IN THIS ISSUE

Monthly Magazinefor Linguists

April 1999

ISSN 1381–3439

Price single issue:Dfl 14,50

Glot Internationalappears monthly,

except June & July.

© Copyright 1999 byHolland Academic Graphics.

All rights reserved

State-of-the-Article

Heidi Harley and Rolf Noyer onDistributed morphology 3“The ‘Martian’ rule of pluralformation is easy to express in aprocess-morphology: instead ofadding an affix, one simply deletesthe final syllable.”

Guest column

Lisa Lai-shen Cheng and RintSybesma on Chomsky’s UniversalGrammar under fire! 10“[L]ets not try to find out how well-in-formed the writers are.”

Book reviews

Two textbooksMaggie Tallerman’sUnderstanding syntaxreviewed by Marcel den Dikken 23Ian Roberts’ Comparative syntaxreviewed by Rint Sybesma 24

Goodies

Linguistic E-conferencesby Daniel Seely 28

Dissertations

Projecting the adjective. The syntaxand semantics of gradability andcomparisonby Chris Kennedyreviewed by Petra Hendriks 11

Prosodic wordsby Sharon Peperkampreviewed by Caroline Wiltshire 15

Word order and clausal structurein Spanish and other Romancelanguagesby Francisco Ordóñezreviewed by João Costa 19

Conference reports

Going Romance 1998 (Utrecht,December 1998)by Josep Quer 2514th Comparative GermanicSyntax Workshop (Lund, January1999)by Kleanthes K. Grohmann 26

Leaving the field

Rint Sybesma on StanislasDehaene’sThe number sense. How the mindcreates mathematics 1→18

The title of Dehaene’s book is so similar to StevenPinker’s The language instinct. How the mindcreates language (1994, Morrow, New York) that itis bound to intrigue anyone who is sympathetictowards the basic arguments for Chomsky’s lin-guistic innateness hypothesis laid out in Pinker’sbook. The differences between the titles may bemore interesting, though: not so much Dehaene’suse of the word sense where Pinker has instinct,as the fact that in Pinker’s title, the word lan-guage occurs twice, while Dehaene uses number inthe main title and mathematics in the subtitle.Does the word creates denote the same process inboth titles? Does the mind “create” mathematics inthe same way it does language?

The number sense is a very accessible andpleasantly written book on all aspects of the rela-tion between humans and numbers: numerosity inother animals, numerosity and simple calculationsin babies, the history of the expression of numberin language, the history of number notation, theneural circuitry necessary for doing arithmaticsand calculations, the localization in the brain,

The number senseHow the mind creates mathematics

by Stanislas Dehaene

1997. Oxford University Press, New York. xii + 274 pp. With index. ISBN: 0-19-511004-8. US$25 (hb.)

“The Homo sapiens brain is to formal calculation what the wing of the prehistoric bird Archaeopteryx was toflying: a clumsy organ, functional but far from optimal.” p. 134

arithmatic savants, the mathematical order of theuniverse, and so on and so forth. Dehaene writesabout all these interesting topics with an enthusi-asm which is very contagious (I don’t think I haveever seen so many exclamation marks in a popularscience book).

Some of the numerous issues related tonumber which Dehaene discusses touch on lan-guage, in one way or another. For instance, De-haene explains the well-known phenomenon thatpeople generally switch to their native languageas soon as they start calculating. The explanationis quite simple. It turns out that when calculatingwe don’t actually do very much real time calculat-ing: for many tasks we heavily rely on our memo-ry, such as for several simple algorithms(carryovers), as well as the multiplication tableswe learned by rote in school. We learn these tablesverbatim, which has the consequence of tyingthem to language — a particular language. Andevery time we dig up the arithmetic tables fromour memory, we go back to the language welearned them in.

Volume 4

ISSUE

Continued on page 18

Reference Glot International, Volume 4, Issue 4, April 1999 2

ColophonEditorsLisa Lai-Shen Cheng

U.C. IrvineRint Sybesma

HIL/Leiden University

Editorial addressHIL/Leiden University | Department of General Lin-guistics | P.O. Box 9515 | 2300 RA Leiden | The Neth-erlands | Phone +31–71–527 2104 | Fax+31–70–448 0177 | e-mail: glot@rullet.leidenuniv.nl |www.hagpub.com/glot.htm

Editorial assistantPatricia de Jonge-mail: glot@hagpub.com

Conference reports editorAndrew SimpsonDepartment of Linguistics, SOAS | University ofLondon | Thornhaugh Street | Russell Square | LondonWC1H OXG | U.K.e-mail: as4@soas.ac.uk

Goodies editorAndrew CarnieDepartment of Linguistics | University of Arizona |Douglas 200 E | Tuscon AZ 85721-0028 | U.S.A.e-mail: carnie@u.arizona.edu

Book review editorPeter SvenoniusDepartment of Linguistics | University of Tromsø |9037 Tromsø| Norwaye-mail: sven@isl.uit.no

Correspondents/Regular contributorsJoão Costa (Lisbon) | Marcel den Dikken (CUNY) |Helen de Hoop (Utrecht) | Claartje Levelt (VU Amster-dam) | Colin Phillips (Delaware) | Christine Tellier(Montréal) | Dylan W.-T. Tsai (Tsing-hua) | GuidoVanden Wyngaerd (Brussels) | Jan-Wouter Zwart(Groningen)

ColumnistsElan Dresher (Toronto) | Neil Smith (London)

Information for contributorsBefore submitting anything, please consult the Guide-lines for contributors on our web site (www.hagpub.com/glot.htm) or, if you prefer to get them by mail or e-mail,write to glot@hagpub.com.Squibs: Squibs are freely submittable; we welcome

submissions on any subject in (“theoretical”) linguis-tics. The length is limited to 1500 words, includingthe references. No footnotes!

Dissertations: If you have recently finished yourdissertation and want it to be reviewed in GlotInternational, you are invited to send a copy to theeditors.

Other: If you are thinking of contributing a State-of-the-Article, a review of a book or a dissertation, a confer-ence report, a goodie, or something else, pleasecontact the editors.

Subscription informationGlot International appears every month, except in Juneand July.

Rates for Volume 4INSTITUTIONAL RATE

electronic/pdf NLG 100.00printed version NLG 120.00

INDIVIDUAL RATE

pdf only NLG 35.00printed version (includes pdf version) NLG 50.00

For printed version, add POSTAGE: Netherlands: NLG

32.50; Europe: NLG 42.50; elsewhere: NLG 47.50; surfaceshipping in three batches (March—August—December):Netherlands: NLG 15.50; elsewhere NLG 20.50; For EUresidents, add VAT (6%).

Individual rates only apply when ordered directly fromthe publisher.

Orders must be prepaid.Preferred modes of payment: postal giro (account

number, see below), international money order orcredit card (American Express, Euro/Master/Access,Visa). Credit card orders must include the cardholder’s complete billing address.

For payment by international bank cheque or banktransfer, add NLG 22.50.

PublisherHolland Academic GraphicsZilverstraat 58 | 2551 KS The Hague | TheNetherlands | Phone: +31–70–3595110 |Fax: +31–70–3595044 | e-mail: astrid@hagpub.com |www.hagpub.comBank relations: ING 657370581 |Postbank/postal giro 6261107VAT number: NL 0588.61.798.B01

AdvertisementsContact the publisher.

PrintingPrinted in the Netherlands by Ars Nova

Contributors to this issue

João CostaUniversidade Nova de Lisboa | FCSH | Av. de Berna26-C | 1050 Lisboa | Portugala.j.costa@mail.telepac.pt

Marcel den DikkenLinguistics Program, Graduate Center | The CityUniversity of New York | 33 West 42 Street | NewYork, NY 10036-8099 | U.S.A.mdikken@email.gc.cuny.edu

Kleanthes K. GrohmannUniversity of Maryland | Department of Linguistics| College Park, MD 20742 | U.S.A.grohmann@wam.umd.edu

Heidi HarleyInstitute for Research in Cognitive Science | Univer-sity of Pennsylvania | 3401 Walnut St., 400A |Philadelphia PA 19119 | U.S.A.hharley@ling.upenn.edu

Petra HendriksCognitive Science and Engineering (TCW) |Department of Dutch | University of Groningen |Grote Kruisstraat 2/1 | 9712 TS Groningen | TheNetherlandsp.hendriks@let.rug.nl

Christopher KennedyDepartment of Linguistics | Northwestern Universi-ty | 2016 Sheridan Road | Evanston, IL 60208 |U.S.A.kennedy@ling.nwu.edu

Rolf NoyerUniversity of Pennsylvania | Department of Linguis-tics | 619 Williams Hall | Philadelphia, PA 19104-6305 | U.S.A.rnoyer@ling.upenn.edu

Francisco OrdóñezDepartment of Spanish, Italian and Portuguese |University of Illinois at Urbana-Champaign | 4080Foreign Languages Building, MC-176, 707 S.Mathews Ave. | Urbana, IL 61801 | U.S.A.fordonez@staff.uiuc.edu

Sharon PeperkampLaboratoire de Sciences Cognitives et Psycholinguis-tique | CNRS-EHESS | 54, Bd. Raspail | 75270 ParisCedex 6 | Francesharon@lscp.ehess.fr

Josep QuerDepartament de Filologia Catalana | UniversitatAutonoma de Barcelona, Facultat de Lletres |Edifici B | E-08193 Bellaterra | Spainjosep.quer@let.ruu.nl

Daniel SeelyDepartment of English | Eastern Michigan Universi-ty | Ypsilanti Michigan 48197 | U.S.A.dseely@emunix.emich.edu

Caroline R. WiltshireProgram in Linguistics | University of Florida,Gainesville | Box 115454 | Gainesville, FL 32611-5454 | U.S.A.wiltshir@lin.ufl.edu

Dissertations reviewed

Christopher Kennedy, Projecting the adjective: thesyntax and semantics of gradability and comparison.Degree date: September 1997. University of Califor-nia, Santa Cruz. Supervisors: Donka Farkas, WilliamLadusaw. 243 pp. Available through Slug Pubs,Department of Linguistics, UCSC; email:slugpubs@ling.ucsc.edu, http://ling.ucsc.edu/slugpubs.Price: $12.00.

Sharon Peperkamp, Prosodic words. Degree date:December 1997. HIL/University of Amsterdam.Supervisors: Geert Booij, Marina Nespor. 231 pp.ISBN: 90-5569-035-x. Available from Holland Academ-ic Graphics www.hagpub.com. Price: HFL 40 (excl.VAT, P&P; for individuals ordering directly fromHAG only).

Francisco Ordóñez, Word order and clausal structurein Spanish and other Romance languages. Degreedate: October 1997. CUNY. Supervisor: RichardKayne. 225 pp. Available from UMI Microfilm, AnnArbor, Michigan.

Books reviewed

Ian Roberts, Comparative syntax. 1997. x+307 pp.With index. London: Arnold. ISBN: 0 340 59286 9(pb.), 0 340 67659 0 (hb.). Price: GBP 16.99 (pb.),45.00 (hb.).

Maggie Tallerman, Understanding syntax. 1998.xiv+226 pp. With index. London: Arnold. Pb. ISBN: 0340 60377 1. Price: GBP 13.99.

BUGGER OFFMy ears areclogged up!

I�m only yawning ...

... so that Ican hear you

better!

artist

: Er

ic B

roer

s

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 3

Whenever a major revision to the architecture ofUG is proposed, it takes some time for sufficientwork to accumulate to allow evaluation of theviability of the proposal, as well as for its broadoutlines to become familiar to those not immediate-ly involved in the investigation. The introductionof Distributed Morphology (henceforth DM) in theearly 1990s, by Morris Halle and Alec Marantz, isa case in point. In the four-year period since thefirst paper outlining the framework appeared, areasonably substantial body of work has ap-peared, addressing some of the key issues raisedby the revision. The goal of this article is to intro-duce the motivation and core assumptions for theframework, and at the same time provide somepointers to the recent work which revises andrefines the basic DM proposal and increases itsempirical coverage. Since the particular issues wediscuss cover such a broad range of territory, wedo not attempt to provide complete summaries ofindividual papers, nor, for the most part, do weattempt to relate the discussion of particular issuesto the much broader range of work that has beendone in the general arena. What we hope to do isallow some insight into (and foster some discussionof) the attitude that DM takes on specific issues,with some illustrative empirical examples.

This article is organized as follows. Section 1sketches the layout of the grammar and discussesthe division of labor between its components. The“distributed” of Distributed Morphology refers tothe separation of properties which in other theo-ries are collected in the lexicon, and in section 1we elaborate on the motivation for this separationand its particulars. Section 2 explicates the me-chanics of Spell-Out, giving examples of competi-tion among phonological forms from Dutch,introducing the notion of f-morpheme and l-morpheme and distinguishing allomorphy fromsuppletion with examples from English. Section 3discusses the operations which are available in theMorphology component, addressing in turn Mor-phological Merger, Impoverishment and Fission,with examples from Latin, Serbo-Croatian, Norwe-gian and Tamazight Berber. We also provide anillustration of the contrast between a “piece-based”theory such as DM and process-based morphologi-cal theories. Section 4 treats the relationship of thesyntax to the morphology, Separationism and itslimitations, the ways in which a mismatch be-tween syntactic terminal nodes and morphosyn-tactic features may arise, and the distinct types ofsyntax/morphology mismatches conventionallyclassified as cliticization. We conclude in Section 5with an agenda for future research.

1. The structure of the grammarThere are three core properties which distin-

guish Distributed Morphology from other morpho-logical theories: Late Insertion, Underspecification,and Syntactic Hierarchical Structure All the WayDown. The grammar, still of the classic Y-type, issketched in (1).

Unlike the theory of LGB (Chomsky 1981) and itsLexicalist descendants, in DM the syntax properdoes not manipulate anything resembling lexicalitems, but rather, generates structures by combin-ing morphosyntactic features (via Move andMerge) selected from the inventory available,subject to the principles and parameters governingsuch combination.

Late Insertion refers to the hypothesis thatthe phonological expression of syntactic terminalsis in all cases provided in the mapping to Phono-logical Form. In other words, syntactic categories

are purely abstract, having no phonological con-tent. Only after syntax are phonological expres-sions, called Vocabulary Items, inserted in aprocess called Spell-Out. It is further worth notingthat this hypothesis is stronger than the simpleassertion that terminals have no phonologicalcontent: as we will see below, there is essentiallyno pre-syntactic differentiation (other than,perhaps, indexing) between two terminal nodeswhich have identical feature content but willeventually be spelled out with distinct VocabularyItems such as dog and cat.

Underspecification of Vocabulary Itemsmeans that phonological expressions need not befully specified for the syntactic positions wherethey can be inserted. Hence there is no need forthe phonological pieces of a word to supply themorphosyntactic features of that word; rather,Vocabulary Items are in many instances defaultsignals, inserted where no more specific form isavailable.

Syntactic Hierarchical Structure All theWay Down entails that elements within syntaxand within morphology enter into the same typesof constituent structures (such as can be dia-grammed through binary branching trees). DM ispiece-based in the sense that the elements of bothsyntax and of morphology are understood asdiscrete constituents instead of as (the results of)morphophonological processes.

1.1. The Lexicalist Hypothesis and DMThere is no lexicon in DM in the sense famil-

iar from generative grammar of the 1970s and1980s. In other words, DM unequivocally rejectsthe Lexicalist Hypothesis. The jobs assigned to theLexicon component in earlier theories are distrib-uted through various other components. Forlinguists committed to the Lexicalist Hypothesis,

this aspect of DM may be the most difficult toaccept, but it is nevertheless a central tenet of thetheory. (For discussion of this issue from a Lexical-ist viewpoint, see Zwicky & Pullum 1992.)

The fullest exposition of the anti-Lexicaliststance in DM is found in Marantz (1997a). There,Marantz argues against the notion of a generativelexicon, adopted in such representative examplesof the Lexicalist Hypothesis as Selkirk (1982) orDiSciullo and Williams (1987), using argumentsfrom the very paper which is usually taken to bethe source of the Lexicalist Hypothesis, Chomsky’s(1970) ‘Remarks on Nominalization’. Marantzpoints out that it is crucial for Chomsky’s argu-ment that, for instance, a process like causativiza-tion of an inchoative root is syntactic, not lexical.Chomsky argues that roots like grow or amusemust be inserted in a causative syntax, in order toderive their causative forms. If their causativeforms were lexically derived, nothing shouldprevent the realization of the causativized stem ina nominal syntax, which the poorness of *John’sgrowth of tomatoes indicates is impossible. Otherlexicalist assumptions about the nature of lexicalrepresentations, Marantz notes, are simply un-proven: no demonstration has been made of corre-spondence between a phonological “word” and aprivileged type of unanalyzable meaning in thesemantics or status as a terminal node in thesyntax, and counterexamples to any simplisticassertion of such a correspondence are easy tofind.

Because there is no lexicon in DM, the term“lexical item” has no significance in the theory, norcan anything be said to “happen in the lexicon”,and neither can anything be said to be “lexical” or“lexicalized”. Because of the great many taskswhich the lexicon was supposed to perform, theterms “lexical” and “lexicalized” are in fact ambig-uous. (For a discussion of terminology, see Aronoff1994). Here we note a few of the more usualassumptions about lexicalization, and indicatetheir status in the DM model:

I Lexical(ized) = Idiomatized. Because thelexicon was supposed to be a storehouse forsound-meaning correspondences, if an expres-sion is conventionally said to be “lexicalized”the intended meaning may be that the ex-pression is listed with a specialized meaning.

DISTRIBUTED MORPHOLOGY

Heidi Harley & Rolf Noyer

Morphosyntactic features:

[Det] [1st] [CAUSE] [+pst][Root] [pl]

etc...

Syntactic Operations(Merge, Move, Copy)

List A

List B

Morphological Operations Logical Form

Phonological Form(Insertion of Vocabulary Items,

Readjustment, phonological rules)

Vocabulary Items/dog/: [Root] [+count] [+animate] ...

/-s/: [Num] [pl] .../did/: [pst] ...

etc...

Conceptual Interface(“Meaning”)

List C Encyclopedia(non-linguistic knowledge)

dog: four legs, canine, pet, sometimes bitesetc... chases balls, in environment “let sleeping ____s

lie”, refers to discourse entity who is better left alone...

cat: four legs, feline, purrs, scratches, inenvironment “the ___ out of the bag” refers

to a secret ... etc...

(1)

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 4

In DM such an expression is an idiom andrequires an encyclopedia entry (see 1.4).There is no “word-sized” unit which has aspecial status with respect to the idiomatiza-tion process; morphemes smaller than word-size may have particular interpretations inparticular environments, while expressionsconsisting of many words which obviouslyhave a complex internal syntax may equallybe idiomatized.

II Lexical(ized) = Not constructed by syntax. Theinternal structure of expressions is demonstra-bly not always a product of syntactic opera-tions. In DM structure is produced both insyntax and after syntax in the Morphologycomponent (labeled Morphological Operationsabove). Nevertheless, because of SyntacticHierarchical Structure all the Way Down,operations within Morphology still manipulatewhat are essentially syntactic structuralrelations. The syntactic component produces arepresentation whose terminal elements aremorphosyntactic features, which is thensubject to operations such as Merger UnderAdjacency, Fission or Fusion, accounting fornon-isomorphic mappings from syntacticterminals to morphophonological constituents.

III Lexical(ized) = Not subject to exceptionlessphonological processes, i.e., part of “lexical”phonology in the theory of Lexical Phonologyand Morphology (Kiparsky 1982 et seq.). InDM the distinction between two types ofphonology — “lexical” and “postlexical” — isabandoned. All phonology occurs in a singlepost-syntactic module. While Lexical Phonolo-gy and Morphology produced many importantinsights, DM denies that these results requirean architecture of grammar which dividesphonology into a pre-syntactic and post-syntactic module (see also Sproat 1985).Rather, post-syntactic phonology itself mayhave a complex internal structure (Halle &Vergnaud 1987).

1.2. The status of Vocabulary Items and thelexical/functional distinctionIn DM, the term morpheme properly refers

to a syntactic (or morphological) terminal nodeand its content, not to the phonological expressionof that terminal, which is provided as part of aVocabulary Item. Morphemes are thus the atomsof morphosyntactic representation. The content ofa morpheme active in syntax consists of syntacti-co-semantic features drawn from the set madeavailable by Universal Grammar.

A Vocabulary Item is, properly speaking, arelation between a phonological string or “piece”and information about where that piece may beinserted. Vocabulary Items provide the set ofphonological signals available in a language forthe expression of abstract morphemes. The set ofall Vocabulary Items is called the Vocabulary.

(2)Vocabulary Item schemasignal ←→ context of insertion

Example Vocabulary Itemsa. /i/ ←→ [___, +plural]

A Russian affix (Halle 1997)b. /n/ ←→ [___, +participant +speaker, plural]

A clitic in Barceloni Catalan (Harris 1997a)c. /y-/ ←→ elsewhere

An affix in the Ugaritic prefix conjugation (Noyer 1997)d. Ø ←→ 2 plu

A subpart of a clitic in Iberian Spanish (Harris 1994)e. /kæt/ ←→ [DP D [LP ____ ]]

Root inserted in a nominal environment (Harley & Noyer1998a)

Note that the phonological content of a VocabularyItem may be any phonological string, includingzero or Ø. The featural content or context ofinsertion may be similarly devoid of information:in such cases we speak of the default or “else-where” Vocabulary Item. Note that the two do notnecessarily coincide — that is, a null phonologicalaffix in a given paradigm is not necessarily thedefault Vocabulary Item. For example, the zeroplural affix inserted in the context of marked

English nouns like sheep is not the English defaultplural.

In early work in DM, Halle (1992) proposed adistinction between concrete morphemes, whosephonological expression was fixed, and abstractmorphemes, whose phonological expression wasdelayed until after syntax. More current work inDM, however, endorses Late Insertion of all pho-nological expression, so Halle’s earlier concrete vs.abstract distinction has been largely abandoned.

Harley & Noyer (1998a) propose an alterna-tive to the concrete vs. abstract distinction; theysuggest that morphemes are of two basic kinds: f-morphemes and l-morphemes, correspondingapproximately to the conventional division be-tween functional and lexical categories, or closed-class and open-class categories.

F-morphemes are defined as morphemes forwhich there is no choice as to Vocabulary inser-tion: the spell-out of an f-morpheme is determinis-tic. In other words, f-morphemes are those whosecontent (as defined by syntactic and semanticfeatures made available by Universal Grammar)suffices to determine a unique phonological ex-pression. One prediction is that Vocabulary Itemsconventionally classified as “closed-class” shouldeither express purely grammatical properties orelse have meanings determined solely by univer-sal cognitive categories (see 2.3 for further discus-sion).

In contrast, an l-morpheme is defined as onefor which there is a choice in spell-out: an l-mor-pheme is filled by a Vocabulary Item which maydenote a language-specific concept. For example,in an l-morpheme whose syntactic position wouldtraditionally define it as a noun, any of the Vocab-ulary Items dog, cat, fish, mouse, table etc. mightbe inserted. Note that because the conventionalcategorial labels noun, verb, adjective etc. are byhypothesis not present in syntax (l-morphemesbeing acategorial), the widely adopted hypothesisthat Prosodic Domain construction should beoblivious to such distinctions (Selkirk 1986; Chen1987) follows automatically.

1.3. The syntactic determination of lexicalcategoriesThe conjecture we have just alluded to, which

we will term the L-Morpheme Hypothesis, (Ma-rantz 1997a; Embick 1997, 1998a, 1998b; Harley1995; Harley & Noyer 1998a, 1998b; Alexiadou1998), contends that the traditional terms forsentence elements, such as noun, verb, and adjec-tive, have no universal significance and are essen-tially derivative from more basic morpheme types(see also Sapir 1921, ch. 5). As noted above, Ma-rantz (1997a) contends that the configurationaldefinition of category labels is already implicit inChomsky (1970).

Specifically, the different “parts of speech” canbe defined as a single l-morpheme, or Root (toadopt the terminology of Pesetsky 1995), in cer-tain local relations with category-defining f-morphemes. For example, a noun or anominalization is a Root whose nearest c-com-manding f-morpheme (or licenser) is a Determiner,a verb is a Root whose nearest c-commandingf-morphemes are v, Aspect and Tense; withoutTense such a Root is simply a participle (Embick1997; Harley & Noyer 1998b). Thus, the sameVocabulary Item may appear in different morpho-logical categories depending on the syntacticcontext that the item’s l-morpheme (or Root)appears in. For example, the Vocabulary Itemdestroy is realized as a noun destruct-(ion) whenits nearest licenser is a Determiner, but the sameVocabulary Item is realized as a participledestroy-(ing) when its nearest licensers are Aspectand v; if Tense appears immediately above Aspect,then the participle becomes a verb such as destroy-(s). However, it is probably the case that manytraditional part-of-speech labels correspond tolanguage-specific features present after syntaxwhich condition various morphological operationssuch as Impoverishment (see 3.2) and VocabularyInsertion.

1.4. Idioms: the content of the EncyclopediaIn DM, the Vocabulary is one list which

contains some of the information which in lexical-ist theories is associated with the Lexicon. Anothersuch list is the Encyclopedia, which relates Vocab-ulary Items (sometimes in the context of otherVocabulary Items) to meanings. In other words,the Encyclopedia is the list of idioms in a lan-guage.

The term idiom is used to refer to any expres-sion (even a single word or subpart of a word)whose meaning is not wholly predictable from itsmorphosyntactic structural description (Marantz1995, 1997a). F-morphemes are typically notidioms, but l-morphemes are always idioms.

(3)Some idiomscat (a fuzzy animal)(the) veil (vows of a nun)(rain) cats and dogs (a lot)(talk) turkey (honest discourse)

The notion of “idiom” in DM, obviously, embracesmore than the conventional use of the term im-plies. Idioms in the conventional sense — that is,groups of words in a particular syntactic arrange-ment that receive a “special” interpretation, forexample kick the bucket, whose meaning is rough-ly ‘die’ — are represented in DM as subparts of theEncyclopedic entry for the Root (or Roots) whichare involved. The Encyclopedia entry for kick, forexample, will specify that in the environment ofthe direct object the bucket, kick may be interpret-ed as ‘die’. The study of conventional idioms hasbeen an important source of evidence for localityrestrictions on interpretation in DM; in particular,following the observations of Marantz (1984), thefact that external arguments are never includedas part of the contextual conditioning of Roots inconventional idioms has led to the proposal where-by external arguments are projected by a separate“little-v” head, not by any Root, and they thus arenot mentioned by Encyclopedia entries for Rootsas a possible interpretive conditioner. (For analternative, non-DM discussion of idioms, seeJackendoff 1997.)

As indicated in the schema in (1) above, the“meaning” of an expression is interpreted from theentire derivation of that expression, including theinformation from the Encyclopedia which is con-sidered extralinguistic. LF does not express orrepresent meaning; LF is merely a level of repre-sentation which exhibits certain meaning-relatedstructural relations, such as quantifier scope. (Therelationship of the Encyclopedia to the Vocabularyis the topic of much current debate, see, for exam-ple, Marantz 1997a; Harley & Noyer 1998a).

2. Spell-OutSpell-Out inserts Vocabulary Items (phonolog-

ical pieces) into morphemes. In the unmarkedcase, the relation between Vocabulary Items andmorphemes is one-to-one, but as we have seen,several factors may disrupt this relation (Noyer1997), including fission of morphemes, removal ofmorphosyntactic features by Impoverishment,local displacements of Vocabulary Items by Mor-phological Merger and post-syntactic insertion ofdissociated morphemes.

Spell-Out works differently depending onwhat type of morpheme is being spelled out, f-morphemes or l-morphemes. Regardless of thetype of morpheme, however, Spell-Out is normallytaken to involve the association of phonologicalpieces (Vocabulary Items) with abstract mor-phemes. Halle (1992) construes Spell-Out as therewriting of a place-holder “Q” in a morpheme asphonological material. This operation is normallyunderstood as cyclic, such that more deeply em-bedded morphemes are spelled-out first.

2.1. Spell-Out of f-morphemes: the SubsetPrincipleEarly work in DM was focused primarily on

the spell-out of f-morphemes. In such cases sets ofVocabulary Items compete for insertion, subject towhat Halle (1997) called the Subset Principle

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 5

(Lumsden 1987, 107 proposes a similar principleand calls it “Blocking”. Halle’s principle is not to beconfused with the Subset Principle of Manzini &Wexler 1987, which deals with learnability issues).

Subset PrincipleThe phonological exponent of a Vocabulary Item is inserted intoa morpheme... if the item matches all or a subset of thegrammatical features specified in the terminal morpheme.Insertion does not take place if the Vocabulary Item containsfeatures not present in the morpheme. Where severalVocabulary Items meet the conditions for insertion, the itemmatching the greatest number of features specified in theterminal morpheme must be chosen.

Below, we give an example from Sauerland(1995).

(4)a. Dutch strong adjectival desinences

[–neuter] [+neuter][–pl] -e Ø[+pl] -e -e

b. Vocabulary ItemsØ ←→ [ ___ , +neuter -plural] / Adj + ____-e ←→ Adj + ____

In Dutch, after syntax, a dissociated morpheme(see section 3) is inserted as a right-adjunct ofthose morphemes which are conventionally la-beled “adjectives”. The Vocabulary Items abovecompete for insertion into this morpheme. In thespecific environment of the neuter singular, Ø isinserted. In the remaining or elsewhere environ-ment -e is inserted. The insertion of Ø in thespecific environment bleeds the insertion of -ebecause, under normal circumstances, only asingle Vocabulary Item may be inserted into amorpheme. Note that the Vocabulary Items aboveare not specially stipulated to be disjunctive exceptinsofar as they compete for insertion at the samemorpheme.

Note that all Vocabulary Items may competefor insertion at any node; there is no pre-insertionseparation of Vocabulary Items into “related”forms which may compete. However, since theinsertion process is restricted by feature content, acertain collection of Vocabulary Items correspond-ing to the traditional notion of a “paradigm” maybe the set under discussion when accounting forthe phonological realization of a given terminalnode. In some theories certain such collectionshave a privileged status or can be referred to bystatements of the grammar (Carstairs 1987;Wunderlich 1996). But in DM, paradigms, likecollections of related phrases or sentences, do nothave any status as theoretical objects, althoughcertain regularities obtaining over paradigms mayresult from constraints operating during languageacquisition.

2.2. Feature Hierarchies, FeatureGeometries and the Subset PrincipleIn some cases it would be possible to insert

two (or more) Vocabulary Items into the same f-morpheme, and the Subset Principle does notdetermine the winner. Two approaches have beenproposed in DM for such cases. Halle & Marantz(1993) suggest that such conflicts are resolved byextrinsic ordering: one Vocabulary Item is simplystipulated as the winner. Alternatively, Noyer(1997) proposes that such conflicts can always beresolved by appeal to a Universal Hierarchy ofFeatures (cf. also Lumsden 1987, 1992; Zwicky1977 and Silverstein 1976). Specifically, the Vo-cabulary Item that uniquely has the highestfeature in the hierarchy is inserted.

(5)Fragment of the Hierarchy of Features1 person > 2 person > dual > plural > other features

Harley (1994), following a proposal of Bonet(1991), argues that the conflict-resolving effects ofthe Feature Hierarchy can be derived from ageometric representation of morphosyntacticfeatures, according to which the Vocabulary Itemwhich realizes the most complex feature geometryis inserted in such situations. See also section 3.2on Impoverishment, below.

2.3. Spell-Out of l-morphemes: competition,suppletion and allomorphyFor l-morphemes there is a choice regarding

which Vocabulary Item is inserted. For example, aRoot morpheme in an appropriately local relationto a Determiner might be filled by cat, dog, house,table or any other Vocabulary Item we wouldnormally call a noun. Harley & Noyer (1998a)note that it is clear that such Vocabulary Items arenot in competition, as are the Vocabulary Itemsinserted into f-morphemes. Rather, these Vocabu-lary Items can be freely inserted at Spell-Out,subject to conditions of licensing. Licensers aretypically f-morphemes in certain structural rela-tions to the Root where the Vocabulary Item isinserted, and as outlined above, these structuralrelations typically determine the traditional notionof category. Nouns are licensed by an immediatelyc-commanding Determiner; different verb classes,such as unergatives, unaccusatives, and transi-tives each are licensed by different structuralconfigurations and relations to various highereventuality projections.

Marantz (1997a) discusses the interestingcase of l-morphemes which undergo apparentallomorphy in different environments, such as therise/raise alternation. These pose a problem inthat they appear to be in competition for insertionin different environments (that is, raise is insertedin the context of a commanding CAUSE head,while rise, the intransitive and nominal variant, isthe elsewhere case). They cannot be separateVocabulary Items, however, for if they were, raiseshould be a separate verb with the properties ofthe destroy class. The absence of nominalizationslike *John’s raise of the pig for bacon, however,indicate that raise is simply a morphophonologicalvariant of the basic intransitive rise root, which isa member of the grow class. That is, in DM, l-morpheme alternations like rise/raise must not bedetermined by competition, as may be the case forallomorphy of f-morphemes, but rather must bethe product of post-insertion readjustment rules.

DM, then, must recognize two different typesof allomorphy: suppletive and morphophonologi-cal. Suppletive allomorphy occurs where differentVocabulary Items compete for insertion into an f-morpheme, as outlined in section 2.1 above. Togive another example, Dutch nouns have (atleast) two plural number suffixes, -en and -s. Theconditions for the choice are partly phonologicaland partly idiosyncratic. Since -en and -s are notplausibly related phonologically, they must consti-tute two Vocabulary Items in competition.

Morphophonological allomorphy occurs wherea single Vocabulary Item has various phonologi-cally similar underlying forms, but where thesimilarity is not such that phonology can be direct-ly responsible for the variation. For example,destroy and destruct- represent stem allomorphs ofa single Vocabulary Item; the latter allomorphoccurs in the nominalization context. DM hypothe-sizes that in such cases there is a single basicallomorph, and the others are derived from it by arule of Readjustment. The Readjustment in thiscase replaces the rime of the final syllable ofdestroy with -uct. (Alternatively such allomorphsmight both be listed in the Vocabulary and berelated by “morpholexical relations” in the sense ofLieber 1981.)

Traditionally it is often thought that there is agradient between suppletion and other types ofmore phonologically regular allomorphy, and thatno reasonable grounds can be given for how todivide the two or if they should be divided at all.Marantz (1997b) has recently proposed that truesuppletion occurs only for Vocabulary Items incompetition for f-morphemes, since competitionoccurs only for f-morphemes. An immediate conse-quence of this proposal is that undeniably supple-tive pairs like go/went or bad/worse must actuallyrepresent the spelling of f-morphemes. The class off-morphemes is as a result considerably enriched,but since the class of f-morphemes is circumscribedby Universal Grammar, it is also predicted thattrue suppletion should be limited to universalsyntactico-semantic categories. Moreover, given

that some independent grounds might in this waydivide suppletive from Readjustment-drivenallomorphy, a theory of the range of possibleReadjustment processes becomes more feasible.

The controversial distinction between deriva-tional and inflectional (Anderson 1982) has noexplicit status in DM. However, the distinctionbetween f-morphemes and l-morphemes perhapscaptures some of the intuition behind the deriva-tional/inflectional distinction, although certainlynot all f-morphemes would normally be considered“inflectional”. DM also distinguishes betweensyntactic and non-syntactic (dissociated) mor-phemes, although again this distinction has nostraightforward analogue in the derivational/inflectional debate.

3. Manipulating structured expressions:morphological operationsIn DM any given expression acquires at least

two structural descriptions during its derivation.In a morphophonological description, an expres-sion’s phonological pieces (its Vocabulary Items)and their constituent structure are displayed. In amorphosyntactic description, an expression’smorphemes and their constituent structure aredisplayed.

(6)The expression cows:

Morphosyntactic description: [Root [+plural]]Morphophonological description: [kaw+ z]

The morphosyntactic structure of an expression isgenerated by several mechanisms. Syntax, usingconventional operations such as head-movement,plays a major role in constructing morphosyntacticstructures, including “word”-internal structure.But in addition, DM employs several additionalmechanisms in a post-syntactic component, Mor-phological Structure.

First, morphemes such as [passive] or [case](in some instances, see Marantz 1991) which, byhypothesis, do not figure in syntax proper, can beinserted after syntax but before Spell-Out. Thesemorphemes, which only indirectly reflect syntacticstructures, are called dissociated morphemes.For a full exposition of the mechanism of dissociat-ed morpheme insertion, see Embick (1997).

Second, the constituent structure of mor-phemes can be modified by Morphological Merger,which can effect relatively local morpheme dis-placements.

3.1. MergerMorphological Merger, proposed first in Ma-

rantz (1984), was originally a principle of well-formedness between levels of representation insyntax. In Marantz (1988, 261) Merger wasgeneralized as follows:

Morphological MergerAt any level of syntactic analysis (d-structure, s-structure,phonological structure), a relation between X and Y may bereplaced by (expressed by) the affixation of the lexical head of Xto the lexical head of Y.

What Merger does is essentially “trade” or “ex-change” a structural relation between two ele-ments at one level of representation for a differentstructural relation at a subsequent level. (Re-bracketing under adjacency is also proposed anddiscussed at length in Sproat 1985.)

Merger has different consequences dependingupon the level of representation it occurs at.Where Merger applies in syntax proper it is essen-tially Head Movement, adjoining a zero-levelprojection to a governing zero-level projection(Baker 1988). Cases of syntactic lowering may bea type of Merger as well, presumably occurringafter syntax proper but before Vocabulary Inser-tion, e.g. the Tense to verb affixation in English(see Bobaljik 1994) or perhaps C-to-I lowering inIrish (McCloskey 1996).

The canonical use of Merger in Morphology isto express second-position effects. Embick & Noyer(in progress) hypothesize that where Mergerinvolves particular Vocabulary Items (as opposed

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 6

to morphemes), the items in question must bestring-adjacent. Such cases of Merger are calledLocal Dislocation. Schematically Local Dislocationlooks like this:

(7)Local Dislocation:X [Y ... ] → [Y + X ...

In Local Dislocation, a zero-level element trades itsrelation of adjacency to a following constituentwith a relation of affixation to the linear head(peripheral zero-element) of that constituent.(Local Dislocation has also received considerableattention outside of DM from researchers workingin Autolexical Syntax, see Sadock 1991.)

For example, Latin -que is a second-positionclitic which adjoins to the left of the zero-levelelement to its right (8) (* represents the relation ofstring adjacency; Q represents dissociated mor-phemes).

By hypothesis, Prosodic Inversion (Halpern 1995)is a distinct species of Merger at the level of Pho-nological Form, and differs from Local Dislocationin that the affected elements are prosodic catego-ries rather than morphological ones.

For example, Schütze (1994), expanding onZec & Inkelas (1990), argues that the auxiliaryclitic je in Serbo-Croatian is syntactically in C, butinverts with the following phonological word byProsodic Inversion at Phonological Form (paren-theses below denote phonological word bounda-ries):

(9)Serbo-Croatian second-position cliticsMorphological structureafter Spell-Out [je [VP [PP U ovoj sobi] klavir]]

Parse intophonological words je (U ovoj) (sobi) (klavir)

Prosodic inversion ((U ovoj)+ je) (sobi) (klavir)In this AUX room piano‘In this room is the piano’

The positioning of the clitic cannot be stated interms of a (morpho)syntactic constituent, since Uovoj ‘in this’ does not form such a constituent.Embick & Izvorski (1995) specifically argue thatsyntactic explanations, including those involvingremnant extraposition, cannot reasonably be heldaccountable for this pattern.

However, it should be emphasized that theextent to which Local Dislocation and ProsodicInversion are distinct devices in the mapping toPhonological Form remains controversial, withmany researchers seeking to reduce the two to apurely prosodic or a purely syntactic mechanism.

3.2. ImpoverishmentImpoverishment, first proposed in Bonet

(1991), is an operation on the contents of mor-phemes prior to Spell-Out. In early work in DM,Impoverishment simply involved the deletion ofmorphosyntactic features from morphemes incertain contexts. When certain features are delet-ed, the insertion of Vocabulary Items requiringthose features for insertion cannot occur, and aless specified item will be inserted instead. Halle &Marantz (1994) termed this the ‘Retreat to theGeneral Case’.

(10)Adjectival suffixes in Norwegian (Sauerland 1995)STRONG [–neuter] [+neuter][–pl] Ø -t[+pl] -e -eWEAK [–neuter] [+neuter][–pl] -e -e[+pl] -e -e

In Norwegian, there is a three-way distinction(t ~ e ~ Ø) in adjectival suffixes in a “strong”

syntactic position, but in the “weak” position onefinds only -e. By hypothesis, it is not accidentalthat the affix -e is the Elsewhere affix in thestrong context, and also appears everywhere inthe weak context. Sauerland’s (1995) Impoverish-ment analysis of the weak paradigms captures thisinsight. He proposes the following set of Vocabu-lary Items:

(11)Norwegian Vocabulary Items/t/ ←→ [ ___ , -pl +neut] / Adj + ____Ø ←→ [ ___ , -pl -neut] / Adj + ____/e/ ←→ elsewhere / Adj + ____

In the weak syntactic position, a rule of Impover-ishment applies, deleting any values for genderfeatures:

(12)Impoverishment[±neuter] → Ø

Impoverishment thus guarantees that neither theVocabulary Items t nor Ø can be inserted, sinceboth require explicit reference to a value for[±neuter]. Insertion of the general case, namely -e,follows automatically.

As we have noted above, in Bonet’s originalproposal (1991) and in several subsequent works(Harley 1994; Harris 1997a; Ritter & Harley1998), morphosyntactic features are arranged in afeature geometry much like phonological features,and Impoverishment is represented as delinking.Consequently, the delinking of certain featuresentails the delinking of features dependent onthem. For example, if person features dominatenumber features which in turn dominate genderfeatures, then the Impoverishment (delinking) ofnumber entails the delinking of gender as well:

(13)Impoverishment as delinking

2

pl

f

2

®

Noyer (1997) rejects the use of geometries of thissort as too restrictive, and proposes instead thatImpoverishments are better understood as featureco-occurrence restrictions or filters of the typeemployed by Calabrese (1995) for phonologicalsegment inventories. For example, the absence ofa first person dual in Arabic is represented as thefilter *[1 dual], and a Universal Hierarchy ofFeatures dictates that where these features com-bine, because [dual] is a number feature and [1] isa (hierarchically higher) person feature, [dual] isdeleted automatically. Calabrese (1994) and(1996) further expand this idea. The use of fea-ture geometries in DM remains an unresolvedissue at this time, but Feature Hierarchies, wheth-er geometric or not, ensure that normally lessmarked feature values persist in contexts of neu-tralization.

Feature-changing Impoverishment, which asa device has approximately the same power asRules of Referral (Zwicky 1985b; Stump 1993),has in general been eschewed in DM. However,Noyer (1998a) discusses cases where feature-changing readjustments seem necessary. It isproposed that such cases always involve a changefrom the more marked value of a feature to theless marked value and never vice versa.

3.3. Fission and Feature DischargeFission was originally proposed in Noyer

(1997) to account for situations in which a singlemorpheme may correspond to more than oneVocabulary Item. In the normal situation, only

one Vocabulary Item may be inserted into anygiven morpheme. But where fission occurs, Vocab-ulary Insertion does not stop after a single Vocab-ulary Item is inserted. Rather, Vocabulary Itemsaccrete on the sister of the fissioned morphemeuntil all Vocabulary Items which can be insertedhave been, or all features of the morpheme havebeen discharged. A feature is said to be dis-charged when the insertion of a Vocabulary Itemis conditioned by the presence of that feature.

However, Noyer (1997) argues that featuresconditioning the insertion of a Vocabulary Itemcome in two types. A Vocabulary Item primarilyexpresses certain features in its entry, but it maybe said to secondarily express certain otherfeatures. This distinction corresponds (approxi-mately) to the distinction between primary andsecondary exponence (Carstairs 1987). Onlyfeatures which are primarily expressed by aVocabulary Item are discharged by the insertion ofthat Item.

For example, in the prefix-conjugation ofTamazight Berber, the AGR morpheme can appearas one, two or three separate Vocabulary Items,and these may appear as prefixes or as suffixes:

(14)a. Tamazight Berber Prefix Conjugation. dawa ‘cure’

singular plural3m i-dawa dawa-n3f t-dawa dawa-n-t2m t-dawa-d t-dawa-m2f t-dawa-d t-dawa-n-t1 dawa-> n-dawa

b. Vocabulary Items/n-/ ←→ 1 pl/->/ ←→ 1/t-/ ←→ 2/t-/ ←→ 3 sg f/-m/ ←→ pl m (2)/i-/ ←→ sg m/-d/ ←→ sg (2)/-n/ ←→ pl/-t/ ←→ f

Some features in the above Vocabulary Item listare in parentheses. This notation denotes that theVocabulary Item in question can be inserted onlyif the parenthesized feature has already beendischarged, whereas the features which are not inparentheses cannot already have been dischargedif insertion is to occur. For example, -m can beinserted only on a verb to which t- ‘2’ has alreadybeen attached. Parentheses are thus used todenote features which are secondarily expressedby a Vocabulary Item, while ordinary features —those which a Vocabulary Item primarily express-es — are not parenthesized.

In a fissioned morpheme, Vocabulary Itemsare no longer in competition for a single position-of-exponence, i.e. for the position of the morphemeitself. Rather, an additional position-of-exponenceis automatically made available whenever aVocabulary Item is inserted (see Halle 1997 for aslightly different view).

A form like t-dawa-n-t ‘you (FEM.PL) cure’ hasthree affixes, t-, -n, and -t. The affixes are addedin an order determined by the Feature Hierarchy.Hence t- ‘2’ is added first, then -n ‘plural’, andfinally -t ‘feminine’. (In the feature-geometricapproach of Harley & Ritter (1998), fission detach-es subtrees of the feature geometry and realizesthem as separate affixes, giving much the sameeffect).

In a form like n-dawa ‘we cure’ there is but oneaffix. By discharging the feature ‘1’, the insertion ofn- ‘1 pl’ prevents the subsequent insertion of -> ‘1’.This illustrates that two Vocabulary Items can bedisjunctive not by competing for the same position-of-exponence, but rather by competing for thedischarge of the same feature. Such cases aretermed Discontinuous Bleeding.

3.4. Morphological processes and thepredictions of a piece-based theoryDM is piece-based inasmuch as Vocabulary

Items are considered discrete collections of phono-logical material and not (the result of) phonologi-cal processes (as in Anderson 1992). NeverthelessReadjustment can alter the shape of individual

(8)Latin -que placement

Morphological structure: [[A Q] [N-Q]] [cl [[A-Q] [N-Q]]]Vocabulary insertion: [[bon i] [puer i]] [–que [[bon ae] [puell ae]]]Linearization: [[bon*i] * [puer*i]] * [–que* [[bon*ae] * [puell*ae]]]Local dislocation [[bon*i] * [puer*i]] * [[[bon*ae]*que] * [puell*ae]]]

good-NOM.PL boy-NOM.PL good-NOM.PL-and girl-NOM.PL

‘Good boys and good girls’

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 7

Vocabulary Items in appropriate contexts. Twofactors thus distinguish DM from process-onlytheories of morphology.

First, since Readjustment can affect onlyindividual Vocabulary Items and not more thanone Vocabulary Item at once, it is predicted that“process” morphology is always a kind of allomor-phy (see also Lieber 1981). For example, Marantz(1992) shows that truncation applies to (Papago)O’odham verb stems to produce a separate stemallomorph; it does not affect more than one Vocab-ulary Item at once.

Second, since processes produce allomorphsbut do not directly “discharge” features, it is com-mon for an allomorph to have several contexts ofuse. For example, in O’odham the truncated verbstem allomorph has several functions, includingbut not limited to its use in the perfective form,and the property of perfectivity is primarily ex-pressed in another morpheme, namely an affix onthe syntactic auxiliary. It is therefore incorrect todirectly equate truncation and the perfective;rather, truncation applies to verb stems whichappear in the perfective. This conception of stemallomorphy conforms to the viewpoint of Lieber(1981).

Since process-morphology can in principleapply to any string, regardless of its morphologicalderivation, it is predicted in that theory that alanguage could mark the category Plural bydeletion of a final syllable, regardless of whetherthat syllable consisted of one or several discretephonological pieces. Consider “Martian” below:

(15)Singular and plural nouns in the pseudo-language ‘Martian’singular pluraltakata taka ‘earthling’takata-ri takata ‘earthling-GEN’laami laa ‘antenna’jankap jan ‘flying saucer’jankap-ri janka ‘flying saucer-GEN’zuuk lorp ‘canal’zuuk-ri zuu ‘canal-GEN’yuun-i yuu ‘antenna waving’merg-i mer ‘canal digging’merg-i-ri mergi ‘canal digging-GEN’

In “Martian”, nominalizations can be formed fromnoun stems by addition of the suffix (-i) andgenitives with the suffix (-ri). Regardless of thederivation of a noun, the plural is always either atruncation of the last syllable of the singular, orsuppletive (zuuk ~ lorp). The truncated formnever occurs anywhere else except in plurals.Number marking has no other expression thantruncation.

The “Martian” rule of plural formation is easyto express in a process-morphology: instead ofadding an affix, one simply deletes the finalsyllable. In DM however, this language couldnever be generated, because processes like “deletethe final syllable” could only be expressed asReadjustments (or morphological relations) whichaffect individual Vocabulary Items.

4. Syntax and morphologyAs noted in section 1, DM adopts a strictly

syntactic account of word-formation; structuring ofthe morphosyntactic feature primitives is per-formed by the syntactic structure-forming opera-tions. Features which will eventually be realizedas a subpart of a phonological word are treated nodifferently from features which will eventually berealized as an autonomous word. The phonologicalrealization of features is accomplished by a distinctset of operations at Insertion and afterwards. Thatis, DM adopts a variety of Separationism.

4.1. SeparationismSeparationism characterizes theories of mor-

phology in which the mechanisms for producingthe form of syntactico-semantically complex ex-pressions are separated from, and not necessarilyin a simple correspondence with, the mechanismswhich produce the form (“spelling”) of the corre-sponding phonological expressions. Lexeme-Morpheme Base Morphology developed by RobertBeard (e.g. Beard 1995) is another example of aSeparationist model, but differs principally from

DM in its endorsement of the “lexeme” as a privi-leged unit in the grammar.

Theories endorsing Separationism are attrac-tive because (a) they allow similar syntactico-semantic forms to be realized in quite differentways phonologically and (b) they permit polyfunc-tionality of phonological expressions: a singlephonological piece (e.g. the English affix -s) mightcorrespond to a set of distinct and unrelated syn-tactico-semantic functions.

Theories endorsing Separationism, on theother hand, are unattractive for exactly the samereasons as above: when unconstrained, they fail tomake any interesting predictions about the degreeto which syntactico-semantic and phonologicalform can diverge. See Embick (1997, 1998a,1998b) for a discussion of how Separationismcould be constrained in DM.

4.2. Morphosyntactic features and terminalnodesIn the early 1990s some linguists looked on

with apprehension at the “explosion” of Infl andthe increasing elaboration of clause structure. It isworth noting that the DM does not necessarilyentail a complex clausal architecture simply be-cause morphosyntactic features are manipulatedby the syntax. In DM, because dissociated mor-phemes can be inserted after syntax, not everymorpheme need correspond to a syntactic termi-nal. Rather it remains as always an open questionwhat the set of syntactic terminal types is and howthese relate to the morphophonological form of anutterance. In addition, fission of morphemesduring Spell-Out in some cases allows multiplephonological pieces to correspond to single mor-phemes, further obscuring the morphosyntacticstructure. Nevertheless, these departures areconsidered marked options within a grammar, andtherefore are assumed to require (substantial)positive evidence during acquisition.

4.3. Theta-assignmentMost work in DM does not recognize a set of

discrete thematic roles. Instead, following theinsights of Hale & Keyser (1993, 1998), thematicroles are reduced to structural configurations. Forexample, Harley (1995) proposes that ‘Agent’ isthe interpretation given to arguments projectedinto the specifier of Event Phrase (see also Travis1994 on ‘Event Phrase’, and Kratzer 1996 forrelated ideas). ‘Theme’ corresponds to the interpre-tation given to any argument projected as a sisterof Root. Unlike Hale & Keyser (1993), however,DM does not differentiate between an ‘l-syntax’occurring in the lexicon and a regular ‘s-syntax’.Both are simply one module, syntax. See alsoMarantz (1997a).

Such an approach is not necessarily entailedby the DM model, however. One could imagine amodel in which there were different types of[Root], corresponding to the verb classes of theworld’s languages, which assigned different sets oftheta roles to elements in certain structural rela-tions to them. What is not possible, in DM, is forone type of [Root] to be mapped onto another via apre-syntactic lexical operation.

4.4. The phonology/morphology/syntaxconnection: clitics“Clitic” is not a primitive type in DM but

rather a behavior which an element may display.Conventionally, clitics are said to “lean” on a“host”; this sort of dependency relation of oneelement on another manifests itself differentlydepending on what the element is and where itsdependency relation must be satisfied. Hencethere is no coherent class of objects which can betermed clitics; instead morphemes and VocabularyItems may show a range of dependencies.

“Leaners” (Zwicky 1985a) are VocabularyItems which cannot form phonological words bythemselves but whose morphemes have no otherspecial displacement properties. For example, theEnglish reduced auxiliary -s (from is) “promiscu-ously” attaches to any phonological host to its left(Zwicky & Pullum 1983):

(16)LeanersThe person I was talking to’s going to be angry with me.Any answer not entirely right’s going to be marked as an error.

Selkirk (1996) analyzes prosodically dependentVocabulary Items as either free clitics (adjuncts tophonological phrases), affixal clitics (adjuncts tophonological words) or internal clitics (incorporat-ed into phonological words). These options areshown schematically below:

(17)Types of phonological cliticsN[ = phrase boundary, T[ = word boundary

N[... free clitic N[ T[ host ] ... ]N[... T[affixal clitic T[host ] ] ... ]N[... T[internal clitic + host ] ... ]

English leaners are typically free clitics, accordingto Selkirk, but other languages exploit otheroptions. For example, Embick (1995) shows that,depending on whether they undergo head move-ment or are simply leaners, Polish clitics behavephonologically as either affixal clitics (allowingtheir host to undergo word-domain phonology), oras internal clitics (preventing their host fromundergoing word-domain phonology on its own).

Second-position clitics, illustrated for Serbo-Croatian in section 3.1, are Vocabulary Itemswhich undergo either Local Dislocation or ProsodicInversion with a host.

Finally, the term “clitic” is sometimes used todescribe syntactically mobile heads, typicallyDeterminers, such as certain Romance pronomi-nals on some accounts. In such cases the depend-ency relation or special behavior is a syntacticproperty of a morpheme. In many cases the Vocab-ulary Items which are inserted into these mor-phemes also show either phonological dependencyas leaners or additional peculiarities of position viaLocal Dislocation or Prosodic Inversion. See Harris(1994, 1997a) and Embick (1995) for case studies.

5. An agenda for future researchThe research program envisioned by Distrib-

uted Morphology encompasses a great manyaspects of the theory of grammar. Thus, the agen-da for future research with which we concludehere touches upon what we feel are some of themost pressing questions in contemporary syntacticand morphological work. We have divided theagenda into three headings.

5.1. Syntactic categories and thearchitecture of grammarAs noted, DM denies that syntactic categories

necessarily stand in any simple relation to tradi-tional parts-of-speech such as nouns and verbs;moreover, DM denies that syntactic categoriesstand in any simple relation to phonological words.Thus, as is also the case with much work in Mini-malist syntax, the DM research program demandsa reassessment of the inventory and bases forsyntactic categories. Related questions include thefollowing. First, the ramifications of the L-mor-pheme Hypothesis (according to which open-classVocabulary Items always instantiate the samesyntactic category) point to the need for continuedstudy of so-called “mixed” categories and the cross-linguistic validity, if any, of traditional part-of-speech labels in universal syntax. Second, how dothese categories relate to universal semanticprimes and to what extent do certain types ofderivational word-formation manipulate suchprimes? This topic is explored extensively in thework of Robert Beard (e.g. Beard 1995), but hasnot yet been properly incorporated into the DMmodel. Third, DM hypothesizes that syntax ma-nipulates only categories defined by features madeavailable by Universal Grammar. This leads to thequestion of whether language-specific features(such as gender or form class) are present in thesyntax at all, or whether such features are una-vailable in syntax proper and are supplied forpurposes of Spell-Out and agreement onlythrough Vocabulary Insertion after syntax (fordiscussion, see Embick 1998b).

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 8

As Aronoff (1994) has most persuasivelyargued, morphology requires the manipulation ofform classes and stem types whose relation tosyntactic properties or configurations is not direct,but mediated by a complex mapping. DM adoptsthis Separationist position by positing a compo-nent of Morphology after syntax which providesfor this mapping. Nevertheless, an importantquestion for future work is whether this mappingis constrained by any interesting universal princi-ples (Embick 1997).

Along with the research program of Hale &Keyser (1993, 1998), DM does not recognize theassignment of theta-roles by “lexical items”. Thus,research in DM continues to explore whethertheta-roles may be dispensed with as primes of thetheory and replaced by a configurational defini-tion of argument roles.

Properties of the Encyclopedia and its relationto grammatical well-formedness raise additionalimportant issues. Marantz (1997b) for examplehas suggested that (phrasal) idioms cannot extendbeyond the Event (v) projection, but it remains anopen question how the Encyclopedia effects thisconstraint on semantic interpretation. A relatedquestion concerns the distinction between whatare conventionally termed “productive” and “non-productive” processes. The earliest work in genera-tive morphology such as Halle (1973) postulated aDictionary which effectively licensed the use ofexpressions formed by non-productive word-formation rules. The question of whether the DMEncyclopedia can or should perform this licensingfunction, or how, if at all, expressions formed bynon-productive mechanisms of the grammar are tobe specially treated, is currently under investiga-tion.

5.2. Spell-OutA number of researchers in DM have accepted

the traditional view that morphosyntactic featureshave markedness properties or are aligned intohierarchies of various sorts. Open questions —which DM in fact shares with all theories of mor-phology — currently include what the set ofuniversal morphosyntactic features is and what, ifany, are their universal markedness properties, aswell as how these are structured in representation(e.g. in a geometry, in a list, or in some other way).

Spell-Out of morphemes may be conditionedby properties in nearby morphemes, and so animportant issue is the syntagmatic (locality) con-straints on Spell-Out, that is, how close structural-ly a morpheme has to be to another to influencethe other’s Spell-Out. Similarly, opinion remainsdivided as to whether the outcome of a competitionof Vocabulary Items for positions may be settled bymeans of a hierarchy of features or can be stipu-lated.

Finally, not all morphemes are present insyntax proper, but some are purely morphological,reflecting syntactic configurations or properties.Which morphemes, then, are inserted after syntaxand what kind of limitations are placed on mor-pheme-insertion?

5.3. OperationsImpoverishment, Fission and Morphological

Merger are the chief novel operations proposed inDM for the Morphological component, and ques-tions remain open about each.

Is Impoverishment constrained to reducemarkedness only, and if not, does it differ funda-mentally from Rules of Referral (Zwicky 1985b;Stump 1993)? What are the syntagmatic (locality)

Alexiadou, A. (1998). An ergative pattern in thesyntax of nominative-accusative languages.Paper presented at the 29th meeting of theNorth Eastern Linguistics Society, Universityof Delaware.

Anderson, S.R. (1982). Where’s morphology?Linguistic Inquiry 13, 571–612.

Anderson, S.R. (1992). A-morphous morphology.Cambridge: Cambridge University Press.

Aronoff, M. (1994). Morphology by itself. Cam-bridge, Mass.: MIT Press.

Baker, M. (1988). Incorporation: A theory ofgrammatical function changing. Chicago:University of Chicago Press.

Beard, R. (1995). Lexeme-morpheme base morphol-ogy. Albany: SUNY Albany Press.

Bobaljik, J.D. (1994). What does adjacency do?[MIT Working Papers in Linguistics 22: TheMorphology-Syntax Connection], edited by H.Harley & C. Phillips, 1–32. Cambridge, Mass.:MITWPL.

Bobaljik, J.D. & S. Wurmbrand. (1997). Prelimi-nary notes on agreement in Itelmen. [MITWorking Papers in Linguistics 30: Papers atthe Interface], edited by B. Bruening, Y. Kang& M. McGinnis, 395–424. Cambridge, Mass.:MITWPL.

Bonet, E. (1991). Morphology after syntax: Pro-nominal clitics in Romance. Doctoral disserta-tion, Massachusetts Institute of Technology.

Calabrese, A. (1994). Syncretism phenomena inthe clitic systems of Italian and Sardiniandialects and the notion or morphologicalchange. The Proceedings of the North EastLinguistic Society 25, Vol. 2, edited by J.N.Beckman, 151–174. Amherst: GLSA, Univer-sity of Massachusetts, Amherst.

Calabrese, A. (1995). A constraint-based theory ofphonological markedness and simplificationprocedures. Linguistic Inquiry 26, 373–463.

Calabrese, A. (1997a). Distributed morphologyand sentential complementation in the Salen-tino dialect of Italian. Harvard Working

constraints on the operation of Impoverishment?Is the mechanism of morpheme fission, in whichpositions are automatically generated as neededfor the insertion of features, really necessary, andif so, under what circumstances do morphemesundergo fission? How many types of Morphologi-cal Merger are there and how do they differ? CanMerger be reduced to a purely syntactic or purelyphonological mechanism?

In the realm of morphophonology, Marantz’sconjecture that true suppletion is limited to f-morphemes prompts a search for non-stipulativecriteria dividing suppletion from Readjustment.Once cases of true suppletion are factored out, thepossibility arises for an interesting theory ofReadjustment allomorphy based on the degree ofrelatedness between allomorphs necessary forthese to be acquired as variants of the same Vo-cabulary Item.

6. ConclusionWe have presented DM’s primary theoretical

assumptions, provided some concrete illustration ofthe implementation of certain of its mechanisms,and proposed an agenda for future research.Although we have touched on a large array oftopics in current morphological theory, we cannotclaim to have fully elucidated the advantages ofDM relative to its competitors, nor have we ex-hausted the historical bases for many of its tenets.Instead, we hope that our exposition will providethe groundwork for an informed discussion ofDM’s contribution to the theory of grammar.Interested readers should consult the followingbibliography of representative works within DMas well as important alternative approaches to theissues that stimulated the DM research program.

A Distributed Morphology Bibliography

Papers in Linguistics, Vol. 6, edited by S.Peter, B. Vaux & S. Kuno. Cambridge, Mass.:Harvard University.

Calabrese, A. (1997b). On Fission and Impover-ishment in the verbal morphology of thedialect of Livinallongo. Manuscript, Universi-ty of Connecticut, Storrs.

Calabrese, A. (1998). Some remarks on the Latincase system and its development in Romance.Theoretical Analysis of Romance Languages,edited by E. Treviño & J. Lema, 71–126.Amsterdam: John Benjamins.

Carstairs, A. (1987). Allomorphy in inflection.London: Croom Helm.

Chen, M. (1987). The syntax of Xiamen tonesandhi. Phonology 4, 109–150.

Chomsky, N. (1970). Remarks on Nominalization.Readings in Transformational Grammar,edited by R. A. Jacobs & P. S. Rosenbaum,184–221. Waltham, Mass.: Ginn.

Chomsky, N. (1981). Lectures on Government andBinding. Dordrecht: Foris.

DiSciullo, A. & E. Williams. (1987). On the defini-tion of word. Cambridge, Mass.: MIT Press.

Embick, D. (1995). Mobile inflections in Polish.The Proceedings of the North East LinguisticSociety 25, Vol. 2, edited by J.N. Beckman,127–142. Amherst: GLSA, University ofMassachusetts, Amherst.

Embick, D. (1996). Causativization in Hupa.Proceedings of BLS 22, 83–94.

Embick, D. (1997). Voice and the interfaces ofsyntax. Doctoral dissertation, University ofPennsylvania.

Embick, D. (1998a). Voice systems and the syntax/morphology interface [MIT Working Papers inLinguistics 32: Papers from the UPenn/MITRoundtable on Argument Structure andAspect], edited by Heidi Harley, 41–72. Cam-bridge, Mass.: MITWPL.

Embick, D. (1998b). Syntax and categories: Verbsand participles in the Latin Perfect. Manu-script, Massachusetts Institute of Technology.

Embick, D. & R. Izvorski. (1995). Participle-auxiliary word-orders in Slavic. Formal ap-proaches to Slavic linguistics: The CornellUniversity Meeting, edited by W. Browne,210–239. Ann Arbor: Michigan Slavic Publica-tions.

Embick, D. & R. Noyer (to appear). MorphologicalMerger and Locality. Paper presented at theMassachusetts Institute of Technology collo-quium series.

Hale, K. & S.J. Keyser. (1993). On argumentstructure and the lexical representation ofsyntactic relations. The View from Building20, edited by S.J. Keyser & K. Hale, 53–109.Cambridge, Mass.: MIT Press.

Hale, K. & S.J. Keyser. (1998). The basic elementsof argument structure [MIT Working Papersin Linguistics 32: Papers from the UPenn/MIT Roundtable on Argument Structure andAspect], edited by Heidi Harley, 73–118.Cambridge, Mass.: MITWPL.

Halle, M. (1973). Prolegomena to a theory of wordformation. Linguistic Inquiry 4, 3–16.

Halle, M. (1990). An approach to morphology.Proceedings of the North East LinguisticsSociety 20, Vol. 1, 150–84. Amherst: GLSA,University of Massachusetts, Amherst.

Halle, M. (1992). Latvian declension. Morphologyyearbook 1991, edited by G. Booij & J. van derMarle, 33–47. Dordrecht: Kluwer.

Halle, M. (1995). The Russian declension. Perspec-tives in phonology, edited by J. Cole & C.Kisseberth, 321–353. Palo Alto: CSLI.

Halle, M. (1997). Distributed morphology: Impov-erishment and fission [MIT Working Papersin Linguistics 30: Papers at the Interface],edited by B. Bruening, Y. Kang & M. McGinn-is, 425–449. Cambridge, Mass.: MITWPL.

Halle, M. & A. Marantz. (1993). Distributed mor-phology and the pieces of inflection. The Viewfrom Building 20, edited by K. Hale & S.J.Keyser, 111–176. Cambridge, Mass.: MITPress.

State-of-the-Article Glot International, Volume 4, Issue 4, April 1999 9

Halle, M. & A. Marantz. (1994). Some key fea-tures of Distributed Morphology [MIT Work-ing Papers in Linguistics 21: Papers onphonology and morphology], edited by A.Carnie & H. Harley, 275–288. Cambridge,Mass.: MITWPL.

Halle, M. & B. Vaux. (1998). Theoretical aspects ofIndo-European nominal morphology: Thenominal declensions of Latin and Armenian.Mír Curad: Studies in honor of Calvert Wat-kins, edited by J. Jasanoff, H.C. Melchert &L. Olivier, 223–240. Innsbruck: InnsbruckerBeitraege zur Sprachwissenschaft.

Halle, M. & J. Vergnaud. (1987). An essay onstress. Cambridge, Mass.: MIT Press.

Halpern, A. (1995). On the placement and mor-phology of clitics. Palo Alto: CSLI.

Harley, H. (1994). Hug a tree: Deriving the mor-phosyntactic feature hierarchy [MIT WorkingPapers in Linguistics 21: Papers on phonologyand morphology], edited by A. Carnie & H.Harley, 275–288. Cambridge, Mass.: MIT-WPL.

Harley, H. (1995). Subjects, events, and licensing.Doctoral dissertation, Massachusetts Instituteof Technology.

Harley, H. & R. Noyer. (1998a). Licensing in thenon-lexicalist lexicon: nominalizations, Vocab-ulary Items and the Encyclopedia [MITWorking Papers in Linguistics 32: Papersfrom the UPenn/MIT Roundtable on Argu-ment Structure and Aspect], edited by H.Harley, 119–137. Cambridge, Mass.: MIT-WPL.

Harley, H. & R. Noyer. (1998b). Mixed nominali-zations, short verb movement, and object shiftin English. Proceedings of NELS 28, edited byP.N. Tamanji & K. Kusumoto, 143–157. Am-herst: GLSA, University of Massachusetts,Amherst.

Harris, J. (1994). The syntax-phonology mappingin Catalan and Spanish clitics [MIT WorkingPapers in Linguistics 21: Papers on phonologyand morphology], edited by A. Carnie & H.Harley, 321–353. Cambridge, Mass.: MIT-WPL.

Harris, J. (1995). The morphology of Spanishclitics. Evolution and revolution in linguistictheory, edited by H. Campos & P. Kempchin-sky, 168–197. Washington: GeorgetownUniversity Press.

Harris, J. (1996). The syntax and morphology ofClass Marker suppression in Spanish. Gram-matical theory and Romance languages,edited by K. Zagona, 99–122. Amsterdam:John Benjamins.

Harris, J. (1997a). Why n’ho is pronounced [li] inBarceloni Catalan [MIT Working Papers inLinguistics 30: Papers at the Interface], editedby B. Bruening, Y. Kang & M. McGinnis,451–479. Cambridge, Mass.: MITWPL.

Harris, J. (1997b). There is no imperative para-digm in Spanish. Issues in the phonology andmorphology of the major Iberian languages,edited by F. Martínez-Gil & A. Morales-Front,537–557. Washington: Georgetown UniversityPress.

Harris, J. (1998). Spanish imperatives: syntaxmeets morphology. Journal of Linguistics 34:27–52.

Harris, J. (1998). Enclitic -n in Spanish. RomanceLingusitics: Theoretical perspectives, edited byA. Schwegler, B. Tranel & M. Uribe-Extebar-ria, 111–127. Amsterdam: John Benjamins.

Harris, J. (to appear). Los imperativos pluralesseguidos de clítico en español. Paper present-ed at IV Encuentro de Lingüística del No-roeste, Universidad de Sonora, Hermosillo,México. To appear in proceedings.

Jackendoff, R. (1997). The architecture of thelanguage faculty. Cambridge, Mass.: MITPress.

Kiparsky, P. (1982). Lexical phonology and mor-phology. Linguistics in the morning calm, Vol.2, edited by I. S. Yang, 3–91. Seoul: Hanshin.

Kratzer, A. (1996). Severing the external Argu-ment from its verb. Phrase structure and thelexicon, edited by J. Rooryck & L. Zaring,109–37. Dordrecht: Kluwer.

Lidz, J. (1998). Valency in Kannada: Evidence forinterpretive morphology. University of Penn-sylvania Working Papers in Linguistics 5.2:Current Work in Linguistics, edited by A.Dimitriadis, H. Lee, C. Moisset & A. Williams,37–63. Philadelphia: University of Pennsylva-nia.

Lidz, J. & W.J. Idsardi. (1998). Chains and Phono-Logical Form. University of PennsylvaniaWorking Papers in Linguistics 5.1: Proceed-ings of the 22nd Penn Linguistics Collo-quium, edited by A. Dimitriadis, H. Lee, C.Moisset & A. Williams. Philadelphia: Universi-ty of Pennsylvania.

Lieber, R. (1981). On the organization of thelexicon. Doctoral dissertation, MassachusettsInstitute of Technology.

Lumsden, J. (1987). Syntactic features: Parametricvariation in the history of English. Doctoraldissertation, Massachusetts Institute of Tech-nology.

Lumsden, J. (1992). Underspecification in gram-matical and natural gender. Linguistic In-quiry 23, 469–486.

Manzini, M.R. & K. Wexler. (1987). Parameters,Binding Theory and learnability. LinguisticInquiry 18, 413–444.

Marantz, A. (1984). On the nature of grammaticalrelations. Cambridge, Mass.: MIT Press.

Marantz, A. (1988). Clitics, morphological merger,and the mapping to phonological structure.Theoretical morphology: Approaches in mod-ern linguistics, edited by M. Hammond & M.Noonan, 253–270. San Diego: Academic Press.

Marantz, A. (1991). Case and Licensing. Proceed-ings of the Eastern States Conference onLinguistics ’91, 234–53. Distributed by OhioState University Department of Linguistics.

Marantz, A. (1992). How morphemes are realizedphonologically. Paper presented at the DI-MACS Workshop on Human Language,Princeton University. Manuscript, Massachu-setts Institute of Technology.

Marantz, A. (1995). A late note on Late Insertion.Explorations in generative grammar, editedby Y. Kim et al., 396–413. Seoul: Hankuk.

Marantz, A. (1997a). No escape from syntax: Don’ttry morphological analysis in the privacy ofyour own Lexicon. Penn Working Papers inLinguistics 4:2: Proceedings of the 21st Annu-al Penn Linguistics Colloquium, edited by A.Dimitriadis, L. Siegel, C. Surek-Clark, & A.Williams, 201–225. Philadelphia: Universityof Pennsylvania.

Marantz, A. (1997b). Cat as a phrasal idiom: stemsuppletion, or the arbitrariness of the sign.Paper presented at the Université de ParisVIII.

McCloskey, J. (1996). On the scope of verb move-ment in modern Irish, Natural Language andLinguistic Theory 14, 47–104.

McGinnis, M. (1995). Fission as feature-movement[MIT Working Papers in Linguistics 27:Papers on Minimalist Syntax], edited by R.Pensalfini & H. Ura, 165–187. Cambridge,Mass: MITWPL.

McGinnis, M. (1998). Case and locality in L-Syntax: Evidence from Georgian [MIT Work-ing Papers in Linguistics 32: Papers from theUPenn/MIT Roundtable on Argument Struc-ture and Aspect], edited by H. Harley, 139–158. Cambridge, Mass: MITWPL.

Minkoff, S. (1993). Plurality, clitics, and morpho-logical merger in Caribbean Spanish [MITWorking Papers in Linguistics 20: Papersfrom the Fifth Student Conference in Linguis-tics], edited by V.M. Lindblad & M. Gamon,177–192. Cambridge, Mass.: MITWPL.

Noyer, R. (1997). Features, positions and affixesin autonomous morphological structure. NewYork: Garland. Revised version of 1992 MITDoctoral Dissertation.

Noyer, R. (1998a). Impoverishment theory andmorphosyntactic markedness. Morphologyand its relation to phonology and syntax,edited by S. Lapointe, D.K. Brentari & P.Farrell, 264–285. Palo Alto: CSLI.

Noyer, R. (1998b). Vietnamese “morphology” andthe definition of word. University of Pennsyl-vania Working Papers in Linguistics 5.2:Current Work in Linguistics, edited by A.Dimitriadis, H. Lee, C. Moisset & A. Williams,65–89. Philadelphia: University of Pennsylva-nia.

Pesetsky, D. (1995). Zero syntax: experiencers andcascades. Cambridge, Mass.: MIT Press.

Ritter, E. & H. Harley. (1998). Sorting out you, meand the rest of the world: A feature-geometricanalysis of person and number. Paper pre-sented at the 21st meeting of GenerativeLinguistics of the Old World, Tilburg, theNetherlands.

Sadock, J. (1991). Autolexical syntax: A theory ofparallel grammatical representations. Chica-go: University of Chicago Press.

Sapir, E. (1921). Language. New York: Harcourt,Brace and World.

Sauerland, U. (1995). The Late Insertion of Ger-manic inflection. Manuscript, Cambridge,Mass.: MIT.

Schütze, C. (1994). Serbo-Croatian second positionclitic placement and the phonology-syntaxinterface [Working Papers in Linguistics 21:Papers on phonology and morphology], editedby A. Carnie, H. Harley & T. Bures, 373–473.Cambridge, Mass.: MITWPL.

Selkirk, E.O. (1982). The syntax of words, Cam-bridge, Mass.: MIT Press.

Selkirk, E.O. (1986). On derived domains insentence phonology. Phonology 3: 371–405.

Selkirk, E.O. (1996). The prosodic structure offunction words. Signal to syntax: Bootstrap-ping from speech to grammar in early acquisi-tion, edited by J.L. Morgan & K. Demuth,187–213. Mahwah, N.J.: Lawrence Erlbaum.

Siegel, L. (to appear). Gerundive nominals andaspect. To appear in the Proceedings of ES-COL 97.

Silverstein, M. (1976). A hierarchy of features andergativity. Grammatical categories in Austral-ian languages, edited by R.M.W. Dixon, 112–171. Canberra: Australian Institute forAboriginal Studies.

Sproat, R.W. (1985). On deriving the lexicon. PhDdissertation, Massachusetts Institute of Tech-nology.

Stump, G. (1993). On rules of referral. Language69:3, 449–479.

Travis, L. (1994). Event phrase and a theory offunctional categories. Paper presented at theCanadian Linguistics Association, Universityof Calgary.

Wunderlich, D. (1996). Minimalist morphology:the role of paradigms. Yearbook of morphology1995, edited by G. Booij & J. van Marle, 93–114. Dordrecht: Kluwer.

Zec, D. & S. Inkelas. (1990). Prosodically con-strained syntax. The phonology-syntax con-nection, edited by S. Inkelas & D. Zec,365–378. Chicago: University of ChicagoPress.

Zwicky, A. (1977). Hierarchies of person. Papersfrom the Thirteenth Meeting, Chicago Lin-guistics Society, edited by W.A. Beach, S.E.Fox & S. Philosoph, 714–733. Chicago: Uni-versity of Chicago.

Zwicky, A. (1985a). Clitics and particles. Language61: 283–305.

Zwicky, A. (1985b). How to describe inflection.Proceedings of the Berkeley Linguistics Socie-ty 11, 372–386. Berkeley, Calif.: UCB.

Zwicky, A. & G. Pullum. (1983). Cliticization vs.inflection: English n’t. Language 59, 502–13.

Zwicky, A. & G. Pullum. (1992). A misconceivedapproach to morphology. Proceedings of WC-CFL 91, edited by D. Bates, 387–398. PaloAlto: CSLI.

Column Glot International, Volume 4, Issue 4, April 1999 10

In the short interview we had with him about ayear ago (Glot International 3.5), Henk vanRiemsdijk warned that theoretical, and especiallygenerative, linguistics is “in considerable danger”in the sense that the interplay of forces in the fieldof linguistics — and academics more generally —does not seem to be geared at leading up to asituation which is favorable to us. The reason is:We don’t have enough friends — or even peoplewho understand what we are doing and why weare doing it. First of all, there are the linguistswho engage in the entirely honorable activity ofstudying language from other than theoreticalperspectives but who, for reasons that elude usmore and more, seem to hate theoretical linguisticsmore than their neighbor’s labradors shitting allover their front yard. Secondly, non-linguists —the general public, politicians, as well as otheracademics who may at some point in their careerbe in charge of allocating research money — alsodon’t seem to have a clear idea of what theoreticallinguistics is all about.

They can hardly be blamed, of course. Howshould they find out if nobody tells them about it?

So we were very happy that NRC Handels-blad, a Dutch daily, published a page-long articleon Universal Grammar in its Science and Educa-tion section of 23 January 1999 (“Wonderlijkewaren woorden wezenloos?”, p. 49). Written bystaff writer Hendrik Spiering, it is a very goodarticle. It has a few minor mistakes (for example,“Generative Grammar is an alternative term forUniversal Grammar”; somewhere else in thearticle the term “Universal Grammar” is confus-ingly used in the context of the grammar of thelanguage of the first Homo sapiens) and there arethings that we would have preferred to seephrased differently (Spiering says quite explicitlythat to Chomsky, structure is the only aspect oflanguage that counts, meaning not playing anyrole of importance whatsoever), but on the whole,this being a newspaper article, we have no realcomplaints. Here is a short summary of the article.

After commemorating Chomsky’s 70th birth-day and mentioning the fact that he is the onlyliving person in the Top Ten of the Arts and Hu-manities Citation Index (lower than Marx, Lenin,Shakespeare, Aristotle, the bible, Plato and Freud,but higher than Hegel and Cicero), Spiering goeson to summarize the essence of the innatenesshypothesis; he does this adequately. The next partof the article is devoted to criticism of the hypoth-esis. Spiering briefly discusses articles in recentissues of Behavioral and Brain Sciences andScience (also mentioning the contribution to thediscussion by Elizabeth Bates and Jeffrey Elmanin Science of 13 December 1996), which report onresearch the results of which give reason to ques-tion the validity of the assumptions underlyingthe innateness hypothesis. Furthermore, Spieringreports that a dissertation soon to be defended inGroningen by Willem de Graaf argues that it isprincipally impossible that something like aninnate Universal Grammar, or, for that matter, thecomplexity of human language, can have arisenas the result of evolutionary development. Spier-ing also visits Amsterdam professor Remko Scha,who tells him that the search for UG has met withso many problems and so much counterevidence,that he has decided to approach language in adifferent way.

The remainder of the article is for the otherside. Spiering mentions another Science article (of

1 January 1999), allegedly supporting the innate-ness hypothesis, and he talks to Eddy Ruys fromUtrecht, who gets ample opportunity to explain inquite some detail why, to him, the generativeenterprise is not so hopeless after all, primarilygoing into the question of parameters and lan-guage variation. Ruys also talks about methodolo-gy. The article ends with a quote from Ruys,saying that he will resign if UG turns out to be“an empty shell”: “I started this work in order toinvestigate a fundamental property of Homosapiens, and not in order to compare accidentalgrammatical structures.”

As we said, a good, fair article: here is ahypothesis, some researchers have reasons tobelieve that it cannot be right in its current form,other researchers have reasons to think it worth-while to investigate some of its consequences, andboth parties get due attention. So if we want tocomplain about the article at all, we can onlycomplain about the fact that the flag of the sub-headline, “Chomsky’s Universal Grammar underfire”, doesn’t cover the complete cargo — but thenagain, headlines are there to attract attention.

Recommended reading material for universityadministrators and others involved in the fundingof scientific research.

Then the letters came.In subsequent weeks, NRC Handelsblad

published seven letters-to-the-editor, responding toSpiering’s article (“Chomsky 1” to “Chomsky 7”; alldates and page numbers given below refer toissues of NRC Handelsblad; all dates are 1999.)With the exception of one, and maybe two (whichwill be left out of the following discussion), theseletters underscore the relevance of Van Riemsdijk’swarning.

The general picture that arises from theseletters is that generative linguistics is a fallacy, apseudoscience, an inbred personality cult (withChomsky as the guru) or, most probably, all of theabove.

For instance, incidentally showing that hedoesn’t seem to understand that, if all is well,science actually progresses, one of the letter writ-ers says that the people who follow Chomsky onlydo that out of personal loyalty — it cannot bebecause of his theories “because they changecontinuously” (“Chomsky 4”, 13 Feb, p. 50). Thesame writer says that he is relieved that “themedia”, after having “created Chomsky as a phe-nomenon”, finally start to “dissociate themselves”from him: after all, “many linguists have [foryears] been fed up with Chomsky’s pretensions toa priori knowing the truth and his contempt for[linguistic] facts.”

Similar sentiments are ventilated by anotherwriter, whose letter is actually so improper and sonasty that we think that NRC Handelsblad wasnot right in printing it. It paints a picture ofChomsky as an incompetent maniac who, as anunfortunate victim of an extremely bad linguisticeducation, has to “keep on pretending that he isoriginal” by constantly creating “new concoctions”(“Chomsky 6”, 20 Feb, p. 48). Regretfully, hecontinues, many talented linguists followed himon his hopeless quest. This writer, by the way, alsoseems to count it against Chomsky as a scientistthat he actually develops his ideas.

The suggestion that the generative enterprisehas nothing to do with science is further raised bystatements like the following: “That fanatic Chom-skyans exist ... is due to the lack of mathematical

knowledge among Arts-and-Humanities people”(“Chomsky 2”, 30 Jan, p. 50). “[Chomsky’s] work iswithout any mathematical elegance” (“Chomsky6”, 20 Feb, p. 48). “These abstract properties,which children would not be able to learn andwhich for that reason have to be innate, aremostly artefacts in the form of algorithms, inwhich Chomsky and his followers cast their gram-mars, without any linguistic necessity or plausibil-ity” (“Chomsky 1”, 30 Jan, p. 50). “Why do[Chomskyan linguists] take a more or less plausi-ble hypothesis for an absolute truth?” (“Chomsky4”, 13 Feb, p. 50). And so on and so forth.

Let’s not go into the question as to what themotives of these people are (although it is obviousthat some of the correspondents are driven by astrong personal dislike of Chomsky). Also, let’s nottry to find out how well-informed the writers are(some more than others, but it is a fact of life thateverybody seems to think that they are linguistsbecause they all speak a language (the sameeverybodies don’t claim to also be urologists de-spite the fact that they have a bladder and all thatand use it too)). Let’s not even ponder the possibili-ty of writing a letter to the newspaper ourselves(to write what? Defend ourselves? Against what?Spiering’s article was fine).

The point is, Van Riemsdijk’s warning mayhave come too late.

A very close relative called us up not too longago. A retired scientist, he is very interested inintellectual debate and he keeps himself up-to-date by reading everything he can get his handson. He has always been very interested in whatwe are doing and we talk about it regularly. Hehad also read the Spiering article, ánd the ensu-ing letters. So he telephoned us and asked (al-though we don’t think we were supposed to reallyanswer):

“So that is what you’ve gotten yourselvesinvolved in?”

Well, at least he didn’t say “implicated”.But it made us think of the university and

science fund administrators and we were secretlyhoping that they are so busy that they don’t havetime to read letters-to-the-editors.

CHOMSKY’S UNIVERSALGRAMMAR UNDER FIRE!

by Lisa Lai-Shen Cheng and Rint Sybesma

diss. Leiden |LOT Int’l Series 9University

Advertisement

Recent dissertations from LOT

Weightless SegmentsRob Goedemans

Weightless Segments deals with a specific question that arises when weregard the stress rules of quantity-sensitive languages. In theselanguages, vowels and coda consonants can, by virtue of theirpresence or absence, influence the weight of the syllable, and thus,indirectly, the location of word stress. It is common knowledge thatonset consonants do not have this capacity. The question that thisbook tries to answer is why.

A phonetic explanation for the weightless behaviour of syllableonsets is sought in the durational behaviour of onset, nucleus andcoda. Assuming that duration is the primary phonetic correlate ofphonological weight, an asymmetry in the durational behaviour ofthese subsyllabic constituents might explain the observed differencesin potential weight. A series of production and perceptionexperiments was conducted to reveal this asymmetry, and to explainit.

A further task that is undertaken in this book is the reanalysis of agroup of, mostly Australian Aboriginal, languages for which somedegree of onset influence in their stress rules has been claimed in thepast (in defiance of the universal rule that onsets do not count). In thesecond part of this book, after a general overview of stress inAboriginal languages, it is shown that these offending languages canall be reanalysed without any reference to onset influencewhatsoever.

Price EUR 28.49, NLG 63.00 (ex. 6% VAT where applicable,

ex. P&P). Individuals ordering directly from HAG are

eligible for a 33% discount.

Holland Academic GraphicsP.O. Box

The Hague

The Netherlands

53292

2505 AG

info hagpub.com

www.hagpub.com

fax: +

@

31 70 3595044

[ s c i e n t i f i c [ d o c u m e n t ] p r o c e s s i n g ]

Dissertations Glot International, Volume 4, Issue 4, April 1999 11

The focus of this thesis is the meaning of gradableadjectives like tall, dense, and bright, and thestructure and interpretation of the complex con-structions in which they appear, which I refer toas degree constructions. Roughly speaking,degree constructions are complex syntactic expres-sions formed out of an adjective and a degreemorpheme — an element of {er/more, less, as,too, enough, so, how, ...}. Some typical examplesare given in (1)–(7): comparatives, equatives, too-and enough-constructions, so...that-constructions,and how-questions.

(1)Mars Pathfinder was less expensive than previous missions toMars

(2)Venus is brighter than Mars

(3)Neptune is not as distant as Pluto

(4)The equipment is too old to be of much use to us

(5)Current spacecraft are not fast enough to approach the speedof light

(6)The black hole at the center of the galaxy is so dense thatnothing can escape the pull of its gravity, not even light

(7)How bright is Alpha Centauri?

Degree constructions, and comparatives in partic-ular, have been the focus of a large body of workon the syntax and semantics of gradable adjec-tives. The syntactic complexity of these construc-tions has been a topic of investigation since earlywork in generative grammar (see e.g. Lees 1961;Smith 1961; Pilch 1965; Huddleston 1967; andHale 1970), and has formed the basis for impor-tant developments in theories of phrase structure,ellipsis, quantification, and long-distance move-ment (see e.g. Bresnan 1973, 1975; Hankamer1973; Jackendoff 1977; Chomsky 1977; Kuno1981; Pinkham 1982; Napoli 1983; Abney 1987;Corver 1990, 1997; Gawron 1995; and Hazout1995). On the semantic side, degree constructionshave provided the empirical foundation for inves-tigations of the meaning of gradable adjectives and,more generally, the expression of ordering relationsin natural language since at least Sapir 1944 (seealso Bartsch & Vennemann 1973; McConnell-Ginet1973; Seuren 1973, 1978; Kamp 1975; Cresswell1976; Klein 1980; Atlas 1984; Bierwisch 1989;Sánchez-Valencia 1994, and others).

This thesis continues in this tradition by usingdegree constructions (and comparatives in particu-lar) as an empirical basis to motivate and developa semantic analysis of gradable adjectives asmeasure functions: functions from objects toabstract representations of measurement or de-grees (cf. Bartsch & Vennemann 1973). Thisanalysis falls within a broader class of “scalaranalyses”, in which the core meaning of a grada-ble adjective is defined in terms of an ordered setof degrees, or scale (see e.g. Cresswell 1976;

Bierwisch 1989; see Klein 1991 for useful over-view). It contrasts with an alternative “vaguepredicate analysis”, in which gradable adjectivesare analyzed as predicative expressions whoseextensions may vary from context to context (seee.g. McConnell-Ginet 1973; Kamp 1975; Klein1980). The first part of the thesis presents anoverview of these two types of approaches, andargues that only a scalar analysis provides anadequate explanation of several important sets offacts, including the anomaly of two types of com-parative constructions — examples involvingincommensurable adjectives, such as (8), andexamples involving adjectives of opposite polarity,such as (9) — and the semantic characteristics of“comparison of deviation” constructions such as(10).

(8)??My copy of The Brothers Karamazov is heavier than it is old

(9)??The Brothers Karamazov is longer than The Dream of aRidiculous Man is short

(10)William is as tall as Robert is short

(11)William’s feet are as wide as Robert’s feet are long

Unlike the more typical equative in (11), (10)entails that the properties predicated of the com-pared objects hold in the absolute. Moreover, (10)does not represent a claim that William and Robertare equal in height (an interpretation parallel tothat of (11), and completely impossible), but ratheran assertion that the extent to which Williamexceeds some standard of tallness is the same asthe extent to which Robert exceeds some standardof shortness (see Kennedy 1997 for additionaldiscussion). The problem with a vague predicateanalysis is that it not only fails to account for theanomaly of examples like (8) and (9), it also doesnot derive the inferences associated with (10).Given an appropriate formalization of degrees,however (see below), a scalar analysis explains theanomaly of (8) and (9) in terms of basic principlesof ordering relations, and also accounts for theinterpretation and inferences of (10).

Although the analysis of gradable adjectivesthat I develop in this thesis falls into the generalfamily of scalar analyses, it differs from traditionalscalar approaches in two fundamental ways: (a) interms of the semantic type of gradable adjectives;and (b) in terms of the analysis of degree construc-tions. Typically, scalar analyses characterizegradable adjectives as relational expressions,specifically, as relations between objects anddegrees. On this view, an adjective like brighttakes two arguments, a degree and an individual,and establishes a relation between them:bright(x,d) is true just in case the degree to whichx is bright is at least as great as d. In principle,this variable should provide a position for a quan-tificational expression to bind; in most scalarapproaches, this is exactly the function of a degreeconstruction. (See, for example, Hellan 1981;Hoeksema 1983; von Stechow 1984a; and Heim1985 and others for analyses of comparatives in

terms of existential quantification over degrees,and see Cresswell 1976; Moltmann 1992; andHendriks 1995 for analyses in terms of universalquantification.) If degree constructions involvequantification over degrees, however, then wewould expect them to interact with other quantifi-cational operators to trigger configurational scopeambiguities. Focusing on the interpretation ofcomparatives, I demonstrate that this expectationis not borne out: the quantificational force intro-duced by a comparative does not participate inscope relations.

To account for the apparently restricted scopalpossibilities of comparatives, I introduce an alter-native, non-quantificational analysis of degreeconstructions, built around the assumption thatgradable adjectives denote measure functions.Specifically, I claim that propositions formed out ofa gradable adjective N contain three primarysemantic constituents (cf. Russell 1905): (a) areference value, which indicates the projectionof the target of predication onto the scale associat-ed with N; (b) a standard value, which is intro-duced either contextually, by a measure phrase(e.g. 2 meters in the telescope is 2 meters long), orby the comparative clause (the complement ofthan or as); and (c) a degree relation, which isintroduced by degree morphology and is assertedto hold between the reference value and thestandard value. To this end, I analyze degreeconstructions as properties of individuals definedin terms of relations between degrees, as shown in(12), where G is a gradable adjective meaning (afunction from objects to degrees), G(x) is thereference value, s is the standard value, and R isa degree relation.

(12)8x[R(G(x))(s)]

This proposal is implemented in the context of thesyntactic analysis of degree constructions devel-oped in Abney 1987; Corver 1990, 1997; andGrimshaw 1991, in which adjectives project ex-tended functional structure headed by degreemorphology. Specifically, the adjective that headsthe extended projection introduces G in (12), thedegree morpheme introduces R, and the compara-tive clause (or measure phrase) introduces s. Thisis illustrated by the tree in (13), which schemati-cally represents the structure of a comparative.

(13)DegP

Deg¢

Deg¢ XP

sDeg AP

n8 8 8G s x G x s[ ( ( ))( )]R

In effect, the analysis defended in this thesisrepresents a kind of decompositional approach toadjective meanings. In traditional scalar analyses,the meaning of a gradable adjective contains ameasure function — in order to determine thedegree to which x is bright, it is necessary todefine a function from x to the scale associatedwith the adjective bright — but this is not the coremeaning of the adjective. Instead, the meaning ofthe adjective is characterized in terms of a meas-ure function plus a relational component, typicallya partial ordering relation (though see Carston1988 and Horn 1992 for relevant discussion). Incontrast, the analysis outlined here removes therelational component from the adjective meaning,locating it in the degree morphology. The result isa division of labor between the adjective — themeasure function — and the degree morphology, adivision that is reflected in the independentlymotivated syntactic structure of the AP/DegPextended projection. Chapter 2 of the thesis pro-vides initial support for this approach — first, byshowing that it accounts for the scopal characteris-tics of comparatives, and second, by demonstratingthat it provides the basis for a general, composi-

PROJECTING THE ADJECTIVE:THE SYNTAX AND SEMANTICS OFGRADABILITY AND COMPARISON

by Chris Kennedyreviewed by Petra Hendriks

Dissertations Glot International, Volume 4, Issue 4, April 1999 12

tional analyses of a broad range of degree con-structions in both the comparative and absoluteform. What remains is to show that the approachextends to an insightful account of degree con-structions headed by too, enough, so and how aswell.

The thesis concludes by addressing the ques-tion: what sort of objects are degrees? Taking threesets of empirical phenomena as a starting point —the anomaly of comparatives formed out of anton-ymous pairs of adjectives (see (8) above), theinterpretation of comparison of deviation construc-tions (see (9)), and the monotonicity properties ofpolar adjectives (see Seuren 1973, 1978; Ladusaw1979; Linebarger 1980; Sánchez-Valencia 1990;Kennedy to appear), I argue that degrees shouldbe formalized not as points on a scale, as stand-ardly assumed, but as intervals on a scale, orextents (cf. Seuren 1978; von Stechow 1984b;Bierwisch 1989; Löbner 1990). Given the assump-tion that antonymous pairs of adjectives mapobjects onto the same scale (for example, both talland short map objects onto a general scale ofheight), an analysis in which degrees correspondto points actually predicts that the anomaloussentence (14) should be logically equivalent to(15), because the degree to which Mike is short isthe same as the degree to which he is tall (cf.Rullmann 1995).

(14)??Carmen is taller than Mike is short

(15)Carmen is taller than Mike

I demonstrate that if degrees are instead formal-ized as extents, and if adjectival polarity is charac-terized in terms of a sortal distinction betweenpositive and negative adjectives based on a struc-tural difference between the objects in their rang-es (i.e., the scalar intervals onto which they projecttheir arguments), then the anomaly of exampleslike (14) is due to the fact that the ordering rela-tion introduced by the comparative morpheme isundefined for extents of opposite polarity. I con-clude by showing that the assumptions aboutextents and adjectival polarity that provide anexplanation for the anomaly of examples like (14)also provide the basis for principled explanationsof the semantic characteristics comparison ofdeviation constructions and the monotonicityproperties of gradable adjectives.

Reviewby Petra Hendriks

Jack Hoeksema once remarked that “[i]f the realmof language is seen as a cosmos, vast, largelyunexplored and sometimes bewildering, then thecomparative construction must be a microcosm,reflecting all the complexity of the whole” (Hoekse-ma 1984, 93). Many linguists and philosophers setout to explore this mysterious microcosm, usingdifferent tools and working from different theoreti-cal perspectives. Yet, the map of the comparativeconstruction still shows several white spots andunknown regions, indicating the sometimes almostintangible nature of comparatives due to its intri-cate interaction with a variety of syntactic andsemantic phenomena. In his dissertation, ChrisKennedy has taken up the challenge to furtherexplore the microcosm of the comparative con-struction, the aim of his research already beingsuggested by the choice of the example sentences,in which space shuttles, space telescopes, MarsPathfinders and other instruments for cosmicexploration feature prominently as objects ofcomparison.

Kennedy’s dissertation concentrates on thesemantic and syntactic representation of gradableadjectives and predicative AP comparatives inEnglish. Other English degree constructions, suchas too constructions and enough constructions,and other types of comparatives, are only touchedupon briefly. Proceeding from an analysis of

gradable adjectives as measure functions,Kennedy provides both a semantic representationand the corresponding syntactic structure for arange of predicative AP comparatives. Part of themotivation for his analysis stems from an investi-gation of the interpretation of constructions in-volving incommensurability, cross-polar anomalyand comparison of deviation, that is, constructionsthat have not received a comprehensive treatmentbefore.

One of the strong points of Kennedy’s disser-tation is his attempt to combine a semantic analy-sis of degree constructions with a syntacticanalysis, thereby showing that a compositionalsemantic analysis of degree constructions is possi-ble while assuming a syntactic structure of degreeconstructions as developed in, e.g., Abney (1987).In this respect, two traditionally more or lessdistinct lines of research are brought together inthis well-written and sensibly organized disserta-tion. On the whole, this dissertation is an excellentpiece of work, offering clear discussions on severaltheoretical issues and providing careful argumen-tation. Although in the rest of this review I willpoint out a number of possible weaknesses of theproposal and mention some points of debate, in myview this dissertation presents a very interestingand stimulating discussion of gradable adjectivesand comparatives.

1. Gradable adjectivesPredicative AP comparatives such as (1) are

characterized by the presence of a gradable adjec-tive.

(1)Jupiter is larger than Saturn (is)

There is no doubt that the gradable adjective largecontributes to the meaning of the comparative in(1), but the question that is central to the first partof Kennedy’s dissertation is what it actuallycontributes. Kennedy answers this question in twosteps. First, he argues that the ordering relationassociated with the meaning of a gradable adjec-tive must actually be part of its meaning insteadof some inherent property of the domain of agradable adjective. In a second step, he arguesthat a degree argument is not part of the meaningof a gradable adjective, in contrast to what isgenerally assumed.

1.1. ScalesThe motivation behind the first conclusion lies

in the unacceptability of comparatives involvingan antonymous pair of adjectives, such as large/small and bright/dim. If the ordering relationassociated with the meaning of a gradable adjec-tive would be some property of the domain of theadjective, the domain of one of the adjectives of anantonymous pair should simply be the inverse ofthe other. Only then is it possible to account forthe generalization that if a is larger than b, then bis smaller than a. However, under this assumptiona sentence like (2) should be acceptable:

(2)*Venus is brighter than Mars is dim

Nevertheless, this sentence is unacceptable, illus-trating cross-polar anomaly. Therefore, Kennedyconcludes that the ordering relation associatedwith the meaning of a gradable adjective must bepart of the meaning of the gradable adjective. Inparticular, the adjective imposes an ordering on itsdomain by relating objects to points, or degrees, ona scale. If adjectives relate objects in their domainto degrees on a scale, a plausible way to accountfor the unacceptability of (2) would be to assumethat every adjective is associated with its ownscale. This would also explain the unacceptabilityof a sentence like (3), which involves two un-related adjectives:

(3)*Morton is as tall as Richard is clever

Because the adjectives are distinct, the scales

associated with the adjectives must also be dis-tinct. Hence, the comparison in (3) is undefined.Thus, the anomaly of (3) is explained by theuncommensurability of tall and clever. The unac-ceptability of (2), on the other hand, ultimatelyreceives a different explanation. Kennedy con-vincingly argues that positive adjectives denotefunctions from objects to positive intervals (i.e.,intervals from the lower end of the scale, usual-ly 0, to some positive point) and negative adjec-tives denote functions to negative intervals (i.e.,intervals from some positive point to ∞) on thesame scale. Since the range of positive adjectivesand the range of negative adjectives are formedby different parts of the same scale, their projec-tions of objects on the scale are distinct and hencecannot be compared.

In the light of the previous examples, it issurprising that the following sentence is accepta-ble:

(4)Fortunately, the ficus was shorter than the ceiling was low, sowe were able to get it into the room

Because (4) is interpretable, Kennedy claims thatthe pair of adjectives in (4) must be associatedwith the same scale or with very similar scales.Under this view, the adjectives short and lowintroduce orderings according to different aspectsof the same basic property, namely some notion oflinear extent. However, if this were correct, wewould expect (5) to be acceptable as well, since it ispossible to measure both the tallness of an individ-ual and the thickness of a book in, for example,inches.

(5)*Morton is taller than this book is thick

Nevertheless, (5) seems as unacceptable as (3).The important difference seems to be that there issome causal relation involved in (4) that is notinvolved in (5), suggesting that the acceptabilityof (4) might be caused by certain pragmatic factorswhich somehow license a mapping between twodistinct scales. However, Kennedy will not want togeneralize this by claiming that every adjective isassociated with its own scale. The assumption thatadjectives of opposite polarity project onto thesame scale is crucial to his detailed and convincingexplanation of a number of different observationswith respect to gradable adjectives and compara-tives, such as the fact that only positive adjectivescan be modified by a measure phrase, the fact thatonly negative adjectives license negative polarityitems and the different interpretation that com-paratives of deviation receive.

1.2. QuantificationAlthough the ordering relation associated

with the meaning of a gradable adjective must bepart of its meaning, Kennedy argues that a de-gree argument is not part of its meaning. Hisargumentation focuses on comparatives. Compar-atives are typically analyzed as quantificationalexpressions, quantifying over degrees introducedby the adjectives. The logical representation of agradable adjective would thus be adjective(x,d),where x is the object of predication and d thedegree introduced by the adjective. However, ifcomparatives are quantificational expressions,they would participate in scope ambiguities withquantified NPs and negation. Kennedy arguesthat this prediction is not borne out. He concludesthat standard quantificational analyses of com-paratives must therefore be wrong in their as-sumption that gradable adjectives introduce adegree argument. Although the examplesKennedy presents do not show any scope ambigu-ities, there exist comparatives that do seem toshow scope ambiguities.

(6)Everyone kissed someone

(7)Everyone ran faster than Jacky expected

Dissertations Glot International, Volume 4, Issue 4, April 1999 13

In the same way as (6) is ambiguous between areading in which the kissed people co-vary withthe kissers and a reading in which there is oneperson that is being kissed by everyone, (7) seemsambiguous between a reading in which Jacky haddifferent expectations about the speed of differentrunners and a reading in which there is onemaximal speed which Jacky expected that norunner would exceed. If this ambiguity indeedexists, and if it cannot be explained in the sameway Kennedy explains the ambiguity of contradic-tory comparatives in counterfactuals, the impossi-bility of other comparatives to participate in scopeambiguities might require some other explanation.

In chapter 2, Kennedy presents anotherargument against this relational analysis ofgradable adjectives, namely the problem of compo-sitionality. A relational analysis of adjectives isbased on the assumption that comparison is apsychological primitive, and that the interpreta-tion of gradable adjectives should be stated interms of such a relation. To refute this assumption,Kennedy presents sentence (8) as an example thatnot all degree expressions involve a notion ofcomparison.

(8)Pug is too stinky to go to the Ritz

According to Kennedy, the best characterization ofthe meaning of this sentence is “the degree ofPug’s stinkiness makes it impossible for him to goto the Ritz” (cf. Moltmann 1992, 301), which doesnot make reference to a comparison relation. Butthis characterization cannot be correct, since itwould also apply to the (very unlikely) situation inwhich Pug is not stinky enough to go to the Ritz.This obviously is not a valid implication of (8). Abetter characterization of the meaning of (8)would be “the degree of Pug’s stinkiness is greaterthan the maximal degree of stinkiness that makesit possible for him to go to the Ritz”, which doesmake reference to a comparison relation.

Kennedy uses the two arguments discussedabove to motivate his conclusion that gradableadjectives do not introduce a degree argument. Toaccount for the interpretation of expressionsinvolving a gradable adjective as relations be-tween objects and degrees on a scale, he has toassume that the extended projection of the adjec-tive must be headed by a phonologically nulldegree morpheme, which introduces both thedegrees and an ordering relation between degreesinto the semantic representation. This way,Kennedy is also able to present a completelyparallel analysis of absolute constructions involv-ing gradable adjectives and comparative construc-tions. A drawback of this analysis of absoluteconstructions is that, figuratively speaking, all thework with respect to the semantics of the construc-tion is done by a null element. Since I am notquite convinced by the presented arguments, theintroduction of this null element makes me feelslightly uncomfortable.

2. ComparativesIn scalar approaches to comparison, compara-

tives are traditionally assumed to compare twodegrees on a scale. The question is how thesedegrees are introduced into the semantic represen-tation if they are not introduced by the adjectives.The only plausible alternative is that these de-grees are introduced by the comparative mor-pheme (i.e., more/-er, less, fewer, as). Kennedycorrectly does not consider the possibility that thecomparative conjunctions than and as introducethese degrees, since comparison is not dependenton the presence of a comparative conjunction,witness (9).

(9)The Sojourner rover is not very long, but it is even less wide

The assumption that the comparative morphemeintroduces the degrees into the semantic represen-tation forms the basis of Kennedy’s semantic andsyntactic analysis of a number of different com-

parative constructions.

2.1. A compositional semanticsComparatives exhibit a whole range of ellipsis

phenomena. The comparative clause, i.e., thecomplement of a comparative conjunction than oras, can be an apparently complete clause (a so-called comparative subdeletion construction), butcan also be a reduced clause (a comparative dele-tion construction) or even a single phrase (whatKennedy terms a phrasal comparative but is alsoknown as a comparative ellipsis construction).These constructions are exemplified in (10), (11)and (12), respectively:

(10) Comparative subdeletionThe Sagan Memorial Station is taller than the Sojourner roveris long

(11) Comparative deletionThe Mars Pathfinder mission was more successful than anyonethought it would be

(12) Comparative ellipsis (phrasal comparative)The Mars Pathfinder mission was less expensive than theViking missions

A semantic analysis of these constructions canproceed along different lines. One option would beto derive the meaning of a reduced comparativefrom the meaning of a full comparative throughdeletion under identity or LF copying. Anotheroption would be to derive the meaning of a re-duced comparative directly, i.e., without syntacticreconstruction. In the latter case, the composition-al burden of the interpretation of the comparativeconstruction rests on the comparative morpheme.It is this latter approach that Kennedy takes, achoice which he motivates through the existenceof comparative constructions in which the deletedmaterial and the antecedent material are clearlynon-identical. Syntactically, Kennedy assumesthat the missing material in both types of clausalcomparatives is the trace of a null operator inSpecCP of the comparative clause; in (10), theoperator is of the category of a degree variableand in (11), the operator is of category DegP.

Because he advocates a direct interpretationapproach, Kennedy has to posit three distinctinterpretations for the comparative morphemes:one for comparative subdeletion constructions, onefor comparative deletion constructions and one forphrasal comparatives. In all cases, the compara-tive morpheme denotes an ordering relation be-tween a reference value, which indicates thedegree to which the subject has the propertydenoted by the adjective, and a standard value,which corresponds to some other degree. Theinterpretation of the comparative morphemedetermines how the standard value is derivedfrom the comparative clause, and in Kennedy’sproposal indirectly reflects the elements containedwithin the comparative clause. As an illustration,compare the interpretation of more in comparativedeletion constructions in (13) with its interpreta-tion in phrasal comparatives in (14):

(13)more2 = 8G8Q8x[MORE(G(x))(Q(G))]

(14)more3 = 8G8y8x[MORE(G(x))(G(y))]

Here, G is the function denoted by the gradableadjective, x the subject of the matrix clause, y anindividual and Q a function from gradable adjec-tives to degrees. In both cases, the comparativemorpheme combines with a gradable adjectivewhich applies to the subject to yield the referencevalue (this step is logically represented by G(x) in(12) and (13)). Furthermore, the comparative mor-pheme in (13) together with the adjective combinewith a function which applies to the gradable adjec-tive of the matrix clause (i.e., Q(G)) to yield thestandard value. In other words, this comparativemorpheme co-occurs with a than-clause that islacking a gradable adjective. The comparativemorpheme in (14), on the other hand, combineswith an individual which is the argument of the

gradable adjective of the matrix clause (i.e., G(x)).Thus, this comparative morpheme co-occurs with athan-phrase containing a DP subject.

A disadvantage of this approach is that thenumber of interpretations for the comparativemorpheme is not limited to three. For every com-parative with a differently structured comparativeclause or comparative phrase, it will be necessaryto posit a new interpretation for the comparativemorpheme. For example, Kennedy’s three inter-pretations for the comparative morpheme do notderive the meaning of the discourse comparativein (9). In this comparative, the degree to whichthe Sojourner rover is wide (which is denoted bythe matrix clause) is compared to the degree towhich it is long (which is not denoted by a compar-ative clause but must be derived from the preced-ing clause). So the standard value of this discoursecomparative is dependent on another adjectivethan the one in the matrix clause and, moreover,there is no material present corresponding toeither Q or y. Yet another interpretation will haveto be introduced for the construction in (15), whichKennedy presents as a comparative deletionconstruction but which obviously differs fromstandard comparative deletion constructions.Clearly, the material lacking from the than-clausein (15) does not correspond to just a gradableadjective.

(15)The telescope was less expensive than I expected

If NP comparatives and sentential comparativesare also considered, the number of required inter-pretations for the comparative morpheme increas-es even more. This explosion of interpretations forthe comparative morpheme is the cost of giving upLF reconstruction as a way to arrive at a semanticrepresentation for incomplete comparatives. Ofcourse, a reconstruction approach would give riseto its own problems with respect to the enormousrange of possibilities of ellipsis in comparativeconstructions. One of these problems will be dis-cussed in section 2.2. A clear advantage ofKennedy’s approach is that it correctly predictsthat, whatever material can be omitted from acomparative construction, the comparative mor-pheme can never be omitted.

2.2. Ellipsis versus direct interpretationA puzzle for a reconstruction approach to

elliptical comparatives is formed by the localdependency of comparative deletion. Kennedyobserves that the interpretation of the missingadjective in comparative deletion comparativesdiffers from the interpretation of the missing VP inVP ellipsis constructions, in that the former exhib-its a local dependency:

(16)The table is wider than this rug is, but this rug is longer thanthe desk is

(17)Marcus read every book I did, and I bought every book Charlesdid

A characteristic of VP ellipsis is that an elided VPcan typically locate its antecedent from any acces-sible VP within recent discourse. Therefore, the VPellipsis example in (17) is ambiguous between areading in which I bought every book Charlesbought (involving a local antecedent) and a read-ing in which I bought every book Charles read(involving a non-local antecedent). In contrast,the missing adjective in the comparative deletionconstruction in (16) only receives a local interpre-tation. Surprisingly, the following comparative,which is minimally distinct from (16) and has asubdeletion construction as its first conjunct, isambiguous.

(18)The table is longer than this rug is wide, and this rug is longerthan the desk is

The second conjunct of (18) has a reading inwhich the rug is longer than the desk is long

Dissertations Glot International, Volume 4, Issue 4, April 1999 14

(reading (i)) and a reading in which the rug islonger than the desk is wide (reading (ii)). Ifcomparative deletion involves reconstruction ofelided material, the first conjuncts in (16) and (18)would be completely parallel at LF. Since (18) isambiguous but (16) is not, this is another argu-ment for a direct interpretation approach to com-parative deletion. Unfortunately, in order toexplain the second reading of (18), Kennedyweakens his initial position substantially by as-suming that the ambiguity of (18) results from thepossibility of two different derivations: one inwhich the second conjunct is a non-ellipticalcomparative deletion construction, and one inwhich the second conjunct is a subdeletion con-struction which has undergone ellipsis underidentity with the first conjunct. The problem withthis newly introduced ellipsis operation, however,is that it does not resemble any other ellipsisphenomenon occurring in coordinate constructionsand should thus be limited to comparative con-structions occurring within coordinate construc-tions. For example, this operation cannot be theprevisouly discussed operation of VP ellipsis, sincethe construction in (18) with its two interpreta-tions is also possible in Dutch, whereas Dutch doesnot allow for VP ellipsis. Moreover, if the structurecorresponding to reading (ii) is the result of anoperation of ellipsis under identity, reading (ii)should be equally available for the followingsentence:

(19)The table is wider than this rug is wide, and this rug is longerthan the desk is

The sentence in (19) differs from (18) in thatlonger in the first conjunct of (18) is replaced bywider in (19). This does not have any effect on theidentity between the supposedly elided constituentand its antecedent, so it should not change thepossible interpretations for the second conjunct.The only effect of this change is that it decreasesthe parallelism between the two conjuncts. Sur-prisingly, reading (ii) seems almost unavailable forthe second conjunct in (19). This strongly suggeststhat what is responsible for the interpretation of(18) and (19) is not a purely syntactic operation ofellipsis under identity, but rather a more freelyapplicable form of ellipsis. Under this latter viewon ellipsis, ellipsis resolution should be determinedby non-syntactic factors as well, such as parallel-ism with respect to the preceding sentence, possi-ble prosodic effects associated with this parallelismand general pragmatic constraints that, for exam-ple, summon to anaphorize text when possible. Infact, such an approach to ellipsis resolution isproposed by Hendriks & De Hoop (1998), based onthe interpretation of elliptical quantified sentenc-es. If this approach to ellipsis is correct and if it istrue that certain comparative constructions receivea direct interpretation, as Kennedy argues, then itmight be possible to eliminate ellipsis under identi-ty altogether.

3. ConclusionKennedy’s dissertation is a careful and de-

tailed study of the semantic and syntactic repre-sentation of gradable adjectives and predicativeAP comparatives. Kennedy argues persuasivelyfor a direct interpretation approach to ellipticalcomparative constructions. This conclusion isimplemented in a syntactic analysis of compara-tives which supports a straightforward composi-tional semantic analysis of these constructions.

ReferencesAbney, S. (1987). The English noun phrase in its

sentential aspect. Doctoral dissertation, Massa-chusetts Institute of Technology.

Atlas, J. (1984). Comparative adjectives and adverbi-als of degree: An introduction to radically radicalpragmatics. Linguistics & Philosophy 7, 347–377.

Bartsch, R. & T. Vennemann (1972). Semantic struc-tures: A study in the relation between syntax andsemantics. Frankfurt: Athenäum Verlag.

Bierwisch, M. (1989). The semantics of gradation.Dimensional adjectives, edited by M. Bierwisch

& E. Lang, 71–262. Berlin: Springer-Verlag.Bresnan, J. (1973). Syntax of the comparative clause

construction in English. Linguistic Inquiry 4.3,275–343.

Bresnan, J. (1975). Comparative deletion and con-straints on transformations. Linguistic Analysis1, 25–74.

Carston, R. (1988). Implicature, explicature, andtruth-theoretic semantics. Mental representa-tions: The interface between language and reality,edited by R. Kempson, 155–181. Cambridge:Cambridge University Press.

Chomsky, N. (1977). On wh-movement. Formalsyntax, edited by P. Culicover, T. Wasow & A.Akmajian. New York: Academic Press.

Corver, N. (1990). The syntax of left branch construc-tions. Doctoral dissertation, Tilburg University.

Corver, N. (1997). Much-support as a last resort.Linguistic Inquiry 28.1, 119–164.

Cresswell, M.J. (1976). The semantics of degree.Montague grammar, edited by B. Partee. NewYork: Academic Press.

Gawron, J.-M. (1995). Comparatives, superlatives,and resolution. Linguistics & Philosophy 18,333–380.

Grimshaw, J. (1991). Extended projection. Manu-script, Brandeis University.

Hale, A. (1970). Conditions on English comparativeclause pairings. Readings in English transforma-tional grammar, edited by R. Jacobs & P. Rosen-baum. Washington: Georgetown UniversityPress.

Hankamer, J. (1973). Why there are two than’s inEnglish [CLS 9]. University of Chicago.

Hazout, I. (1995). Comparative ellipsis and LogicalForm. Natural Language and Linguistic Theory13, 1–37.

Heim, I. (1985). Notes on comparatives and relatedmatters. Manuscript, University of Texas, Aus-tin.

Hellan, L. (1981). Towards an integrated analysis ofcomparatives. Tübingen: Narr.

Hendriks, P. (1995). Comparatives and categorialgrammar. Doctoral dissertation, University ofGroningen.

Hendriks, P. & H. de Hoop (1998). On the interpreta-tion of semantic relations in the absence ofsyntactic structure. Prepublication TCW-1998–3,University of Groningen.

Hoeksema, J. (1983). Negative polarity and thecomparative. Natural Language and LinguisticTheory 1, 403–434.

Hoeksema, J. (1984). To be continued: The story ofthe comparative. Journal of Semantics 3, 93–107.

Horn, L. (1992). The said and the unsaid [SALT 2].Ithaca: Cornell University Linguistics.

Huddleston, R. (1967). More on the English compara-tive. Journal of Linguistics 3, 91–102.

Jackendoff, R. (1977). X-bar syntax. Cambridge, Mass:MIT Press.

Kamp, H. (1975). Two theories of adjectives. Formalsemantics of natural language, edited by E.Keenan, 123–155. Cambridge: Cambridge Uni-versity Press.

Kennedy, C. (1997). Comparison and polar opposition[SALT 7]. Ithaca: Cornell University Linguistics.

Kennedy, C. (to appear). On the monotonicity of polaradjectives. Perspectives on Negation. Amster-dam: Benjamins.

Klein, E. (1980). A semantics for positive and compar-ative adjectives. Linguistics & Philosophy 4.1, 1–45.

Klein, E. (1991). Comparatives. Semantik: Ein inter-nationales Handbuch der zeitgenössischenForschung, edited by A. von Stechow & D.Wunderlich. Berlin: Walter de Gruyter.

Kuno, S. (1981). The syntax of comparative clauses[CLS 17]. University of Chicago.

Ladusaw, W. (1979). Polarity as inherent scope rela-tions. Ph.D. dissertation, University of Texas,Austin.

Lees, R. (1961). Grammatical analysis of the Englishcomparative construction. Word 17.2, 171–185.

Linebarger, M. (1980). The grammar of negativepolarity. Ph.D. dissertation, MassachusettsInstitute of Technology.

Löbner, S. (1990). Wahr neben Falsch: Duale Opera-toren als die Quantoren natürlicher Sprache.Tübingen: Niemeyer.

McConnell-Ginet, S. (1973). Comparative construc-tions in English: A syntactic and semanticanalysis. Doctoral dissertation, University ofRochester.

Moltmann, F. (1992). Coordination and comparatives.Ph.D. dissertation, Massachusetts Institute ofTechnology.

Napoli, D. (1983). Comparative ellipsis: A phrasestructure analysis. Linguistic Inquiry 14.4, 675–694.

Pilch, H. (1965). Comparative constructions in Eng-lish. Language 41.1, 37–58.

Pinkham, J. (1982). The formation of comparativeclauses in French and English. Doctoral disserta-tion, Indiana University.

Rullmann, H. (1995). Maximality in the semantics ofwh-constructions. Ph.D. dissertation, Universityof Massachusetts, Amherst.

Russell, B. (1905). On denoting. Mind 14, 479–493.Sánchez-Valencia, V. (1994). Polarity, predicates, and

monotonicity. Manuscript, University of Gronin-gen.

Sapir, E. (1944). Grading: A study in semantics.Philosophy of Science 11, 93–116.

Seuren, P.A.M. (1973). The comparative. Generativegrammar in Europe, edited by F. Kiefer & N.Ruwet. Dordrecht: Riedel.

Seuren, P.A.M. (1978). The structure and selection ofpositive and negative gradable adjectives. Papersfrom the Parasession on the Lexicon [CLS 14],edited by D. Farkas, W. Jacobson & K. Todrys,336–346. University of Chicago.

Smith, C. (1961). A class of complex modifiers inEnglish. Language 37, 342–365.

Stechow, A. von (1984a). Comparing semantic theo-ries of comparison. Journal of Semantics 3, 1–77.

Stechow, A. von (1984b). My reaction to Cresswell’s,Hellan’s, Hoeksema’s, and Seuren’s comments.Journal of Semantics 3, 183–199.

This book examines the syntax and semantics of Salishdeterminers and quantifiers in detail; one of the findings isthat Salish, though it possesses a robuß syßem of DP-internal quantification, lacks quantificational determiners.

Lisa Matthewson proposes an account of diƒerences in thedeterminer and quantification syßems of Salish andEnglish. In particular, Salish and English exemplifyopposite settings of a Common Ground Parameter, whichmeans that Salish determiners may not access the commonground of the discourse. This parameter also derivesseveral other diƒerences between Salish and Englishdeterminers, such as the absence of a definitenessdißinçion in Salish.

Lisa MatthewsonDeterminer systems and quanti cational strategiesïEvidence from Salish (World Theses, 1)

“This book is an indepth ßudy of the quantificational component of the

grammar of an Interior Salish language. It is an important contribution, not

only to the ßudy of quantification in universal grammar, but to Salish

linguißics as well, because it documents significant variation in the syntax of

quantification within the language family.”

Professor Eloise Jelinek, University of Arizona

1998. ISBN 90-5569-041-4. xvi + 380 pp. Hardbound.

Price: NLG 119.40 (ex. 6% VAT where applicable, ex. P&P)Individuals ordering directly from Holland Academic Graphics are

eligible for a 33% discount.

Holland Academic Graphics

s c i e n c e p u b l i s h i n g

P.O. BoxThe Hague

The Netherlands

532922505 AG

mail hagpub.comwww.renemulder.comfax: +

@

31 70 3595044

Advertisement

René Mulder

Dissertations Glot International, Volume 4, Issue 4, April 1999 15

Summaryby the author

From a phonological point of view, morphologicalwords, i.e. syntactic atoms, do not necessarilybehave as a unit. For instance, derivational affix-es and compound members can be treated inde-pendently by phonological word-level rules. Theprosodic word has been defined in order to accountfor the non-isomorphy between morphology andphonology. Prosodic words are typically character-ized as being the domain of word stress, phonotac-tics and segmental word-level rules. This thesisdeals with various aspects concerning the defini-tion of the prosodic word in the realm of deriva-tion, compounding, and cliticization. In addition, itaddresses several morphological issues; given thelimitations on the length of the present articlethough, I will leave these aside.

The prosodic word is but one element in aseries of hierarchically ordered phonological con-stituents known as the prosodic hierarchy (Selkirk1981, 1986; Nespor & Vogel 1986). The StrictLayer Hypothesis (SLH) determines the geometryof this constituent structure, as follows:

(1) Strict Layer Hypothesis (Selkirk 1984; Nespor & Vogel1986)

a. A given non-terminal unit is composed of one or moreunits of the immediately lower category

b. A unit of a given level is exhaustively contained in thesuperordinate unit of which it is part

The first clause concerns prosodic domination; itrequires each prosodic constituent to directlydominate constituents of the immediately lowercategory only. The second clause concerns theformation of well-formed prosodic trees, in that itdemands each string to be parsed exhaustivelyinto non-overlapping domains. Both clauses ap-pear to be problematic with respect to the forma-tion of prosodic words.

As to the first clause, the prosodization ofaffixes, clitics, and compound members can induceviolations of the requirements on prosodic domina-tion. Specifically, some of these elements neitherincorporate into an adjacent prosodic word norform an independent prosodic word. An example isprovided by prefixation in Spanish.

In Spanish, words cannot begin with [s]followed by another consonant; a rule of e-epen-thesis applies at the left edge of underlying /sC/-clusters, as is shown in (2a). Crucially, I show thatwhereas the process does not generally applyword-internally, see (2b), it does apply at the leftedge of the base of productively formed prefixedwords, as in (2c).

(2)a. estable ‘stable’

esnob ‘snob’b. instrucción ‘instruction’

obstaculo ‘obstacle’c. inestable ‘unstable’

biescalar ‘biscalar’

Contrary to Cressey (1978) and Harris (1983,1986), I argue that e-epenthesis does not refer to amorphological constituent. Consider, for instance,the multiply derived word inestabilidad ‘unstabili-ty’. There are two reasons for attributing themorphological structure shown in (3) to this word.First, its meaning is ‘the state of not being stable’,rather than ‘not the state of being stable’. Second,in- subcategorizes for adjectives, not for nouns.

(3)N

A

A

/in stabl idad/

Suppose that e-epenthesis applied to the embed-ded adjectival stem /stabl/. The rule would thenprecede suffixation. Given that many suffixes inSpanish are stress-shifting, it would also be or-dered before stress assignment. Consequently, wewould predict that epenthetic /e/ can surface withstress, contrary to fact (Harris 1986).

Alternatively, I propose that epenthesis ap-plies at the left edge of the prosodic word, whichcrucially contains a stem and any suffixes, to theexclusion of any prefixes. Thus, estabilidad formsa single prosodic word, which does not incorporatethe prefix in-. Given the requirement of prosodicminimality (McCarthy & Prince 1986), the prefixcannot form an independent prosodic word either.In fact, it does not bear main word stress. Alterna-tively, I propose that it adjoins to the base prosodicword, as in (4).

(4)PW

PW

in estabilidadF

In this structure, a prosodic word dominatesanother prosodic word. Moreover, the prefix sylla-ble is not dominated by a foot.

Similarly, I argue that compounding andcliticization can also give rise to recursion and theskipping of levels in the prosodic constituentstructure. A constrained account of when and howthese marked prosodic structures occur cruciallyinvolves the decomposition of the first clause of theSLH into separate, violable, constraints, as pro-posed in Selkirk (1995).

(5) Prosodic dominationa. LAYEREDNESS: No Ci dominates a Cj, j > ib. HEADEDNESS: Any Ci must dominate a Ci-1c. NONRECURSIVITY: No Ci dominates another Cid. EXHAUSTIVITY: No Ci immediately dominates a Ck, k<i-1

In particular, by ranking NONRECURSIVITY andEXHAUSTIVITY below constraints on the alignmentof morphological and prosodic structure or onfaithfulness of input representation, recursion andthe skipping of levels, respectively, result.

In recent optimality-theoretic work, constraintsrequiring surface identity between paradigmaticallyrelated forms such as a base and a word derivedfrom it, are invoked to explain the non-coherentcharacter of certain affixes, such as the Spanishprefixes (Burzio 1994; Benua 1995; Kenstowicz1996). Essentially, the base of the affixed wordbehaves as if the affix were not there, thus remain-ing phonetically identical to its form in isolation; ifthe affix were to cohere phonologically, either word-internal phonological processes would apply at thejuncture with the affix, or word-edge processeswould fail to apply at this juncture. In either casethe base would be made distinct from its surfaceform in isolation. Cohering affixes, in contrast,attach to bases that are not existing words. There is,therefore, no effect from the isolated form of thesebases in the derived words.

I argue that this approach is too restricted intwo ways (cf. Peperkamp 1997). On the one hand,paradigmatic identity effects can arise in derivedwords of which the morphological base is not anoccurring word. On the other hand, the distinctionfound within a single language between wordsthat are subject to paradigmatic identity effectsand those that are not cannot always be attribut-ed to a distinction between stem-based and word-based morphology. Spanish e-epenthesis is anexample of the former case. In fact, the presenceof epenthetic /e/ in inestabilidad cannot be due toa paradigmatic identity constraint, since estabili-dad, the only independently occurring wordembedded within it, is not its morphological base(cf. (3)). An example of the latter case is providedby suffixation in Dutch. Syllabification of a base-final consonant applies across the boundary ofcohering suffixes in Dutch, as in (6a), but isblocked across the boundary of non-coheringsuffixes, as in (6b).

(6)a. groen+ig groe.nig ‘greenish’b. groen+achtig groen.achtig ‘greenlike’

Morphological subcategorization for stems orwords does not interfere with the phonologicalbehavior of suffixes. Rather, it is the phonologicalform of the suffix itself that determines whether itis cohering or non-cohering; suffixes which areeligible for prosodic word status are non-cohering,while suffixes that fail to satisfy requirements onprosodic words are cohering (Booij 1995). Paradig-matic identity constraints have nothing to sayabout this distinction, since suffixes — whethercohering or non-cohering — do not occur as inde-pendent outputs.

In many languages, the distinction betweencohering and non-cohering affixes coincides withthat between derivational suffixes and prefixes.That is, in these languages, prefixes, as opposed tosuffixes, fail to incorporate into the prosodic wordto which they attach (Booij & Rubach 1984; Ne-spor & Vogel 1986; Cohn 1989). I show that thesame left-right asymmetry occurs with cliticiza-tion, in that in several languages, proclitics arephonologically less coherent than enclitics. Thisasymmetry provides an argument against thenecessity of the clitic group, introduced by Hayes(1989) and Nespor & Vogel (1986) as a constituentof the prosodic hierarchy. In fact, given the sym-metrical nature of the clitic group, phonologicalasymmetries between proclisis and enclisis cannotbe accounted for by making reference to thisconstituent. Rather, I propose that clitics are baresyllables that can attach at various prosodic levels.This varied — but constrained — set of possibleprosodizations allows to account not only forasymmetries between proclisis and enclisis, butalso for crosslinguistic variation found in thebehavior of clitics and their hosts with respect tostress assignment (cf. Peperkamp 1995, 1996).

Postlexical resyllabification constitutes theother potential problem for the SLH. Recall thatthe second clause of the SLH requires prosodicconstituents to be properly nested within theconstituent that dominates them. Syllabification,therefore, should not cross prosodic word bounda-ries. In many languages, however, phrasal resyl-labification applies across prosodic words.Consider, for instance, the Italian phrase baraperto ‘open bar’, which is syllabified[ba.ra.per.to]. I argue that neither extraprosodici-ty nor ambisyllabicity are feasible alternatives toan account involving lexical syllabification fol-lowed by postlexical resyllabification. I then pro-pose that resyllabification induces prosodicrestructuring, such that all syllables are properlypart of a single prosodic word. Specifically, I pro-pose that of two possible restructurings, shown in(7b), the latter one is correct. In fact, there is noevidence that resyllabification induces the restruc-turing of two prosodic words into one, with asingle main stress.

PROSODIC WORDS

by Sharon Peperkampreviewed by Caroline Wiltshire

Dissertations Glot International, Volume 4, Issue 4, April 1999 16

(7)a. lexical prosodic structure

(bar)PW (aperto)PWb. postlexical prosodic structure

*(ba.ra.per.to)PW*(ba.)PW (ra.per.to)PW

Thus, postlexical resyllabification results in theformation of postlexical prosodic words, whichdiffer minimally from the lexically built prosodicwords from which they derive. As a consequence,syllables are properly nested within prosodicwords, both lexically and postlexically, and no ill-formed prosodic trees result.

In order to account for prosodic restructuring,I propose a generalized Proper Nesting constraint,which involves the alignment of prosodic constitu-ents. In fact, the decomposition by Selkirk in (5)involves the first clause of the SLH only. Thesecond clause translates into the Proper Nestingconstraint, defined in (8).

(8)Proper nestingAlign(Ci, L/R; Cj, L/R)where Ci and Cj are categories of the prosodic hierarchy and Ciimmediately dominates Cj

Notice, finally, that Proper Nesting is unviolableand hence should be universally undominated.Violation of Proper Nesting would, indeed, giverise to geometrically ill-formed structures.

Reviewby Caroline Wiltshire

1. IntroductionFor over a decade, the prosodic hierarchy has

motivated a vast and productive line of research,accounting for phonological alternations throughprosodic domains, templates, and phonotactics.Peperkamp’s dissertation situates an analysis ofprosodic structure in a constraint based OptimalityTheory framework (Prince & Smolensky 1993),and advances the research program with aninsightful examination of the prosodic word levelof the hierarchy.

Peperkamp approaches the prosodic wordfrom all angles: inside out (derivation), outside in(compounding), above (phrases, clitic groups), andbelow (syllabification). The result is a work thatbrings together a variety of data around a com-mon theme: mismatches between prosodic wordsand morphological words. Such mismatches areconstrained by the tenets of the Strict LayerHypothesis (SLH), broken into violable compo-nents following Selkirk (1995). Taking advantageof the new structures made possible by minimalviolations of the SLH, Peperkamp shows that notall prefixes have the same prosodic structure, andthat their place in the prosodic word can be dis-tinct from that of suffixes; parallel analyses aregiven for clitics and for words in compounds. Herparallel analyses connects the asymmetries ofprefixes and suffixes with that of proclitics andenclitics, and with the strength of the left edgeboundary of prosodic words.

Peperkamp provides neat solutions to somefamiliar thorny problems, along with new insightsinto English, Dutch, and Romance phonology andmorphology. She also sketches previous work ineach area and compares her analyses to the alter-natives, providing interesting arguments againstOutput-Output constraints, as illustrated in hersummary. Peperkamp’s dissertation is valuablereading for phonologists, morphologists, and thoseinterested in Romance languages in particular,although one need not be a Romance linguist tobenefit from her conclusions.

I found her arguments convincing and thor-ough, and many of her solutions elegant; I sketcha few of the best below in sections 2–4. The onlygeneral critique I offer is that Peperkamp does notalways push her own approach as far as shecould. Like many a phonologist switching para-digms from rules to constraints in the 1990s, herdescriptions begin with a rule-based formulation,

reflecting a tendency to conceptualize phonologyin terms of a sequence of events in a derivation. Asa result, what sometimes seems a simple solutionin prose becomes more complex when formalizedin OT, and the formalization seems almost anafterthought. By beginning with the basic tenetsof violability and parallelism and pushing thepotential for mismatches a little further, I thinkthe work begun in this dissertation will have farreaching consequences (section 5).

2. DerivationPeperkamp’s analysis of the distinction be-

tween cohering and non-cohering affixes as dis-tinct prosodic word structures can be illustratedwith one of her examples from Italian, in whichsuffixes are cohering and prefixes are non-coher-ing. A cohering affix joins inside a prosodic wordwith the stem, while a non-cohering affix remainsoutside the stem’s prosodic word boundary, eitheras an independent prosodic word or inside a recur-sive prosodic word:

(1)a. Suffixes, cohering: [fix-es]PWb. Prefixes, non-cohering: Independent PW: [pre]PW[fix]PWc. Prefixes, non-cohering: Recursive PW: [ pre [fix]PW]PW

While strict adherence to the SLH results inoptions (1a) and (1b), breaking down the SLH intoseparate violable constraints provides Peperkampwith a third option, (1c), which violates NONRECUR-SIVITY due to the one prosodic word inside another.Non-cohering affixes therefore need not formindependent prosodic words themselves, contraNespor & Vogel (1986).

Peperkamp analyzes Intervocalic s-voicing(ISV) in Northern Italian to show the need for thisthird option. For Italian, stress and the presence oflow mid vowels, which indicate a primary stressedsyllable, are used as diagnostics for the prosodicword. Using these criteria, Peperkamp dividesprefixes into the structures (1b) and (1c) on princi-pled grounds. ISV applies within monomorphemicwords, before suffixes, and at the ends of prefixes,but not to word-initial /s/ following a vowel-finalword or prefix. Peperkamp’s insight is that, giventhe correct prosodic structure for prefixes, thereare no prosodic words beginning with [zV], aconstraint generally true throughout Italian.Ranking this constraint *PW[zV above a constraintagainst intervocalic voiceless [s] leads directly tothe desired results; an intervocalic /s/ which is notPW-initial is voiced (2a-c), while those that arePW-initial are not voiced (2d,e).

(2)a. PW-internal: [asola]PW a[z]ola

‘button-hole’b. PW-internal(with suffix): [cas-ina]PW ca[z]ina

‘little house’c. Prefix-final: [dis[onesto]PW]PW di[z]onesto

‘dishonest’d. PW initial: [la]PW[sirena]PW la[s]irena

‘the siren’e. PW-initial (with prefix) [a[sociale]PW]PW a[s]ociale

‘asocial’

Thus the prefix-final /s/ in dis- can voice, since itis not at the left edge of a prosodic word; the /s/ insociale cannot be voiced because it begins a pro-sodic word.

Nespor & Vogel (1986) had argued that ISV isa word span rule. In order to make their analysiswork, however, they required a definition of theprosodic word that includes all vowel-final prefixesas independent words as in (1b), even if monosyl-labic or monomoraic, in violation of prosodic mini-mality. As independent prosodic words,monosyllabic prefixes also should be stressed, butthey neither have mid-low vowels nor do theytrigger “raddoppiamento sintattico” like otherstressed syllables. In Peperkamp’s analysis, thedefinition of prosodic words is not altered to ac-count for the behavior of vowel-final prefixes; thepresence of the left prosodic word bracket separat-ing them from the stem accounts for their non-cohering behavior. Her approach is thus able tocapture the surface-true constraints as directly aspossible.

In addition to her analyses of affixation inItalian, Peperkamp examines Spanish segmentalrules and syllabification, to show that prefixes donot incorporate into the base prosodic word as dosuffixes. A common point to both discussions isthat coherence in prosodic word structure is inde-pendent of coherence in syllabification. That is,while a prefix may be outside a prosodic word forsegment-based rules such as ISV, it may be syllab-ified with the following prosodic word. Due to herconstraint on Proper Nesting, which she claims isinviolable, Peperkamp analyzes such cases asrequiring two levels of syllabification (but seesection 5 below). Peperkamp’s basic point, thatcoherence and non-coherence are not monolithic,holds true in either case and advances our under-standing of affix behavior.

3. CompoundingIn compounding, two or more words or lex-

emes are concatenated to form another word, butnot all compounds have the same internal phono-logical structure. Peperkamp argues for threeforms of prosodic structure for compounds: singleprosodic words, two adjacent prosodic words, or arecursive prosodic word, parallel to the analysis ofaffixes in (1). In her discussion, Peperkamp touch-es on a wide range of data, both within Italianand cross-linguistically, but the focus is on themorphology and phonology of Italian compounds.In particular, Peperkamp examines the headproperties of compounding in Italian, and comes tothe conclusion that productive compounding isexocentric or rightheaded; structures that appearto be left-headed are not true compounds, andhence there are no word+root compounds.

Prosodic structure is derived directly from themorphological structure. Root+root compoundsbehave like monomorphemic words both morpho-logically and phonologically, and hence aremapped to a single prosodic word. Such wordshave a single primary stressed syllable and under-go ISV even at the beginning of the second root.Word+word compounds, if productively formed,constitute two adjacent prosodic words, as shownby the presence of the low-mid vowels which arerestricted to stressed syllables and the failure ofISV to apply at the boundary. Familiar com-pounds, her term for compounds which are lexical-ized and frequent, are given the recursive wordstructure of (1c), thereby explaining why theyhave only a single primary stressed vowel but failto undergo ISV internally:

(3)a. Root + Root compounds:

[root-root]PWcromo[z]óma ‘chromosome’filó[z]ofo ‘philosopher’

b. Root/Word+Word compoundsProductive:[[word]PW [word]PW]PWc[f¢]pri sélla ‘saddle cover’r[7¢]ggi lúme ‘lamp stand’

c. Familiar:[word [word]PW]PWc[ò]pri l[é]tto ‘bedspread’r[è]ggi s[é]no ‘bra’

The recursive prosodic word structure of thefamiliar compounds allows for only one mainstress; hence the vowels in the initial word cannotbe low-mid, since low-mid vowels appear onlyunder primary stress. Both compounds in (3b,c)have an intervocalic initial prosodic word bracket,preventing ISV as in the case of prefixes. Theparallels with affixation are both neat and con-vincing, and extend to her discussion of cliticiza-tion as well.

4. CliticizationPeperkamp argues against the need for the

clitic group as a level in the prosodic hierarchy,located below the phonological phrase and abovethe prosodic word. While the clitic group has beensupported by Hayes (1989) and Nespor & Vogel(1986), Peperkamp’s arguments take the classicform: the use of the clitic group results in a theorythat is both too strong and too weak. First, since

Dissertations Glot International, Volume 4, Issue 4, April 1999 17

the levels of the prosodic hierarchy generallydominate elements exactly one level below, theclitic group should dominate prosodic words. Butgiving a clitic or clitic cluster the status of a pro-sodic word makes many incorrect predictions. Forexample, clitics are usually not subject to wordphonotactics, especially minimality requirements,and they often do not bear word stress, similar toher argument against treating non-coheringprefixes as prosodic words. Second, the clitic groupresults in a theory which is too weak to account forclitics which take a phrasal host, as argued forHausa in Inkelas (1989). Furthermore, the cliticgroup provides no means for distinguishing procli-tics from enclitics, though these show the sameasymmetries as prefixes and suffixes, or for distin-guishing a sequence of host plus clitics from asequence of clitics, though these show distinctstructural properties. Peperkamp also arguesaway the data that motivated the clitic group aseither not belonging to truly prosodic processes orprosodic but better reanalyzed without the cliticgroup. Supporters of the clitic group may not findall these arguments equally convincing, butPeperkamp makes a strong case for reconsideringthem.

Peperkamp then analyzes enclisis in threevarieties of Italian, Neapolitan, Lucanian, andStandard, to show that a clitic group is neithernecessary nor sufficient (following Peperkamp1996). These three varieties show different stresspatterns as a result of adding one or two encliticsto a verb. In Neapolitan, stress appears on theverb as it would in isolation, on one of the lastthree syllables; a single enclitic is not stressed, buta sequence of two clitics is stressed on the first.Lucanian stresses the penultimate syllable of thecombination of verb + enclitic(s), regardless of thenumber of enclitics added or the position of stresson the verb in isolation; stress can thus appear offthe verb entirely, if the verb has two enclitics.Standard Italian stresses the verb in the sameposition as in isolation; regardless of how manyclitics appear after the verb, no clitics are stressed.

(4)Verb V + one enclitic V + 2 enclitics

a. Neapolitan: cónt6 cóntal6 cóntatíll6tellIMP tellIMP-it tellIMP-youREFL-it

b. Lucanian: vínn6 v6nníll6 vinn6míll6sellIMP sellIMP-it sellIMP-me-it

c. Standard: pórta pórtami pórtamelobringIMP bringIMP-me bringIMP-me-it

Peperkamp analyzes the three cases using two ofthe constraints from the Selkirk (1995) version ofthe SLH: NONRECURSIVITY and EXHAUSTIVITY. Bas-ing her analysis on a two level system, in whichprosodic structure is built on lexical items beforeclitics are added, she also refers to a constraint forpreserving lexical prosodic structure:

(5)FAITHFULNESS: do not modify lexically built prosodic structure(p. 189)

Ranking these three constraints differently gives adifferent system depending on which constraint islowest and therefore violable; the three varieties ofItalian illustrate each ranking:

(6)a. PW-adjunction

[[[host]PWclitic(s)]PW]PPh [[[cónta]PWtíll6]PW]PPhNeapolitan: NONRECURSIVITY violation tolerated (PW insidePW)

b. PW-incorporation[[host-clitic]PW]PPh [[vinn6-míll6]PW]PPhLucanian: FAITHFULNESS violations tolerated (host alone isno longer a PW)

c. PPh-incorporation[[host]PWclitic]PPh [[pórta]PWmelo]PPhStandard Italian: EXHAUSTIVITY violations tolerated (PPhdominates a clitic which is not part of a PW)

Given the structures above, the stress generaliza-tion is the same in all three varieties: disyllabicfeet (which do not cross prosodic word boundaries)get stress at the right edge of a prosodic word. InNeapolitan, both the internal prosodic word andthe external prosodic word are stressed, resulting

in stress remaining on the verb and appearing onany sequence of clitics large enough to form a foot.In Lucanian and Standard Italian, there is onlyone prosodic word, hence only one stress in therightmost foot. The analysis is an elegant exampleof the factorial typology: each constraint rankingresults in a language. Her constraints also make aprediction: in a right branching recursive struc-ture, stress could only fall on both the inner andthe outer prosodic word, not solely on the outerprosodic word, and this seems to be right.

However, Peperkamp’s account relies on aconstraint, FAITHFULNESS, which is not surface-based; she also assumes that prosodic/metricalstructure is built first lexically, before clitics areadded post-lexically. In Peperkamp’s approach,this means that Lucanian words would havestructure built lexically and erased post-lexicallywithout a trace, as FAITHFULNESS is lowest ranked.Peperkamp is technically correct that OT as aframework does not make a claim about the use oflevels in phonology, but OT does attempt to limitderivationality. To extend the non-derivationalityof the approach, we could reanalyze the Neapoli-tan and Standard Italian examples, in whichFAITHFULNESS is active, using surface based align-ment constraints. Following Selkirk (1995), we canuse a family of constraints relating the edges oflexical words with the edges of prosodic words:

(7)ALIGN-LEX/L: Align (Lex, L; PW, L) } Align the L/R edge of

every lexical word withALIGN-LEX/R: Align (Lex, R; PW, R) } the L/R edge of some PW

ALIGN-PW/L: Align (PW, L; Lex, L) } Align the L/R edge ofevery PW with the

ALIGN-PW/R: Align (PW, R; Lex, R) } L/R edge of some lexicalword

For Neapolitan, EXHAUSTIVITY and ALIGN-LEX/Rdominate NONRECURSIVITY:

Tableau (1) Neapolitan

Candidates Exhaustivity Align-Lex/R NonRecursivity

[[cónta]PWtill6]PPh *![[conta-tíll6]PW]PPh *!

☞ [[[cónta]PWtíll6]PW]PPh *

In Lucanian, ALIGN-LEX/R ranks lowest of thethree:

Tableau (2) Lucanian

Candidates Exhaustivity NonRecursivity Align-Lex/R

[[vínn6]PWmill6]PPh *![[[vínn6]PWmíll6]PW]PPh *!

☞ [[vinn6-míll6]PW]PPh *

In Standard Italian, EXHAUSTIVITY violations aretolerated in favor of satisfying the other twoconstraints:

Tableau (3) Standard Italian

Candidates NonRecursivity Align-Lex/R Exhaustivity

[[[pórta]PWmélo]PW]PPh *![[porta-mélo]PW]PPh *!

☞ [[pórta]PWmelo]PPh *

Peperkamp needs alignment in any case, to keepclitics from forming their own prosodic words (p.186); these same constraints will do the work ofpreventing lexical words from accepting clitics intotheir foot structure. While Peperkamp doubts thatthere are empirical differences between the use ofsimultaneous alignment constraints and the use oflevels (lexical/post-lexical) with potentially differ-ent constraint rankings, she does note that Luca-nian enclitics seem to allow for restructuring of theword, in violation of her FAITHFULNESS, while procli-tics may not. This would be exactly the sort ofempirical difference in favor of the simultaneousalignment approach; alignment at the left edgecan rank higher than alignment at the right edge,while FAITHFULNESS allows for no such distinctionsunless levels of derivation are further multiplied.Furthermore, the simultaneous use of constraintsrather than a sequence of levels prevents differentrankings at different levels of the grammar, allow-ing for a tighter theory.

5. Violability and parallelismAlthough the most interesting analyses result

from breaking the SLH into ranked and violableconstraints, Peperkamp balks at one type of ill-formedness: violations of Proper Nesting, in whichan element at one level is dominated by more thanone unit in the layer above. This can be represent-ed with an improper graph or improper bracket-ing, as shown:

(8)PW PW

F F F F F [ [ ] ]FF F FFPW PW

Since syllabification often crosses prosodic wordboundaries, Peperkamp proposes that prosodicword boundaries are minimally relocated in suchcases so that each syllable is properly nestedwithin a prosodic word.

However, McCarthy & Prince (1993) giveseveral tenets of OT, among them:

(9)a. Violability Constraints are violable; but violation is

minimalb. Parallelism Best satisfaction of the constraint hierarchy

is computed over the whole hierarchy andthe whole candidate set. There is no serialderivation (McCarthy & Prince 1993, p. 4, andfn. 3)

By declaring Proper Nesting inviolable, in contrastto (9a), Peperkamp is also forced to rule out themaximally parallel, single-level analyses advocat-ed in (9b). Instead, lexical syllabification obeyingprosodic word boundaries is followed by post-lexical resyllabification, which also restructuresprosodic word boundaries. There is, however, noexplicit evidence for this proposed shift in prosodicword boundaries due to resyllabification. Presum-ably such evidence would take the form of a word-bounded post-lexical process to prove that asegment has joined a different prosodic word andnot just a different syllable, and none springs tomind. The shift in word boundaries may also beproblematic if it creates subminimal prosodicwords, such as the ba in (7b) in Peperkamp’ssummary above. Finally, the insistence on theinviolability of Proper Nesting forces her to usetwo levels, an otherwise unnecessary complicationin the grammar.

The shift of word boundaries due to resyllabi-fication also seems to make wrong predictions. Inan analysis of Tamil in Wiltshire (1998), I notedthat words end with sonorants in one dialect ofTamil. For obstruent-final stems before vowel-initial words, Peperkamp’s account predicts thatthe resyllabification of the obstruent across aword-boundary would result in a restructured,and therefore acceptable, prosodic word. However,stem-final obstruents must be followed by anepenthetic vowel, regardless of the following word,see (10). Only in sonorant-final words do we getresyllabification, since these satisfy the word-final-sonorant constraint without restructuring, as in(11):

(10)Obstruent Final No resyllabification Resyllabification

(No restructuring) (with restructuring)/sappaz/ /ellaam/ [s%ppa2Y]PW[j7llã] *[s%ppa]PW[27llã]food all ‘all kinds of food’

(11)Sonorant Final Resyllabification

(No restructuring)/maram/ /ellaam/ [m%r6m]PW[7llã]tree all ‘all the trees’

Extending her use of a set of violable constraintsrather than a monolithic SLH, we need violationsof Proper Nesting as well, so that resyllabificationneed not undo prosodic word boundaries.

Another example to illustrate the advantageof the alignment approach can be found closer tohome. “Raddoppiamento sintattico” in Italiandoubles consonants after words ending in stressedvowels, and applies regardless of whether thefollowing word begins with a single consonant or aconsonant cluster of rising sonority. It does not,

Dissertations Glot International, Volume 4, Issue 4, April 1999 18

however, apply before words beginning in sCclusters:

(12)a. [p]ulita ‘clean’ cittá[pp]ulita ‘a clean city’b. [t]riste ‘sad’ cittá[tt]rieste ‘a sad city’c. [sp]orco ‘filthy’ cittá[sp]orca ‘a filthy city’

In an OT analysis (Wiltshire & Maranzana, toappear), we propose that the lengthening is moti-vated by a requirement that stressed syllables areheavy, PKPROM, and epenthesis is limited by aconstraint against surface consonants which arenot present in the input, DEP-IO(C). Combinedwith ONS, the constraint that syllables begin withconsonants, we can easily handle (12a):

Tableau (4) Input: /cittá pulita/ ‘a clean city’

Candidates PkProm Ons Dep-IO(C)

cittá]F[pulita *!cittáp]F[ulita *!

☞ cittáp]F[pulita *(p)

The same ranking would, however, incorrectlyresult in [(cit)(tát)]PW[(ris)(te)], with resyllabifica-tion and prosodic word restructuring of the initial/t/ of triste:

Tableau (5) Input: /cittá triste/ ‘a sad city’

Candidates PkProm Ons Dep-io(C)

cittá]F[triste *!L cittát]F[riste

cittát]F[triste *(t)!

Rather than appealing to levels of syllabification,with the resyllabification of triste blocked by somekind of FAITHFULNESS to lexically built structure,we use an alignment constraint on the left edge ofthe word, so long as syllable readjustments do notinvoke prosodic word readjustments:

Tableau (6) Input: /cittá triste/ ‘a sad city’

Candidates PkProm Align-Lex/L Dep-io(C)

cittá]F[triste *!cittát]F[riste *!

☞ cittát]F[triste *(t)

If prosodic word boundary readjustment is trig-gered by resyllabification, we have no way to ruleout the second candidate. Again, we need to con-sider candidates with violations of Proper Nestingin order to choose the correct output.

We are now also equipped to handle theapparently exceptional case of word initial sCclusters. An [sC] cluster is not a highly desirableonset, and word internally such clusters are bro-ken into coda /s/ plus onset /C/ (as in pas.ta).Onsets generally satisfy a constraint that thesonority difference between members is at leastfour degrees on the scale proposed by Davis(1990), a constraint we call *<4DIFSON. Not only isthis constraint motivated by the word internalseparation of sC clusters into separate syllables,but it also results in the failure of “raddoppiamen-to sintattico” due to its ranking above ALIGN-LEX/L:

Tableau (7) Input: /cittá sporca/ ‘a filthy city’

Candidates PkProm *<4DIFSON Align-Lex/L Dep-IO(C)

cittá]F[sporca *! *cittás]F[sporca *! *(s)

☞ cittás]F[porca *

In an account in which lexically built structuremust be preserved due to a high ranking FAITH-FULNESS constraint, we could derive cittáttristethrough two stages of syllabification, but wewould also derive *cittássporca. Conversely, ifFAITHFULNESS ranks low, we can derive cittásporcaby violating it, but we would also derive*cittá-triste. If we allow for the misalignments oflexical and prosodic words but limit it throughconstraints, we can derive both cases with a singlesyllabification and eliminate FAITHFULNESS. AddingProper Nesting to the set of SLH constraints is avaluable idea, so long as it too is violable.

6. ConclusionI enjoyed reading this dissertation and hope

that it reaches the wide audience that it deserves.

Peperkamp offers both new data and fresh in-sights into familiar data, and raises questions forfuture research, such as the universality of theasymmetry between left and right edges of prosod-ic words. Her proposals are significant and broadreaching, her arguments well-reasoned and thor-ough. I look forward to seeing her future work.

ReferencesBenua, L. (1995). Identity effects in morphological

truncation. Papers in Optimality Theory,edited by J. Beckman, L. Walsh Dickey & S.Urbanczyk, 77–136. Amherst, Mass: GLSA.

Booij, G. & J. Rubach (1984). Morphological andprosodic domains in lexical phonology. Phonol-ogy Yearbook 1, 1–27.

Booij, G. (1995). The phonology of Dutch. Oxford:Clarendon Press.

Burzio, L. (1994). Principles of English stress.Cambridge: Cambridge University Press.

Cohn, A. (1989). Stress in Indonesian and brack-eting paradoxes. Natural Language andLinguistic Theory 7, 167–216.

Cressey, W. (1978). Spanish phonology and mor-phology: A generative view. Washington, D.C.:Georgetown University Press.

Davis, S. (1990). The onset as a constituent of thesyllable: Evidence from Italian. Papers fromthe 26th Regional meeting of the ChicagoLinguistic Society, edited by M. Ziolkowsky,M. Noske & K. Deaton, 71–80. Chicago: CLS.

Harris, J. (1983). Syllable structure and stress inSpanish: A nonlinear analysis. Cambridge,Mass: MIT Press.

Harris, J. (1986). Epenthesis processes in Spanish.Studies in Romance languages, edited by C.Neidle & R. Nuñez Cedeño, 107–122. Dor-drecht: Foris.

Hayes, B. (1989). The prosodic hierarchy in meter.Rhythm and meter, edited by P. Kiparsky & G.Youmans, 201–260. New York: AcademicPress.

Inkelas, S. (1989). Prosodic constituency in thelexicon. Ph.D. dissertation, Stanford. Pub-lished by Garland, NewYork (1990).

Kenstowicz, M. (1996). Base-identity and uniformexponence: Alternatives to cyclicity. Currenttrends in phonology: Models and methods.Volume one, edited by J. Durand & B. Laks,363–393. Salford: ESRI.

McCarthy, J. & A. Prince (1986). Prosodic mor-phology. Manuscript, University of Massachu-setts, Amherst, and Brandeis University.

McCarthy, J. & A. Prince (1993). Generalizedalignment. Technical Report #7 of the RutgersCenter for Cognitive Science, Rutgers, N.J.

Nespor, M. & I. Vogel (1986). Prosodic phonology.Dordrecht: Foris.

Peperkamp, S. (1995). Enclitic stress in Romance.Papers from the 31st Regional Meeting of theChicago Linguistic Society. Volume 2: Theparasession on clitics, edited by A. Dainora, R.Hemphill, B. Luka, B. Need & S. Pargman,234–249. Chicago: CLS.

Peperkamp, S. (1996). On the prosodic representa-tion of clitics. Interfaces in phonology, editedby U. Kleinhenz, 102–127. Berlin: AkademieVerlag.

Peperkamp, S. (1997). Output-to-output identityin word level phonology. Linguistics in theNetherlands 1997, edited by J. Coerts & H. deHoop, 159–170. Amsterdam: John Benjamins.

Prince, A. & P. Smolensky (1993). OptimalityTheory. Technical Report #2 of the RutgersCenter for Cognitive Science, Rutgers, N.J.

Selkirk, E. (1981). On prosodic structure and itsrelation to syntactic structure. Nordic prosodyII, edited by T. Fretheim, 111–140. Trondhe-im: TAPIR.

Selkirk, E. (1984). Phonology and syntax. Therelation between sound and structure. Cam-bridge, Mass: MIT Press.

Selkirk, E. (1986). On derived domains in sen-tence phonology. Phonology Yearbook 3, 371–405.

Selkirk, E. (1995). The prosodic structure of func-tion words. Papers in Optimality Theory,

edited by J. Beckman, L. Walsh Dickey & S.Urbanczyk, 439–469. Amherst, Mass: GLSA.

Wiltshire, C. (1998). Extending ALIGN constraintsto new domains. Linguistics 36:3, 423–467.

Wiltshire, C. & E. Maranzana (to appear). Gemi-nates and clusters in Italian and Piedmontese:A case for OT ranking. Formal perspectives onRomance linguistics, edited by M. Autier, B.Bullock & L. Reed. Amsterdam/Philadelphia:John Benjamins.

Another example has to do with people whosuffered brain damages. One such patient “withmassive damage to the left hemisphere” (p. 193),Dehaene reports, suffered from a major readingdeficit: it was very hard for him to read aloudwords or numbers. As to numbers, he could onlyread them out if they had a meaning beyondbeing a number. For instance, he could onlypronounce the string 504 after associating it witha certain model of Peugeot, i.e., through its non-quantitive meaning. Findings like this surely havesignificance for neuro-linguistic research.

There are many more linguistically interest-ing things in this book (one section is called “Thecost of speaking English”).

But how about the similarities and differencesbetween language and mathematics — or, for thatmatter, the language instinct and the numbersense? Dehaene does not deal with this questionexplicitly; although he mentions Pinker once(p. 104) he does not seem to be aware of Thelanguage instinct. So let’s see what he says aboutthe number sense and mathematics.

Dehaene makes the point that our brain, likethat of animals such as rats and chimpanzees, isequipped with what he calls an accumulator,which enables us, and the rats and the chimps, toperceive, memorize and compare numerical mag-nitudes. This accumulator enables us (humansand the other animals) to estimate how numeroussome events are, but does not allow us to computetheir exact number — beyond 2 or 3, everythinggets fuzzy. This is the number sense.

In addition to this accumulator, humans havebeen endowed with supplementary competence:the ability to create and manipulate complexsymbol systems (making it possible to do advancedarithmetics). This ability is partly related to someother innate properties, like mechanisms forindividuating objects as well as intuitions aboutsets, continuous quantities, iteration, logic andgeometry of space. Mathematics “can be character-ized as the progressive formalization of theseintuitions. It is the purpose to make them morecoherent, mutually compatible and better adaptedto our experience of the external world” (p. 246).

Mathematics, then, does not arise naturally inthe same way as we acquire our mother tongue.For one thing, formal calculations have to betaught in school (“the human brain ... has notevolved for the purpose of formal calculation” —p. 134) and you have to keep practising, otherwiseyou lose it (p. 164) (sounds similar to a secondlanguage!). However, Dehaene argues emphati-cally, this does not mean that mathematics is“unnatural”: mathematics is the way it is, becauseit is founded on our basic intuitions concerningset, number, space, time and logic. In the sensethat mathematics is shaped by these intuitions, itis “created” by the mind, as it is mentioned in thetitle of this fine book.

Rint Sybesma

Leaving the field

The number sense

Continued from p. 1

Dissertations Glot International, Volume 4, Issue 4, April 1999 19

Summaryby the author

This dissertation explores different aspects of wordorder and clausal structure in Romance, withspecial emphasis on Spanish. These aspects arelooked at in light of the highly constrained theoryof word order proposed in the antisymmetry ap-proach of Kayne (1994).

The approach of antisymmetry consists ofpositing a universal Spec-Head-Complementorder. This rigid theory of phrase structure isdesigned to capture certain cross-linguistic asym-metries to the left such as position of the specifier,agreement patterns and head movement. Theorder is derived from the Linear Correspond-ence Axiom (LCA), which maps hierarchicalstructure into linear order. Some striking conse-quences of the LCA are that it makes unavailablecertain mechanisms such as right adjunction andmultiple adjunctions to the same head or specifier.

The development of antisymmetry is particu-larly significant for our understanding of Ro-mance post-verbal subjects because standardanalyses of these cases have assumed a rightadjunction of the subject to the VP (e.g. Rizzi1995). In this study, an alternative to this analysis— one compatible with antisymmetry — is ex-plored in the context of VSO and VOS alternationsin Spanish. Taking as a primitive the VSO order,following Sportiche’s (1988) VP internal subjecthypothesis, I contend the VOS order is betterunderstood as a scrambling of the object to the leftof the position of the subject. The crucial argu-ment in favor of this alternative is the fact thatthere are certain asymmetries between the VSOand VOS orders that show that the object c-com-mands the subject in the VOS order. Thus, anobject quantifier can bind a pronoun in the subjectin the VOS order but not in the VSO order. Also,principle C effects obtain in the VOS order but notwith VSO. Thus objects in VOS have to be in amore hierarchically prominent position thansubjects, contrary to what a right adjunctionanalysis would predict.

Other types of asymmetries show that thereare restrictions on the nature of objects that canappear in the VOS order. Specifically, wh-objectsin situ and non-specific indefinite objects areallowed in VSO but not in VOS. I point out thatthese restrictions and the same asymmetric bind-ing effects parallel facts attested in languagessuch as German, Hindi and Korean in which anindependently motivated scrambling analysis haslong been accepted for SOV-OSV alternations.Thus, a scrambling approach for the VOS order inSpanish is not only empirically more adequate butalso more general.

My analysis of Spanish leads me to exploreother Romance languages in which the rightadjunction analysis has been proposed, includingCatalan, French and Italian. These languageshave a more restricted distribution of post-verbalsubjects than Spanish since subjects must followmost complements, predicates and some adverbs.Moreover, the subject in that position is alwaysmarked prosodically (e.g., in Catalan and Italianpost-verbal subjects are narrowly focused).

First, I conclude that the difference betweenSpanish and these other Romance languages isthat Spanish contains an extra projection whichallows the subject to remain neutral with respectto prosody. In Spanish, the verb simply movesabove the subject in this extra functional projec-tion by head movement. This movement is parallelto the one that yields the VSO order in Irish andArabic. As in those languages, no restrictions occuron what can follow the post-verbal subject. For thenarrowly focused nature of post-verbal subjects inCatalan and Italian, I explore an analysis inwhich the subject moves to an external focusposition, above IP. The movement of the subject tothis focus position is followed by the movement ofthe whole IP to a higher inflectional projection.

This analysis is in the same spirit as DenDikken’s (1995) account of Heavy NP shift inEnglish, in which the movement of the object isfollowed by that of the whole predicate VP to theleft. One of the consequences of this analysis isthat the need for a special rule of extraposition iseliminated in favor of a combination of leftwardmovements. This combination can explain therestrictions on the distribution of what can appearafter a post-verbal subject through parallelismswith Germanic and other leftward processes inRomance.

Finally, I elaborate further on the proposedanalysis of Spanish. I observe that contrary to thecase of the VSO order, in the VOS order the sub-ject is narrowly focused in Spanish. Following myanalysis for Catalan and Italian, I suggest thatsubjects in VOS structures in Spanish have beenmoved to focus projection. This movement is fol-lowed by a scrambling of the object to the left andfinal movement of the IP to a higher projection.

Another aspect of clausal structure I investi-gate consists of structures with overt wh-move-ment in Spanish and Catalan. It is shown that theobligatory post-verbal positioning of subjects inthese structures cannot be explained by recourseto a required overt head movement of the verb toC as in Rizzi’s (1991) V2 analysis. First of all, theV2 proposal is incompatible with the assumptionsin Kayne (1994) that the sequence clitic–verb isnot subject to head movement. Kayne’s approachcombined with movement of the verb would leadto the expectation that the sequence verb–clitic ispossible. This sequence is indeed found in impera-tives and infinitives — where a verb movementapproach is feasible — but not in cases of wh-movement.

The overt V-to-C approach also predicts thatthe subject following the verb in interrogatives isin SpecIP. However, various tests — such as theinsertion of the subject between the auxiliary andpast participle, the licensing of floating quantifi-ers, and the possibility of insertion of the subjectbetween verb in C and the objects in VP in Italianand Catalan — show that subjects in these lan-guages are not in a SpecIP position. Finally, anovert V-to-C approach makes it impossible toexplain why languages with overt complementiz-ers in interrogatives still have obligatory inversioneffects.

Instead of sticking to V-to-C, I conclude thatpre-verbal subjects in Spanish and Catalan are

always topicalized. Combining a complex CPapproach (Rizzi 1995) with recent ideas on com-plementizer recursion developed by Browning(1996) and Watanabe (1992), I conclude that atopic subject conflicts in features with the +wh-complementizers of interrogatives. Finally, I lookat some complex wh-words and note that they donot require inversion. I propose that these complexwh-words are topicalized.

Since the conclusions concerning wh-move-ment were based on the idea that subjects aretopicalized in Spanish, I motivate this assumptionin the last part of my dissertation. The typicalassumption in a pro-drop language like Spanish isthat covert as well as overt subjects occupy apreverbal position at Spell-Out in which theirCase and agreement properties are satisfied. Ipresent evidence against this claim. Instead, Ishow that pre-verbal overt subjects pattern withleft dislocated DOs and IOs in a wide range ofsyntactic contexts, such as ellipsis and extractionof quantifiers. In these contexts, sentences with“silent subjects” differ from sentences with overtones. I also show that quantified pre-verbal sub-jects have the same restriction on possible inter-pretations as other left dislocated complements. Iconclude that overt pre-verbal subjects are neces-sarily left dislocated.

In order to account for the left dislocatednature of subjects, I propose the elimination ofAgrS as a functional projection. Instead I proposethat person agreement morphemes should beconsidered clitics (as in Taraldsen 1992), and therelation between agreement and overt subjects isone of clitic doubling. Evidence in favor of theseclaims comes from the striking parallelisms be-tween standard clitic-doubling constructions andagreement-subject constructions. Specifically, bothcases pattern similarly in relation to the determi-nation of binding in certain cases of mismatches inperson between the doubling DP and the clitic.

The clitic nature of agreement is also motivat-ed morphologically. Following the morphologicaldecomposition analysis of clitics (Harris 1995), Iobserve that 1pl agreement has the same morpho-logical shape as the object clitic. Striking evidencecomes from the fact that this agreement contains aplural morpheme which I contend is the sameplural morpheme found in the nominal system. Ishow that this plural morpheme amalgamateswith the same plural morpheme in 1pl imperativesin Spanish.

I take agreement to be an argumental cliticthat absorbs theta role and Case. Movement of thedoubling DP subject to a pre-verbal position can-not, therefore, be driven by agreement or Caserequirements. Instead, movement of the subject toa pre-verbal position must be motivated by dis-course considerations, as is typical in left disloca-tions. Finally, the obligatory enclitic nature of thisagreement can be explained by Guasti’s (1991)and Rizzi & Roberts’ (1989) idea that clitic agree-ment has a subcategorization frame which re-quires a verbal host. Clitic agreement moves to ahigher inflectional projection as in other cases ofenclises and the verb left adjoins to it followingantisymmetry.

Reviewby João Costa

‘Word order and clause structure in Spanish andother Romance languages’ is a very importantdissertation. It argues for an antisymmetric analy-sis of several word order phenomena in Spanishand other Romance languages. One of the crucialaspects of this dissertation is that it indeed arguesfor antisymmetry, rather than just assuming it, asis so often done. The author provides severalconvincing arguments for Kayne’s (1994) frame-work, showing that it is not just a different way ofaccounting for word order facts, but a necessarymechanism for a proper description of the data helooks at. In particular, the arguments provided foranalyzing VOS in Spanish as an instance of objectscrambling across the subject are worth mention-

WORD ORDER AND CLAUSALSTRUCTURE IN SPANISH ANDOTHER ROMANCE LANGUAGES

by Francisco Ordóñezreviewed by João Costa

Dissertations Glot International, Volume 4, Issue 4, April 1999 20

ing. Ordóñez shows that this analysis is necessary,since it is the only one able to account for the factthat the object is structurally higher than thesubject.

In this review, I will discuss some aspects ofthe dissertation that seem to me to be controver-sial. It should nevertheless be clear from thediscussion that none of these aspects cruciallycompromise the important results achieved byOrdóñez regarding the representation of wordorder variation in Spanish.

1. The position of postverbal subjects andneutral interpretationWhile presenting the arguments for analyz-

ing VOS orders in Spanish as an instance ofscrambling of the object across the subject, Or-dóñez notes that VSO may have a neutral inter-pretation. In other words, this word order may bean instance of sentence-focus and be uttered as ananswer to what happened?.

According to Ordóñez, the neutral interpreta-tion of VSO arises as a consequence of movementof the subject out of VP to the specifier position ofa functional projection called NeutP, which isbelow TP and above VP:

(1)[TP V [NeutP S tV [VP tS tV O]]]

In this section, I would like to question the ade-quacy of linking the neutral interpretation of thesubject position with a specific functional projec-tion.

As far as I can see, there are at least fourproblems with this assumption:

I In sentence-focus interpretations, not only thesubject is in its neutral position.

II Cross-linguistically, VSO is not neutral.III In some varieties of Spanish, the preverbal

position of the subject is also neutral.IV It is not clear to what extent it is theoretically

desirable to codify the distinction betweenneutral word orders and marked word ordersin terms of functional projections.

Let me address each of these issues separately.

1.1. Neutrality in sentence-focusinterpretationsAccording to the proposal put forward in this

dissertation, the VSO word order in Spanish isneutral because the subject occupies the specifierposition of NeutP. In frameworks allowing for anidentification of discourse functions in terms ofassociation with discourse-related functionalprojections, this is not unnatural.

However, it seems to overlook the discoursefunction of the other constituents of the sentence.In neutral contexts, not only the subject is in itsneutral position, but also the verb and the object.In other words, there is a difference between aVSO sentence, in which all elements are in theirneutral position, and a OVS sentence, in whichthe object is no longer in its neutral position.Following Ordóñez’ proposal and pushing the ideathat neutral interpretations arise as a conse-quence of movement to SpecNeutP, it ought to beassumed that in VSO sentences, all elements moveto or through NeutP. This is not problematic forthe verb, which moves through the head Neut onits way to T. However, it seems to be problematicfor the object, since there is no position for it tomove to, and yet it receives a neutral interpreta-tion. Note that it would not work to propose asecond NeutP for the object to move to, yielding arepresentation like (2):

(2)[TP V [NeutP S tV [NeutP O tV [VP tS tV tO]]]]

This type of representation makes the predictionthat, like recursive Topic Phrases in the work ofRizzi (1995), subject and object may land in eitherSpecNeutP. If that were the case, however, bothVSO and VOS should be neutral, which is not true.

Another problematic fact is that neutralinterpretations are often taken to be ambiguousbetween sentence-focus, VP-focus and object-focus. If that is the case, object-focus may not beexplained in this fashion, since it is not clear howto derive the neutral interpretation of the objectalone.

1.2. VSO is not neutral in other languagesAs Ordóñez discusses, VSO orders are not

allowed in Catalan, French and Italian. He pro-poses that the difference between Spanish and theother Romance languages is that the latter lackthe functional projection NeutP. This prevents thederivation of the VSO word order, since the land-ing site for the subject does not exist. Accordingly,a neutral VSO word order is not possible in theselanguages, because there is no way to generateVSO.

The corollary of this idea is that if a languagepermits VSO, NeutP is active in this language andVSO may arise as a neutral interpretation. How-ever, if one looks at a related language like Portu-guese, it may be seen that VSO word orders arepossible, but inadequate in neutral contexts (Costa1997a,1998):

(3)a. O Paulo partiu a janela

‘Paulo broke the window’b. Partiu o Paulo a janela

(4)a. O que é que aconteceu?

‘what happened?’b. O Paulo partiu a janela

*Partiu o Paulo a janela

The only context in which VSO is legitimate iswhen both the subject and the object are focused:

(5)a. Ninguém partiu nada

‘no-one broke anything’b. Partiu o Paulo a janela

From these data, it can be seen that the onlyneutral word order in Portuguese is SVO. UnderOrdóñez’ account, there is no obvious way toexplain this difference between Spanish andPortuguese. On the one hand, it must be assumedthat Portuguese does not lack NeutP, in order forVSO to be possible. On the other hand, the factsshow that VSO does not correspond to a neutralposition for subjects.

1.3. SVO may be neutral in SpanishAs noted in several works on Spanish syntax

(e.g. Hernanz & Brucart 1987), SVO is also aneutral word order. It is not clear from the litera-ture whether the two options (VSO and SVO)coexist, or whether there is a dialectal split be-tween speakers who utter VSO or SVO as neutral.The relevant aspect for Ordóñez’ argument is thatpreverbal subjects may be neutral. This may notbe explained if the neutrality of subjects is ex-plained in terms of NeutP.

There are three ways of solving this problem.The first would be to assume that the differencebetween neutral SVO and neutral VSO has to dowith the relative position of NeutP in the hierar-chy of functional projections. In neutral SVO, itwould dominate TP, while it would be dominatedby TP in neutral VSO. This type of explanation ofword order differences would be in line with theproposals concerning parametric variation interms of selection properties of functional heads(e.g. Ouhalla 1991). A problem for such proposalwould be to explain the evidence put forward forthe crosslinguistic similarities in the hierarchy ofdiscourse-related functional projections (e.g. Rizzi1995; Kiss 1995). One would have the burden ofexplaining why such similarities break down inSpanish.

The second potential approach to this prob-lem, very much in the spirit of Ordóñez’ ownproposal, would be to assume that preverbalsubjects in Spanish are instances of left-dislocation(see also Barbosa 1995, among others). Since

subjects tend to be topics (cf. Lambrecht 1994;Givón 1984), their neutral interpretation would bederived. This proposal would apparently solve theproblem, but it makes NeutP an unnecessaryfunctional projection. The neutral reading ofsubjects is derived independently of the associa-tion with a functional projection responsible forneutral readings. Moreover, this would not explainwhy in sentence-focus contexts a subject in Spec-TopP (or left-dislocated) has a focus interpretation.

The third potential approach would not associ-ate the neutral reading of sentences with anyspecific functional projection, but rather with thelack of discourse-marked operations. In otherwords, the neutral interpretation of unmarkedword orders of each language would be a conse-quence of the fact that no process associated withdiscourse (e.g. topicalization or focus-movement)applies. The neutral word order of each languageis decided by the syntax alone. Assuming that theneutrality of SVO and VSO in Spanish is not trulyoptional, but reflects the (dialectal) coexistence oftwo grammars, the neutral word order arisesbecause subjects must check features overtly inSpecIP in the SVO variant, but not in the VSOdialects. The only problem of this approach is thatit misses a generalized analysis for null subjectlanguages, according to which case features of thesubject are uniformly weak (Barbosa 1995), andthe preverbal subject position is an instance ofleft-dislocation. If arguments are found for the A-status of preverbal subjects in some null subjectlanguages (cf. Costa 1997a, 1998, among others),this analysis may be maintained (see also sec-tion 3).

1.4. Marked vs. Neutral as differentfunctional projectionsThe representation of discourse functions in

terms of functional projections in syntax is widelyattested in the literature (Brody 1990 and subse-quent work). Although it is questionable whetherthis is desirable for all discourse functions (see e.g.Reinhart 1995 for discussion of information focus),the functional projections are commonly used aslanding sites for XPs that are marked with respectto discourse: they are either topicalized or contras-tive focused constituents. Under this type of ap-proach for the syntax-discourse interface, thereare two ways of encoding neutral interpretations:either the set of constituents receiving a neutralinterpretation is the complement of the set ofconstituents moved to topic and focus positions, orthe neutral interpretation is itself the result ofmovement to a designated discourse-related func-tional projection. The former approach is the oneproposed at the end of the previous paragraph,the latter is Ordóñez’.

There seems to me to be a conceptual problemwith the movement approach. As noted above,neutral interpretations often arise in instances ofsentence-focus. If that is the case, this shouldmean that all the constituents of the sentence (orthe sentence as a whole) should move to a focus-related functional projection. Independently ofwhether or not there is empirical evidence for thistype of movement, there is no evidence left forNeutP. If neutral interpretations are instances ofsentence-focus and if discourse functions areencoded in the syntax, as Ordóñez assumes,NeutP should be some version of FocusP.

In short, the functional projection NeutPseems problematic, and there are alternative waysto derive the neutral reading of postverbal sub-jects that would not compromise Ordóñez’ findingsregarding the clause structure of Spanish.

2. VOS in Spanish vs. VOS in French,Catalan and ItalianIt is argued in this dissertation that the im-

possibility of obtaining VSO word orders in Ital-ian, French and Catalan follows, if it is assumedthat NeutP is not projected in this languages. LikeSpanish, these three languages allow VOS wordorders, but it is proposed that the representation ofVOS is not derived via scrambling of the object.Instead, it is argued that subjects move to a focus

Dissertations Glot International, Volume 4, Issue 4, April 1999 21

projection at the left periphery, and the remnantTP moves to the specifier of a functional projectionto the left of the subject. This proposal is in accord-ance with antisymmetry, and it is worth mention-ing that Ordóñez does a good job showing theinadequacy of the right-adjunction approach.

In this section, I would like to note threeaspects of this analysis that seem to be problemat-ic:

I The cross-linguistic variation in terms of theavailability of NeutP;

II The arguments against scrambling;III The role of FocusP.

2.1. Crosslinguistic variationThe proposal put forward in the dissertation is

that VSO is lacking in French, Catalan and Ital-ian, because NeutP is not projected. Since thisfunctional category is not available, the subjectmay not raise to this position.

In the preceding section, I have presentedsome arguments against the existence of NeutP,even in languages allowing VSO. For instance, inthe case of Portuguese, the neutral word order isnot VSO, hence NeutP could not be assumed to beprojected in between TP and VP. If this functionalcategory is not available in this language, theprediction made by Ordóñez is that Portugueseshould lack VSO word orders, like French, Catalanand Italian. This, however, is not true (cf. thesentence in (5)). It thus seems to be difficult to linkthe availability of VSO with the neutral interpre-tation of VSO.

2.2. Arguments against scramblingThe alternative analysis of VOS in languages

with V-to-I movement still within antisymmetry isrefuted by Ordóñez: the scrambling analysis.According to this analysis, VOS would be derivedvia scrambling of the object to the left of thesubject (which may stay within VP). This type ofanalysis has been proposed for Dutch by Koster(1993).

Ordóñez presents two arguments againstextending the analysis of Dutch to Catalan,French and Italian. The first argument is the factthat VOS is a marked prosodic construction inRomance, while SOV is the least marked focuspattern in Dutch. The second argument is thatscrambling is traditionally assumed to be anoptional operation, while it must be assumed to beobligatory in Catalan, French and Italian. Sincethere is no parallelism with respect to these twoproperties, Ordóñez assumes that the two sets oflanguages must not undergo a similar analysis. Iwould like to note that these two counterargu-ments are not as strong as he would like them tobe.

Let us consider first the argument based onintonation. There are two ways to discard thisargument. On the one hand, it is not clear thatthe markedness of intonation may be used as anargument for or against unifying the analysis oftwo languages. If sentence stress is defined interms of embedding (cf. Cinque 1993; Reinhart1995), it is expected that the most embeddedconstituent receives the sentence main stress. Inthat case, SOV in Dutch is unmarked because theobject is the most embedded constituent. If thisapproach is true, the prosodic markedness of VOSin Romance is surprising: the subject shouldreceive unmarked sentence stress. The crucialaspect is that, either in the scrambling analysis orin the light predicate raising analysis, the subjectis the most embedded constituent. In other words,the lack of parallelism in terms of intonation hasnothing to do with the syntactic process involved.If sentence stress is defined in terms of directional-ity (Nespor & Vogel 1986, among others), the roleplayed by the verbal head is crucial. Prosodicconstituents may take into consideration theplacement of the head, and the placement ofsentence stress may differ because of the differentposition of the verb, which must be assumed bothin the scrambling analysis and in the light predi-cate raising analysis. If this approach is right, the

difference in intonation follows from the cross-linguistic difference in terms of verb placementrather than in terms of strategies for moving theobject across the subject.

On the other hand, scrambling in Dutch andGerman does create a more marked intonation. Ifone looks at the difference between S-Adv-O-Vand S-O-Adv-V, it is often reported that the latteris more prosodically marked (see for instanceReinhart 1995; Neeleman & Reinhart 1996, whotreat scrambling as an operation for defocusingthe object). If scrambling of the object across thesubject in Romance is compared with scramblingof the object across an adverb in Germanic, in-stead of with the movement from postverbal topreverbal position, there is no big differencebetween the two sets of languages: in both, theresult is a more prosodically marked structure. Thecrucial difference between Dutch and Romance,besides the position of the verb, would be theobligatory nature of subject movement out of VPin Dutch.

Let us now consider the argument based onoptionality of scrambling. Again, there are twoways of discarding this counterargument. First ofall, if, unlike what I proposed in the precedingparagraph, scrambling in Romance is to be com-pared with the movement of the object frompostverbal to preverbal position in Dutch, likeOrdóñez does, then the movement is obligatoryboth in Romance and in Dutch. In other words,scrambling is only optional in Dutch from thepreverbal position to a higher position.

Second, if the comparison is made betweenthe standard scrambling in Dutch and the move-ment across the subject in Romance, the lack ofoptionality in the latter may be linked with thenature of the movement. If scrambling in Ro-mance is triggered by some strong feature inAgrOP or PredP, no optionality is predicted. Theoptionality of Dutch may be explained in terms ofdefocusing, as in Reinhart (1995). The problem ofoptionality is thus reduced to a problem of featurestrength. A similar solution for the lack of optional-ity in English, compared to Germanic and Ro-mance, has been proposed in Costa (1997b). Inshort, the problem of optionality may not be con-sidered if it is not proven that an analysis in termsof obligatory movement to an intermediary posi-tion is ineffable. The lack of parallelism does notprovide evidence against an uniform analysis inthe same way that the optional vs. obligatorycharacter of wh-movement in French and English,respectively, does not provide evidence against anuniform analysis of wh-movement in these twolanguages.

If the counterarguments for the scramblinganalysis are not as strong as Ordóñez suggeststhey are, it is worth reconsidering a more detailedcomparison between this option and the lightpredicate raising alternative. Especially since thescrambling analysis involves a more simple repre-sentation of the clause and straightforwardlyexplains the fact that the object c-commands thesubject in VOS word orders (cf. section 3.8.1 of thedissertation).

2.3. The role of FocPFor the proposal based on light predicate

raising to work, the left periphery of the sentenceis crucially involved. The derivation of VOS pro-ceeds in terms of movement of the subject toSpecFocP and subsequent movement of the rem-nant TP to the specifier of a functional projectionabove FocP.

The first intriguing aspect of this derivation isthe fact that focused subjects move to this position.In most analyses resorting to FocP (Brody 1990;Kiss 1995; Rizzi 1995, among others), this projec-tion at the left periphery of the sentence is re-served for contrastive foci. This is not the case forpostverbal subjects in French, Catalan and Ital-ian. These subjects are not necessarily contrastive,but rather information foci (cf. Légendre 1998,among others). This mix of discourse functionsassociated with functional projections also comesup in the discussion of Spanish. Ordóñez suggests

that the VS orders may arise either by movementof the subject to NeutP (VSO) or via light predi-cate raising of the remnant TP (VOS). The differ-ence is due to the status of the subject: whether itis neutral or focused. Now, if neutrality is to bereduced to sentence focus, as suggested in sec-tion 1, there is no reason for the subject to end upin different discourse-related functional projec-tions.

Another unclear aspect is that in most analy-ses resorting to FocP, this functional projection issupposed to be able to host wh-phrases. Thisexplains why contrastive foci are in complementa-ry distribution with wh-movement (cf. e.g. Rizzi1995). Surprisingly, wh-movement is sometimesthe factor enabling VOS orders in the Romancelanguages studied in this thesis.

Finally, for the analysis to be clearer, onewould like to know the status of the functionalprojection dominating FocP, where the remnantTP is moved to. Following most studies on the leftperiphery, this position should be topic-related.But, if that is the case, it is not clear why (6) isimpossible:

(6)*Deu à Maria, o livro, o Paulo*gave to Maria the book Paulo

In some Romance languages, like Portuguese,multiple topic constructions are allowed, though abit marked:

(7)(?)O livro, à Maria, o Paulo deu(?)the book to Maria Paulo gave

In these constructions, the order of the topicalizedconstituents is free:

(8)(?)À Maria, o livro, o Paulo deu

Now, if remant TP movement is to be identifiedwith movement of the topical part of the sentence,it is expected that an object may be preposed to aposition before the remnant TP (cf. (9)) or after.The latter, however, is not possible, as was shownin (6).

(9)(?)O livro, deu à Maria o Paulo

If the functional projection above FocP is not topic-related, one would like to know what its nature isand what the trigger is for moving the remnantTP.

In short, neither the light-predicate-raisinganalysis for VOS orders is exempt of problems, norcan the scrambling analysis be as easily discardedas Ordóñez suggests. Hence, a more thoroughcomparison between the two might be helpful andfruitful. The type of work that Ordóñez performsfor excluding the right-adjunction analysis wouldbe useful for deciding between scrambling andlight-predicate raising.

3. Pro-drop and left-dislocated subjectsAs in many current analyses (Barbosa 1995;

Alexiadou & Anagnostoupoulou 1995, amongothers), Ordóñez attempts to establish a relationbetween pro-drop and left-dislocated subjects. Theassumption is that preverbal subjects in nullsubject languages are always instances of left-dislocation.

In this final section, I would like to show thatPortuguese, a null subject language, does notcorroborate these findings: preverbal subjects inPortuguese, unlike those in Spanish, according toOrdóñez’ description, exhibit A-properties. Thiscan be seen in the following set of data.

A-binding. First, it should be noted thatpreverbal definite subjects are A-binders: they areable to bind an anaphor from this position:

(10)Todos os coelhos comeram a sua cenouraall the rabbits ate POSS carrot‘every rabbit ate his carrot’

Dissertations Glot International, Volume 4, Issue 4, April 1999 22

In (10), the QP todos os coelhos is able to bind thepossessive anaphor sua contained in the directobject. Note that there is a contrast between left-dislocated direct objects and in-situ direct objectsconcerning the possibility of binding into anindirect object, which is not expected if the samemechanism evoked for subjects would apply:

(11)a. Os jornalistas deram todos os livrosi aos seusi autores

the reporters gave all the books to POSS authorsb.??/?/*Todos os livros, os jornalistas deram aos seus autores

A-bar minimality effects. If preverbal subjectsare left-dislocated, one expects to get A-bar mini-mality effects if there is A-bar extraction across apreverbal subject, in accordance with the findingsof Rizzi (1991). This is not attested in Portuguese,as the examples in (12) illustrate:

(12)a. Que livros o Paulo leu?

which books Paulo readb. Esses livros, o Paulo leu

those books Paulo read

These examples show that neither wh-movementnor topicalization induce A-bar minimality effects.

Multiple topicalization. In EuropeanPortuguese, it is possible to have multiple topicali-zation, but this is a slightly marked construction:for example, (13b) is felt as more marked than(14a), and it is necessary to introduce a prosodicbreak between the second preposed PP and theverb:

(13)a. Sobre o tempo, falei com o Pedro

about the weather talked-1SG with Pedrob.?/??Sobre o tempo, com o Pedro, falei

The contrast between (13a) and (13b) is not redu-plicated in the constructions involving a preverbalsubject and a preposed constituent: neither amarked sentence is obtained nor is it necessary tointroduce a prosodic break in between the subjectand the verb:

(14)Com a Maria, o Pedro falouwith Maria, Pedro talked-1SG

The difference noted is further confirmed by thefact that in true cases of multiple topicalization,the order of the two preposed constituents is notrigid: there is a counterpart of (13b) with thereverse order of PPs, for which the grammaticalityjudgements do not change:

(15)?/??Com o Pedro, sobre o tempo, falei

with Pedro about the weather talked-1SG

Note that, with preverbal subjects, changing theword order between subject and preposed PP doesnot yield a good result, unless the subject preced-ing the PP is a hanging topic (cf. Duarte 1987), inwhich case it can be reduplicated by another NPin the normal preverbal subject position, as in(16b):

(16)a. ?? O Paulo, com a Maria, falou rapidamente

Paulo with Maria talked quicklyb. ? O Paulo, com a Maria, esse sacana falou rapidamente

Paulo with Maria that jerk talked quickly

The fact that there is a contrast between sentenceswith preverbal subjects and sentences with multi-ple topics provides further evidence against theidea that preverbal subjects are left-dislocated inEuropean Portuguese.

The differences of behavior between preverbalsubjects in Portuguese and Spanish casts somedoubt on the link established between left-dislocat-ed subjects and pro-drop. Specifically to Ordóñez’analysis, the impossibility of having subjects inSpecIP must not be linked to verbal agreementmorphology, since it does not differ in these twolanguages:

(17)Spanish Portuguese1sg -o 1sg -o2sg -s 2sg -s3sg -Ø 3sg -Ø1pl -mos 1pl -mos2pl -is 2pl -is3pl -n 3pl -n

4. ConclusionAs the reader may have noticed, the skeptical

aspects of this review focused on technical detailsof the analysis proposed in the dissertation. Thishopefully shows that the general proposal andanalysis are very interesting. As mentioned at thebeginning, this dissertation is an important contri-bution to our understanding of the syntax ofSpanish and Romance, and provides substantialevidence in favor of antisymmetric syntax. It istherefore strongly recommended to syntacticiansin general and, in particular, to people working onRomance and antisymmetry.

ReferencesAlexiadou, A. & E. Anagnostoupoulou (1995). SVO

and EPP in null subject languages and Ger-manic. FAS Papers in Linguistics, Potsdam.

Barbosa, P. (1995). Null subjects. Doctoral disser-tation, Massachusetts Institute of Technology.

Brody, M. (1990). Remarks on the order of ele-ments in the Hungarian focus field. Ap-proaches to Hungarian, vol. 3, edited by I.Kenesei. Szeged: JATE.

Browning, M.A. (1996). CP recursion and that-teffects. Linguistic Inquiry 27, 237–257.

Cinque, G. (1993). A null theory of phrase andcompound stress. Linguistic Inquiry 24,239−398.

Costa, J. (1997a). Focus in situ: evidence fromPortuguese. To appear in Probus.

Costa, J. (1997b). Scrambling in European Portu-guese [Proceedings of SCIL 8]. Cambridge,Mass: MITWPL.

Costa, J. (1998). Word order variation. A con-straint-based approach. Doctoral dissertation,HILUniversity.

Dikken, M. den (1995). Extraposition as intraposi-tion, and the syntax of English tag questions.Manuscript, Vrije Universiteit Amsterdam.

Duarte, I. (1987). A Construção de Topicalizaçãona Gramática do Português. Doctoral Disser-tation, University of Lisbon.

Givón, T. (1984). Syntax: a functional-typologicalintroduction, Vol 1. Amsterdam: John Ben-jamins.

Guasti, M.T. (1991). Incorporation, excorporationand the lexical properties of causative heads.The Linguistic Review 8, 209–232.

Harris, J. (1995). The morphology of Spanishclitics. Evolution and revolution in linguistictheory, edited by H. Campos & P. Kemchinsky.Washington, D.C.: Georgetown UniversityPress.

Hernanz, M.L. & J.M. Brucart (1987). La Sintax-is. Editorial Crítica.

Kayne, R. (1994). The antisymmetry of syntax.Cambridge, Mass: MIT Press.

Kiss, K. (1995). Discourse configurational lan-guages. New York: Oxford University Press.

Koster, J. (1993). Predicate in corporation and theword order of Dutch. Manuscript, Universityof Groningen.

Lambrecht, K. (1994). Information structure andsentence form. Cambridge University Press.

Légendre, G. (1998). Focus in French stylisticinversion. Paper presented at the Workshopon Inversion in Romance, University of Am-sterdam.

Neeleman, A. & T. Reinhart (1996). Scramblingand the PF-interface. The projection of argu-ments: lexical and compositional factors,edited by W. Geuder & M. Butt. Stanford:CSLI.

Nespor, M. & I. Vogel (1986). Prosodic phonology.Dordrecht: Foris.

Ouhalla, J. (1991). Functional categories andparametrization. New York: Routledge.

Reinhart, T. (1995). Interface strategies. Manu-

script, OTS/Utrecht University.Rizzi, L. (1982). Issues in Italian syntax. Dor-

drecht: Foris Publications.Rizzi, L. (1991). Relativized minimality. Cam-

bridge, Mass: MIT Press.Rizzi, L. (1995). The fine structure of the left

periphery. Manuscript, University of Geneva.Rizzi, L. & I. Roberts (1989). Complex inversion in

French. Probus 1, 1–30.Sportiche, D. (1988). A theory of floating quantifi-

ers and its corollaries for constituent struc-ture. Linguistic Inquiry 19, 203–238.

Taraldsen, T. (1992). Subject-Verb agreement inCeltic and Romance. Proceedings of NELS 23.University of Massachusetts, Amherst: GLSA.

Watanabe, A. (1992). Larsonian CP recursion,factive complements and selection Proceedingsof NELS 23, 523–537. University of Massa-chusetts, Amherst: GLSA.

Holland Academic Graphics

ISBN 90-5569-058-9 | xiv+449 pp. | 1998 | Diss. Leiden University | HIL

P.O. BoxThe Hague

The Netherlands

532922505 AG

info hagpub.comwww.hagpub.comfax: +

@

31 70 3595044

[ s c i e n t i f i c [ d o c u m e n t ] p r o c e s s i n g ]

LO T I n t e r n a t i o n a l S e r ie s , 14

Advertisement

João Costa

Word Order VariationA Constraint-Based Approach

Recent dissertations from the LOT series

Word Order Variation is an investigation of severalphenomena of apparent free variation. The mainhypothesis is that many cases of apparent optio-nality are instances of interactions betweensyntax and discourse. It is argued that this type ofinteraction may be formalized within the theoreti-cal framework of Optimality Theory.

It is argued that monosyllabic adverbs provideevidence for short-verb-movement and for obliga-tory object movement in English. These adverbsand those that are ambiguous between mannerand subject-oriented readings provide empiricalevidence against right-adjunction. Relationsbetween prosody and syntax are first identified inthe distribution of this class of words. The structu-ral positions of subject, verb and object in Portu-guese are studied. It is claimed that postverbalsubjects are in Spec,VP, and that objects mayscramble as in Germanic. The apparent free wordorder in this language is claimed to be discourse-conditioned: sentence-final position is reserved forfocused elements, in accordance with rules ofsentence stress assignment. The prosodic-basedapproach to focus is explicitly compared with LF-movement approaches. It is claimed that the latterare not well motivated. The difference betweendiscourse-configurational languages and langua-ges with a rigid word order is explained in Opti-mality-theoretical terms. Differences betweenadjunction and A-movement in the distribution ofnominal complements follow from relativerankings between Case and economy. The OTapproach is expanded by looking at patterns ofvariation in unmarked contexts. It is argued thatvariation at the base is the reflex of differentrankings between (dominated) syntactic con-straints. The difference between VSO and SVO isconsidered in detail. At the end of the dissertation,the approach defended is compared with alterna-tive parametric approaches. Criteria are establis-hed for the evaluation of the optimality theoreti-cal approach.

Price EUR 41.79, NLG 92.10 (ex. 6% whereapplicable, ex. P P)Individuals ordering directly from Holland AcademicGraphics are eligible for a 33% discount.

VAT&

Book reviews Glot International, Volume 4, Issue 4, April 1999 23

Understanding syntaxby Maggie Tallermanreviewed by Marcel den Dikken

Linguistics — at least, of the sort practised by thereaders of this journal — is a fundamentallycomparative enterprise. On this, we all agree. Yet,while introductory textbooks to morphology andphonology often draw on data from a wide varietyof languages, introductions to syntax typicallyconfine themselves to just a handful of languages— Andrew Radford’s (1997) recent book being anextreme case in point: it uses nothing but Englishexample material. To this, Maggie Tallerman’s newintroduction, Understanding syntax, is a spectacu-lar exception, addressing as it does data from nofewer than 83 different languages, from all sortsof different language families and parts of theworld (which are all carefully registered in thebook’s language index). This fact alone wasenough to arouse my curiosity and interest intothis new textbook. So I read it, in one go — and Iwas not disappointed. Tallerman’s strongly com-parative approach to introducing the basic con-cepts of syntax has materialised in a very readableand interesting book, and the author’s excitementabout language data and their analysis is sure tolight up a spark in her readers.

The fact that Tallerman does not confine herattention to English, or Germanic, or Indo-Euro-pean, has interesting consequences beyond kin-dling students’ interest in syntax — it also leadsher to address issues and construction types whichgarden-variety textbooks typically find no occasionto discuss. Thus, in the discussion of lexical catego-ries (which, by the way, makes the very interest-ing and effective move of using garden-path‘headlinese’ examples like Revived ferry sale fearsdog islanders (p. 32) to introduce the differencebetween verbs and nouns), the author stressesthat “we shouldn’t think that just because, say,English and Italian have an open class of adjec-tives, then all languages must have one” (p. 49).Tallerman touches upon such things as languagevariation and change (pp. 8–11), grammaticalisa-tion (p. 54), serial verbs (pp. 79–81), the head-marking/dependent-marking dichotomy(pp. 103–9) and the differences between (split)ergative and accusative languages (pp. 151ff.),she acknowledges sign language (p. 22), and inher chapter on syntactic processes, she does notconfine her attention to just passive, wh-move-ment and topicalisation, but also pays attention toantipassive, applicative and causative formation— though, surprisingly, there is no mention madeat all of any head-movement processes (like nounincorporation, verb incorporation, or even simplesubject-auxiliary inversion).

Not all of these discussions are equally en-lightening, though — some are truly detailed andinstructive, but others remain too shallow to tellyou much. A good example is a comparison of thesections on antipassive and causative. Of thesetwo phenomena, the former is typologically re-stricted, found basically in ergative languagesonly, but causativisation is robustly represented allaround the globe. Yet, while the discussion onantipassives (pp. 185–91) really fills you in on

most everything there is to know about theseconstructions, the section on causatives (pp. 195–99) is remarkably poor by comparison. To illustratethis latter point, consider what Tallerman tells usabout the French faire à construction: all she saysis that the causative and causativised verbs “be-have generally as a single verbal unit and not aspredicates in separate clauses” (197), mentioningonly the fact that an English type ECM causativeis impossible with faire in support of this state-ment. The claim that in faire à causatives “the twolexical verbs are actually both inside a singleclause, and share a single set of arguments ratherthan each having their own arguments as they doin English” (197) — controversial like most anyclaim in the literature; yet none of its controversialstatus seeps into Tallerman’s discussion of Frenchcausatives — would have been easily illustratedwith the aid of a simple example of clitic climbing,which would have had the additional advantageof introducing the notion of clitics. A missed oppor-tunity, therefore; and an unfortunately shallowdiscussion of causative constructions as a result —a rather miserable successor to the rich andthoughtful section on antipassives earlier inchapter 7.

This is not the only point where Tallermanmisses a good opportunity to do just a little bitmore than what she has already managed to do.Another good example comes up in section 5.3.3,at the end of her discussion of constituent struc-ture and the diagnostics for it. There she raisesthe interesting question of whether all languageshave the same constituents, and answers “no, theyapparently don’t” (138), going on to mentionBasque and Hungarian as examples of languagesin which there seems to be no VP. She leaves thediscussion there, at a point at which the uninitiat-ed reader will no doubt get a sense of bewilder-ment — I can vividly imagine students crying out:‘So what about this constituent structure thing?Can languages just take it or leave it? I don’tfollow!’ And of course they are right in raisingsuch questions. So Tallerman should not havebegged them; she should have addressed them outin the open, in some way. And since she alreadymentions the case of Hungarian in her discussion,one way that would spring to mind is to follow upthe classic claim that this language apparentlylacks a VP with a brief visit to more recent evi-dence (due, in particular, to Kenesei; see e.g.Kenesei 1998) showing that the lack of a VP inHungarian may in fact be merely apparent, andthat there is in fact positive evidence (from focuspatterns) for the existence of a VP in this lan-guage — a view that is absolutely commonplacenowadays. Taking this tack would have been anexcellent opportunity to show that, despite appar-ent evidence to the contrary, syntacticians mayavail themselves of inventive ways of proving aparticular point.

More seriously, there are also occasions in thebook where missing an obvious opportunity goeshand in hand with serious misrepresentation ofthe facts. An example comes up in Tallerman’sdiscussion of cross-linguistic word-order patterns.There, she brings up in passing the case of Ger-man and Dutch, which she says “have SOV order

in embedded clauses but have unmarked SVOorder in main clauses” (148). That’s all she saysabout Dutch and German; and of course everyonevaguely familiar with the facts of these languagesor the discussion in the syntactic literature aboutthem will know that it is flatly false — Dutch andGerman do not have ‘unmarked SVO order inmain clauses’: though in subject-initial construc-tions with a finite main verb, Verb Second (ahead-movement operation; recall from the abovethat Tallerman systematically ignores syntacticprocesses affecting heads) effectively results in asurface SVO pattern, the main verb continues tofollow its object whenever V2 leaves it untouched.And that’s very easy to show; a simple comparisonof an example with a finite main verb (Jan kustMarie ‘Jan kisses Marie’) and one with a finiteauxiliary and a non-finite main verb (Jan heeftMarie gekust ‘Jan has Marie kissed’) will do thetrick — the trick, that is, of proving the quotedclaim false, and of giving readers a chance to toyaround with language data.

Yet, although the above remarks criticise someof its concrete executions in the book, and thoughoccasionally it results in little more than decorative“sprinkling” of the text with exotic language facts(“As we’d expect, co-ordination occurs in manyother languages; examples (70) and (71) showconjoined VPs in Persian and Malagasy”, 136), allin all Tallerman’s decision to use data from manydifferent languages in her introduction to syntaxworks out very well. It is, I believe, the outstand-ing feature of the book — and a highly commend-able one at that.

Almost inevitably, Tallerman’s “zest for theexotic” occasionally makes way for a more ‘classic’Fapproach, when she takes one language (Eng-lish, of course) as the model. An example is herdiscussion of diagnostics for distinguishing mainand embedded clauses on pp. 73ff., which is allabout English and mentions exceptions to thediagnostics posed by other languages only in themargin. It is in passages like these that Taller-man’s book looks just like a typical ‘IntroducingEnglish Grammar’ type book — which leads me toraising the question: what is this book’s reader-ship?

The book was clearly written with a nativeEnglish-speaking readership in mind — the exam-ple just mentioned shows this, and also the factthat exercises interwoven with the text and at theend of each chapter (many of which feature verynice problem sets) occasionally invite readers toconstruct an argument on the basis of Englishsentences to be constructed and judged by thereaders themselves. So I suppose that this bookwas intended for use in an Anglo-Saxon context.In any case, as a primary textbook for a course, itcould only work in university systems in whichthere still is time in the curriculum of a (General)Linguistics department for a basically a-theoreticalintroduction course to the major concepts of syntaxlike the one provided by Tallerman’s book — abook which, as the author herself intimates onp. 209, may serve as a stepping-stone to a courseon syntactic theory. In university systems which,sadly, leave no room for such a course, perhaps themost fruitful way of using this book in an intro-duction to syntax course — the way that I will tryit out myself — is as supplementary reading, forthe students to study on their own (the book isself-explanatory enough not to require the assist-ance of a tutor), in tandem with a more theory-oriented introduction which will allow theinstructor to introduce the technicalities which thepresent book does not go into.

For there is quite a bit left to be learnt orbrushed up once you have gone through Taller-man’s book. It fills you in about heads and phras-es, and about the dependency relations betweenphrases. But the notion ‘government’, firmlyrooted in grammatical tradition, is not mentioneda single time. And it tells you about the existenceof complementisers (which are loosely “defined”over twenty pages after the term is first intro-duced), and about the idea that the sentence is aprojection of the complementiser (CP) (and that,

TWO TEXTBOOKS

Maggie Tallerman’s Understanding syntaxreviewed by Marcel den DikkenIan Roberts’ Comparative syntaxreviewed by Rint Sybesma

Book reviews Glot International, Volume 4, Issue 4, April 1999 24

by the same token the noun phrase might beheaded by the determiner, though the DP hypoth-esis is not actually embraced: “The issue of wheth-er D or N heads the ‘noun phrase’ is acontroversial one in modern linguistics”, 99). Butwhile it develops the structure of the sentence as aCP, it leaves the structure between C and VPunderdeveloped, resurrecting the old S–node onp. 131. And while it introduces X–bar structure onp. 140, the discussion there (which is based on anempirical argument which evaporates if the DPhypothesis hinted at on p. 99 is espoused) is en-tirely confined to the noun phrase, without anymention being made of the X–bar theoretic struc-ture of other phrases.

Organisation-wise, the discussion of syntacticstructure leaves something to be desired, too. Withphrases already introduced, labelled bracketingcontinues at first to be confined just to the head ofthe phrase; and when tree structures are firstpresented (on p. 122, long after bracketed stringswere introduced; the fact that tree diagrams areequivalent to bracketed structures is mentionedonly in passing, in the summary at the end of thediscussion), their branches at first receive nolabels, the labels — which had long been familiarto the reader from the preceding discussion —popping up again on p. 126. In general, the book’sexposition in the domain of phrase structure couldbe much improved, reducing the partial repetitionsand potential sources for confusion which the textis riddled with in its present form.

These are, ultimately, minor wrinkles whichcan easily be ironed out in a revised reprint of thebook. The same is true for the fact that, thoughthe book’s prose is generally unbiased towards anyparticular theoretical approach, there are occa-sional lapses — most notably when it comes todisplacement, which is referred to on severaloccasions with the aid of the generative ‘move-ment’ metaphor without ‘movement’ being intro-duced as a theoretical notion. Equally easy toremedy is the sometimes annoying habit of inter-weaving a whole series of illustrations with therunning text, which does not make it easy toretrieve examples while searching back and forth.But otherwise, the book really leaves very little tobe desired on the formal side of things. The book isadmirably flawless in that department, featuringvirtually no typos or other imperfections — thespontaneous use of “different to” on p. 64 is (inten-tionally?) funny in the light of the remark madeon p. 2 about prescriptive grammars, which “mighttell you not to say different than or different to”.

Understanding syntax is an absolute pleasureto read, thanks to its casual, down-to-earth style.It is user-friendly and accessible, and takes itsreaders seriously — with one curious exception, inthe discussion of dominance on p. 128, whereTallerman warns her readers not to be “fooled bythe fact that [a node] is drawn higher up in thetree”: being drawn higher and dominating are notthe same thing, mind you! Here I believe sheunderestimates her readership — I have taughtthe introduction to syntax at various levels in avariety of academic contexts, but I have neveronce come across a single student who was “fooled”this way. In the general case, though, the booksmartly invites its readers to think along by paus-ing every once in a while and asking them ques-tions (which receive immediate answers in the textthat follows), and at the same time also gives themvery useful pointers and hints — like instructingthem how to tackle examples from other languag-es (“start at the bottom and work upwards, read-ing the translation first, then examining the gloss,then looking at the original”, 14–15), presentingthem with an apt characterisation of a specifier as“an adjunct that has a fixed position within thephrase” (99), and telling them explicitly that‘[t]here is absolutely no rule of ‘once a constituent,always a constituent’” (118), with reference to thepair The students wondered how simple textbookscould be obtained and The students wonderedhow simple textbooks could be.

Most importantly, Understanding syntaxsucceeds in giving the reader a taste of the “awful

lot of syntax out there in the world” (208), as theauthor puts it in her conclusion, while at the sametime putting the message across that, despite allthis language variation, “the overwhelming homo-geneity which exists between languages is farmore impressive” (210). And almost en passant,the syntactic novice learns about such importantthings as lexical categories, semantic roles, phrasestructure and syntactic processes (passive, anti-passive, applicative, causative, wh-fronting,topicalisation), and gets a good feel of how toargue a point and how to motivate an analysis.

ReferencesKenesei, I. (1998). Adjuncts and arguments in

VP–focus in Hungarian. Acta LinguisticaHungarica 45, 61–88.

Radford, A. (1998). Syntactic theory and thestructure of English: A minimalist approach.Cambridge University Press.

Comparative syntaxby Ian Robertsreviewed by Rint Sybesma

Comparative syntax, “a new introduction to thePrinciples and Parameters theory of syntax” (theblurb says), is a good text book. It consists of 5chapters, an introduction, a glossary, a list ofabbreviations (T, GEN), three indices (subject,author, language) and, of course, a bibliography.Each chapter closes off with a very helpful sectioncalled “Further reading”, a number of exercisesand an “Appendix” which offers inquisitive stu-dents a chance to get slightly deeper into one ofthe subjects dealt with in the chapter (for exam-ple, the appendix to the chapter on Case andagreement is called “Formal relations”, the one tothe locality chapter is entitled: “Syntactic scopeand logical scope”). In addition, towards the end ofchapters 1–4, a section called “Parameters dis-cussed in this chapter” presents an overview of thesubjects discussed from a language variationperspective. This turns out to be very useful forthe last chapter of the book, chapter 5, “Principles,parameters and language acquisition” (with 17pages the shortest chapter in the book; no appen-dix, exercises or “Further reading”). BecauseRoberts has been introducing points of languagevariation explicitly as “parameters” throughoutthe book, he is able to discuss the issue of parame-ter setting in language acquisition in much moreconcrete terms than is the case in any other textbook I have seen.

The core of Comparative syntax is made up ofchapters 2–4 on the central subjects of main-stream theorizing: Case and agreement (chap-ter 2, 61 pages, not counting the appendix, theexercises and the “Further reading”), Binding(chapter 3, 53 pages) and Locality (chapter 4, 75pages).

All three chapters present admirably compre-hensive overviews not only of all the importantissues relevant to the subject matter, but also ofthe most influential theories dealing with them.Impressively, Roberts generally manages to findthe right middle road between giving a purelyhistorical overview and dealing with things the-matically.

To give an example, chapter three on Bindingstarts out from a descriptive overview of the differ-ences and similarities between pronouns andanaphors in English, after which the notion ofbinding is introduced, stressing the relevance of c-command which had already been introduced inchapter 1. Roberts goes on to explain very patient-ly how to determine what the binding domain foran element is, not avoiding fundamental questionssuch as why tense or subjecthood would play arole in this matter. After Principles A and B havebeen put in place, we enter the domain of PRO,which quite automatically leads us to the otherempty categories and Principle C. We learn every-thing about DP movement and wh-movement andthe discussion of cross-over enables us to see why

the Binding theory really is a theory of A-binding.By this time, we have covered about 25 pages. Therest of the first part of the chapter deals with nullsubjects in languages like Italian and Chinese,closing off with the [± anaphor, ± pronoun] classi-fication of both empty and non-empty elements.The remainder of the chapter on binding is devot-ed to two subjects. First, Roberts discusses long-distance anaphora, with examples from Icelandicand Chinese, exploring the possibilities of analyz-ing anaphors from a movement perspective, evenin English. Secondly, he presents an excellentsummary of Reinhart and Reuland’s theory ofreflexivity.

Similarly, chapter 4, “Locality”, offers anoverview of most issues relevant to the subject oflocality and most theories trying to deal withthem. After summarizing systematically all wehave seen on the different types of movement inprevious chapters (head, DP and wh), Robertsformulates the common factors as Move Alpha.After that, we are introduced to a number ofislands, the notions of subjacency and successivecyclicity, the ECP and the Argument-Adjunctasymmetries (as well as the that-trace effect).After catching a glimpse of Kayne’s connectedness,we immerse ourselves into Chomsky’s Barriers,Rizzi’s Relativized Minimality and, finally, Manzi-ni’s Locality.

A similar overview could be given for chapter2, the issues and theories related to Case andagreement.

EvaluationIt is clear from the above that Roberts covers quitea lot of ground, and I am impressed by the way hedoes it. Here are a number of reasons why I thinkthis is an attractive text book.

First, it is not only about English. Indeed,true to its title, it shows how certain insights arederived at by comparing different languages.(Although I must add that the comparative angleis not consistently chosen as a way to makeprogress; I guess that certain subjects lend them-selves for a treatment in comparative terms morereadily than others — word order vs. cross-over,for instance.)

The second attractive feature of this book isthat it puts things in a historical perspective —certain problems have been with us for decadesand several theories have been proposed to dealwith them, none of them totally successful, and,most importantly, none of them coming out of no-where: Roberts shows that new theories oftencapitalize on certain aspects of their predecessors.And that some of these theories are new does notnecessarily mean that they will prove definitive.For example, Feature Checking, in chapter 2(“Case and agreement”), is presented as justanother, be it a rather novel, way of looking atcertain phenomena.

Next, also in relation to the two previouspoints, the book very much reflects that the sub-ject matter of the book is an enterprise in progressas-we-speak, lively and exciting. (The disadvan-tage is, of course, as often, that at some points thestudents are left behind in bewilderment, becausenothing seems certain.)

In addition, the style of the book is informaland quite pleasant, with an occasional joke. (Forinstance, explaining the difference between ab-stract Case and morphological case, Robertswrites: “[W]e could be completely wrong aboutabstract Case (perish the thought!), ...” (p. 56).)

Finally, the book is quite difficult, but, afterputting some effort in it, (some of) my studentsfound working through the book rather reward-ing.

Some drawbacksNot everything is wonderful, though. I have somecriticisms too, one quite fundamental — but I’llstart with some lesser points of critique.

Above I praised the historical depth, but insome cases it leads to confusion on the part of thestudents. To give two examples: Roberts’ overviewof the Barriers framework is too long and too

Conference reports Glot International, Volume 4, Issue 4, April 1999 25

detailed, and much of it is never referred to again;similarly, the summary of the reasoning behindthe LGB PRO Theorem involves too many con-cepts and ideas that are not fully explained andnot relevant elsewhere, that it does not reallyserve any useful purpose.

Secondly, in some respects the book is a bitunbalanced; in some cases Roberts squats downand is extremely patient, explaining every littlestep of the process, while in others he puts on hisseven league boots and runs home, losing thestudent readers one after another along the way.For instance, Roberts spends a full two pagesexplaining the difference between morphologicalcase and abstract Case, even going the extra mileof giving the full singular six case paradigm of theLatin dominus, while Manzini’s very involvedtheory of locality is done in a mere eight pages: hisoverview of her “categorial indices” and “address-es”, her way of dealing with weak and strongislands, her definitions of barrier and G-marking,her treatment of the that-trace effect and the wayshe unifies subjacency, antecedent-governmentand proper-head government leaves the poorstudent reader behind gasping for air.

Next, writing a text book, one obviously has tomake choices. However, two pages on Control is abit little. Hardly anything on wh-in-situ (or partialwh-movement, for that matter) in a book calledComparative syntax is also quite disappointing, Ithink. V-second also does not get very much atten-tion. (But, I guess, in these things, some of it mayjust come down to taste.)

This point of missing subjects, however, bringsme to the more fundamental drawback of thisbook, which can be introduced with the question:Who is this book for? The blurb on the back of thebook says that it wants to “take students with abasic knowledge of syntax up to a point wherethey are able to read the primary literature andunderstand the latest theoretical developments.”In the “Introduction”, Roberts says that the “idealbackground for this book would be the first fourchapters of Ouhalla (1994)” (p. 8).

That is not very practical; curriculum-wise, itdoes not make very much sense to do phrasestructure grammar, transformations and X-bartheory from Ouhalla’s thorough and very basicintroduction, and then, when the more difficultstuff comes up, switch to a higher gear and dobinding, theta theory and everything else fromRoberts. The result of Roberts’ choice is that Com-parative syntax lacks a chapter on the basics. Sofar I have not said anything about chapter 1 ofthe book, entitled “Categories and constituents”.The reason is that it is baffling. In the first 21pages, this chapter introduces phrase structurerules, X-bar-theory, the theory of lexical catego-ries, functional categories and Kayne’s LCA andthe consequences for our ideas on word ordertypology! The following 15 pages run us, at anequally high speed, through Head-movement(HMC, minimal c-command!), CP, V-second phe-nomena in some Germanic languages, and theSplit-Infl Hypothesis, including AgrO!

And a subject like The (T-) Model is not eventalked about at all. (It is briefly mentioned onp. 99, in chapter 2.)

In my course, I had to provide for these intro-ductory basics myself.

If the first chapter were to be rewritten andsplit into two, dealing with its subjects in a waythe other subjects in the book are dealt with,Comparative syntax would be a very good high-level introduction into principles and parameterssyntax.

ReferencesOuhalla, J. (1994). Introducing Transformational

grammar. From rules to principles andparameters. London: Edward Arnold.

Going Romance 1998by Josep Quer

The twelfth edition of Going Romance (Conference onRomance Languages) took place at Utrecht Universi-ty, the Netherlands, between December 10 and 12,1998. Most of the papers presented in the mainsession dealt with a great variety of syntactic andsemantic issues in Romance Linguistics. They weregenerally situated in the broader generative gram-mar tradition, although other theoretical perspec-tives were also present.

The conference was opened by Raffaela Zanut-tini (Georgetown University), who presented jointwork with Paul Portner on the syntax and seman-tics of exclamatives, within a wider research projecton clause types. They defended the view that thesemantics of exclamatives can be reduced to that ofinterrogatives (roughly, a set of alternative proposi-tions), but with a widening of the domain that is theresult of the exclamative force. This allowed for anexplanation of similarities that exclamative andinterrogative sentences display at the C level. How-ever, they also argued for the existence of an extraCP layer in the case of exclamatives on the basis ofthe evidence provided by Paduan (Nothern Italiandialect).

Three papers in the conference addressed differ-ent issues related to negation. AdamPrzepiórkowski (Universität Tübingen) offered ananalysis in Situation Semantics of the ambiguity ofpreverbal Italian n-words in interrogative sentenceswhich was closely tied to the ambiguity of negationbetween a propositional (‘it is not the case that...’)and an eventuality reading (‘it is the case that not...’).In his account, preverbal n-words always expressnegation, but in questions they can be interpretednon-negatively as a consequence of the fact that thenegated eventuality reading is indistinguishable fromthe propositional one in semantic (though not inpragmatic) terms. This allowed him to dispense withstandard ECP accounts of the asymmetries betweenpre- and postverbal n-word licensing.

Hans Georg Obenauer (CNRS, Paris), develop-ing Kayne’s (1998) ideas about movement of negativephrases in English, proposed that French objectpersonne moves overtly as a consequence of itsquantificational force. He based his claim on linearorder evidence provided by lower adverbs like bienwhen co-occurring with personne. Cases where bienappears to the left of the negative object are ex-plained by movement of bien (or a projection contain-ing it). The different distribution of Italian nessunowas argued to follow from its non-quantificationalstatus. In other Romance varieties showing French-type negation negative objects were suggested tohave the same type of overt movement as objectpersonne.

Caterina Donati (UHSR/Università di Urbino)offered an account of expletive negation in Italiancomparatives that rejected previous analyses defend-ing the presence of real negation in such cases. Onthe basis of evidence provided by less-comparatives,the scopal and movement properties of than-clausesand the morphology of expletive negation, she arguesthat expletive negation cannot be viewed as arisingfrom underlying negation. Rather, it constitues thescope marker of the focus operator that is postulatedin comparatives, as the comparison construction hasstrong focus properties.

Topics in the syntax and semantics of DPsconstituted the focus of several talks. CarmenDobrovie-Sorin (URA 1028-CNRS, Université Paris7) presented a new approach to Romanian (andSaxon) genitives and to Hebrew Construct State

Nominals. As opposed to genitives introduced by aparticle, structural genitive is argued to be assignedto SpecDP, which can be to the right of N, but only ifD is filled with the definite article or if D is empty.The head N is assumed to denote a function thatapplies to the individual denoted by the DP inSpecDP and yields another individual, the one denot-ed by the whole possessive DP. The (in)definitenessspread which has been proposed for the ConstructState in Hebrew and which Dobrovie-Sorin explicitlyrejects is made to follow from this functional analysisin those cases where the genitives are definite.However, for indefinite genitives of Construct State,the [–def] feature is not inherited from the genitive,but is arguably contributed by a null indefinite articlethat is merged in the D position of the head N.

Alexandra Cornilescu (University of Bucha-rest) presented evidence from Romanian againstGrimshaw’s (1990) claim that all nominalizationssuppress the external argument of the correspondinga-structure of the verb. She showed that the subjectof an event-nominal of the supin type, as opposed tothe infinitival type, is not a modifier but an argu-ment. The difference is made to follow from theaspectual type of the nominalizing suffix: while theinfinitival suffix is [+Telic], the supine one is [–Telic].Given that only one Genitive argument can beprojected in Romanian nominalizations, the objectmust be present as an event measure with theinfinitive, but in the supin nominalization it is suffi-cient to lexicalize the subject in order to identify theactivity. Crucially, it is assumed that the [+Telic]feature of the event nominal entails the projection ofa lexical object DP, as its checking is contingent onthe checking of the structural Genitive Case.

María Luisa Rivero (University of Ottawa)examined the crosslinguistic patterns of impersonalse in Romance and Slavic languages. In order toaccount for all the uses of se, she proposed analyzingit as an ambiguous pronominal anaphor that has noreflexivizing function and is not referentially inde-pendent. The simplex anaphor se was suggested tohead a DP in all cases, while the empty head Nwould be open to different specifications: a non-relational/one-argument N with arbitrary/humanfeatures (impersonal se), a relational/two-argumentN coupled to an identity relation (reflexive se) or arelational/two-argument N coupled to a disjoint (non-identity) relation (reciprocal se). The impersonal uses(in the Romance and Slavic varieties that have it) andthe null object impersonal use (present in Slavic andabsent in Romance) are viewed as the consequence ofthe subject or object orientation of the pronominalanaphor, which is executed by movement to subjector object related positions that remedy the featuredeficiency of se.

Judy B. Bernstein (Syracuse University)concentrated on the derivation of certain word orderpatterns in Romance DPs where focused demonstra-tives, demonstrative reinforcers, possessives andsome quantifiers appear in the right periphery of theDP (Sp. el libro de matemáticas ese). Drawing on theparallelism between verbal and nominal projections,she argued that these orders involve movement(leftward “scrambling”) of the whole projection con-taining the head N, adjectives and complementsacross the element ultimately occupying the rightperiphery. As a consequence of this, the rightmostelement receives focus, which must be seen as thetrigger for this just apparently optional movement.By contrast, the non-neutral intermediate positionsof demonstratives and possessives (el libro ese dematemáticas) was argued to be derived by N-raising.

Word order issues were discussed extensively intwo papers, which independently argued for remnant

Going Romance 1998

Utrecht University, 10-12 December 1998by Josep Quer

14th Comparative Germanic Syntax Workshop

Lund University, Sweden, 8-9 January 1999by Kleanthes K. Grohmann

Conference reports Glot International, Volume 4, Issue 4, April 1999 26

movement in Spanish. Jon Franco (Universidad deDeusto) suggested an analysis of word order possibili-ties in Spanish small clauses that appeals to leftwardremnant movement. Unlike French or Italian, thesmall clause predicate can appear adjacent to themain verb, but the author argues that the incorpora-tion and the extraposition accounts of this patternare untenable mainly on empirical grounds. Theorder under examination would be the result ofmoving the small clause projection (AGRAdjP) overtlyto the Spec of an inner TopicP (located below AgrSP/TP) after the DP subject of the SC has vacated it andmoved to SpecAgrOP. Consequently, apparent freeword order in Spanish would follow from this innertopicalization strategy.

Reineke Bok-Bennema (Groningen University)presented an analysis of finite verb movement inSpanish that crucially appeals to remnant movementof the VP to SpecI. This allows for an explanation ofthe linearization of V with respect to different sortsof adverbials, as well as the fact that the auxiliarypied pipes the participle in the perfect without re-course to incorporation. The sequences where anadverbial intervenes between the auxiliary and theparticiple are argued to be cases of stylistic splitderiving from movement of V out of the remnant VPin SpecI to a higher I′ head. The proposal impliesthat object stranding reduces to movement of theobject to an object licensing position prior to VPmovement to SpecI. Light complements are shown tobe unable to vacate the VP before it moves, so theyget pied piped.

Sequence of tense phenomena were treated intwo different papers. Alessandra Giorgi (Universitàdi Bergamo) and Fabio Pianesi (IRST, Trento)devoted their talk to the peculiar temporal dependen-cies of the complements of fictional predicates likesognare ‘dream’ in Italian. Unlike with propositionalattitude predicates, the embedded events of dreamcan be temporally unrelated with respect to thematrix one. This property shows up most noticeablywith the imperfective past, which also appears inother contexts not requiring temporal anchoring.Other indicative tenses are possible as well, but theygive rise to “prophetic readings” about actual statesof affairs. They argue that fictional predicates do notestablish a relationship between a subject and thecontent of the subordinate clause, and in this senseare different from ordinary propositional attitudeverbs.

James Higginbotham (University of Oxford)proposed that sequence of tense is obligatory in(Italian and English) complement clauses because thetense information in Infl must be copied onto C andanaphora is obligatory. When [−past] moves from Inflto C, both copies get interpreted and a double accessreading arises: the [–past] tense is interpreted twice,anaphorically in C and non-anaphorically in theembedded Infl. If it is [+past] that moves, only thecopy in C is interpreted. For double-access cases, itmust be assumed that [–past] can be anaphoric to[+past] only from C. Additionally, the [+past] mor-pheme must be taken as either expressing anteriori-ty or triggering anaphora. Crucially, this kind ofmovement of Infl to C must be absent from relativeclauses.

The interpretation of mood contrasts was ad-dressed by Josep Quer (Universitat Autònoma deBarcelona), who defended the hypothesis that moodmarking (indicative/subjunctive oppositions) overtlyconveys information about the models where proposi-tions are interpreted. Specifically, he argued thatmood shift signals a change in the model of evalua-tion. This is clearly observed in the domain of clausalcomplementation: as opposed to weak intensionalpredicates, which introduce epistemic(-like)/veridicalmodels, strong intensional ones take clauses in thesubjunctive, as they contribute non-veridical models.Mood shift is also triggered by the model shift thatderives from a change in the individual anchor of amodel. The same kind of explanation extends tomood contrasts in relative clauses and adjunct claus-es like concessives.

In their talk, Ricardo Etxepare (LEHIA) andKleanthes K. Grohmann (University of Maryland)proposed that root infinitives in (adult) Spanish are tobe analyzed as containing an affixal null modal,

drawing a parallelism between these cases andsubjunctive clauses. On the basis of distributionalevidence (adverb placement, left dislocation, etc.), itwas argued that the complex head formed by the nullmodal and the infinitival raises overtly to C. Theabsence of root infinitivals in child Spanish wasargued to follow from the fact that subjunctivemorphology, which is acquired very early on, blocksthem. Certain differences with English and Germanroot infinitivals were also discussed.

Jacques Lamarche (Université du Québec aMontréal) provided an explanation for transivitycontrasts in French and English between verbs ofManner of Motion (marcher vs. walk) and Directedmotion (entrer vs. enter), which analyzes them asfollowing from the inflectional properties specific toeach language. While in French rich subject agree-ment was suggested to block the identification of theManner lexical constant by any other (potentiallyagentive) argument (Marie marche (*le chien)), it wasalso argued to make a non-lexical Agent of a verb ofDirected Motion close enough to the Theme in orderto establish a substantive link with the affectedentity (Marie entre la voiture). By way of contrast,the lexical Manner constant of English walk providesa link between the non-lexical Agent and the Theme(Mary walks the dog). However, the Interior constantof enter relates to the Goal rather than the Theme,and such a link is therefore unavailable (*Maryenters the car). Lack of (syntactic) agreement doesnot interfere with these configurations in English.

Jenny Doetjes (HIL/Leiden University) offereda detailed analysis of the notion of frequency asinstantiated in French. She proposed decomposingfrequency into two subcomponents: relative quantity(RQ) and iteration (IT). Frequency expressions willthen instantiate different specifications for suchvalues: (a) a frequency adverb like souvent is [+RQ],[+IT]; (b) a degree adverb like beaucoup is [+RQ], [–IT], and (c) an iterative adverbial like trois fois is [–RQ], [+IT]. Such distinctions are shown to besustained by the evidence that different contextsprovide as to the restrictions in the readings theyimpose. Habitual contexts accommodate (a) and (b)because they require [+RQ] readings. Relationalreadings are only compatible with (a), as they requireboth [+RQ] and [+IT]. Finally, (a) and (c) can havescope over indefinites, because they are [+IT].

The conference was followed by a one-day work-shop on the structure and acquisition of the lexicon.

ReferencesGrimshaw, J. (1990). Argument structure. Cambridge,

Mass.: MIT Press.Kayne, R. (1998). Overt and covert movement.

Syntax 1, 128–191.

14th Comparative Germanic SyntaxWorkshopby Kleanthes K. Grohmann

If you have never been to Lund, go there! It is abeautiful town in Sweden’s southwest whose Depart-ment of Nordic Languages hosted this year’s CGSW.A tightly packed programme turned out to be a greatsuccess not only due to the high level of performanc-es by the speakers but also by the organizers. Theidyllic atmosphere all around certainly contributed,and so did a dinner party that can hardly be beaten. Iwill begin this report with the three invited talksbefore briefly considering the other talks in the orderthey were presented.

Liliane Haegeman (U of Geneva) considered“Negation in West Flemish and the derivation of SOVword order.” On the empirical evidence of IPP-constructions in West Flemish, Haegeman proposeda generalized V-to-I movement mechanism within anantisymmetric approach to Germanic clause struc-ture: the interplay of sentential negation, functionalheads in the middle field and remnant categorymovement could lead to an analysis of West German-ic clause structure that maintains the antisymmetric,head-initial approach on the one hand (Kayne 1994and much follow-up work) and generalized V-to-Imovement on the other. Ken Safir (Rutgers U)talked about “Derived complementarity in the pat-tern of anaphora” which argued for exactly that: a

derivation of the complementarity that (potentially)dependent elements exhibit, not a consequence ofbinding theoretical stipulations. As such, Safir decom-poses Principle B into three different parts whichinteract to reproduce Principle B without strictuniversality due to cross-linguistic differences regard-ing morphological inventories. Picking up Reinhart &Reuland’s (1993) splitting of Principle B into a largedomain expressing non-coargument relations and asmaller one for coargument reflexives, he rejects theirReflexive Interpretation Principle and replaces it bythe Form to Interpretation Principle which, moreo-ver, replaces the original obviative principle, giving itmore power. Safir then considers in some depthdifferent types of reflexives and reciprocals amongthe Germanic languages. Finally, Jan-WouterZwart (Groningen University) combined “Bareargument structure and the syntax of middles.” Hebasically argued that primitive relations of predica-tion are licensed in unambiguous phrase structures(within a view of argument structure à la Hale &Keyser 1998). This analysis takes care of the debatewhether subjects of middles have raised from theirbase-position, generated as the internal argument ofthe middle verb, or not. Zwart starts off by consider-ing non-argument middles for which he develops ananalysis that involves base-generation of the adverbin the internal argument position of the middle verb.He then proposes that argument middles are in factnon-argument middles, at least in Dutch. Thussyntactically raising internal arguments is superflu-ous, and so is a flexible relation between lexicalconceptual structure and syntactic structure. Empiri-cal support for the collapse of argument and non-argument middles comes from certain propertiescommon to both, such as aspectual, complementa-tion, diathesis and selection restrictions.

Thórhallur Eythórsson (U of Manchester)talked about “VP order and clausal architecture indiachrony.” Challenging the V2-approach to OldEnglish that enjoyed much attention in recent years,he argued for IP-internal V-movement independent ofV2. The evidence comes from the very early stages ofGermanic and takes diachronic developments intoaccount. These data also shed light on the issue ofVP-internal ordering of arguments. ThorbjörgHróarsdóttir (U of Tromsø) proposed an interestinganaylsis of “VP-preposing in Icelandic.” In a Kayneanantisymmetric model, differences between the Ger-manic OV- and VO-languages might be accounted forby obligatory argument raising in the former (à laZwart 1997 and related work) or by some sort ofverbal movement in the latter (Hinterhölzl 1998,Kayne 1998 and others). Hróarsdóttir adopts thelatter approach, arguing that V-movement takesplace in VO-languages, and proposes displacement ofthe remnant VP across complements. ShalomZuckerman (Groningen) considered “The underlyingstructure of embedded participles in Dutch” andargued that the underlying order should be auxiliary-particle; the data come from markedness/preferencejudgements across dialects on the one hand, andfrom child language on the other. Structurally,Zuckerman assumes an approach where the partici-ple moves obligatorily to the left of the auxiliary,while the auxiliary may then undergo optionalraising, thus deriving either order. He then considersvarious alternatives to implement optionality ofsyntactic operations into the minimalist program.

Ute Bohnacker (Lund U) talked about “Rootinfinitives in bilingual child Icelandic-English” andsuggests to take RIs are not defective constructions orsyntactic misanalyses on the part of the child butrather an independent construction type made availa-ble in the grammar (of the child). The subject of thisstudy, a child of Icelandic parents growing up in Eng-land, exhibits an extremely extended period of RI-use.This long RI-stage and the nature of her languageprovide further evidence for Bohnacker’s view of RIs.Tying in the phenomenon with various aspects ofverbal syntax, Bohnacker concludes that the late RIs inthe child’s Icelandic production are novel imperativeswith a preverbal English-like subject. “Optimal strate-gies for identifying D” were Øystein AlexanderVangsnes’ (U of Bergen) concern, evolving from thepuzzle that pre-adjectival articles in definite NPs pose,namely that they are obligatory in Mainland Scandina-vian and Faroese, but not in Icelandic. A look at the

Conference reports Glot International, Volume 4, Issue 4, April 1999 27

element, much on a par with dialectal use of ‘for’ inEnglish (Belfast, Ottawa), stemming from raising,control, negation and ECM-constructions. The finaltalk, “Prepositions as bare infinitival complementiz-ers in some Westgermanic languages,” was deliveredby Reimar Müller (Tübingen U) who turned thetraditional derivation of (many) complementizers inthe Germanic languages from prepositions around.Rather than looking at for...to-type constructions (asin the previous talk), Müller considered complemen-tizers in Westgermanic dialects (synchronic anddiachronic) which select bare infinitival comple-ments. Müller analyses these constructions as CPswith a bare infinitive arguing that historically theseCPs appear independently in Frisian and Swabian.He employs a very careful study of data from thesetwo dialects in their history to show infinitival gram-maticalization accounting for present-day verbal,adjectival, prepositional and sentential to-infinitives(Frisian) and prepositional adjuncts, purposive con-structions and nominal infinitives (found in olderstages of Swabian).

I’m sure I can talk for all participants expressingmy thanks to the Lundian linguists for a very suc-cessful workshop which was desperately needed. Inthis respect let me note that the number of submit-ted abstracts was fortunately much higher than thelast time (see the report in Glot International 3.1) butthere could have been more. The community canonly hope for ample submission of abstracts for thenext workshop which will be held in Groningen toensure another CGSW of high quality.

ReferencesArad, M. (1996). A minimalist view of the syntax-

lexical semantics interface. University College ofLondon Working Papers in Linguistics 8, 215–242.

Beck, S. (1996). Quantified structures barriers forLF-movement. Natural Language Semantics 4,1–56.

Belletti, A. & L. Rizzi (1988). Psych-verbs and thetatheory. Natural Language and Linguistic Theory6, 291–352.

Bonet, E. (1994). The Person-Case Constraint: amorphological approach. The MorphosyntaxConnection [MIT Working Papers in Linguistics22], edited by H. Harley & C. Phillips, 33−52.

data reveals that related questions involve AP/NP-movement in Icelandic definite NPs (versus demon-strative NPs), adjectival incorporation in NorthernSwedish dialects and a correlation between lack ofagreement and co-occurrence of pre-nominal posses-sives and definite articles. Vangsnes proposes that allfour cases are optimal solutions in the respectivegrammars with respect to meeting the requirementthat D be identified by an appropriate element. Hethen considers the most appropriate element in consid-erable details, concluding that the optimality hinges ondiffering lexical properties of classes of constituents,defined either by lexical semantics or morphologyacross the dialects.

Peter Svenonius (U of Tromsø) considered“The expression of negation in Germanic” languagesand especially the typological differences yielded bydifferent settings of two parameters: whether or notto move negative XPs to the left yields languagesexhibiting Negative Concord (West Flemish versusdialects of English) and Double Negation (German,Dutch and Scandinavian versus Standard English).Pointing to some problems of a strict adherence tothe Neg Criterion (Haegeman & Zanuttini 1991),Svenonius replaces it by two separate clauses whichforce (i) the Neg-head to be licensed by an negativeexpression in its checking domain (universal) and (ii)each negative expression to be licensed in the check-ing domain of the Neg-head (parameterized). Jóhan-nes Gísli Jónsson (U of Iceland) spoke on “Caseabsorption with st-verbs in Icelandic” where thesubjects may bear thematic but not idiosyncraticCase (cf. Yip, Maling & Jackendoff 1987). A number ofcorrelations can be found in the behaviour of theseverbs with respect to nominative and dative subjects(medio-passives, reflexive/reciprocal meaning, incho-ative meaning) which lead Jónsson to consider DativeSubstitution (or “Dative Sickness” as can be read inIcelandic newspapers): there exists a strong tendencyto replace accusative by dative on Experiencer sub-jects, or to replace either by nominative if the subjectis Patient or Theme (Nominative Substitution).OtherCase patterns in Icelandic further support the Case-distinction that he argues for (thematic versusidiosyncratic). And Cedric Boeckx (U of Connecti-cut) ended the first day with “Quirky agreement inIcelandic, English, and elsewhere” in which heargued for a proper understanding of the fact thatonly partial agreement (number, not person) existsbetween a verb and a nominative element when aquirky subject is present, resulting in a properunderstanding of Quirky Case and more tentatively,inherent Case. Boeckx adopts Bonet’s (1994) Person-Case Constraint under which object agreement islimited to third person if dative agreement takesplace. This accounts for the above mentioned limitedagreement pattern found. Quirky subjects, on theother hand, fully agree (if only rather abstractly) withthe verb and should thus be assigned full subjectstatus. One aspect of his approach is that agreementand Case are only one feature (see George & Kornfilt1981), as also recently endorsed by Chomsky (1998).As such, Boeckx can easily do without agreementprojections.

Ursel Luhde (U of Wales) started the firstsection of the second day after Ken Safir’s talk about“Aspectual features and argument structure” inwhich she extends recent work on aspectual features(e.g. Arad 1996): three aspectual features — OR‘Originator’ (the prototypical agent), EM ‘EventMeasurer’ (traditional theme/patient) and DEL‘Delimiter’ indicating the endpoint of the event) —are associated with three verbal heads in a VP-shell,As such, theta-roles are not primitives in the gram-mar but derived configurationally. Empirically, thisapproach intends to not only account for alternativeauxiliary selection for verbs of motion (across Ger-manic and Romance languages, for example), but alsoderive the observations that a resultative phrase canonly be predicated of the direct object, a subject of anergative verb is the underlying object and others.Elin Bech and Tor Åfarli (U of Trondheim) consid-ered “The syntax of two types of object experiencerverb in Germanic,” namely the difference of irritate-type verbs (denoting feelings or emotions) and ap-pear-type verbs (denoting judgements or acts/statesof imagining) where only the former verbs assign anexternal theta-role; this analysis accounts for the

differences in passive constructions or topicalization.The data from Norwegian are compared and con-trasted with German and Old Norse correlates; thesetwo types do not correspond to any of the groups ofpsych-verbs in Italian as described in Belletti andRizzi 1988. Anna-Lena Wiklund (U of Umeå) inves-tigated “Morphosyntactic parasites and the path toPF” concentrating on supine constructions in Swed-ish which are parasitic in that they only appear inthe context of an adjacent superordinate supine verb.This is analysed as copying before the word getsassigned PF-features, as the phonological form of theparasitic supine is identical to its non-parasiticcounterpart. On the other hand, this copying opera-tion must follow LF-interpretation because theparasitic is interpreted as an infinitive (not as aparticiple). Wiklund also considers underspecificationin general, proposing that LF only acknowledgespositively specified values (hence the possible inter-pretation of supines as infinitives). PF-realization ofunderspecified verbs depends largely on language-specific “Vocabulary” (taken to be a raw lexicon). Theentire approach is an extension of Distributed Mor-phology (Halle & Marantz 1993), much in the spirit ofZwart (1997), which furthermore splits the Lexiconand insertion into explicit components of Vocabularyand Morphology.

Mark de Vries (U of Amsterdam) proposed toanalyse “Extraposition of relative clauses as specify-ing coordination.” He capitalized on 11 properties ofextraposed relative clauses that apply exactly alike toinstances of specifying coordination. Arguing againstan analysis of extraposition as rightward movement,adjunction or stranding, de Vries favours a coordinat-ed structure to underlie these cases where a deletionoperation gets rid of the relative clause in the lowerconjunct. He also tackles arising problems such asthe loss of a promotion analysis of relative clauses,unlimited coordination of unequal categories, theabsence of syntactic dependencies between the firstand second associate and the fact that there is no“base position” needed for binding. As a result theunderlying coordinate structure is identical to thatfound in specifying coordination and/or asyndeticcoordination. Kleanthes Grohmann (U of Mary-land) considered “Multiple interrogatives, discourserestrictions and quantifier interaction.” He onceagain (see the SCIL 10-report in Glot International3.9/10 and also Grohmann 1998) presented theobservation that questions with two wh-expressionsin German underlie “Discourse-Restricted Quantifica-tion” according to which a possible set of referentsmust have been established in the context prior toasking which he still considers evidence that bothwh-elements must move overtly into a position inwhich this special type of discourse-linking can belicensed; TopP in the C-domain is the candidatesuggested, arguing explicitly for wh-topics. Empiricalevidence comes from the “Beck-effects” (Beck 1996)which can be generalized into a ban on topicalizationof non-topicalizable material: increasing quantifiersmay intervene between the two wh-elements, de-creasing quantifiers may only follow both. Theextension to earlier proposals concern a principledjustification of DRQ and possible parametrizationwhich ties in syntactic clause types with semanticsentence types. As such, all derivations must contin-ue at LF where the force-expressing C-head (cf.“ForceP” in Rizzi 1997) specifies a construction as adeclarative or interrogative, for example; in thesecases, the lower wh-topic raises to SpecCP at LF tounambiguously categorize the multiple wh-questionas such (a sub-type of interrogatives).

Kersti Börjars (U of Manchester), Kate Burr-idge (La Trobe U, absent) and Sue Spence (U ofManchester) found “For...to-constructions in varietiesof German and English: Consequences for clausestructure” which are particularly interesting as theGerman zu or ze ‘to’ is part of the inflectional system,unlike fer or fir (found in some German dialects)which under standard assumptions should correlateto the English complementizer ‘for’ in these construc-tions. However, evidence from German “dialects”suggest complications of the matter: on the onehand, both fir and ze are often optional in Luxem-bourgish; on the other hand, zu is lost completely inPennsylvania German. There seems to be goodevidence that fer in these variants is an inflectional

Goodies Glot International, Volume 4, Issue 4, April 1999 28

IntroductionThe LINGUIST Network (http://linguistlist.org/)held its first on-line conference, “Geometric andThematic Structure in Binding” in the fall of 1996.By all accounts this was a successful endeavor,and we hope that there will be many future con-ferences. This first conference taught us a numberof lessons that may be of some value in the plan-ning and implementation of future electronicgatherings. We review some of our experiencewith online conferencing in these notes.

Why on-line conferencesOne of the goals of the conference was to takeadvantage of the potential of the Internet toencourage scholarly interchange. Electronic con-ferences have at least 3 potential advantages. Oneof these was obvious from the first, but the othertwo were discovered in the course of the confer-ence.

First, the potential audience of linguists whocan “attend” an electronic conference is muchlarger than that for a traditional conference. Themedium allows linguists to be actively involvedwherever they may be, regardless of financial,geographical or political constraints. As a result,nearly fifty countries were represented at thismeeting. There were also some 525 subscribers, avery impressive number given the very specificfocus of the conference (and this does not countthe number of researchers who went directly tothe conference Web Site but who where not officialsubscribers).

Second, the medium offers ways to overcomeor minimize some disciplinary constraints. Many ofthe “attendees” subscribed in order to access up-to-the-minute research in a subdiscipline outsidetheir own, research that they would not normallyread. The hypertext format allowed us to provideancillary information about the presenters and thetheoretical context of the papers. The conferenceorganizer established hyperlinks to relevanthome-pages, to bibliographic information aboutcited works, and to definitions of key terminology.Such information, we hope, enriched the confer-ence for everyone, but it seems to have beenespecially useful to non-specialists.

Third, on-line conferences allow the immedi-ate and permanent archiving of the papers pre-sented, and of the commentary as well. As one ofthe participants mentioned, this makes the confer-ence proceedings a useful teaching tool: he pointedout that teachers can ask students to read andcomment on the papers, then later compare theirown comments to the archived discussion. Thepapers and discussion are, of course, also availableto scholars who want to consider the papers morecarefully than a traditional conference allows.

Basic structure of the conferenceThe organization of the conference was quitesimilar to a regular one in some respects. Forexample, a call for abstracts was made, and theusual procedure for reviewing and acceptingabstracts was followed. One salient difference wasthat all correspondence — including submission ofabstracts and papers — was electronic. In fact,this meeting did not require a single sheet ofpaper. Unlike most linguistic conferences, howev-er, the authors of the successful abstracts wereasked to provide a finished version of their paperssome time before the conference took place. Thesepapers were put on the Web, as well as sent outvia e-mail to the “attendees”, all of whom were puton a special e-mail list called “linconf”. Discussionwas carried out by e-mail, but all comments were

archived and immediately translated into hyper-text format so that they could also be made availa-ble on the Web.

Because the electronic medium requires extrareading and discussion time, and because theparticipants were from many different time zones,this conference lasted for three weeks.

To allow thoughtful consideration of the workpresented, the conference was divided into threesessions each with three papers, roughly onesession per week; a keynote address by Prof.Howard Lasnik brought the total number ofpresentations to ten. The original idea was tomake all of the three papers of a session availableat the beginning of the week, have a two-dayreading period, and then open up the floor foropen discussion, with the guidance of a moderator,for the rest of the week. At the end of the confer-ence, there would then be discussion of all confer-ence papers. As we will see below, however, inactual practice, more time was needed for bothreading and discussion.

The linguistic theme of the conference wasnarrowly focused in order that the meeting be ofmanageable size and scope; and, furthermore, theprimary conference organizer has some expertisein binding theory and it seemed useful to select atopic that we knew fairly well — indeed, this wasimportant for everything from establishing thereview board to editing papers for final presenta-tion.

In the interest of space, the specifics of thelinguistic theme won’t be reviewed here, but alldetails are available at LINGUIST sites, e.g.:

http://linguistlist.org/linconf/http://www.emich.edu/~linguist/linconf/http://www.sfs.nphil.uni-tuebingen.de/

linguist/linconfhttp://www.philol.msu.ru/linguist/linconf/

Subscriber commentsAt the end of the meeting, we encouraged com-ments from subscribers in order to better deter-mine which components of the conference workedand which didn’t. Some thirty subscribers wrote inand their comments and suggestions provedvaluable. A summary follows.

On the positive side, comments can be dividedup into those that apply to on-line conferences ingeneral, and those relevant to this meeting inparticular. Common general themes were:

• On-line conferencing is a good idea and shouldbe continued.

• Such conferences do save money and time.• Because there is more time to read and digest

complex material, questions/comments andresponses can be more thoughtful than at aregular conference.

Comments specific to this first meeting include:

• That it was well-organized.• Papers and comments were of good quality.• It was a good opportunity for non-specialists to

get exposure to some of the latest developmentsin generative syntax; exposure that might notbe practical otherwise.

As for the negative side, the central generalizationwas this:

• A regular conference is more focused in thesense that attendees do not have classes and theother normal responsibilities of the professionfor the duration (usually 2 or 3 days) of the

meeting. This means that conferees can concen-trate entirely on the conference. Because an on-line meeting must take place over asubstantially longer period of time, however,such focus is not possible. This makes readingall the papers and comments potentially diffi-cult.

Representative comments on this point include:

• I would have participated more if I had time toread everything. The disadvantage of an on-line conference is that life does not stop duringthe conference, as it does for a real conferencethat you travel to.

There were a number of interesting suggestionsfor future meetings. One was to have more infor-mation available about subscribers so that every-one could keep better track of who wascommenting. Another is that meetings be keptopen such that people could make comments onpapers for a longer period of time.

ConclusionsOver all, we think it clear that this first meetingwas a success and we hope there will be manymore such meetings in the future. One conferencecurrently under development is on EndangeredLanguages. This conference will serve as a start-ing point for a larger project that the moderatorsof LINGUIST (Anthony Aristar, Helen Aristar-Dry,and Andrew Carnie) have in mind. LINGUISTwould like to serve as a central repository forelectronic grammars, dictionaries, and corpora. Inparticular, LINGUIST would like to focus primari-ly on endangered and minority languages. Theconference would serve as a starting point in thatendeavour. And it may have a special session onelectronic methods & standardizations for gram-matical information, a session on the role of gram-mars in language preservation, and sessions onlanguage reports (similar to, but hopefully not incompetition with, the endangered sessions at theLSA annual meeting). This conference would bescheduled for summer 1999.

AcknowledgementThanks to Anthony Aristar, Helen Alistar-Dry, andAndrew Carnie for their help in preparing thisreport.

LINGUISTIC E-CONFERENCES

by Daniel Seely