, and movement transformational operations word- formation. Lexical , of rejects position claims words formed combining objects stems, afxes) principles. But except one important point, Distributed is not different lexical . Notably the two morphemes lexical objects intrinsic words, morphology . Paradigm Function Morphology morphemes by taking and that words have constituent (except claims the same form. For Distributed , composite paradigms would be no level; it derive English comparative paradigm. , makes available two devices for with Lexical Paradigm Function Morphology a unied fashion a empirical The rst device the UBSET RINCIP lexical insertion (Halle corresponds Paradigm Func- Morphology' s and XPRESS I VE NE S S (8) exponent a vocabulary a position if the matches a the the terminal morpheme. does take the vocabulary contains the morpheme. several vocabulary meet conditions the the of the terminal morpheme must be Distributed s device blocking effects the syntactic mechanism merger lowering). , the synthetic and is the Mixed paradigms merger . effect positing merger — in itself — Subset Principle hand, merger both unmotivated Consider the discussion of the English comparative & Noyer They more is lowered the adjective synthetic comparative, t i smart+ i provided familiar prosodic satised, 7 Note that more omissible the comparative synthetic: This better , at least expensive can't mean `This better, least expensive'. 5 is smarter than Bill. is more Bill. is Bill. d. more smart than Bill. the lowering analysis, Embick Noyer claim blocked there intervening element ( in apparently indicates a locality work. mo-st amazingly smart . . amazingly smart-est However, constituent Noyer' s argument (12) amazingly] smart [amazingly smart] is shown by the fact modies just the adjective adverb immediately follows it. The data in how most in the adverb, and has the adjective that follows. most member , most member , data in their follow assumption there no “lowering” and that (like adverbs its immediately Since failed argument Noyer offer, they have no a lowering analysis synthetic comparatives. Nor, far I know , argument this ever for any merger any the types arguments merger empirical, but non-lexicalist versions Minimalist syntax. Since implemented a way lexicalist assumptions, I there no a theory makes merger a alone a The burden should devolve solely on version the Subset Principle I have argued, 2 Periphrasis and mixed paradigms: the Latin perfect passive A lexicalist analysis T o better assess relative of the approaches morphology particular, a more complex composite conjugation Latin is subject for there explicit couched both Paradigm Function Morphology & Spencer Distributed (Embick I propose show that, in spite of the chasm that two, both rather difculties, that lexicalist provides a analysis which avoids those difculties. The the Latin passive is a formation a in inectional paradigm. following table shows the 3.Sg forms laudat and passive ¯ atur praised' the different tense/aspect 6 Non-P P assive ¯ atur `is praised' Past ¯ ¯ ¯ Future ¯ ¯ P Present ¯ laud ¯ atus/ ¯ a/um Past ¯ averat laud ¯ atus/ ¯ a/um Future ¯ averit laud ¯ atus/ ¯ a/um Lexical suggests the following straightforward (15) morphology passive inections a for this be below). construction completes the syntax. CONOMY forms forms there no synthetic What “passive”? well-known passive in Latin several functions, but detransitivizing some Passive morphology marks passive verbs. fact, nearly such verbs have obligatorily passive inection. However, a few verbs, `makes' perdit `destroys' don't take passive in the Let' s ACTIV A TANTUM passive supplied by active verbs: `becomes, made', destroyed'. verbs just passive but function (unaccusative) intransitives with no implied hand, passive is way marking lexical reexives reciprocals, belonging to the verbs grooming, equipping, Kiparsky passive marks intransitives transitive motion verbs (inchoatives), e.g. vertitur `turns, `revolve', congregitur mov ¯ etur `moves'. Some intransitives have active vehit movet `moves'. Finally , perplexingly , passive is a active but passive verbs, DEPONENT VERBS verbs shown in Non-P Present ¯ atur `exhorts' Past ¯ ¯ Future ¯ P Present ¯ atus/ ¯ a/um Past ¯ atus/ ¯ a/um Future ¯ atus/ ¯ a/um 7 verbs not just unaccusatives, such but verbs any (18)  unergatives: `complains'  transitives: sequitur `follows', ¯ atur  psych-verbs: ¯ ¯ atur `wonders' SEMI O NT verbs, have passive inection the , g ¯ ¯ sus `he has rejoiced; gaudet active form. passive inection Latin is a conjugational — we'll Passive] — lexically specied, verb inectional or unspecied. into V erb 1. Unspecied: verbs may active passive e.g. ¯ a- [+Passive]: e.g. ¯ a- `exhort' [–Passive]: , perdi- `destroy' 1. Unspecied: indifferent Part. -ns [+Passive]: passive 3.Sg. Passive [–Passive]: active 3.Sg. Active verbs unspecied Passive]. verbs any ectional they receive [–Passive] or unspecied their argument remains unmodied. [+Passive] inections trigger or more the the verb' s argument in forming passives, reexives, inchoatives, further, idiosyncratic, the verb. ¯ ans ¯ atur praised' Deponent verbs [+Passive], a [+Passive] unspecied inectional but with a [–Passive] ¯ ans `exhorting' ¯ atur `exhorts' Activa (like facit) [–Passive], and a [–Passive] or with unspecied inectional but with a passive ¯ ens done' endings undifferentiated (unspecied) Passive], hence logically compatible with types verbs, nonnite. They include the supine, the 3.p. imperative Sg. ¯ o, Pl. ¯ o, ¯ ¯ o use'. 8 From regular verbs (stems unspecied Passive]): ¯ ans (not `being ¯ atum ¯ o From (stems specied [+Passive]): ¯ ¯ ans `encour- rati ¯ ¯ ¯ ¯ atum exhort', possess' From (stems specied [-Passive]): perditum destroy', perdend ¯ o `by destroying', faciend ¯ o exceptional They more marking the Passive]. particular, verbs g ¯ av ¯ sus `rejoices', ausus have be listed inher- passive. listing is anyway , be formally irregular, Paradigm Function Morphology (one-word) forms extending the morphology was rst B ¨ V Chapman elaborated the framework Paradigm Function Morphology by Sadler Sadler two arguments against deriving forms , kind of blocking I proposed above. rst argument forms the Latin compositional, the express perfect(ive) assumption behind this argument, words combine , is Numerous consist of syntactic but noncompositional, verb-particle up even the the were fully noncompositional, it doesn't follow a morphological formation. But any argument go through more immediate that both its main compositionally derived from meanings its a `relative which the (and, , the participle meaning the a compositional its fact, a misnomer — rather, synthetic a portmanteau. the participle called temporal meaning those uses combine a a perfect. A example the Finnish the 1.Sg. forms puhu- Afrmative Negative (don't) Past spoke, didn't Perf. puhu-nut have spoken' Past 6 puhu-nut spoken' Negation expressed the , for number, combine a non-nite verb form, (which has same imperative), the 9 . Since negation + = negated past, follows = past. the formally present + , the the formally + -nut, the the Reichenbach-style treatment the a relative past exploits such a com- 8 Finnish morphemes in have the following temporal unmarked): event E extends over a t extends over “now” Past : E extends over a t w .r “now” (be-Present : E extends over a t 1 w a t 2 extends over “now” (Present) Past : E extends over a t 1 w a t 2 w “now” spoken = Pres(Past(speak)), spoken = Past(Past(speak)). Whenever par- the , status a tense obvious, Sanskrit, overt constructions shows a auxiliary combination with a tense form, line with the semantic decomposition Marathi formed combining the imperfect tense the the Deo, number gender and . Imperfect PL M dh ¯ ¯ F dh ¯ ¯ M dh ¯ ¯ ¯ ¯ F dh ¯ av-l-¯ ¯ ¯ M dh ¯ ¯ a ¯ F dh ¯ av-l-¯  ¯ ¯ N dh ¯ ¯ av-l-¯ PL M dh ¯ ¯ ¯ ¯ F dh ¯ ¯ ¯ ¯ M dh ¯ ¯ a ¯ ¯ ¯ a ¯ ¯ a-t F dh ¯ av-l-¯  ¯ ¯ ¯ a ¯ ¯ a-t M dh ¯ ¯ a ¯ ahe ¯ ¯ ahe-t F dh ¯ av-l-¯  ¯ ahe ¯ ¯ a ¯ ahe-t N dh ¯ ¯ ¯ av-l-¯  ¯ Steever Ch. 3 Dravidian parallels. Latin meaning: is a (expressing tense telic use of the like the imperfect, relative of the present sequence perfective contrasts 8 For the evidence, see Kiparsky the — the tense and activities — expressed the Imperfect (27) Imperfect: E extends over a extends over a t w .r “now”. meaning the Latin seems be indistinguishable a tense. meaning results (leaving feature uninter- Since cannot function nite verbs, a nite be participle is a predicate. Since nite verbs have tense have be which unmarked remain 9 simple meaning the Latin available with telic is due to blocking by a competing tense, imperfect. imperfect intrinsically and it pre-empts extended words, no the Latin its plain to telic is simply blocking by the XPRESS IV E S S Kiparsky further of the the different meanings the perspective. I argue there that a universal category dened above, that variety its meanings (existential, universal, resultative, stative distinct just pragmatically that they emerge from alternative of the s event the parameters dene a complex category , then follows morpho- logically simple synthetic a portmanteau folds categories . makes immediate the the (14): missing synthetic per- passive would express PRESENT PAST PASSIVE would have made the only triple portmanteau well-known, morphological complexity is one of the factors that determine distribution of paradigms. A passive obviously not impossible, but a the paradigm, is a likely for to . rst argument Sadler & Spencer, invalid if the compositional the Sadler & s argument verbs “active meaning”. be derived from a which passive content. construction part the the verb expresses of not of argument the premise that ACTIVE PASSIVE categories than categories form. For sketched above, no “active meaning” sive any relevant ASSIVE any At level of morphology , Passive] is a level of (but imperfectly) a a verbal s lexical that way highest subject) syntactic arguments. Lexical reexives, 9 Sanskrit, have the niteness requirement, past participles function as past tenses. and inchoatives lexical argument structures (Blumen- Syntactically , passive have no a s passive determined rules/constraints its syntax. Just a noun but is formally the rst declension, the argument a verb but is formally active passive. there no “passive is and uncontroversial modern , and that there no “passive meaning” uncontroversial These prove Latin they make a verbs (20)–(23). But, no syntactic or semantic Passive], Sadler & have no argument. Starting the the passive, Sadler & conversely , periphrastic block the synthetic They framework Paradigm Function Morphology , argue that Latin data framework. this paradigms Morphemes, not exist lexical have no intrinsic (Stump to regular realizational allows “transderivational” EFERRA L) make the outputs of Sadler & a “constructional” treatment the forms”, themselves “devoid meaning”. their combination means ERFECT SPECT is the diametrical of our compositional a primitive semantic Sadler & s Rules can morphology , so forms become exponents of paradigms. works like this. The perfective passive a subtype of the PREDICA T I VE ADJECTIVE CONSTRUC T N () they dene the rule below . This subtype, the PREDICA T I VE PARTICIP CONSTRUCT I , dened rule Rule denes the passive participle, a rule of denes ppc exponent the Perfective Passive, verbs passive . (M-features =Complement:[ AP . . . A[Subj . . . . ]] V T ype:Copula] predicative-participle =Complement:[ AP . . . A[m-Vform:PassPart,]] V T [m-Aspect:Imperfective] [Vform:PassPart] ) [m-Class:1/2] Given a verbal lexeme, m-feature  (excluding [m-V oice:Passive, m-Asp:Perfective]), [m-V oice:Passive, m-Asp:Perfective, ](X) = If lexeme L marked [Class:Deponent], then for , if ([Class:Deponent:Semi] & ([Class:Deponent:Full],   [m-V oice:Active] ) [m-V oice:Passive] Realization rule of introduces more Passive and overrides Thus Treating a is several First, important cross-linguistic  built on the participle the some verb form).  a present , a auxiliary way  As many is the the imperfect. Secondly , implies verbs have the same morphology passive verbs. is a rule But, have they don't. The fact nent verbs, like active verbs, but unlike passive verbs, have gerunds, imperatives. Third, notion synthetic the fact tributional which determine respective distribution always more the synthetic forms. the inventory of synthetic forms motivated by constraints words, but far I know the inventory of forms motivated by any on the forms themselves. means the inventory of forms blocking by synthetic forms, and sensibly terms the forms themselves. English comparative a simple this situation, inasmuch distribution of the synthetic governed prosodic the distribution of the afx -er. much more Sanskrit, synthetic blocked impossible missing forms by formed combining verb the a verb derived afxing - ¯ to the several where synthetic unavailable. V owel-initial roots superheavy ( ¯ VC VCC) not would the form, phonologically deviant, obvious and . ¯ . (middle) * ¯ . e ¯ . ¯ . cakre `has Secondly , monosyllabic unsufxed roots undergo polysyllabic or or more derivational sufxes (causative instead. (30) (active) ¯ am ¯ asa, or cintay ¯ am . ¯ ara that the the are distributed a different way these periphrastic constructions. synthetic a semantics, formation the the meaning. `knows', ¯ (formally perfects) known', ¯ feared' ¯ . ¯ ar-a known', ¯ am . ¯ ar-a feared' Sadler & s would apply Sanskrit they would have to be formed the morphology by a Rule the the derived formed circumstances would synthetic destroys phonological distribution of would superheavy But excellent phonological they should avoid superheavy vowel-initial syllables can't shows forms periphrastic forms, Sadler & claim, way Distributed Morphology (2000) developed a comparably elaborate the Latin Distributed . T behind it Late . F eature or or properties vocabu- the syntax; the morphology . Embick, derived from same syntactic difference dergo a process Merger, postsyntactically the main verb, adjoining the T to Asp. The merged the unmerged For verbs, they derived from same syntactic passive verbs, they difference passives a different in the syntax. The following derivation make . P assive of A small verbs both synthetic periphrastic perfects. This to expected variation the acceptability the synthetic 14 T T Asp Asp vP p �v Asp v p P t . . . t. . . Merger (movement T p �v TP T T Asp vP Asp T v p P p �v Asp T Agr t . . . t. . . T a of this kind is implement. Somehow movement aspect T that passives not undergo Merger remain when T , must that always passive Embick three formal nds difculties.  Solution 1: A Merger . the from syntactic same way any verb can from complement). But follow Merger, they This two syntax, with V ocabulary clearly  Solution 2: uninterpretable Merger . below v v external argument (in passives unaccusatives). For verbs, a the v' s complement Roots and a imposed on they be Root is massive syntactic complications.  Solution 3: Roots in the syntax. The this solution is a is incompatible with the Insertion (32)). Regardless is Distributed empirical weaknesses, interestingly enough parallel the Paradigm Function Morphology identied . above, ask it is the the I argued that is due to the compositional the Feature on the hand, merely this. The verbs have the same inectional passive verbs simply like active verbs, but unlike passive verbs, have gerundives, following examples. distinction, would be blocked “merging” categories question only in passives. not how this differences passives prove s they Sadler & Finally , the Sanskrit with late insertion. The merger the phonology and meaning a but structure syntactic movement, vid ¯ . v ¯ a  dam . ¯ knew is exactly kind of situation whose existence Distributed is exclude. T owards his briey criticizes what a lexicalist treatment the Latin would look like. of one argument: since deponent verbs have passive , lexicalism they should have passive the difference passives a node the This is not so. The [+Passive], a (quasi-conjugational) affects argument is afxation. [+Passive] a stem, its effect the verb' s argument overridden the verb' s lexical . is simply the `derived environment effect', was expressed that rules blocked they the lexical derived environment effect even more common morphology . For example, nouns inherently to the rst (stems in - ¯ a) may or feminine — of formed sufxing - ¯ a invariably serva `female ser- vant'). Similarly , Sanskrit, the sufx makes causatives verbs, k ¯ ar-ay-a=ti make' - `make', but verbs have , necessar- causatives, e.g. W ords like nauta cint-ay-a-ti — could listed — formally verbs, and accommodated lexicalist morphology suggested above They the afxes to their stems. s sole lexical falls 3 The main nding of this study is a blocking, ing in its own provides inection. Blocking organizes expressions paradigms a competition faithfulness XPRES S I VE NE S S ) markedness CONOMY From perspective, paradigms that forms not invalidate lexicalist morphology , but provide new This was argued justied empirically Latin verb inection. larger is extent the two evidence Paradigm Function Morphology Distributed . how follow and how strong the evidence a theory falsied it imposes a analysis a allows which impossible any don't have the rst of falsication. For Sadler & I have shown question consequences the question; moreover Paradigm Function Morphology Distributed lexical , that matter) have so far formulated enough even allow a Arguably we the of falsication, though. Sadler & have shown Paradigm Function Morphology express grammatical analyses periphrastic forms forms. seems be the way forms always periphrastic forms. shows Distributed ogy divorces combinatoric in a way have the analysis mixed paradigms like those of Latin Sanskrit. A lexical morphological one advocated here blocking allows phonology-morphology So, the respective Paradigm Function Morphology Distributed must be lexical remains a morphology . 17 References NDREWS A VERY . Unication Natural Language and Lin- Theory B ¨ ORJARS , ERSTI NIGEL INCENT AND AROL HAPMAN Paradigms, periphrases, and inection: a Y Morphology 1996, 155-120. E, DA . categories the Latin Linguistic Inquiry 31.185-230. H, ORRIS Prolegomena to a W ord-Formation. Linguistic Inquiry IPARSKY , AUL T in F oundations of Language 4:30-57. IPARSKY , AUL Lexical Phonology . Linguistics in the Morning Calm Korea, Seoul: Hanshin. IPARSKY , AUL and event V Y South Asian Languages and Linguistics IPARSKY , AUL Event and In David Beaver, Casil- Mart´ Stefan The Construction of . CSLI Publications. IPARSKY , AUL 2002b. Disjoint T ypology of Pronouns. Kaufmann Stiebels More W ords Studia Grammatica Berlin: Akademie V IPARSKY , AUL MS. Blocking injunctive. KA R B , NDREW . variation, Spanish progressive MS, LOSKE Y , JAMES K HALE . inection modern Natural Language and Linguistic Theory AUL ERMANN Sprachgeschichte. T ¨ ubingen: Niemeyer . OSER WILLIAM. Blocking of lexical Ivan Sag (eds.), Lexical Matters CSLI Publications. S, OUISA AND NDREW PENCER . Syntax an exponent Y Morphology 2000, S, ANFORD Analysis to Oxford: University Press. S , REGORY . morphology . A paradigm structure approach. Cambridge: CUP . UNDERLIC H, D . Minimalist morphology: the role of paradigms. In: Booij & van Marle (eds.) Y Morphology 1995, Dordrecht: Kluwer .