IN THIS ISSUEjlawler/glot3-5.pdfState-of-the-Article Glot International, Volume 3, Issue 5, May 1998...

IN THIS ISSUEState-of-the-Article

Marc van Oostendorp onSchwa in phonological theory 3“In general it seems that the requirementson schwa syllables are much stronger thanthose on syllables headed by othervowels.”

Column

Neil Smith onAcquired whining 9“To return at the end to Woody Allen, it isclear that the quotation I started with wasinspired by his admiration for Chomsky.”

Dissertations

The syntax of Past Participles. Agenerative study of nonfinite con-structions in Ancient and ModernItalianby Verner Egerlandreviewed by Yves D’Hulst 10

Direct object scrambling in Dutchand Italian child languageby Jeannette Schaefferreviewed by Astrid Ferdinand 13

Books

Base rules phoneticallyby Joan Mascaróreviewing Grounded phonologyby Diana Archangeli and DouglasPulleyblank 16

Announcement23

Goodies

Signal analysis and digital signalsby Stefan Frischreviewing WinSAL-Vby Ingolf Franke 24

Conference reports

GLOW Special:ReportsGLOW Hyderabad, India, January20–23 (by Georges Tsoulas) 19GLOW Tilburg, April 15–18, includ-ing the workshops (by Ileana Paul,Marc van Oostendorp, and Jan-Wouter Zwart) 20Interview with Henk van Riemsdijk 18(by Lisa Cheng and Rint Sybesma)“We’re a great field, but it is small and inconsiderable danger.”

ConSOLE 6, December 15–17,University of Lisbon (by João Costa) 22

Extra

The number of death 25A Linguistic Mystery in EightInstallmentsby Chris Sidney TappanChapter 4. Towards some hypotheses“Ah, Nichael,” Paul said. “I know him, he isone of my informants. He is interestingbecause his phonology is all screwed up. Didyou notice?”

Interview

“What’s right about X-bar syntax isthe X and the bar. What’s wrong iseverything else!”An interview with James McCawley1→26

WITH JAMES McCAWLEY,UNIVERSITY OF CHICAGO

by Lisa Cheng & Rint Sybesma

What do you think about frameworks and theo-ries? Wouldn’t it be much more efficient if every-body worked within the same framework, usingthe same theory?

As an epistemological anarchist in the tradition ofFeyerabend, I am delighted to have multipleframeworks being used. Even if I regard a certainframework as completely crackpotted, that doesn’tmean that it’s not a resource that might be valua-ble to me. I share Feyerabend’s idea that anyframework is going to lead its adherents to comeup with facts that quite likely you wouldn’t comeup with so easily in your own framework. So Iregard the multiplicity of frameworks as highlydesirable.

Working within a particular framework todaydoesn’t imply that you have to work within thesame framework tomorrow. It doesn’t prevent youfrom stepping outside the framework. I teachcourses that deal with frameworks other than thesort of very revisionist kind of transformational

grammar that is represented in most of the stuffthat I am doing myself. When I teach a course onRelational Grammar or Montague Grammar, Idescribe it as a tour, not a sight-seeing tour but ashopping tour. That way I hope to give peoplesome idea of the kinds of facts that Montaguegrammarians and Relational grammarians areled to look at. I do comparisons and show howessentially the same thing is done in differentframeworks. But I also point out cases where thevariant way of looking at things has yieldedinteresting results. Take, for example, in Rela-

Monthly Magazinefor Linguists

May 1998

ISSN 1381–3439

Glot Internationalappears monthly,

except June & July.

© Copyright 1998 byHolland Academic Graphics.

All rights reserved.

Price single issue:Dfl 14,50

5Volume 3

ISSUE

Continued on page 26

“What’s right about X-bar syntaxis the X and the bar. What’swrong is everything else!”

INTERVIEW

Reference Glot International, Volume 3, Issue 5, May 1998 2

Contributors to this issue

João CostaDepartamento de Linguistica Geral e Romanica | Univer-sidade de Lisboa | Faculdade de Letras | Cidade Universi-taria | 1699 Lisboa Codex | [email protected]

Verner EgerlandDepartment of Romance Languages | University of Lund| Sölvegatan 7 | S-223 62 Lund, [email protected]

Astrid FerdinandA. van Scharnlaan g47 | 6226 EL Maastricht | TheNetherlands

Stefan FrischIndiana University | Speech Research Laboratory |Psychology Building | Bloomington, IN 47405 | [email protected]

Yves D’HulstLeiden University | Department of Italian | P.O. Box 9515| 2300 RA Leiden | The [email protected]

Joan MascaróDepartament de Filologia Catalana | Universitát Autóno-ma de Barcelona | 08193 Bellaterra | Catalonia | [email protected]

Marc van OostendorpHIL/Leiden University | Department of Linguistics | P.O.Box 9515 | 2300 RA Leiden | The [email protected]

Ileana PaulDepartment of Linguistics | McGill University | 1001Sherbrooke West | Montréal, Québec | Canada H3A 1G5

Jeannette SchaefferDepartment of Foreign Literatures and Linguistics | BenGurion University of the Negev | P.O. Box 653 | BeerSheva | [email protected] or [email protected]

Chris Sidney Tappan137 Gardner Road | Brookline MA 02146 | U.S.A

Neil SmithUniversity College London | Department of Linguistics |Gowerstreet | London WC1E 6BT | [email protected]

George TsoulasDepartment of Language and Linguistic Science | Univer-sity of York | Heslington, York-YO1 5DD | England | [email protected]

Jan-Wouter ZwartUniversity of Groningen | Department of Linguistics |Oude Kijk in ’t Jatstraat 26 | 9712 EK Groningen | [email protected]

Dissertations reviewed

Verner Egerland, The syntax of Past Participles. A genera-tive study of nonfinite constructions in Ancient and ModernItalian. Lund University. 1996. Supervisors: Kjell-ÅkeGunnarsson and M. Rita Manzini. 352pp. Lund UniversityPress. [Études Romanes de Lund 53.] ISSN: 0347–0822.ISBN: 91–7966–368–0. No price given.

Jeannette Schaeffer, Direct object scrambling in Dutch andItalian child language. UCLA. March, 1997. Supervisors:Nina M. Hyams (chair), Hilda Koopman, DominiqueSportiche, Terry Au. 158pp. Available through: UCLAWorking Papers in Linguistics: [email protected]: US$10.

Books reviewed

Diana Archangeli & Douglas Pulleyblank, GroundedPhonology. Cambridge, Mass., MIT Press. 1994. 502pp.,with index. ISBN 0–262–01137–9. GBP 46.50.

Goodies

WinSAL-V, developed by Ingolf Franke. Trier, Germany:Media Enterprise, 1996. PC software on CD-ROM. CDROM version 349DM, with video option 399DM + 30DM S& H (outside Germany). Discount for students. On-lineordering information: <http://www.media-enterprise.de/home_eng.htm> (English); <http://www.media-enterprise.de/home.htm> (German).

Colophon

EditorsLisa Lai-Shen ChengRint SybesmaHolland Institute of Generative Linguisticsaddress: Department of General Linguistics | LeidenUniversity | P.O. Box 9515 | 2300 RA Leiden | TheNetherlands | phone: +31 71 5272104 | fax: +31 715272615 | e-mail: [email protected]

Conference reports andannouncements editorJan-Wouter ZwartGroningen University | Department of Linguistics |Oude Kijk in ’t Jatstraat 26 | 9712 JK Groningen |The Netherlands | e-mail: [email protected]

Goodies editorAndrew CarnieDepartment of Linguistics | Harvard University |77 Dunster Street | Cambridge MA 02138 | USA |email: [email protected]

Correspondents/Regular contributorsEulàlia Bonet (Barcelona) | João Costa (Lisbon) |Marcel den Dikken (Tilburg University) | PaulaFikkert (Konstanz) | Helen de Hoop (Utrecht) |Colin Phillips (Delaware) | Christine Tellier (Mon-treal) | Dylan W.-T. Tsai (Tsing-hua) | Guido VandenWyngaerd (Brussels) | Shohei Yoshida (YokohamaNational University)

ColumnistsElan Dresher (Toronto), Neil Smith (London)

Information for contributorsLinguists who want to contribute a State-of-the-Article

or a Review are invited to contact the Editors.Linguists who have recently finished their disserta-

tion are invited to forward a copy to the Editors.Everybody is welcome to contribute to Glot Inter-

national. But before you send in anything, please,contact the Editors.

Glot International prints conference announcements,calls for papers, conference programs, etc.

Subscription informationGlot International appears every month, except inJune and July. Subscriptions are for the printed orthe electronic (pdf) version.

Prices for Volume 3 (1998):INSTITUTIONAL RATE NLG 120.0INDIVIDUAL RATE

pdf only NLG 035.0printed version NLG 050.0

POSTAGE

surface mail Netherlands NLG 032.5other countries NLG 037.5

air mail Europe NLG 042.5other countries NLG 047.5

surface shipping in three batches(March—August—December)

Netherlands NLG 015.5other countries NLG 020.5

Subscriptions at the individual rate must be ordereddirectly from the publisher.

Subscriptions at the institutional rate may be orderedfrom your bookseller or subscription agent, ordirectly from the publisher.

Prices exclude 6% VAT (EC only).Orders must be prepaid.Payments preferably by giro, credit card (American

Express, Euro/Access/ Master, Visa) or interna-tional money order. Credit card orders mustinclude the complete cardholder’s address. Thereis an additional charge of Dfl 17,50 for paymentsby international bank cheque or bank transfer.

PublisherHolland Academic GraphicsJulius Röntgenstraat 15 | 2551 KS The Hague |The Netherlands | phone +31 70 4480203 |fax +31 70 4480177 | e-mail: [email protected] |http://www.hag.nl | bank acc. ABN AMRO 49 88 22796 | postal giro 6261107 | VAT nr. NL0588.61.798.B01

AdvertisementsContact the publisher

PrintingPrinted in The Netherlands

Objectpro inEnglish!

NEXT MONTH!

State-of-the-ArticleJonathan Bobaljik on Floating quantifiers

Guest columnHelen Dry onThe LINGUIST list: A personal history

DissertationsYuko Yoshida’s On pitch accent phenomena inStandard Japanesereviewed by Nancy Ritter

Book reviewsZ¦ eljko Boškovic’s The syntax of nonfinitecomplementation. An economy approachreviewed by Marcel den Dikken

And Chris Sidney Tappan’s Murder mysteryThe Number of Death continues...“He’s not creepy, he just does not have the ability todistinguish important things from the unimportantdetails. I think you are betting —”

State-of-the-Article Glot International, Volume 3, Issue 5, May 1998 3

1. IntroductionIf a language has schwa in its vowel inventory,

this segment usually plays special role in thephonology of the language. It can only occur in asimple type of syllable; or it is invisible for thestress system; or it is epenthetic; or it is the resultof reduction; etc. Linguistic theory has to explainthis special behaviour of schwa: why is it exactlythis segment which behaves in exactly this way inso many languages? A lot of subtheories of phonolo-gy are at stake — syllable structure theory, stresstheory, the theory of (sub)segmental representation,and of interaction between segments — and thequestions surrounding schwa therefore will proba-bly not be resolved until the perfect theory of pho-nology has been discovered.

On the other hand, schwa offers an excellenttestcase for many theories of phonology. It thereforeshould come as no surprise that the vowel has beenstudied intensively within various branches ofgenerative phonology. This article offers an over-view of some of the ideas that have been proposedand some of the facts that have been discoveredover the past thirty years. None of these analysesseems perfect in the sense that it can handle all therelevant facts; but together the different propsalsgive us a rather precise picture of the behaviour ofschwa.

It is my aim in this paper to present in a moreor less systematic way those facts about schwawhich I consider to be most characteristic. I alsobriefly present some of the more influential theoriessurrounding this vowel within the generativeparadigm.

I find it useful to distinguish pretheoreticallybetween three types of schwa:

* e-schwa; this is the type of schwa that alternateswith zero. The e in the name of this vowel refers toepenthesis because this is one of the possible sourc-es of the alternation. It is of course also possiblethat this alternation is caused by deletion of schwain certain environments, and in some (monostratal)theories one may deny the role of vowel epenthesisor deletion altogether and claim that the occurrenceor non-occurrence of schwa in a certain position is amatter of phonetic interpretation (Charette 1991).In any case, if an alternation is observed, thiscounts as a diagnostic for E-schwa.* r-schwa; this is the name for a schwa that alter-nates with a full vowel. The r stands for reduction,but again, it is also possible that the alternation isthe result of a ‘fortition’ of schwa; and in principle,also nonderivational accounts are available.* s-schwa; this is stable schwa, which is a restcategory from a descriptive point of view: if there isno reason to call a schwa e-schwa or r-schwa, I callit s-schwa. S-schwa is usually already present inthe underlying structure, but this is not a distinc-tive property of the vowel: also e-schwa can beunderlyingly present (when the alternation is theresult of deletion), and so can r-schwa (when thealternation is the result of fortition).

I do not think the taxonomy just presented shouldbe awarded any theoretical status: a schwa may

change its behaviour over the years, and indeed wealso find cases where a vowel is e-schwa and r-schwa at the same time (as in the French paradigmappeler ["ple~"p6le] ‘to call’ j’appelle [�"p7l] ‘I call’).

The classification is presented here because ithelps to distinguish the various roles that schwa canplay in the phonology of a language. Before we canstudy the behaviour of schwa in natural language, itmay be useful to go into the question of how thesegment should be represented phonologically.

2. The representation of schwaSchwa is special not only from the point of view

of phonological theory. We find that it is a specialvowel also if we look at it from an articulatory pointof view: it could be described as a ‘targetless vowel’for which no inherent articulatory target has beenspecified, or as a vowel which targets a neutralvocalic position, ‘the mean tongue-tract variableposition for all the full vowels.’ (Cf. Browman andGoldstein 1992).

From a standard autosegmental/metricalpoint of view, we have several options to representthe emptiness of schwa. We can divide theseoptions roughly into two sets. On the one hand,we can award schwa the status of an emptyprosodic position without (auto)segmental materi-al (an empty mora, an empty nucleus, an empty V-slot in a CV-model). On the other hand, we canassign some minimal segmental specification tothis vowel. In the feature geometric model ofMcCarthy (1988), for instance, major class fea-tures constitute the root node. Schwa can then beseen as a bare [–cons] root, or as a root node withan empty (vocalic) place node. It is hard to findgood empirical tests that could distinguish be-tween these options. Many things depend on theother theoretical choices we make vis à vis thestructure of syllables and segments. In the re-mainder of this article, I will arbitrarily assumethat schwa could be pictured phonologically as abare root node, in (1):

(1)[−cons]

Schwa can be literally seen as the maximallyunmarked vowel: it is marked for being a vowel([−cons]) but for nothing else. I believe that, in theideal case, as many phonological properties aspossible should be made to follow from the interac-tion of this fact and general principles of phonology.Ideally, no linguistic rule or constraint shouldspecifically refer to schwa.

I will take the fact that many languages do nothave a schwa in their inventory as an example.Within constraint-based phonology, we could ana-lyse this fact as a consequence of a specific con-straint against schwa:

(2)*SCHWA: Don’t have schwa

Yet it seems more reasonable to see this as a specif-ic consequence of something more general. Forinstance the set of constraints that have been calledFILL within Optimality Theory (Prince and Smolen-sky 1993):

SCHWA IN PHONOLOGICALTHEORY

Marc van Oostendorp

The schwa is a special vowel: it is often the result of reduction or epenthesis. It isindispensible, yet, invisible for stress systems. In the following article, Marc vanOostendorp gives an overview of the different contexts in which we find schwa(and why) and of the different ways phonological theory has been trying to accountfor the distribution, characteristics and behaviour of this intriguing vowel.

(3)FILL: Empty segments are disallowed.

In Van Oostendorp (1995) it is proposed that thebehaviour of schwa can be understood in terms of arelation between segmental material and prosodicstructure. If a segment has certain features, itshould project to certain positions (say, the head ofa syllable or a foot), and if a segment appears in thehead position of a certain syllable or a foot, it shouldhave certain (vocalic) features. This is regulated bythe following general constraint scheme:

(4)Where F=a (vocalic) feature, P=a prosodic positionPROJECT (F, P): If a vowel is [+F], it should appears in position P.PROJECT (P, F): If a vowel appears in position P, it should be [+F].

Van Oostendorp (1995) shows that particularsettings for the arguments F and P can explain thebehaviour of many vowels in a number of languag-es. Schwa is a special case of this. Since schwa doesnot have any vocalic feature at all, it is restricted toa very small number of positions. In certain lan-guages this number may be so limited, that it isnull.

3. E-schwaOne of the most prominent characteristics of

schwa is that it alternates with zero. As a matter offact the term shvarabakti vowel was initially meantto refer to the epenthetic vowel in Tiberian Hebrew.Epenthesis — insertion of non-underlying material— is of course one possible source of the schwa-zeroalternation that defines e-schwa. Another source isdeletion of schwa.

If schwa is a very simple vowel, there arereasons to assume that this vowel is particularlyeligible for both epenthesis and deletion. In the caseof epenthesis, it is the most economical choice tomake if we have to insert a vowel: no materialneeds to be epenthesized beyond the feature [−con-sonantal]. In the case of deletion, it is simpler todelete only the specification [−consonantal] than todelete this feature alongside others (which is need-ed in case of the deletion of a full vowel). It is thuspredicted that if a language (i) allows schwa on thesurface (of a given subphonology) and (ii) has avowel-zero alternation (within that same subpho-nology), schwa will be among the alternating vow-els. I have no evidence that would falsify thisprediction; I know of no study which tries to sys-tematically verify it either.

Assuming for the sake of simplicity that therecan be no constraints on underlying forms (this is ofcourse the standard assumption within OptimalityTheory), we have evidence for deletion if there arecertain phonologically well-defined contexts inwhich schwa does not surface. We have evidence forepenthesis on the other hand if there are phonologi-cally well-defined contexts in which we always finda schwa. Interestingly, there are languages in whichwe find evidence both for schwa epenthesis and forschwa deletion: there are certain contexts in whichschwa always occurs, there are other contexts inwhich schwa never occurs and there may even becontexts in which it is unpredictable whether wefind a schwa or not.

In languages of this type, epenthesis anddeletion often seem to be defined in complementarycontexts. These languages are of particular interestto constraint-based theories of phonology, becausethey often involve some form of ‘rule conspiracy’:epenthesis and deletion seem to conspire to attainsome absolutely well-formed (syllable) structure. Arather well-known instance of such a language isFrench. The distribution of schwa in French hasbeen a topic of debate during virtually any period oftime within the history of generative phonology(Anderson 1982, Basbøll 1981, 1988, Charette 1991,Dell 1973/1985, Durand 1976, 1986, 1990, Morin1982, 1988, Noske 1982, 1988, 1992, Van Oosten-dorp 1995, Schane 1965, 1968, 1974, Selkirk 1978,1982, Tranel 1981, 1987, 1994). The facts of Frenchare rather complicated, and a lot of dialectal andprobably even idiolectal variation is involved. Yet,the French facts are quite typical of the behaviourof e-schwa in natural language. Noske (1992)presents a useful classification of the facts ofFrench e-schwa, which I will take as my guideline


here.In the following discussion I distinguish between

‘static’ and ‘dynamic’ pieces of evidence. ‘Static’evidence for a constraint (or rule) is distributional:we never find schwa in a certain position, thereforethere should be a constraint or a morpheme struc-ture condition against schwa in that position (or arule deleting it there); we always find a schwa in acertain position, therefore there should be a con-straint forcing it to be present (or a rule epenthesiz-ing it). In order to be able to accept ‘static’ evidencefor an output constraint, we have to assume someversion of OT’s ‘Richness of the Base’ (Prince andSmolensky 1993): there are no constraints on under-lying representations, and everything can be input tothe grammar.

‘Dynamic’ evidence for a rule or a constraint isbased on alternations. A schwa in form A correspondsto a zero in exactly the same environment in themorphologically related form B. I take this to meanthat there is a constraint against zero in A, oragainst schwa in B (or that there is an epenthesisrule applying to A or a deletion rule applying to B).

Type A: Prevocalic schwa deletion. The evidencefor this type of process is that schwa cannot befollowed by a full vowel. It is actually hard to findevidence for the rule in French, because we need acontext where /x6#v/ surfaces as [X#V], and schwa isdeleted at the end of a word or phrase in many casesanyway. The alternation of the masculine singulardeterminer le is sometimes cited as evidence for therule. This clitic surfaces as /l6/ before a consonant-initial stem (le camerade ‘the comrade’ [l6kamrad])and as /l/ before a vowel-initial stem (l’ami ‘thefriend’ [lami]). The problem is, however, that asimilar alternation applies to the feminine singulardeterminer la (la camerade — l’amie), but we haveno reason to postulate a general rule of a-deletionbefore a full vowel.

The other dynamic pieces of evidence usuallygiven are equally dependent on other assumptions.For instance, it is often assumed that the femininesingular suffix is a schwa (autre ‘other, FEM.’ /otR/ +/6/ [otR]). The problem is that this schwa neversurfaces at the end of a phrase, and in those environments in which it does surface (autre femme ‘otherwoman’ [otR6fam]), this may well be due to schwaepenthesis (alternation type D discussed below).Although the indirect phonological evidence, which isbased on the process of liaison but which I cannotreview here (cf. Encrevé 1988 and references citedthere), is favourable to the hypothesis that thefeminine suffix is schwa, there is no really strongphonetic evidence that the schwa is present at all, orthat it is ever deleted specifically in the contextbefore a vowel.

On the other hand, we do have evidence that theconstraint against schwa in front of a vowel is opera-tive in French; although schwa can be rather freelydistributed over all positions in the French word,there are no words in which it immediately precedesa vowel (either a full vowel or schwa). French sharesthis property with many other languages, includingDutch, English, German and Indonesian. In Dutchwe actually have some pieces of dynamic evidence: ifthe schwa-final word (elite [elit6]) is followed by avowel-initial suffix (-air [7r]), the schwa gets deleted(elitair [elit7r] ‘snobbish’) (Booij 1995).

I propose that this behaviour of schwa can beunderstood within Optimality Theory in the follow-ing way. As we have seen above, we need to posit aconstraint against the occurrence of schwa. Thisconstraint outranks most faithfulness constraints onschwa, but it is itself dominated by several well-formedness constraints on syllable structure. Sche-matically:

(5)Well-formedness >> *SCHWA >> Faithfulness.

This ranking implies that schwa never surfacesunless it is required to do so by the syllable structureconstraints. This in turn means that it is almostimpossible to say whether we see deletion or epen-thesis at work in a given case: schwa is deletedwherever it is not necessary because of *SCHWA andinserted wherever it is necessary because of high-ranking Well-formedness.

If an underlying schwa is both preceded andfollowed by a consonant, it needs to surface in orderfor the word to satisfy general constraints on sylla-ble structure, in particular constraints againstcomplex onsets (*[lkamrad]-[l6kamrad]). Yet ifschwa is followed by a full vowel, there is no neces-sity for it to surface ([lami] is perfectly well-formedas far as syllable structure is concerned). This maybe the reason behind the deletion of schwa.

A famous exception to the deletion of schwabefore full vowels is formed by a class of words witha so-called h aspiré: a consonant position that wasprobably filled by a /h/ in earlier stages of thelanguage but that is presently phonetically empty,although phonologically it still behaves as a full-fledged consonant:

(6)mets le dessus [m7l6dsy] ‘put it on!’retourne la [rturnla] ‘return it!’mets le dehors [m7l6d6fr] ‘put it outside!’rehausse la [r6osla] ‘raise it again’

It is standardly assumed that this behaviour can beunderstood by assuming some more or less abstractconsonant in this position. The schwa is thereforenot adjacent to the full vowel in cases like these;hence, it is not subject to the deletion process.

Type B: Postvocalic schwa deletion. A similarline of reasoning may help to explain why schwadoes not surface on the righthand side of a fullvowel. Also in this context, there is no necessity forschwa to surface. The syllable structure of [ami] isactually better than the structure of hypothetical[ami6], because the latter has one extra onsetlesssyllable.

Again it is rather hard to find dynamic evi-dence for a deletion rule — the evidence that iscited by Noske 1993 for instance involves onceagain the feminine suffix which reputedly getsdeleted after a full vowel: risée ‘laughed at, FEM’/rise/+/6/→[rise], but as I have already pointed outabove, word-final schwa gets deleted in many othercases as well — yet there is some static evidence inthat the relevant constraint is at work in the lan-guages just mentioned: we don’t find any (mono-morphemic) words in which a schwa immediatelyfollows a full vowel.

The question may now arise why not all vowelsget deleted when they occur next to another vowel. Ithink that the reason for this is that deletion ofschwa is less costly than deletion of other vowels,due to the fact that the schwa literally is a substruc-ture of all the other vowels. If it is allowed to deletea full vowel in a certain context, this implies that itis allowed to delete a root node in that context;hence, also schwa will be deletable in that context.

Type C: schwa deletion in a ‘two sided opensyllable’. A schwa which occurs in an open syllablethat is preceded by another open syllable may getdeleted. In practice this means that an underlyingschwa occurring in the context ...V$C__$C(C)V...can disappear on the surface. In contradistinctionto the two previous cases, this type of alternation iseither optional or otherwise subject to stylisticconditions.

The following relatively famous examples serveto illustrate this point.

(7)Henri devrait partir [ãrid6vR7paRtiR]‘Henri would have to leave’

[ãridvR7paRtiR]Jacques devrait partir‘Jacques would have to leave’ [�akd6vR7paRtiR]

In this case it looks as if underlying schwa’s shouldbe deleted in principle, but deletion is blocked if itwould result in an unwellformed cluster of conso-nants, in particular one which cannot form a rea-sonable syllable structure. Certain additionalstipulations are necessary, however, since [dvr] isnot otherwise a wellformed onset cluster of French,yet the syllabification that most speakers wouldassign to Henri devrait partir seems to be[ãri.dvR7.paR.tiR]. This can be understood in severalways. It is possible for instance that there is still a‘trace’ of schwa between [d] and [v]; this is thesolution adopted in Charette (1991) modulo techni-cal details. It is also possible that there is a distinc-

tion between two rounds of syllabification (e.g.lexical and postlexical); again modulo technicaldetails, this seems the type of solution favoured byNoske (1993).

Type D: Schwa epenthesis in the environmentCC]_[C. A schwa is (optionally) inserted betweentwo words, if the first word ends in a cluster, andthe second word starts with a consonant. In thiscase syllable structure seems the driving force(rather than a force blocking deletion); clusters ofconsonants are avoided also in this case:

(8)un contact pénible [œ« kf«takt(6)penibl]‘a painful contact’un index formidable [œ« 7«d7ks(6)ffRmidabl]‘a terrific index’

Maybe we can understand these examples in thefollowing way: complex codas are dispreferred inFrench. A schwa is needed to support the finalconsonant in a cluster. However, schwa can neversurface at the end of a phrase in French, as we willsee below. It also cannot appear (is deleted) before avowel. The only case in which we therefore see thesupporting schwa appear on the surface is phrase-internal before a consonant. The fact that we haveepenthesis in these cases is therefore readily under-stood; the fact that the epenthetic vowel is schwaalso is readily understood. Schwa is the simplestvowel of all; therefore it is more economical to insertthis vowel than it is to insert any other.

Charette (1991) observes that within com-pounds, stress also seems to play a role in thebehaviour of schwa. A schwa has to surface if thefirst component ends in a consonant cluster and thesecond is monosyllabic; if the second is polysyllabic,the schwa cannot surface (according to Charette, asimilar pattern is found in Québec French in phras-es as well).

(9)porte-clefs ‘keyring’garde-fou ‘railing’porte-manteau ‘coat rack’garde-manger ‘meat safe’

Since stress is on the final syllable of the compound,it does not seem unreasonable to suppose that itplays a role in these examples. The question is ofcourse why schwa should surface when it is imme-diately followed by a stressed syllable. Charette’sanswer to this is that French feet are iambic, andthat the unstressed syllable in the foot should be aschwa. These assumptions are not uncontroversial,but as far as I know alternatives to Charette’sproposal are yet to be worked out.

Type E: Schwa deletion in phrase-initialsyllables. The existence ot this property, men-tioned by Noske (1993), seems subject to debate: notall speakers of French agree on the data. In anycase, some speakers allow alternations such as thefollowing:

(10)revenez (demain) [r6v6ne]~[rv6ne] ‘demain’te fais (pas de bil) [t6f7]~[tf7] ‘don’t worry’

Syllable onsets [rv] or [tf] are not normally allowedin French. It is possibke that the initial syllable ofthe word allows a more complex structure (as itdoes in Polish, cf. Rubach and Booij 1990).

According to Charette (1991) this type ofdeletion is restricted to bisyllabic words: cheval‘horse’ may be pronounced as [šfal] , but chevalier‘knight’ is [š6valje]. In this case, it looks as ifstressed syllables allow more complex onsets. Onthe other hand, the explanation given here seems torun counter to the explanation of porte-clefs vs.porte-manteau just given. In this case, Frenchspeakers do not try to build a ‘perfect’ iamb [š6val].

Type F: Schwa deletion in phrase-final sylla-bles. This final type of ‘alternation’ is one of themost important ones: schwa never surfaces phrase-finally (except in Midi French, Durand 1986).

(11)je vois l’autre [�6vwalotR] ‘I see the other’voilà mon oncle [vwalamõnõkl] ‘there is my uncle’la terre est plate [lat7r7plat] ‘the earth is flat’la route est longue [laRut7lõg] ‘the road is long’

It is not immediately clear why schwa should be


deleted in final position. One possibility, suggestedin Van Oostendorp (1995), is to extend the scope ofthe constraint FINALC to the phrase. The constraintwould thus say that all phrases have to end in aconsonant. Deletion of full vowels under this viewwould be blocked by the fact that this would involvedeletion of too many features. Deletion of schwa, onthe other hand, is less costly, and therefore, it isallowed in this particular configuration.

The solution of Government Phonology istechnically a little bit different. According toCharette (1991), a domain-final empty nucleus isexceptionally licensed to remain empty in French.In any case, it seems as if something special needsto be said about the end of the phrase in French. Nospecial theory is needed, however, in order to de-scribe why it is exactly schwa that behaves asspecial in this particular spot.

4. R-schwaSchwa not only alternates with zero in lan-

guages of the world: it also often alternates with fullvowels. There is an instance of this in French:schwa alternates with [7] in the following contextsin this language (Dell 1973/1985):

* when schwa is the final phonetically expressedsegment of a word ([paqu[7]t])* when schwa phonetically occurs in a word-inter-nal or word-final closed syllable

It seems reasonable to assume that underlyingschwa turns to [7] in certain environments: namelywhen it is stressed or when it is in a closed syllable.This is not surprising; schwa disfavours stressedpositions or closed syllables also in other languages,as we will see below. This thus seems to be aninstance of fortition rather than reduction. We mayof course wonder why schwa turns to [7] rather thanto some other vowel; a good theory of segmentalstructure should provide us with an answer to thisquestion.

Reduction to schwa is well-known from thestudy of Germanic languages. A very simple theoryof vowel reduction in English is presented in Chom-sky and Halle (1968). The reduction rule in thiswork basically amounts to (12):

(12)Unstressed vowels turn to schwa.

Syllable structure plays an important role in theanalysis of E-schwa; in the case of R-schwa it isoften stress that seems to trigger the alternation.

Within Optimality Theory, there are two possi-ble ways to explain (12). In Alderete (1995) it isproposed that in principle all vowels turn to schwa.Schwa is the unmarked vowel and Alderete as-sumes that there is a constraint which requires allvowels in the word to be unmarked (I will call thisconstraint TURNSCHWA for the sake of simplicity, butit can of course be seen as an instance of the gener-al constraint *STRUCTURE). On the other hand thereare position-specific faithfulness constraints thatspecifically require stressed syllables to be faithfulto the underlying structure. These position-specificconstraints I will shorthand here as FAITHFUL-STRESS.

An analysis for the English word apron wouldnow run along the following lines.

(13)/aprfn/ FAITHFULSTRESS TURNSCHWA FAITHFUL

aprfn **!apr6n * *6pr6n *! **

Alderete (1996) claims that FAITHFULSTRESS con-straints can also be put to use in analyses of phe-nomena other than reduction. For instance, heshows that the fact that epenthetic vowels usuallyavoid stressed positions (even if they are not schwa)can be understood in terms of this constraint. It isworse to epenthesize into a stressed position andviolate both FAITHFULSTRESS and FAITHFUL, than toepenthesize into an unstressed position in whichcase only FAITHFUL will be violated.

Yet there are also problems connected to anapproach based on position- specific faithfulness.We have already seen evidence for the fact that weneed to posit a constraint against schwa (e.g.*SCHWA, (2)) within the universal inventory of

constraints, if only because there are languageswhich do not have schwa. We then have a rankingFAITHFULSTRESS >> *SCHWA >> FAITHFUL within ourfactorial typology. This however gives us a resultwhich is probably absurd: a language in whichunderlying schwa can only surface in a stressedposition:

(14)/6pr6n/ FAITHFULSTRESS *SCHWA FAITHFUL

6¢pr6n **!6¢pren * *épren *! **

More in general, a FAITHFULSTRESS account impliesthat there are more contrasts possible in stressedpositions than in unstressed positions. While this istrue in general, it is not true in the case of schwa.The prediction that there are languages in whichwe can only have schwa in a stressed position doesnot seem to be borne out.

An alternative is to make TURNSCHWA ratherthan faithfulness into a position-specific constraint,or a set of such constraints. We then get some formof projection constraints, as proposed in Van Oost-endorp (1995):

(15) PROJECT-FOOT

a. Unstressed vowels turn into schwab. Schwa wants to be unstressed

These constraints fit into the general theory ofProjection outlined in (4): (12b) says that a vowel inthe head position of a foot needs to have someminimal feature structure; (12a) says that vowelswith a certain feature structure desire to be in thehead of a foot. If they are not, they turn into schwa.An analysis of English apron would run along thefollowing lines.

(16)/aprfn/ PROJECT-FOOT (a) PROJECT-FOOT (b) FAITHFUL

áprfn *!ápr6n *6¢pr6n *! **

A language that only allows schwa in stressedpositions is not part of the typology in this system:if we rank *SCHWA very highly, we get a language inwhich schwa does not appear at all; if we rankFAITHFUL very highly, we get a language in whichschwa can in principle occur in any position.

Another advantage of this approach is it that itexplains why there are languages in which underly-ing schwa’s are just as stressless as the epentheticones or the ones which are the result of reduction.(Underlying schwas will be the topic of the nextsection.) There are also problems with this ap-proach, however: most importantly, the fact thatgenerally stressed positions do license more featurecontrasts than unlicensed ones, and that epentheticvowels are unstressable requires an explanation.

Regardless of which of the two lines of researchwill turn out to be most fruitful, the analysis ofEnglish vowel reduction should be more sophisticat-ed. Burzio (1994) argues, quite convincingly in myview, that the traditional view on reduction inEnglish (every unstressed vowel gets reduced toschwa) is too simple. It causes all sorts of problems,because it forces us to assign stress in an extremelycomplicated way and furthermore to introduce theformal device of destressing rules into the theory inorder to undo some of the results of the stressingrules. Burzio (1994) proposes to simplify the stressrules. As a result of this, the conditions on reductionget more complicated. Burzio’s main point is thatreduction is generally blocked in unstressed closedsyllables ending in an obstruent, but not in un-stressed syllables ending in a sonorant; in this way,he can get rid of the so-called ‘sonorant destressing’rules.

I find Burzio’s arguments convincing becausethe difference between closed and open syllables,and the sonority of a following consonant also playa role in the behaviour of other types of schwa: wehave already seen that schwa cannot occur in closedsyllables in French, and I will give an example of S-schwa in Dutch below. It therefore seems morenatural to say that unstressed vowels resist reduc-tion because they occur in a closed syllable, than tosay that they are stressed, and possibly laterdestressed. There is no independent evidence forthe latter assumption.

DutchI will now turn to a language in which reduction issubject to even more factors than it is in English.This language is Dutch (the following discussion isbased on Booij 1981, 1995, Kager 1989, Martin1969, Van Oostendorp 1995). Factors that presuma-bly play a role in Dutch vowel reduction are:

i. Frequency:Vowels in more frequent words tend to get reducedmore easily than those in non-frequent ones.ii. Differences between sociolects and dialects, anddifferences betwee style levelsCertain dialects/sociolects/style levens show morereduction than others.

It is hard to see how factors (i) and (ii) could possi-bly be taken into account within generative gram-mar; presumably, this is the reason why they areusually ignored in the literature (but see Hinskenset al. (1997) for several recent attempts to integratesome of these concepts into phonological theory). Wewill not go into these factors either. Other factorscan be analysed more easily.

iii. Syllable type:Vowels in open syllables are easier to reduce thanthose in closed syllables; in some idiolects (e.g. theone of Booij 1981, and the one of the author of thisarticle, but not in the one of Kager 1989) reductionin closed syllables is only possible if the followingconsonant is deleted.

(17)/b7nzin6/ [b6zín6] (*[b6nzín6]) ‘petrol’/k"ntor/ [k6tór] (*[k6ntor]) ‘office’

This is the same restriction as the one that Burzio(1994) claimed to exist for English. Yet in the litera-ture Dutch there is less controversy that the pur-ported restriction on open syllables belongs to thetheory of reduction rather than to the theory ofstress assignment. The reason for this is thatreduction is optional in Dutch; apparently it is moreattractive to say that reduction can optionally applyto invariably assigned metrical structures than it isto say that the stress rules themselves are optional.

I think that the restriction can be understoodin the following way. A closed syllable is morecomplex than an open syllable. A minimal vowel isnot sufficient to license such a complex structure. Aprojection constraint relating the quality of thevowel to the rhyme structure of the syllable inwhich that vowel occurs is therefore to be preferred.

iv. Stress position:In Dutch — as in English and apparently any otherlanguage in which reduction is dependent on theposition in the metrical structure — only un-stressed vowels get reduced. As a matter of fact, wecan distinguish between two types of unstressedvowel (Van Zonneveld 1985). Vowels in so-called‘semi-weak’ positions — immediately following astressed position — are easier to reduce than thosein ‘weak’ positions.

(18)/fonolo?i/ formal style [fònolo?í]

less formal style [fòn6lo?í]informal style [fòn6l6?í]impossible *[fònol6?í]

Similar facts can be found in other languages aswell (Burzio 1994:113). In English, tat[6]magouchiseems preferable to tatam6gouchi. Jacobs (1989)observes a similar patterning for syncope in thehistory of French:

(19)Latin: similitudinem → simlitudinem → Old French:sembletume ‘resemblance’

The precise way to describe the difference betweenthe two types of unstressed position is dependent onour precise assumptions regarding metrical struc-ture. If we assume, with Burzio (1994) that feet canbe ternary, we should distinguish between ‘foot-internal’ and ‘peripheral’ unstressed syllables. If wework in a theory in which only binary feet areallowed, we should distinguish between weaksyllables in a foot and syllables which are somehowleft unfooted by the parsing algorithm.

Vowel reduction to low vowels in Russian(/stol-á/ — stal-á) only occurs in the syllable


directly preceding the stressed syllable (hence-forth: the pretonic syllable henceforth). All vowelsreduce to [6] in unstressed, non pretonic syllables(the following facts have been taken from Alderete1995).

(20)Nom. Sg. stól slóv-oGen. stal-á slóv-aDat. stal-ú slóv-uInstr. stal-óm slóv-omLoc. stal-é slóv-ev[ó]d6j nom. pl.v[a]d-á nom. sing.v[6]davóz ‘water carrier’

Here it seems that the position next to (on thelefthand side of) a stress is less likely to undergofull reduction than vowels in other positions.

v. Position in the word:Vowels in absolute word-initial or absolute word-final position do not reduce. This observation isabsolutely surface-true in Dutch. No matter whatthe style of speech is, we do not find reduction ofthe vowels that are absolutely word-initial orabsolutely word-final. The /e/ is reduced quiteeasily in Dutch, for instance, but not if it is thefirst or the last segment of the word, as the follow-ing segments are intended to show:

(21)/plezir/ [pl6zír] ‘fun’/e?al/ [egál] (*[6?al]) ‘even’/tffe/ [tffe] (*[tf¢f6]) ‘toffee’

We may think of this as some kind of alignmenteffect of phonological to morphological structure:the edges of output words should correspondexactly to their morphological specifications.

An alternative explanation for the word-initial vowels is that syllables consisting of onlyschwa is disallowed; a syllable needs some mini-mal specification (Cohen et al. 1959). An argu-ment for this alternative may be that vowels nextto /h/ also cannot be reduced:

(22)/plezir/ [pl6zír] ‘fun’/helas/ [helas] (*[h6las]) ‘alas’

We will see in the next section that also underly-ing schwas cannnot occur in absolutely word-initial position; this is true for many otherlanguages as well, including German, Indonesianand French.

There is also an alternative explanation forthe non-reduction at the right edge of the word.Some authors (e.g. Brink 1970) have claimed thatDutch schwa is always the result of reduction:words such as mode ‘fashion’ are underlyinglyspecified as [mode]; in that case reduction isobligatory in these environments and the wordtoffee [tffe] should somehow be marked as anexception (it is true that there is only a handful ofwords in Dutch ending in unstressed [e]).

vi. Vowel quality:The quality of the underlying vowel plays a role inthe choice between reduction and non-reductionin many languages. In English, for instance, highvowels are not as likely to undergo reduction asnon-high vowels (Chomsky and Halle 1968). Midvowels in Dutch are easier to reduce than highand low vowels; front vowels are easier to reducethan back vowels; rounded vowels are easier toreduce than unrounded ones. This is illustrated inthe following table (from Kager 1989):

(23)Reduces in Weak position Semi-weak position/e/ formal formal/a/ semi-formal semi-formal/o/, /i/ semi-formal informal/u/, /y/ informal excluded

Interestingly, Dressler (1973) found more or lessthe same hierarchy of segments for Breton vowelreduction to schwa. As far as I am aware, thestructure of this hierarchy is still in need of anexplanation.

5. S-schwaR-schwa can be a non-underlying schwa which

is the result of a reduction operation; e-schwa canbe a non-underlying schwa which is the result ofminimal vowel epenthesis. Both types of schwa can

also be underlying; the alternation in the case of r-schwa can then be seen as a case of fortition. Thealternation in the case of e-schwa can be analysedas a result of deletion. As a matter of logics, we alsoexpect underlying schwas which do not alternate;these vowels I call s-schwa.

Even though s-schwas are stable in the sensethat they occur both in the underlying and in thesurface structure of a word, they still often dis-play some special behaviour. The factors involvedmay be familiar by now: s-schwa does not occur inan unstressed position or in a syllable that is toocomplex.

We have seen that schwa in French cannotoccur in a closed syllable. In Dutch, there are alsoseveral restrictions on the syllable shape in whicha schwa can occur. For one thing, we do not findschwa in an onsetless syllable. The static evidencefor this is that the vowel does not occur as thefirst segment of a word (*[6?al]), and that it doesnot occur immediately after a vowel (*[hi6t]). Thisrestriction seems quite common in languages ofthe world: it can be found also in Indonesian,Slavic languages, and French. On the other hand,Dutch schwa also cannot occur in a syllable with acomplex onset (*[papavr6]). I have found no otherlanguages yet in which this restriction seems tohold (except maybe related Germanic languagessuch as German and certain dialects of Norwe-gian, Kristoffersen 1995).

The restrictions on schwa are not just restric-tions on the onset; there are also restrictions on therhyme of a schwa syllable. Other than full (short)vowels (harp ‘id.’ [h"rp], Belg [b7lx] ‘Belgian’, ramp[r"mp] ‘disaster’), schwa cannot be followed by twotautosyllabic consonants, except if the secondconsonant is coronal (*[ad6rp], *[kat6lx], *[ad6mp],anders ["nd6rs] ‘different’). Marked coda’s, just likemarked onsets, are not allowed.

Even within the class of schwa syllables witha simple coda, we find a clear asymmetry. Schwahas a preference for syllables closed by a sono-rant: words such as hennep, tinnef, etc. are veryrare. It is not uncommon for languages to requirethat codas can only contain a sonorant. There isno evidence, however, that syllables headed by afull vowel are subject to this requirement inDutch.

In general it seems that the requirements onschwa syllables are much stronger than those onsyllables headed by other vowels. In Van Oosten-dorp (1995) I proposed that this is due to the theoryof Projection in (4): schwa as an (almost) emptyvowel does not have sufficient material to license allof the syllabic nodes that are necessary in order toget complex onsets or codas. Simple vowels are notallowed to project complex syllables.

Schwa in stressIn most languages that have (underlying) schwa,the vowel is left unstressed. Roughly speaking,this can happen in two ways: schwa can prefera-bly end up in the weak position in a foot; thisseems to happen in a language like Dutch, inwhich stress always falls on the penultimatesyllable if the final syllable contains a schwa(ballade [b"lád6], *[b"¢lad6], *[b"lad6¢]); if the finalsyllable contains a full vowel, more (lexical)variation is possible (pánama, pyáma, chocolá).Schwa thus is not completely invisible for stress;under the standard assumption that Dutch feetare trochaic, we should say that schwa can onlyoccur in the weak syllable of a foot.

Alternatively, schwa can behave as if it iscompletely invisible for the stress algorithm; aninstance of this is Indonesian. In this language(Cohn 1989, Cohn and McCarthy 1994) primarystress is usually on the penultimate syllable of theword (a−b) secondary syllable is on the first sylla-ble and on every odd syllable counting from theright-hand side of the word. An exception to thisare words which contain a schwa. These can bebest understood as if schwa were not present atall:

(24)a. cát ‘print’b. cári ‘search for’c. bicára ‘speak’

d. bìjaksána ‘wise’e. kòntinuási ‘continuation’f. òtobìográfi ‘autobiography’g. àmerikànisási ‘Americanization’a′. b6rí ‘give’b′. s6t6láh ‘after’c′. gám6lan ‘Indonesian orchestra’d′. apárt6men ‘appartment’e′. c6rìt6ra ‘story’f′. p6r6mpúan ‘woman’g′. dìf6rensiási ‘differentiation’h′. dìv6rsifikási ‘diversification’

It is unclear, at least to me, how exactly thesefacts should be understood. We may understandwhy schwa cannot occur in the stressed position ofa foot, but this does not clarify why it cannotoccur in the unstressed position of a foot. Particu-larly intriguing are the cases in (g′) and (h′).There is no secondary stress on a full vowel inthese cases: we do not find [dìf6rénsiási]. Accord-ing to Cohn (1989) and Cohn and McCarthy(1994) the reason for this is that stress only seesfull vowels: a trochaic foot is therefore built overthe first two full vowels of the word, i.e. the footstructure of this word is [(dìf6ren)si(ási)]. It thuslooks as if in this case schwa is not strong enoughto project a grid mark at all, not even one whichappears in the weak position of a foot.

Schwa remains unstressed also in mostdialects of Yupik (Hayes 1995). Different dialectsof Yupik choose different means to escape anunstressed schwa. Stress is generally on thesecond syllable of the word in Yupik; the stressedvowel is lengthened (/qayani/ — [qaya:ni]). If thissyllable contains schwa, the vowel is deleted inthe General Central dialect of Yupik, as if it wantsto escape a stress position. Stress is then assignedto the first syllable of the word.

(25)/qayapigkani/ [qayáápixkani] ‘his own future kayak’/qan6qa/ [qánqa] ‘my mouth’

GCY schwa therefore should be classified as an e-schwa; it is however different from the cases of e-schwa we have seen above because the triggeringfactor for the alternation is not syllable structure,but rather stress. A pattern that is fairly similarto the one of GCY can be found in Alutiiq Yupik.

In other dialects of Yupik, schwa does getstressed, but this stress does not trigger lengthen-ing, as it does on other vowels. Different dialectsof Yupik may solve this problem in different ways.In the Norton Sound dialect, for instance, theconsonant following the schwa gets lengthened:/at6pik/ — [at6¢ppik]. In this way, the stressedsyllable is still heavy, even though schwa needsnot be lengthened in order to achieve this result.In the Central Siberian dialect of Yupik, on theother hand, nothing happens to the stressedschwa-headed syllable at all: /at6pik/ — [at6¢pik].

6. ConclusionSchwa is special in the sense that it has a

more limited distribution than other vowels: forinstance it can often only occur in a rather simplesyllable type, and only in an unstressed position.Schwa is also special in the sense that it is thetarget of reduction and deletion; furthermore it iseasier to delete schwa than it is to delete a fullvowel. Phonology should be able to describe all ofthese facts.

In order to fully understand the behaviour ofschwa, we need a fully developed theory of sylla-ble structure, of metrical structure, of segmentalstructure, and of the way in which these differentdimensions of phonological structure can interact.Inversely, while developing these subtheories, wesharpen our view of schwa. Most of the factsmentioned above could probably not have beenraised without sufficient understanding of phono-logical structure, phonological derivation andphonological constraint interaction. Some of thefacts mentioned here are still rather problematicfor any phonological theory known to the authorof this overview; chances are that new facts maybe discovered while phonologists are developingtheir theories. I suspect that we will not have asatisfying theory of schwa untill we have a satis-fying Theory of Everything.


A schwa bibliography

Anderson, S. (1982) ‘The analysis of French schwa;Or, how to get something from nothing’. Lan-guage 58 121−138.

Archangeli, D. (1984) Underspecification in Yawelm-ani phonology and morphology. Dissertation,Cambridge, Mass: MIT.

Barry, W.J. (1992) ‘Comments on chapter 2: ‘ “Tar-getless” schwa; An articulatory analysis”’ InPapers in Laboratory Phonology II: Gesture,Segment, Prosody, G. Docherty and R. Ladd,eds. Cambridge, UK: Cambridge UniversityPress.

Barry, W.J. (1995a) ‘Schwa vs. schwa+/r/ in German’.Phonetica 52, 228−235.

Barry, W.J. (1995b) ‘Variation in schwa + /r/ inGerman’. In Proceedings of International Con-gress of Phonetic Sciences 1995. Stockholm:ICPhS.

Basbøll, H. (1981) ‘Metrical theory and the Frenchfoot’. In Phonologica 1980; Akten der vierteninternationalen Phonologietagung, W. Dresslerand O. Pfeiffer, eds. pp. 12−36.

Basbøll, H. (1988) ‘Sur l’identité phonologique duschwa français et son rôle dans l’ accentation etdans la syllabation’. In Verluyten (1988).

Basbøll, H. (1990) ‘Distinctive features, syllablestructure, and vowel space’. In Certamen phono-logicum II: Papers from the 1990 cortona pho-nology meeting, P.B. Bertinetto, M. Kenstowicz,and M. Loparco, eds. Torino: Rosenberg andSellier. pp. 3−18.

Bell, A. (1978) ‘Syllabic consonants’. In Universals ofhuman language, vol. II: Phonology, Greenberg,ed. Stanford: Stanford University Press. pp.153−201.

Bennink, C. (in prep.) Harmony and disharmony inHungarian and Turkish. Dissertation, Tilburg:Tilburg University.

Berendsen, E (1986) The phonology of cliticization.Dissertation, Utrecht: University of Utrecht.Also appeared at Dordrecht: Foris.

Berendsen, E. and W. Zonneveld (1984) ‘Nederlandseschwa-invoeging op z’n Deens’. Spektator 14,166−17.

Bergem, D. van (1993) ‘Acoustic vowel reduction’.Speech Communication 12, 1−23.

Bergem, D. van (1995) ‘A model of coarticulatoryeffects on the schwa’. Speech Commun. 14,143−162.

Bethin, C. (1992) Polish syllables. Columbus, Ohio:Slavica.

Bobaljik, J. (1997) ‘Mostly predictable: Cyclicity andthe distribution of schwa in Itelmen’. Manu-script, Harvard. Available at ftp://ruccs.rutgers.edu/pub/OT/TEXTS/archive/208−07971/208−07971.ps

Booij, G. (1976) ‘Klinkerreductie in het Nederlands’.Leuvense Bijdragen 65, 461−470.

Booij, G. (1981) Generatieve fonologie van het Neder-lands. Utrecht & Antwerpen: Het Spectrum.

Booij, G. (1992) ‘Fonologische en fonetische aspectenvan klinkerreductie’. Spektator 11, 295−301.

Booij, G. (1983) ‘French C/0 alternations, extrasyl-labicity and Lexical Phonology’. The LinguisticReview 3, 102−134

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

Bouchard, D. (1981) ‘A voice for “e muet”’. Journal ofLinguistic Research 1, 17−47.

Brink, D. (1976) Problems in phonology; A generativephonology of Dutch. Dissertation, University ofWisconsin.

Brockhaus, W. (1995) Final devoicing in the phonolo-gy of German. Tübingen: Niemeyer.

Browman, C. and L. Goldstein ‘“Targetless schwa”:An articulatory analysis’. In Papers in Laborato-ry Phonology II: Gesture, segment, prosody, G.Docherty and R. Ladd, eds. Cambridge, UK:Cambridge University Press.

Burzio, L. (1989) ‘Prosodic reduction’. In Studies inRomance Linguistics, C. Kirschner and J.Decesaris, eds. John Benjamins, Amsterdam.

Burzio, L. (1994) Principles of English stress. Cam-bridge, UK: Cambridge University Press.

Charette, M. (1991) Conditions on phonologicalgovernment. Cambridge, UK: Cambridge Uni-versity Press.

Chomsky, N. and M. Halle (1968) The sound patternof English. Cambridge, Mass: The MIT Press.

Clements, N. and E. Sezer (1982) ‘Vowel and conso-nant disharmony in Turkish’. In The structure ofphonological representations, Vol. II, H. van derHulst and N. Smith, eds. Foris, Dordrecht.

Cohn, A. (1989) ‘Stress in Indonesian and bracketingparadoxes’. Natural Language and LinguisticTheory 7, 167−216.

Cohn, A. and J. McCarthy (1994) “Foot alignmentand apparent cyclicity in Indonesian”. Manu-script, Cornell University and University ofMassachusetts.

Dell, F. (1973/1985) Les règles et les sons. Paris:Herrmann. First edition 1973, second, revisededition 1985.

Dell, F. (1976) ‘Schwa précédé d’un groupe obstru-ante-liquide’. Recherches Linguistiques 4,75−111.

Dell, F. (1979) ‘On French phonology and morphologyand some vowel alternations in French’. Studiesin French linguistics 1.3.

Dell, F. (1995) ‘Consonant clusters and phonologicalsyllables in French’. Lingua 95, 5−26.

Donegan, J. (1978) On the natural phonology ofvowels. Dissertation, Ohio State University.Published by Garland Press, 1985.

Dressler, W. (1973) Allegroregeln rechtfertigenLentoregeln; Sekundäre Phoneme des Breto-nischen. Inssbruck: Innsbrucker Beiträge zurSprachwissenschaft 9.

Dressler, W. (1975) ‘Methodisches zu Allegro-Regeln’. InPhonologica 1972, W. Dressler and F. Mares, eds.München: Fink Verlag. pp. 219−234.

Durand, J. (1976) ‘Generative phonology, dependen-cy phonology, and Southern French’. Lingua etStile 11, 3−23.

Durand, J. (1990) Generative and nonlinear phonolo-gy. London: Longman.

Durand, J., C. Slater and H. Wise (1988) ‘Observa-tions on schwa in Southern French’. In FrenchSound Patterns: Changing Perspectives, C.Slater, J. Durand and X. Bate, eds. Associationfor French Language Studies no. 32, Universityof Essex.

Encrevé, P. (1988) La liaison avec et sans enchaîne-ment; Phonologie tridimensionelle et usages dufrançais. Paris: Éditions du Seuil.

Fidelholtz, J. (1979) ‘Word frequency and vowelreduction in English’. Proceedings of the ChicagoLinguistics Society 11, 200−213.

Fouché, P. (1959) Traité de prononciation française;2e édition. Paris: Klinksieck.

Giegerich, H. (1987) ‘Zur Schwa-Epenthese imStandarddeutschen’. Linguistische Berichte 112,449−469.

Giegerich, H. (1992) English phonology; An introduc-tion. Cambridge, UK: Cambridge UniversityPress.

Gimson, A.C. (1980) An introduction to the pronunci-ation of English. London, Arnold Publishing.

Gussenhoven, C. (1978) ‘Het Nederlandse diminu-tief-suffix: schwa-insertie anders bekeken’. DeNieuwe Taalgids 71, 206−211.

Gussmann, E. (1991) ‘Schwa and syllabic sonorantsin a nonlinear phonology of English’. AnglicaWratilaviensa 17, 27−39.

Haas, W. de (1986) ‘Partial syllabification and schwaepenthesis in Dutch’. Gramma 10, 143−161.

Hall, T. (1989) ‘German syllabification, the velarnasal and the representation of schwa’. Linguis-tics 27, 807−842.

Hall, T. (1992) Syllable structure and syllable-relatedprocesses in German. Tübingen: Niemeyer.

Halle, M. and K.P. Mohanan (1985) ‘Segmentalphonology of Modern English’. Linguistic In-quiry 16, 57−116.

Halle, M. and J.-R. Vergnaud (1987) An essay onstress. Cambrdige, Mass: The MIT Press.

Hamans, C. and R. Noske (1988) ‘The analysis ofGerman schwa’. Wiener Linguistische Gazette,Supplement/Beiheft 6, 17−19.

Harris, J. (1990) ‘Reduction harmony’. Paper pre-sented at the 13th GLOW Colloquium, London:SOAS.

Harris, J. (1994) English sound structure. Oxford:Blackwell.

Hart, J. ’t (1969) ‘Fonetische steunpunten’. DeNieuwe Taalgids 62, 168−174.

Hayes, B. (1995) Metrical stress theory: Principlesand case studies. Chicago: University of ChicagoPress.

Hawkins, W. (1985) ‘Patterns of vowel loss in Ma-cushi Carib’. International Journal of AmericanLinguistics 16, 87−90.

Hoeksema, J. (1985) ‘Wortels, stammen en deschwa’. TABU 15, 150−153.

Hooper, J. (1976) An introduction to natural genera-tive phonology. New York: Academic Press.

Hulst, H. van der (1984) Syllable structure andstress in Dutch. Dordrecht: Foris.

Hulst, H. van der, and J. van Lit (1987) ‘Lettergreep-structuur’. GLOT 10, 165−195.

Issatschenko, A. (1974) ‘Das “schwa mobile” und“schwa constans” im Deutschen’. In Sprachsys-tem und Sprachgebrauch; Festschrift für HugoMoser, U. Engel and P. Grebe, eds.. Düsseldorf:Schwann. pp. 141−171.

Jacobs, H. (1989) Non-Linear studies in the histori-cal phonology of French. Dissertation, Nijmegen:Catholic University of Nijmegen.

Jong, D. de and M. Hietbrink (1994) ‘The morphonol-ogy of the French prefix RE’. In Linguistics inthe Netherlands 1994, R. Bok-Bennema and C.Cremers, eds. John Benjamins, Amsterdam. pp.43−52.

Kager, R. (1989) A metrical theory of stress anddestressing in English and Dutch. Dordrecht:Foris.

Kager, R. (1990) ‘Dutch schwa in moraic phonology’.In Papers from the 26th Regional Meeting of theChicago Linguistic Society, M. Ziolkowski, M.Noske and K. Deaton, eds. CLS, Chicago. Vol-ume II: The parasesion on the syllable in phonet-ics and phonology. pp. 189−214.

Kager, R., E. Visch and W. Zonneveld (1987) ‘Neder-landse woordklemtoon (Hoofdklemtoon, bij-klemtoon, reductie, voeten)’. GLOT 10,197−226.

Kahn, D. (1976) Syllable-based generalizations inEnglish phonology. Dissertation, Cambridge,Mass: MIT. Also appeared with Garland Press.

Katz, D. (1987) A grammar of the Yiddish language.London: Duckworth.

Kaye, J. (1990) ‘Government in phonology: The caseof Moroccan Arabic’. The Linguistic Review 6,131−160.

Kaye, J., J. Lowenstamm and J.-R. Vergnaud (1985)‘The internal structure of phonological elements:A theory of charm and government’. PhonologyYearbook 2, 305−328.

Kaye, J. and Y. Morin (1977) ‘Il n’ y a pas de règlesde troncation voyons!’ Manuscript Montreal:UQAM.

Keating, P. (1988) ‘Underspecification in phonetics’.Phonology 5, 275−292.

Kenstowicz, M. (1994) ‘Syllabification in Chuckchee:A constraints-based analysis’. Manuscript,Cambridge, Mass: MIT. Available at ftp://ruccs.rutgers.edu/pub/OT/TEXTS/archive/30−10941/30−10941.ps

Kleinhenz, U. (1991) ‘Die Alternation von Schwa mitsilbischen Sonoranten im Deutschen’. Magister-arbeit, Köln: Institut für Deutsche Sprache undLiteratur.

Kohler, K.J. (1990) ‘Segmental reduction in connect-ed speech in German: Phonological facts andphonetic explanations’. In Speech productionand speech modelling, M. Hardcastle, ed. Kluw-er Academic Publishers, Dordrecht.

Koopmans-Van Beinum, F. (1980) Vowel contrastreduction: An acoustic and perceptual study ofDutch vowels in various speech conditions.Amsterdam: Academische Pers.

Koopmans-Van Beinum, F. (1982) ‘Akoestische enperceptieve aspecten van klinkercontrastreduc-tie en de rol van de fonologie’. Spektator 11,284−294.

Koopmans-Van Beinum, F. (1993/1994) ‘What’s in aschwa?’ IFA proceedings 1992, 16, 53−62.

Kristoffersen, G. (1992) The syllable structure ofNorwegian. Dissertation, Tromsø University.

Ladefoged, P, and I. Maddieson (1996) Vowels of theworld’s languages. Oxford: Blackwell.


Lahiri, A. and J. Koreman (1988) ‘On foot typology’.NELS 18, 286−299.

Lapointe, S. and M. Feinstein (1982) ‘The role ofvowel deletion and epenthesis in the assignmentof syllable structure’. In The structure of phono-logical representations Vol II, H. van der Hulstand N. Smith, eds. Dordrecht: Foris. pp. 69−120.

Lass, R. (1976) English phonology and phonologicaltheory; Synchronic and diachronic studies.Cambridge, UK: Cambridge University Press.Cambridge Studies in Linguistics 17.

Lass, R. (1984) Phonology; An Introduction to BasicConcepts. Cambridge University Press, Cam-bridge.

Lindblom, B. (1963) ‘Spectographic study of vowelreduction’. Journal of the acoustic society 35,1773−1781.

Lodge, K. (1987) ‘French phonology again: Deletionphenomena, syllable structure and other mat-ters’. In Explorations in Dependency Phonology,J. Anderson and J. Durand, eds. Dordrecht:Foris. pp. 133−168.

Lyche, C. (1979) ‘French “schwa deletion” in NaturalGenerative Phonology’. Nordic Journal ofLinguistics 2, 91−111.

Martin, A. (1968) ‘Klinkerreductie: Een casus’.Manuscript, Universiteit Utrecht.

Martinet, A. (1972) ‘La nature phonologique d’ecaduc’. In Papers in linguistics and phonetics tothe memory of Pierre Delattre, A. Valdman, ed.The Hague: Mouton. pp. 239−399.

Mascaró, J. (1987) ‘Vowel reduction as deletion’,Handout Going Romance Conference.

Meinhold, G. (1989) ‘Das problematische [!]’. In VonLauten und Leuten; Festschrift für Peter Mar-tens zum 70. Geburtstag, G. Slembek, ed. Frank-furt am Main: Scriptor Verlag. Sprache undSprechen volume 21.

Melvold, J. (1990) Structure and stress in the phonol-ogy of Russian. Dissertation, Cambridge, Mass:MIT.

Morin, Y.-C. (1974) ‘Règles phonologiques à domaineindéterminé: Chute du cheva en français’.Cahiers de linguistique 4, 69−88. Presses del’Université du Québec, Montréal.

Morin, Y.-C. (1978) ‘The status of mute e’. Studies inFrench linguistics 1, 312−345.

Morin, Y.-C. (1982) ‘Cross-syllabic constraints andthe French “e-muet”’. Journal of linguisticresearch 2, 41−56.

Morin, Y.-C. (1983) ‘De l’ouverture des [e] du moyenfrançais’. Revue québécoise de linguistique 12,37−61.

Morin, Y.-C. (1985) ‘French data and phonologicaltheory’. Linguistics 25, 815−843.

Morin, Y.-C. (1988) ‘De l’ajustement du cheva ensyllabe fermé dans la phonologie du français’. InVerluyten (1988).

Moulton, W. (1962) ‘The vowels of Dutch; Phoneticand distributional classes’. Lingua 11, 294−312.

Neidle, C. (1979) ‘Syllable structure, markednessand the distribution of the “mute e” in French”.Manuscript.

Nijen Twilhaar, J. (1990) Generatieve fonologie en destudie van Oostnederlandse dialecten. Amster-dam: P.J. Meertens-Instituut.

Nijen Twilhaar, J. (1994) ‘Genus en fonologischgedrag van nomina op schwa’. In Dialectfonolo-gie, G. Booij and J. van Marle, eds. Amsterdam:P.J. Meertens-Instituut.

Nikiema, E. (1989a) ‘Vocalic epenthesis reanalyzed:The case of Tangale’. In Current approaches toAfrican linguistics, J. Hutchison and V. Manfre-di, eds. Dordrecht: Foris. pp. 47−51

Nikiema, E. (1989b) ‘Gouvernement propre et licen-ciement en phonologie: Le cas du Tangale’. InLangues orientales anciennes philologie etlinguistique 2, D. Peeters, ed. Paris: Louvain.pp. 225−251.

Nooteboom, S. (1972) Production and perception ofvowel duration; A study of durational propertiesof vowels in Dutch. Dissertation.

Nord, L. (1986) ‘Acoustic studies of vowel reductionin Swedish’. Speech transmission laboratoryquarterly progress and status report 4, 19−36.

Noske, R. (1988) ‘La syllabification et les règles dechangement de syllabe en français’. In Ver-luyten (1988).

Noske, R. (1992) A theory of syllabification and

segmental alternation; With studies on thephonology of French, German, Tonkawa andYawelmani. Dissertation, Tilburg University.Also appeared at Tübingen: Niemeyer.

Oostendorp, M. van (1995) Vowel quality and phono-logical projection. Dissertation, Tilburg Univer-sity.

Oostendorp, M. van (1997) ‘Style levels in conflictresolution’. In Variation, change and phonologi-cal theory, F. Hinskens, R. van Hout and L.Wetzels, eds. Amsterdam: John Benjamins. p.207−228.

Piggot, G. (1995) ‘Epenthesis and syllable weight’.Natural Language and Linguistic Theory 13,283−326.

Piggot, G. and R. Singh (1984) ‘The empty node: Ananalysis of epenthesis’. McGill working papersin linguistics 2, 30−55.

Piggot, G. and R. Singh (1985) ‘The phonology ofepenthetic segments’. Canadian Journal ofLinguistics. 1, 64−109.

Prince, A. and P. Smolensky (1993) Optimalitytheory; Constraint interaction in generativegrammar. Manuscript Rutgers and Colorado.

Prokosch, E. (1939) A comparative germanic gram-mar. Philadelphia: Linguistic Society of Ameri-ca/University of Pennsylvania.

Pulgram, E. (1961) ‘French /6/: Statics and dynamicsof linguistic subcodes’. Lingua 10, 305−325.

Pulleyblank, D. (1988) ‘Vocalic underspecification inYoruba’. Linguistic Inquiry 19, 233−270.

Ramers, K.H. (1988) Vokalquantität und -qualitätim Deutschen. Tübingen: Niemeyer.

Rappaport, M. (1984) ‘Stress and ultra-short vowelsin Tiberian Hebrew’. Proceedings of WCFL.

Rennison, J. (1980) ‘What is shwa in AustrianGerman? The case for epenthesis and its conse-quences’. Wiener Linguistische Gazette 24,78−91.

Rennison, J. (1993) ‘Empty nuclei in Koromfe: A firstlook’. Linguistique Africaine 11, 35−65.

Rennison, J. (1994) ‘Syllables, variable vowels andempty categories’. In Phonologica 1992, W.U.Dressler, M. Prinzhorn, and J. Rennison, eds.Torino: Rosenberg and Sellier.

Rennison, J. (1997) ‘Syllables in Koromfe’. GurPapers/Cahiers Voltaiques 2. Universität Bay-reuth.

Rialland, A. (1985) ‘Schwa et syllabes en français’. InStudies in Compensatory Lengthening, L. Wet-zels and E. Sezer, eds. Dordrecht: Foris.

Roca, I. (1994) Generative phonology. London:Routledge.

Ross, J.R. (1972) ‘A reanalysis of English wordstress’. In Contributions to generative phonology,M. Brame, ed. Austin: University of Texas Press.

Rowicka, G. (1985) ‘English word stress and emptyvowel slots’. Paper read at the Tarnawy Confer-ence on Contrastive Linguistics; cited in Szpyra(1996).

Rubach, J. (1986) ‘Abstract vowels in three-dimen-sional phonology: The yers’. The LinguisticReview 11, 124−154.

Rubach, J. and G. Booij (1990) ‘Syllable structureassignment in Polish’. Phonology 7, 121−158.

Schane, S. (1965) The phonological and morphologi-cal structure of French. Dissertation, Cam-bridge, Mass: MIT.

Schane, S. (1968) ‘On the abstract character ofFrench “E-muet”’. Glossa 2, 150−163.

Schane, S. (1974) ‘There is no French TruncationRule’. In Linguistic Studies in Romance Lan-guages, R. Campbell et al., eds. Washington,D.C.: Georgetown University Press. pp. 89−100.

Selkirk, E. (1978) ‘Comments on Morin’s paper: TheFrench foot: On the status of “mute e”’. Studiesin French Linguistics 1, 24−28.

Sievers, E. (1878) ‘Zur Accent- und Lautlehre derGermanischen Sprache’ Beiträge zur Geschichteder Deutschen Sprache und Literatur 5, 63−163.

Sojijärvi, A. (1965) ‘Der Mokschamordvinische6-Vokal im Lichte der Sonagramme’. Publica-tiones Instituti Phonetici Universitatis Hels-ingiensis 20, Helsinki.

Spencer, A. (1985) ‘A non-linear analysis of vowel-zero alternations in Polish’. Journal of Linguis-tics 22, 249−280.

Stemberger, J. (1993) ‘Glottal transparency’. Phonol-ogy 10, 107−138.

Steriade, D. (1995) ‘Underspecification and marked-ness’. In Handbook of Phonological Theory, J.Goldsmith, ed. London: Blackwell. pp. 114−174.

Szpyra, J. (1995) Three tiers in Polish and Englishphonology. Post-doctoral dissertation, Lublin:Marie Curie-Sk odowska University Press.

Szpyra, J. (1996) ‘“Incomplete” vowels in English’. InA Festschrift for Edmund Gussmann from hisfriends and colleagues, H. Kardela and B.Szymanek, eds. Lublin: The University Press ofthe Catholic University of Lublin. pp. 269−292.

Trager, G.L. and H.L. Smith (1951) An outline ofEnglish structure. Norman, Oklahoma: Batten-brug Press. Studies in Linguistics, OccasionalPapers, 3.

Tranel, B. (1974) ‘Le cas de l’effacement facultatif duschwa en français: Quelques implicationsthéoriques’. Recherches linguistiques à Montréal1.

Tranel, B. (1981) Concreteness in generative phonolo-gy. Berkeley: University of Colorado Press.

Tranel, B. (1984) ‘Floating schwas and closed sylla-ble adjustment in French’. In Phonologica 1985,W. Dressler et al., eds. Cambridge, UK: Cam-bridge University Press.

Tranel, B. (1985) ‘On Closed Syllable Adjustment inFrench’. In Selected papers from the 13th lin-guistic symposium on Romance languages, L.King and C. Maley, eds. John Benjamins, Am-sterdam/Philadelphia. pp. 377−405.

Tranel, B. (1987) ‘French schwa and non-linearphonology’. Linguistics 25, 845−866.

Tranel, B. (1988) ‘A propos de l’ajustement de e enfrançais’. In Verluyten (1988).

Tranel, B. (1994) ‘French liaison and elision revisit-ed: A unified account within Optimality Theory’.Manuscript Irvine: University of California.

Trommelen, M. (1983) The syllable in Dutch; Withspecial reference to diminutive formation. Dor-drecht: Foris.

Velde, H. van der (1996) Variatie en verandering inhet gesproken Standaard-Nederlands. Disserta-tion, Nijmegen: Katholieke Universiteit Nijmeg-en.

Verluyten, S. (1985) ‘Prosodic structure and thedevelopment of French schwa’. In Papers fromthe 6th International Conference on HistoricalLinguistics, J. Fisiak, ed. Amsterdam: JohnBenjamins, and Poznán: Adam MickiewiczUniversity Press. pp. 549−559.

Verluyten, S. (1988) La phonologie du schwafrançais, Amsterdam: John Benjamins.

Verluyten, S. and M. Bertels (1987) ‘Observationaladequacy and formal complexity: A comparativeevaluation of five phonological theories onFrench schwa deletion’. Manuscript, Unicversityof Antwerp.

Visser, W. (1994) ‘Schwa-appendixen in het Fries’. InDialectfonologie, G. Booij and J. van Marle, eds.Amsterdam: P.J. Meertens-Instituut.

Visser, W. (1997) The syllable in Frisian. Disserta-tion, Vrije Universiteit Amsterdam. Appeared atThe Hague: HAG.

Vijver, R. van de (1998) The iambic issue; Iambs as aresult of constraint interaction. Dissertation,Leiden University. Appeared at The Hague:HAG.

Wells, J. (1982) Accents of English. Cambridge, UK:Cambridge University Press.

Wiese, R. (1986) ‘Schwa and the structure of wordsin German’. Linguistics 24, 696−724.

Wiese, R. (1988) Silbische und lexikalische Phonolo-gie; Studien zum Chinesischen und Deutschen.Tübingen: Niemeyer.

Yip, M. (1987) ‘English vowel epenthesis’. NaturalLanguage and Linguistic Theory 5, 463−484.

Zonneveld, R. van (1985) ‘Word rhythm and theJanus syllable’. In The structure of phonologicalrepresentations II, H. van der Hulst and N.Smith, eds. Dordrecht: Foris.

Zonneveld, W. (1993) ‘Schwa, superheavies, stressand syllables in Dutch’. The Linguistic Review10, 59−110.

Column Glot International, Volume 3, Issue 5, May 1998 9

According to Woody Allen “sentence structure isinnate but whining is acquired” (Sunshine1993:190). The first proposition will be reasonablyfamiliar to readers of Glot International, but Iimagine that the second is less expected and, inpart for that reason, mildly amusing. So whatmakes Woody funny? What makes anyone funnycome to that? The literature on humour is notori-ously unfunny, but a little theoretical insight isbreaking into even this murky area, so it’s worthtrying to see how some of it works.

Everyone agrees that incongruity is a crucialelement in most forms of humour, where incon-gruity ultimately reduces to contradiction. Thisraises two problems: first, there is nothing funnyin asserting the contradiction that ‘Sentencestructure is innate and sentence structure is notinnate’; second, how ‘acquired whining’ contrivesto be contradictory is not immediately obvious.These problems suggest that the contradictionhas to creep up on you unawares: it has to beimplicit rather than explicit; and it has to besomething you are led to construct indirectly foryourself: it cannot work without your involuntarycollaboration.

In recent work exploiting Relevance Theory(Sperber & Wilson 1986/95), Carmen Curcó(Curcó Cobos 1997) has spelt out some of theissues involved in making these intuitive assump-tions precise. She starts with the important pointthat what is essential is not the incongruity itselfbut the effects achieved by processing incongruity.Indeed, she makes a strong case for the claim thatthe analysis of humour requires no special theo-retical apparatus which is not independentlyneeded for the analysis of normal communication.Processing utterances is an inferential activity,carried out in conformity with the communicativeprinciple of relevance, which gives you a guaran-tee that what is being said is going to have cogni-tive effects without putting you to gratuitouseffort. In order to achieve such effects you areoften constrained to provide an appropriate con-text on your own responsibility. If you ask wheth-er Husbands and wives is worth seeing, and aretold “all Woody Allen’s films are great”, you areforced to supply the implicated premise that thefilm is by Allen, because that is the only way theanswer can be construed as relevant.

Consider in this context the following excerptfrom a policy document put out by the Irish Minis-try of Transport shortly after Ireland had joinedthe European Union.

For a two-year experimental period, beginning at midnight on31st December, cars will drive on the right (as in the rest ofEurope), instead of the left (as in the UK). If the experiment isa success, buses and lorries will thereafter also drive on theright.

As you read, you naturally interpret ‘cars’ as ageneric term for vehicle, because this is the onlyrational meaning in a context where rules of theroad are concerned and hence will be mutuallymanifest to speaker and hearer. That is youconstruct a scenario including the tacit assump-tion “All vehicles will drive on the right”. Whenyou get to the punch-line, you are forced to supplyan implicated premise to the effect that ‘cars’excludes other types of vehicle, hence that “Not allvehicles will drive on the right”. The writer hasdeliberately not made this explicit but, by tradingon the presumption that processing the utterancewill be relevant, he gives you no option but toaccess this new, contradictory, premise: he haslured you into making incongruous, contradictory,assumptions. The point of the analysis is not toshow why the particular example is funny, but (if

it is funny) how the effect is, in part, achieved.Incongruity can be disguised, and the nature

of one kind of disguise can be revealed usinganother idea from Relevance. Part of the interpre-tive process involves treating something whichshould be in the ‘foreground’ as though it were inthe ‘background’ (Sperber & Wilson 1986:202ff.;for discussion, see Smith 1989, ch. 14). A secondexample from Woody Allen (Sunshine 1993:255;cited in Curcó Cobos 1997:214) can be used tomake the point):

There is no question that there is an unseen world. Theproblem is how far is it from midtown and how late is it open?

As you process any utterance, you make anticipa-tory hypotheses about what is coming next:whether a hypothesis is in the foregrounddepends on the syntactic structure of the sentencebeing uttered, the order in which the constituentsare introduced and, crucially, whether it is rele-vant in its own right. What is in the backgroundis then information that can be taken for granted.With the current example, it is obvious that thetopic is the supernatural which presumably refersto a domain which is not located in space and timethe way the physical world is. Accordingly, whenyou have got as far as “The problem is ...” youexpect to hear next a putative solution to a ‘prob-lem’ about the supernatural. What you get is ananswer which presupposes something of incon-testable relevance: it would overturn all ourconceptions of the occult if it was true, but some-thing which is presented as though it were justpart of common background knowledge: that theunseen world is physically located near midtown.

It is presented this way, but the propositionbeing put forward is not actually being endorsedby the speaker. Another crucial strand in themanipulation of humorous effects depends on thenotions of dissociation and metarepresentation.Intuitively Woody Allen is distancing himself fromthe claim that the unseen world is really accessi-ble from midtown, and is indulging in a degree of(self-)mockery in the way he describes it. Thisdissociation is in turn dependent on the possibili-ty of using a representation not just to describe astate of affairs, but to interpret (or metarepre-sent) another representation of some state ofaffairs. The classic case of such ‘interpretive use’is found in irony. If I say “I do admire the presi-dent’s honesty”, it is likely to be clear that I amintending to ridicule anyone defending the propo-sition that the president is honest, (even if no-oneactually has) and that I am simultaneously disso-ciating myself from such a proposition, perhapswith contempt.

Dan Sperber (1994) has emphasized thesurprising complexity of metarepresentation ineveryday conversation, arguing that speakers andhearers regularly entertain eighth and ninthorder metarepresentations: If John says to Marythat there is an unseen world, then:

John intends — Mary to know — that John intends — Mary tobelieve — that John believes — that there is an unseen world.

Assuming that Mary is not so credulous, she willconstruct a representation of the kind:

John says — there is an unseen world, but I believe — there isnot an unseen world — So John is mistaken.

If Mary believes that John doesn’t really believethere is an unseen world, she will construct themore complex representation:

John says — there is an unseen world, but I believe — thatJohn believes — that there is not an unseen world — So, Johnis lying.

It is a further step or two on the same progressionbefore we get to the situation where the conclu-sion is that John is joking or being ironic:

Mary believes — that John believes — that Mary believes —that John believes — that there is not an unseen world — So,John may be joking (or demented).

Curcó Cobos argues that to get a humorous inter-pretation, a contradictory proposition must beembedded in a metarepresentation of at leastfourth order. If she is right, it is not surprisingthat some poor people consistently miss the pointof jokes and witticisms, and we are probably allpainfully familiar with situations where we can-not really tell whether someone is joking or not.Some people, specifically those with autism, are inan even worse situation: they are inherentlyincapable of ever understanding jokes. Autists arecustomarily supposed to lack a “theory of mind”,entailing that they cannot form metarepresenta-tions, they cannot impute beliefs to others and,crucially, jokes are in principle beyond them. Thisleads to the incomprehension in the face of jokesand pretence that Ianthi Tsimpli and I described(Smith & Tsimpli 1996) in “Putting a banana inyour ear” in an earlier issue of this journal.

To return at the end to Woody Allen, it isclear that the quotation I started with was in-spired by his admiration for Chomsky (see also,“The Whore of Mensa” in Allen 1975). It is intrigu-ing that the respect is mutual, and Chomsky citesWoody Allen in his most serious work with thesame incongruous and slightly humorous effect.For example in Rules and representations (Chom-sky 1980:117) he refers to the fact that “‘Theuniverse is constantly expanding’ … the discoverythat turned Woody Allen into a neurotic”. I leaveit to the reader to work out the correct depth ofmetarepresentational embedding that an appro-priate understanding of this remark requires.

ReferencesAllen, W. (1975) Without Feathers. New York,

Random House.Chomsky, N. (1980) Rules and Representations.

Oxford, Blackwell.Curcó Cobos, C. (1997) The Pragmatics of

Humorous Interpretations: A Relevance-Theoretic Approach. University CollegeLondon PhD thesis.

Smith, N.V. (1989) The Twitter Machine. Oxford,Blackwell.

Smith, N.V. & I.-M. Tsimpli (1996) ‘Putting abanana in your ear’. Glot International2,1/2:28.

Sperber, D. (1994) ‘Understanding verbal under-standing’. In J. Khalfa (ed) What isIntelligence? Cambridge, CUP.

Sperber, D. & D. Wilson (19952) Relevance:Communication and Cognition. Oxford,Blackwell.

Sunshine, L. (ed.) (1993) The Illustrated WoodyAllen Reader. London, Jonathan Cape.

ACQUIRED WHINING

by Neil Smith

Dissertations Glot International, Volume 3, Issue 5, May 1998 10

Summary by the author

The thesis from which the below discussion hasbeen extracted attempts to analyze participialsyntax. The major aim is to reveal the functionalstructure that may be projected by a participialverb and to analyze the mechanisms of AccusativeCase checking that are operative in the participialclause. A further aim is to work towards a maxi-mally unified theory of Accusative checking in theparticipial clause and Nominative checking in thefinite clause. For reasons of space, this presenta-tion is limited to some ideas concerning the nodesof AgrO and Tense, leaving aside the discussion ofAspect nodes which however occupies an impor-tant part of the original thesis. The analysis isinspired by the minimalistic framework of Chom-sky (1993) but diverges on some points from thedevelopments put forth by Chomsky (1995), forinstance in maintaining the assumption of Agree-ment nodes. The present analysis is mainly builton a comparison between different historicalstages of Italian while modern languages aretaken into consideration only to a limited extent.The reason for this is that certain interesting dataattested in Old Italian do not appear to be reflect-ed in modern languages, at least to my knowl-edge.

1. Word order and agreement in 14thcentury participial clausesIn Italian texts dating up to the mid 14th

century, word order in the compound tense isoptional, in the sense that the DP object is attest-ed both to the right and to the left of the participi-al V:

(1)... come che tu abbi perduti i tuoi denari ...even if you have lost[masc.pl.] your money[masc.pl.](Decameron; II:5)

(2)... co’ denari avresti la persona perduta.with your money you would have your life[fem.sing.]lost[fem.sing.](Decameron; II:5)

In the relevant period agreement on the participleis highly regular: the participle agrees with theobject regardless of where the object appears inrelation to the noun — to the right as in (1) or tothe left as in (2). In many of the late 13th centurytexts agreement appears to have been almostexceptionless both in (1) and in (2). From the mid14th century onwards, the modern word order (1)is established and the construction illustrated in(2) disappears, i.e. (2) disappears with the mean-ing of a compound tense. The word order of (2)may also correspond to a different construction,instantiating have of possession followed by aresultative small clause; as such, the word orderof (2) is quite grammatical also in Modern Italian,although perhaps not with a predicate like perd-ere ‘lose’. At the same time, the mid 14th century,

3. The fact that (2) and (4) are attested is a signthat Italian in the relevant period maintainedSOV tendencies, although SVO is the predominantpattern. A natural way to account for this is to saythat the Accusative feature was strong in the senseof Chomsky (1993) and required overt checking,hence, overt movement of DP(O) to SpecAgrO.

4. In a framework such as that of Chomsky(1993), it is not possible to assume, however, thatmovement of DP(O) was optional. In order toarrive at a stronger hypothesis, we may thenassume that overt object shift as in (2) and (4) wasobligatory in Old Italian. As for (1) and (3), Isuggest that they should be accounted for on thebasis on assumptions that have been common inthe discussion on so called free subject inversionin the finite clause (cf. Rizzi 1982, and studiesthat follow his suggestions). As an example of thistype of analysis, let us take a case of subjectinversion with an ergative predicate as in (7) and(8):

(7)Gianni vieneJohn comes

(8)Viene Giannicomes John ‘John comes’

According to the classical analysis, the lexicalsubject in (8) is copied by an expletive element,pro, that occupies the specifier of Infl; cf. Rizzi(1982), (1986) and Chomsky (1986). Consider thestructures in (9):

(9)

Let us suppose that this solution should be gener-alized to the case of Accusative checking in theparticipial clause. The structures correspondingto (1) and (2) in Old Italian are given in (10):

(10)

The advantage of this analysis is that it allows usto generalize a well-established approach toNominative Case checking (and its relation toword order and agreement) in finite clauses to theequivalent phenomenon of Accusative Case check-ing (and its relation to word order and agreement)in participial clauses. The account diverges fromChomsky (1995) insofar as it maintains the as-sumption that Agreement nodes are projected inthe syntax. Moreover, Chomsky (1995, 341) claimsthat SpecAgrS can host both an expletive and aDP, whereas the specifier of AgrO only hosts a DP.My account treats AgrO and AgrS as perfectlyalike in this respect; in other words, expletive-associate dependencies can be both Nominativeand Accusative.

This assumption also accounts for other wordorder variations in Old Italian absolute participleclauses. The modern PAC can host only one lexi-cal argument; the argument carries Nominative ifthe verb is ergative and Accusative if the verb istransitive. In the ancient texts two lexical argu-ments in PAC are often attested:

(11) Modern Italian:*Conosciuta Elena Gianni, ...known Helen John, ...

cases where the participle shows agreement withthe object, as in (1), become rare and in some textsfrom the late 14th century are only occasionallyattested.

The same pattern is observed in participialabsolute clauses (henceforth: PAC); until the mid14th century, the internal argument of a transi-tive PAC is found both to the left and to the rightof the verb:

(3)... serrata la cella con la chiave, dirittamente se n’andò ...locked[fem.sing.] the cell[fem.sing.] with the key, immediatelyhe went away(Decameron; I:4)

(4)... il re di Francia, molte triegue fatte con gli alamanni, morì ...the king of France, many truces[fem.pl.] made[fem.pl.] with theGermans, died(Decameron; II: 8)

By the mid 14th century, the word order in (3)becomes established, Verb-Object, while thatrepresented by (4), Object-Verb, begins to disap-pear from the texts. (3) corresponds to the wordorder of a transitive PAC in Modern Italian. Atthe same time, agreement on the participle is lost;the new word order and agreement patterns thatare established from the late 14th century on-wards are thus illustrated in (5) and (6):

(5) compound tense:... e aver trovato la casa mia...and having found[–Agr] my house[fem.sing.](I Lucidi; I:2)

(6) absolute participle:... serrato le finestre se ne andarono...closed[–Agr] the windows[fem.pl.] they went away(Cene: Introd.)

The correct generalization appears to be thatrightward agreement with a DP object on theparticiple was obligatory as long as the DP objectcould appear also in a leftward position. Participi-al agreement is lost when the position of theobject is established to the right of the participle.These observations have some interesting conse-quences for the theory of syntax. Consider some ofthem:

1. Under the assumption that the rightwardposition of the object is a complement and theleftward one is a specifier, the data lend supportto the Spec-head-agreement approach to participi-al agreement, outlined by Kayne (1989).

2. The fact that a shift in word order is attested inthe compound tense and in the absolute participi-al construction at the same time suggests that thesurface change was due to some change withinthe participial clause; indeed, it is unlikely thatthe shift from (1)/(2) to (5) had anything to dowith, say, a change in Case assigning properties ofthe governing V have, given that a similar shifttakes place from (3)/(4) to (6).

THE SYNTAX OF PASTPARTICIPLES. A GENERATIVESTUDY OF NONFINITECONSTRUCTIONS IN ANCIENTAND MODERN ITALIAN

by Verner Egerlandreviewed by Yves D’Hulst

IP

Spec I’DP(S)j

Infl VPVi

ti tj

IP

Spec I’proj

Infl VPVi

ti DP(S)j

AgrOP

Spec AgrO’DP(O)j

AgrO VPVi

ti tj

AgrOP

Spec AgrO’proj

AgrO VPVi

ti DP(O)j


(12) 14th century:ed Enea presa la lancia, con grande vigore uno si levò control’altro ...and Enea taken the lance, with great force one stood upagainst the other ...(I fatti di Enea; LVII)

Word order in the Old Italian PAC with twoarguments could be VOS, SVO, and VSO; I havenot attested *SOV, *OSV, or *OVS. The generali-zation that emerges is thus that when there is asubject carrying Nominative, the verb must be tothe left of the object. These patterns are explainedon the above hypothesis in terms of expletivechains. If the absolute participial clause can Casemark two lexical arguments, the conclusion in theminimalistic framework of Chomsky (1993) is thatboth AgrS and AgrO are projected in the absoluteclause in the Old Italian PAC. If, then, pro may belicensed in both SpecAgrS and SpecAgrO, thusallowing the DP associates to remain in situ, thefollowing possibilities are predicted:

1. VOS: DP(O) moves overtly to SpecAgrO; DP(S)stays in situ, being copied by an expletive pro inSpecAgrS; the verb raises overtly to AgrS.

2. SVO: DP(S) moves overtly to SpecAgrS; Vraises to AgrS; DP(O) moves overtly to SpecAgrOor stays in situ, being copied by an expletive proin SpecAgrO.

3. VSO: The verb raises overtly to SpecAgrS;DP(S) and DP(O) stay in situ, both being copiedby expletive pros in SpecAgrS and SpecAgrOrespectively.

If Nominative is to be checked in the clause, Iassume that the verb must reach AgrS. It thenfollows that, if there is a Nominative argument toCase check, the verb must raise higher than theobject that does not move beyond SpecAgrO.

2. The relevance of Tense for structuralCase checkingThe fact itself that only one lexical argument

is allowed in the Modern Italian PAC, whereasthe ancient construction could host two, is ofcourse interesting. This difference is presumablyrelated to another property of absolute participles:in Modern Italian, PAC cannot host negation. Inthe ancient texts, cases of negated absolute parti-ciples are attested:

(13) Modern Italian (from Belletti 1990, 95):*Non arrivata Maria, Gianni tirò un sospiro di sollievo.not arrived Mary, John was relieved

(14) 14th century:... , il papa non dimenticato lo sdegno preso contro alla partebianca di Firenze,... , the pope not forgotten the resentment against the white partof Florence,non volle che soggiornasse e vernasse invano.did not want him to stay and spend the winter in vain.(Cronica; VIII: 49)

According to Zanuttini (1991) negation dependson the presence of Tense. Belletti (1990) hasargued that the Modern Italian PAC lacks a Tensenode. If by contrast in the Old Italian PAC such anode is present, the difference between (13) and(14) can be accounted for straightforwardly. Afurther assumption made by Belletti is that inthe Modern Italian PAC the absence of Tenseforces the verb to raise to Comp. Thus, for Bellet-ti, V-raising to Comp is necessary for Case rea-sons, under the assumption that a tense feature isgenerated in Comp. Now, it is clear that in OldItalian the participial V did not raise as far as toComp. This can be seen from the fact that, where-as a lexical argument cannot cooccur with a filledComp in Modern Italian, this was indeed possiblein Old Italian:

(15) Modern Italian (from Belletti 1990, 99):*Benché partita Maria, ci divertimmo ugualmente.*although left M., we amused ourselves anyway

(16) 14th century:... dicono li soprascritti savi che, bene che quelli cotali uominidiventati animali, ...the above mentioned learned say that, although those menbecome animals, ...la mente dentro rimaneva loro umana, ...their mind inside remained human(I fatti di Enea: XXVII)

According to a line of thought in generative gram-mar, Tense is crucial not only for Nominative Casebut also for Accusative Case (see, among others,Belletti 1990, Watanabe 1993, and Hoekstra &Roberts 1993). In the framework of Watanabe, thechecking of structural Case against a functionalhead must be followed by overt raising of thesame head to a higher functional head. For theparticipial clause, I assume the functional struc-ture argued for by Kayne (1993):

(17)... [CP C [AgrSP AgrS [TP T [AgrOP AgrO [VP V]]]]]

The checking of Accusative in AgrOP must befollowed by raising of AgrO to T, and the checkingof Nominative in AgrSP must be followed byraising of AgrS to C. Without entering into thedetails of the so called layered Case theory (seeWatanabe 1993 and Chomsky 1995, 385 f.n. 49), itis clear that the comparison between absoluteparticiples in Old and Modern Italian lendssupport to the general thesis. If in Modern Italianthere is a Comp node but not a Tense node, itfollows that only one structural Case can bechecked; either Accusative, if the verb is transi-tive, or Nominative, if the verb is ergative. If inOld Italian both T and C are projected, it followsthat structural Case in such a grammar can bechecked on two arguments.

Finally, in Modern Italian the pattern illus-trated in (6) above is replaced by (18):

(18)Chiuse le finestre, se ne andarono.closed[fem.pl.] the windows[fem.pl.], they went away

Thus, agreement on the participle again becomesobligatory although the lexical argument appearsto the right of the participle. This is explained onthe assumption that verb and object agree inAgrOP, after which the verb obligatorily raisesbeyond AgrOP to Comp. Belletti (1990) advancesthis hypothesis for the ergative PAC in ModernItalian. If, however, the solution is generalized tothe transitive PAC (which has been suggested byKayne 1989 and Cinque 1990), both transitiveand ergative absolute participle constructions arederived from the same hypothesis.

ReferencesBelletti, A. (1990) Generalized Verb Movement.

Aspects of Verb Syntax. Rosenberg & Sellier,Torino.

Chomsky, N. (1986) Knowledge of Language.Praeger, New York.

Chomsky, N. (1993) ‘A Minimalist Program forLinguistic Theory.’ In K. Hale & S. J. Keyser,eds., The View from Building 20: Essays inLinguistics in Honour of Sylvain Bromberger.MIT Press, Cambridge, Massachusetts.Reprinted in Chomsky 1995.

Chomsky, N. (1995) The Minimalist Program.MIT Press, Cambridge, Massachusetts.

Hoekstra, T. & I. Roberts (1993) ‘Middle Construc-tions in Dutch and English.’ In E. Reuland &W. Abraham, eds., Knowledge and Language.II: Lexical and Conceptual Structure. KluwerAcademic Publishers, Dordrecht.

Kayne, R.S. (1989) ‘Facets of Romance Past Parti-ciple Agreement.’ In P. Benincà, ed., 85–103;Dialect Variation and the Theory of Grammar.Foris, Dordrecht.

Kayne, R. S. (1993) ‘Toward a Modular Theory ofAuxiliary Selection.’ Studia Linguistica 47,3−31.

Rizzi, L. (1982) Issues in Italian Syntax. Foris,Dordrecht.

Rizzi, L. (1986) ‘Null Objects in Italian and theTheory of pro.’ Linguistic Inquiry 17, 501−557.

Watanabe, A. (1993) Agr-Based Case Theory andIts Interaction with the A-bar System. MITDoctoral dissertation.

Ancient Texts Cited:14th century:Cronica di Giovanni Villani. Scelta, introduzione

e note di Giovanni Aquilecchia. Einaudi,Torino 1979.

Decamerone, Giovanni Boccaccio. A cura di VittoreBranca. Arnoldo Mondadori, Milano 1985; Ied. ”Oscar classici” 1989.

I fatti di Enea, Guido da Pisa. A cura di FrancescoFòffano, Sansoni, Firenze 1900; nuova pre-sentazione di Franca Ageno, 1968.

16th century: Le Cene, Antonfrancesco Grazzini (Il Lasca). A

cura di Riccardo Bruscagli, Salerno Editrice,Roma 1976. (I Novellieri Italiani, volume 27)

I Lucidi, Agnolo Firenzuola. A cura di AdrianoSeroni. Sansoni Editore, Firenze 1958; I ed.“Le Betulle” 1991.

Review by Yves D’Hulst

1. IntroductionEgerland’s dissertation gives a detailed and

interesting survey of agreement and word orderpatterns in past participle constructions, bothclausal and absolute, in medieval, renaissanceand modern Italian. The comparative scope of thestudy is extended to other Romance languages(especially Spanish) and to Germanic (Scandina-vian) languages. Egerland tackles virtually allaspects that one might think related to his mainissue: the correlation between word order andagreement, the assignment and checking of case,expletives, affectedness, aspectual interpretation,adjectival phrases, possessive interpretation,functional structure, etc. One important topic isleft out of the overall discussion: the possiblecauses of historical change, although some sug-gestive speculations are put forward.

In this review, I will concentrate on threeissues. The first one relates to the data and ispurely philological. The other two focus on thebulk of Egerland’s dissertation: the interaction ofcase, word order and agreement in compoundtenses and the syntax of participial absoluteclauses.

2. The dataBecause Egerland deals with older stages of

Italian, his data are necessarily corpus-based.Throughout his dissertation, Egerland treats hisdata with philological care, meticulously basinghis analyses only on examples that can be inter-preted correctly without having to appeal tointonation or punctuation.

However, one critical philological note is inorder concerning one of the central issues of thisstudy: agreement of past participles. As the au-thor points out, the quantitative data show thatpast participle agreement was the rule until thefirst half of the 14th century, independently ofword order. Subsequently, agreement becamemore marginal whenever the object followed theparticiple. In both periods of time, there appearsto be some variation among authors and excep-tions are occasionally found.

The problem with this generalization, correctas it may be, is that a large amount of ancientItalian literature, which this corpus is built upon,did not reach this century straight-away. Fre-quently, the original autographs got lost andmodern critical text-editions are based on copiesdating from the 15th or 16th centuries. Consideringthat philological concern about text editions is afairly recent attitude and that renaissance copy-ists had little reluctance manipulating the origi-nal texts at will, one might wonder whether thechanges that Egerland has detected in the late14th century texts actually took effect at thattime. This doubt affects at least three out of fivetexts in Egerland’s corpus for the relevant period


(Il Pecorone, Il trecentonovelle and Il novelliere):exactly the three texts where both OV order andagreement in VO construction score extremely low.

Of course, even if the linguistic change wereto have occurred at a later stage, the fact remainsthat it occurred and that one has to account for it.

3. Case and agreement in compound tensesEgerland’s analysis of case checking in the

three stages of Italian involves two parameters:one concerning the strength of the case feature onthe noun and one concerning the locus of casechecking. Generalizing, the first parameter gov-erns movement of the object, the second governsagreement of the past participle.

To be more precise, a language where accusa-tive case is strong, requires the object to moveovertly to a position where the case feature can bechecked. According to Egerland’s study, medievalItalian is such a language, but neither Renais-sance Italian, nor modern Italian are. as for thelocus parameter, languages chose to check case inAgrOP or in a functional projection below AgrOP,namely Asp(ect)P. With the first setting, pastparticiple agreement is obligatory whenever theobject passes through SpecAgrOP. This is alwaysthe case for medieval Italian where the accusativecase feature is strong. With a different setting forthe strength parameter, one variety of renais-sance Italian still triggers agreement if movementof the object is forced on independent grounds,namely in operator contexts (e.g. wh-interroga-tives, relatives). If, on the other hand, the locusfor case checking is lower down in the structure,past participle agreement is not triggered. This isthe case for the other variety of renaissanceItalian and for modern Italian as well. The differ-ent settings for the accusative case parametersare summarized in (1) and their effects in (2).

Egerland’s analysis is certainly very convincingfrom a descriptive point of view and has a numberof interesting consequences from a comparativepoint of view. Yet some theoretical issues still arein search for a satisfactory explanation.

3.1. Choice of parametersFrom a comparative point of view, one expects

the number of possible grammars to be coexten-sive with the number of settings of these twoaccusative case parameters. More specifically, onecould imagine a language where Accusative isstrong and the locus for checking is not AgrOP butAspP, i.e. a language with overt movement of theobject, but no agreement of the past participle. AsEgerland points out, none is to be found in theRomance area, but Germanic languages withscrambling of the object could very well instanti-ate this setting.

3.2. Case assigning propertyOn several occasions, Egerland notes that his

analysis of accusative case closely resembles thetraditional view on nominative case, a welcomeresult. One issue where nominative and accusa-tive are argued to be alike is precisely the locusparameter. Just like nominative can be checkedeither in SpecTP or in SpecAgrSP, languages maychoose to check case in SpecAspP or in SpecAgrOP.However, the parallelism breaks down in onerespect: the choice for the checking of nominativecase is between the specifier position of the projec-tion with the case-assigning feature, T, and thespecifier position of the agreement projection of T.For accusative case, things seem somewhat differ-ent: in the standard account, the category with

the case-assigning feature is V and the choice forchecking is between the specifier positions of twoother categories, Asp and AgrO. The problem thusreduces to the following: is there a principled wayto account for the possible checking sites of case?If both AgrO and Asp can check accusative case,why can’t other (functional) categories as well?

A possible way to answer this question, wouldbe to resort to head movement: case can bechecked in the specifier position of whateverfunctional projection that hosts a case-assigningproperty, either inherently (T) or through move-ment (AgrS, AgrO, Asp). This hypothesis impliesthat the choice for accusative case checking inparticiple constructions is restricted to the specifi-er positions of the functional categories to whichthe participle is attracted, but not further. Underthis view, it remains unclear, however, how onecan prevent accusative to be checked higher up inthe functional structure of non participial clauses.

There is, of course, an equally reasonable andless problematic alternative. In the literature, ithas been suggested that past participles do nothave a case assigning property (see among othersRoberts 1987). One could imagine, then, that Aspis endowed with a case assigning feature supple-menting the defective participial V. Under thisview, the parallelism with nominative case wouldbe fully restored and the possible checking sitesfor case would be restricted in a principled way:case is checked in the specifier position of theprojection with the case-assigning feature (T, Asp)or in the specifier position of its agreement projec-tion (AgrS, AgrO).

3.3. Null expletivesFinally, let us go back to the summary in (1)

and the settings of medieval Italian: accusativeneeds to be checked overtly and the locus forchecking is SpecAgrOP. With these settings, oneexpects, of course, the object always to moveovertly to SpecAgrOP. Assuming Egerland’s viewthat the past participle raises only as far as AgrO,one also expects the object always to precede thepast participle, as in (3a). However, even in medi-eval Italian, the predominant pattern is the onewhere the object follows the past participle, as in(3b) (in Egerland’s corpus there are 99 cases withVO order and 30 with OV).

(3)a. ... co’ denari avresti la persona perduta.

with your money you would have your life lost(Decameron II:5)

b. ... come che tu abbi perduti i tuoi denari ...even if you have lost your money(Decameron II:5)

Egerland’s account of the pattern in (3b) is similarto what has been proposed in the literature forsubject inversion (see Rizzi 1982): case is licensedthrough a chain headed by an expletive. So onceagain, there appears to be a striking similaritybetween the properties of nominative and accusa-tive DP’s. The question, however, is whetherexpletives of the kind can still be maintained inthe minimalist framework Egerland adopts. Thehypothesis that case can be checked through anexpletive in SpecAgr(O)P seems to completelyundermine the idea that some features have to bechecked in overt syntax. This is so because theexpletive would only carry those features thatneed overt checking. In other words, languagescould have the option to carry out in overt syntaxwhat other languages do in LF: checking of fea-tures by moving only the features, not the lexicalcategory they are associated to.

It seems worthwhile to consider the alterna-tive where the DP overtly moves to the case-checking position (SpecAgr), both for subjects andobjects (4a). Under this view the difference be-tween OV and VO (or SV and VS) results fromsubsequent movement of either the DP, the verb,the XP containing the verb, or a combination ofthe preceding:

(4)a. [XP Spec X [YP Spec Y [AgrP DPi Vj-Agr [ ... [ (ti ) tj (ti)]]]]]b. [XP Spec X [YP Spec Vj-Agr [AgrP DPi tj [ ... [ (ti ) tj (ti)]]]]]c. [XP Spec X [YP DPi Vj-Agr [AgrP ti tj [ ... [ (ti ) tj (ti)]]]]]d. [XP [AgrP ti Vj-Agr [ ... [ (ti ) tj (ti)]]]k X [YP DPi Y tk]]

The alternation between structures (4b) and (4c)is reminiscent of what has been argued for in theliterature on the interaction between scramblingand definiteness in Germanic. The alternationbetween (4a) and (4d) recalls recent proposals onsubject inversion and should be evaluated withrespect to information packaging of the clause.

4. Participial absolute clausesModern Italian participial absolute clauses

(henceforth PAC) allow the inaccusative subjectand the transitive object to be expressed overtly(see Belletti 1990, from which the examples beloware drawn):

(5)a. Arrivata Maria, Gianni tirò un sospiro di sollievo

arrived Maria, Gianni was relievedb. Conosciuta Maria, Gianni ha subito cambiato il suo stile

di vitaknown Maria, Gianni has immediately changed hislifestyle

Both construction types, display two curiousproperties. Constructions like (5a) impose a strictVS order, witness (6a). Furthermore, they chal-lenge the fairly standard view that only finitesentences have the possibility of licensing nomi-native case. Constructions like (5b), on the otherhand, impose agreement with the object in situ, aproperty which is lacking in ordinary auxiliary +past participle constructions in present-dayItalian, and they do not allow the (transitive)subject to be expressed overtly, witness (6b).

(6)a. * Maria arrivata, Gianni tirò un sospiro di sollievob. * Conosciuta Gianni Maria, Federica si è sentita tradita

known Gianni Maria, Federica felt herself betrayed

According to Belletti (1990), absolute past partici-ples only license case if they move to C, eitherovertly or covertly. This hypothesis straightfor-wardly accounts for the word order facts observedin (5–6a) and for the restriction on the overtrealization of the subject in (6b).

With this background in mind, Egerlandconsiders PAC’s in medieval, renaissance andmodern Italian, showing that both the word orderrestriction in inaccusative PAC’s and the subjectrestriction in transitive PAC’s have dramaticallychanged during the ages, as shown in the summa-ry in (7).

(7)Restriction Medieval Renaissance Modern

Italian Italian Italian

a. Strict V1 no no / yes yesb. Overt S and O yes yes no

Egerland’s account of the different settings ofmedieval, renaissance and modern Italian withrespect to the word order property in (7a) isbased, as expected, on the assumption that thereare differences in verb-raising between the threegrammars. Egerland adopts Belletti’s hypothesisthat there is a very strong tendency towards verb-raising to C in modern Italian. As for medievalItalian, he assumes that the verb only moves toAgr. This hypothesis is supported by the fact that(structural) subjects can precede the verb while(structural) objects cannot. Interestingly, renais-sance Italian appears to require the verb to moveto a position between Agr and C: while (structur-al) subjects generally follow the verb, as in mod-ern Italian, pronominal subjects precede the verb,

(1)Accusative case Medieval Renaissance Italian Renaissance Italian Modernparameters Italian Type 1 Type 2 Italian

a. Strength yes no no nob. Locus of checking AgrOP AgrOP AspP AspP

(2) Agreement Medieval Renaissance Italian Renaissance Italian Modern

Italian Type 1 Type 2 Italian

a. Ordinary DP’s yes no no nob. Operators yes yes no no


a pattern which is reminiscent of the variationPoletto (1993) has observed in the clausal syntaxof northern Italian dialects. An interesting conse-quence of Egerland’s analysis is that there ap-pears to be a gradual shift in the raising optionsof the verb form medieval to modern Italian, a factwhich, both from an empirical and a theoreticalpoint of view, deserves more attention.

Egerland assumes that the properties in (7b)follow from a difference in the functionalstructure of PAC’s in medieval and renaissanceItalian with respect to modern Italian. Morespecifically, he assumes that, in the older periods,PAC’s had a fully fledged functional structure,including a Tense projection, and that modernItalian lacks (at least) this Tense projection. As aconsequence, medieval and renaissance Italianhave two case assigners, T and V, while modernItalian only has one, V; hence, medieval andrenaissance Italian license two overt DP’s (see(12) in the author’s summary) and modern Italianonly one. The hypothesis that TP has disappearedfrom modern Italian PAC’s is well supported:several phenomena that, in one way or other, arerelated or have said to be related to the presenceof TP are attested in both medieval and renais-sance Italian, but are typically lacking in modernItalian. Perhaps the clearest case is that of nega-tion: possible in medieval and renaissance PAC’s(see (14) in the author’s summary), but not so inmodern Italian.

Egerland’s proposal that modern Italianabsolute past participles exhibit a strong tenden-cy towards V-to-C raising and lack a tense projec-tion is close in spirit to what Belletti (1990)argued for. However, Egerland’s and Belletti’sproposal differ in two fundamental respects: theaccount of obligatory past participle agreementand the mechanism of case assignment (checking).

In Belletti’s account, the two issues are tight-ly related: agreement is required in order to makecase available under government: either nomina-tive for the subject of inaccusative verbs (5a) oraccusative for the object of transitive verbs (5b).In Egerland’s account, the link between agree-ment and case assignment / checking is preservedonly for inaccusative verbs, i.e. for the checking ofnominative of the subject. As for the object oftransitive verbs, Egerland maintains the reasona-ble claim that accusative need not be checkedovertly. As a consequence, the object should beable to remain in situ and past participle agree-ment should not be available. This conclusion

appears to be correct for renaissance Italian butnot for modern Italian.

Egerland’s solution to the agreement paradoxin modern Italian is both innovative and appeal-ing. His analysis is based on the following as-sumptions: (i) past participle constructions like(5b) are always passive; (ii) passive predicatesrequire agreement with their internal argumentand (iii) absolute past participle constructionsalways project a delimitedness feature (AspP),which overtly checks a nominal feature. As aconsequence, the inner object will move to Spe-cAspP in order to check the nominal feature ofAsp+del and move further up to SpecAgrOP toinsure agreement with the passive predicate.

(8)[AgrOP SpecAgrO [AspP SpecAsp+del [VP V NP]]]

The assumption that PAC’s involve a delimitednessprojection in modern Italian but not in earlierstages has a very welcome consequence: it ex-plains the fact that PAC’s in medieval and renais-sance Italian are not subject to specific lexicalrestrictions while they are in modern Italian: onlyverbs which are compatible with a delimitedinterpretation seem to be admitted in PAC’s, wit-ness the following contrast:

(9)a. Finita la casa, l’architetto andò in pensione

finished the house, the architect retiredb. * Temuta la sconfitta, il generale rinunciò al suo piano

feared the defeat, the general give up his plan

The drawback of this assumption is that in Eger-land’s framework the projection of the delimited-ness feature implies lexical suppression of theexternal theta-role, hence a passive analysis forconstructions like (5b). A passive analysis seemsnot quite adequate, considering that PAC’s oftransitive verbs can contain an anaphor, as in (10)(from Egerland 1996: 262).

(10)Elogiata solo se stessa, Maria restò del tutto isolatapraised only herself, Maria remained completely isolated

Egerland argues that examples like (10) need notconstitute counterevidence against a passiveanalysis and that the anaphor could be bounddirectly by the matrix subject. However, theargumentation does not seem quite satisfactory.One expects a short distance anaphor like sestesso to be bound within the PAC, as the interpre-

tation of sentences like (10) suggests: the onlyavailable interpretation for (10) is the one wheresome person x, in casu Maria, praised him- orherself, not that some arbitrary person praisedsomeone else. Note, furthermore, that even inReinhart & Reuland’s (1993) approach to binding,one cannot account for the properties of (10): if areflexive predicate can lexically satisfy bindingrequirements, one would not expect (quasi-)oblig-atory coindexation of the anaphor with the matrixsubject.

5. ConclusionThere is little doubt that the data Egerland

has worked with, offer interesting contrasts in theparticipial syntax of three stages of Italian. Thisholds both for the puzzling data of participles incompound tenses and for those in absolute claus-es. Because of their complexity, the data consti-tute a strong test case for the descriptive power ofa linguistic theory and Egerland undoubtedly hasgone a long way towards showing the minimalistprogram’s qualities of managing such complexdata. At times, it is true, the analyses give theimpression that the theory is there to play with;and perhaps the main reason for this is the factthat Egerland did not undertake any generalattempt to search for the causes of variation inthe three stages of Italian. This most certainlywould have offered a better basis for evaluatingsome of his proposals and could have strength-ened the theoretical underpinnings of his analy-ses. Nevertheless, the overall picture remainsthat of a very solid descriptive work where thevarious word order phenomena and agreementpatterns nicely fall into place.

ReferencesBelletti, Adriana (1990). Generalized Verb Move-

ment. Turin: Rosenberg & Sellier.Chomsky, Noam (1995). The Minimalist Program.

Cambridge, Mass.: The MIT Press.Poletto, Cecilia (1993). La sintassi del soggetto nei

dialetti italiani settentrionali. Padua: Uni-press.

Reinhart, Tanya & Eric Reuland (1993). Reflexivi-ty. Linguistic Inquiry. 25.4. 657–720.

Rizzi, Luigi (1982). Issues in Italian Syntax.Dordrecht: Foris.

Roberts, Ian (1987). Implicit and DethematizedSubjects. Dordrecht: Foris.

DIRECT OBJECT SCRAMBLING INDUTCH AND ITALIAN CHILDLANGUAGE

by Jeannette Schaefferreviewed by Astrid Ferdinand

Summaryby the author

Over the past decade many theories of first lan-guage acquisition have utilised the concept of“Parameter Setting and Resetting” to explainlinguistic development (cf. Hyams 1983; 1986among others). Despite the fact that the idea ofParameter Setting and Resetting is conceptuallyattractive, it turns out that, to the extent that wecan formulate parameters such as the “Head-final/Head-initial Parameter or the Pro-dropParameter, they are set extremely early, usuallybefore the onset of speech. Therefore, languagedevelopment cannot solely be explained by Pa-

rameter Setting and Resetting. The question thenremains as to how development should be ac-counted for. A key to solving the so-called ‘develop-mental problem’ can be found in the ‘ModularityHypothesis’, as first proposed by Fodor (1982).According to the Modularity Hypothesis thelanguage faculty is divided up into a lexicon, acomputational system (syntax, semantics, phonol-ogy), a pragmatic system and possibly moremodules. Uneven development or growth withinthe separate modules and the interaction betweenthem can explain why children’s grammars devel-op the way they do.

In this thesis I argue that developmentwithin the pragmatic component of language

causes development within syntax. The mainhypothesis is formulated in (1):

(1) Hypothesis(i) specificity is not always grammatically marked in the

grammar of 2-year old children;(ii) (i) is possible because 2-year olds lack pragmatic

Discourse Rule.

In order to test this hypothesis I chose two syntac-tic processes which are purported to involvespecificity in the literature about adult language,namely object scrambling and object clitic place-ment in Dutch and Italian child language. I arguethat while object scrambling and object cliticplacement are close to adultlike by age 3, at age 2object scrambling and object clitics are optionaleven where they are obligatory for adults.

I carried out an elicited production task with49 Dutch and 35 Italian children between theages of 2 and 6. The Dutch children were testedon their knowledge of object placement withrespect to negation and adverbs. The Italianchildren were tested on their knowledge of objectclitic placement in present tense and passatoprossimo (present perfect) constructions. Themain results indicate the following. Dutch 2-yearolds hardly use any adverbs, but fail to movespecific objects over negation 70% of the time(though this movement is obligatory for adults).Furthermore, Italian 2-year olds omit objectclitics 64% of the time in obligatory contexts.Dutch three year olds fail to scramble specific


objects over negation at a rate of only 28%. Italianthree year olds omit overt object clitics in obligato-ry environments at a rate of 15%.

I assume the following sentence structure forDutch and Italian (DiscP = Discourse Phrase; SpP= Specificity Phrase’):

(2)[CP [AgrSP [DiscP (adverbs) [TP [SpP [NegP (Dutch) [AgrOP[VP]]]]]]]]

The Dutch 2-year olds’ lack of object scramblingover negation (70%) and the Italian 2-year olds’dropping of overt object clitics (64%) provideevidence for the hypothesis that children’s nomi-nal expressions are optionally marked with re-spect to specificity, which is elaborated on below.

In the adult grammar scrambling and cliticplacement are driven by specificity. Sportiche (1992)proposes a unification theory of object clitic place-ment (in Romance) and object scrambling (inDutch), in which specific objects (XPs) and clitics(Xs) need to be licensed under spec-head agreementin a functional phrase (SpP in (1)) just below TP. Iextend this analysis and claim that there are twotypes of specificity: ‘discourse-related specificity’and ‘non-discourse-related specificity’. Discourse-related specific object DPs must have an antecedentin the discourse and are licensed in SpecDiscP; non-discourse-related specific object DPs are licensed inSpecSpP. Object clitics (inherently specific) arebase-generated in the head of SpP and licensed by apro object (specific because of its pronominal char-acter) that has moved from its VP-internal positionto SpecSpP. Specific overt object XPs move fromtheir VP-internal position to SpecSpP and arelicensed by an inherently specific null clitic, base-generated in the head of SpP. This is illustrated forItalian (3a) (underlying structure of clitic place-ment) and for Dutch (3b) (scrambling):

(3)a. [AgrSP Gianni [TP [SpP proi [Sp la [AgrOP [VP mangia ti

]]]]]]John it- eats

‘John is eating it’b. dat [AgrSP Jan [DiscP [TP [SpP het boeki [Sp 0clit [niet

that John the book not[AgrOP [VP ti heeft gelezen]]]]]]]

has read‘that John didn’t read the book’

I propose that in (2a) the finite verb mangiaincorporates the clitic when it head-moves to AgrSin order to check its Tense and Agreement fea-tures, so that la+mangia end up in AgrS. Further-more, the Dutch object het boek in (2b) moves onto SpecDiscP if it is discourse-related.

Returning to the experimental results. Ipropose that a nominal expression is marked forspecificity only if the distinction between dis-course- and non-discourse-relatedness is made. Ifthis distinction is not made, (the syntactic fea-ture) specificity is lacking. If the object (whetherovert or pro) is not marked for specificity, it can-not move up to SpecSpP where specificity islicensed under spec-head agreement. In the caseof an overt object, this will result in a ‘non-scram-bled’ object, i.e. one which remains below NegP. Inthe case of a pro object, this will result in objectclitic-drop: the specificity of the clitic cannot belicensed, and therefore the clitic is not spelled out.

Although the analysis just proposed accountsfor the syntactic differences between early childgrammar and adult grammar, it is not immediatelyclear how children interpret object DPs without a[+specific] feature. I argue that these object DPs areinterpreted referentially, just as in adult language,but that children arrive at this referential interpre-tation in a different way than adults. Let us firstconsider the adult interpretive mechanisms.

As is shown in (4), I follow Hyams (1994) andHoekstra and Hyams (1995) in the claim thatevery DP contains a D-chain, coindexing theNoun, the Determiner and an operator in SpecDP,which makes referentiality visible.

(4) D-chain

[DP Opi [Di the [NumP [NP [Ni tree]]]]

Furthermore, I propose that there is anR(eferentiality)-Chain, which coindexes the objectDP with an operator in SpecCP, which hooks upthe object DP to the relevant antecedent in thediscourse. This is illustrated in (5):

(5) R-chain

the tree [CP Opi ... [ DiscP [DP Opi ... [Di the [Ni tree]]]...[SpP...]]]

This mechanism — which I refer to as ‘grammati-cal interpretation’ — accounts for the referentialinterpretation of ‘normal’ definite DPs, such as thetree, the girl, the book.

On the other hand, the much smaller, andmore exceptional class of definite DPs such as thesun, the president, John refer to their referents inthe (model of the) world directly, that is, withoutan R-chain, as illustrated in (6):

(6) No R-chain

[CP ...[ DiscP ...[SpP [DP Opi ... [Di the [Ni sun]]]...]]]

This mechanism does not make use of an R-chainand a linguistic antecedent, and therefore it isreferred to as ‘direct interpretation’.

Returning to the child data, I propose that 2-year old children have both options to interpretdefinite DPs such as the tree. This is illustrated in(7) and (8), respectively.

(7) Grammatical interpretation

the tree [CP Opi ... [DP Opi [Di the [Ni tree]]]

(8) Non-grammatical direct interpretation

[DP Opi [Di the [Ni tree]]]

If they choose to specify specificity on the DP,interpretation takes place in an adultlike fashion,as in (7). However, if the [+specific] feature isabsent, the DP is interpreted directly, as in (8),similar to the way adults interpret DPs such asthe sun.

Now why do 2-year old children have accessto both interpretive options for DPs such as thetree, as opposed to adults only one? I claim thatthis is due to a missing rule in the child’s prag-matic system, namely the ‘Discourse Rule’, whichis stated in (9):

(9) Discourse RulePreceding linguistic discourse and knowledge of interlocutorMUST be taken into account.

If children do not have the Discourse Rule, theydo not always take the discourse into account.This implies that in these cases they do not distin-guish between discourse-related and non-dis-course-related, resulting in the absence of the[+specific] feature. Consequently, movement toeither SpP or DiscP is not prompted and does nottake place.

In contrast, adults, who do have the Dis-course Rule, always take discourse into accountand thus always distinguish between discourse-related and non-discourse-related, resulting in themarking of specificity, and thus movement of the[+specific] DP to SpP or DiscP.

Concluding, I have shown that Dutch andItalian 2-year old children optionally mark specif-icity, resulting in optional object scrambling andoptional object clitic placement, but that by theage of 3 children perform roughly adultlike in thisrespect. I furthermore argued that this develop-ment within syntax is due to the acquisition of arule in the pragmatic system, namely the Dis-course Rule. This provides support for an acquisi-tion theory that exploits the ModularityHypothesis to explain language development.

Topics that remain for future research in-clude the following: a) can the proposed analysisof the behaviour of [+specific] objects also accountfor the behaviour of [+specific] subjects in childand adult language? and b) is it true that 2-yearold children do not distinguish between discourse-related specific DPs such as the tree, and non-

discourse-related specific DPs such as the sun, asis predicted by the analysis proposed here.

ReferencesFodor, Jerry A. (1983) The Modularity of Mind. MIT Press,

Cambridge, Massachusetts.Hoekstra, Teun and Nina Hyams (1995). The syntax and

interpretation of dropped categories in child language:a unified account. Paper presented at WCCFL, 1995.

Hyams, Nina (1983). The acquisition of parameterizedgrammars. Doctoral dissertation, City University ofNew York.

Hyams, Nina (1986). Language Acquisition and the Theoryof Parameters. Dordrecht: Reidel.

Sportiche, Dominique (1992). Clitic Constructions. Manu-script UCLA.

Reviewby Astrid Ferdinand

1. IntroductionThree different hypotheses have been put

forward in the literature concerning the amountof syntax the human child disposes of when start-ing to create his or her first multi-word utteranc-es. The Lexical Thematic Hypothesis claims thatthere is no syntactic structure. According to theFull Competence Hypothesis the opposite is true:the complete structure is available. The PartialCompetence Hypothesis occupies a position be-tween these two extremes.

In Schaeffer’s dissertation, the Full Compe-tence Hypothesis is defended, in order to explainyoung children’s linguistic behavior with respectto direct object scrambling in Dutch and directobject clitic placement in Italian. I will discussSchaeffer’s methods of research, her analysis ofthese phenomena in two-year-old Dutch andItalian children and I will try to determine towhat extent this analysis supports the Full Com-petence Hypothesis.

2. MethodsIt is known from earlier literature that, in

contexts triggering scrambling in adult Dutch,young Dutch speaking children do not scramble(Hoekstra and Jordens 1994) or optionally scram-ble (e.g. Barbier 1993, Schaeffer 1995) the directobject. Schaeffer remarks that the empirical basisof these claims suffers from several weaknesses.To obtain more and more reliable data, Schaeffertested children from 2 to 6 years old, with anelicited production task.

The tests were designed and carried out withcare. Since the researchers visited the children’sday-care center or school repeatedly before theactual testing began, the children were familiarwith them and with the toys used in the tests.According to Schaeffer, they performed the taskswith pleasure. This is an important detail, since itshows that although the scrambling/non-scram-bling sentences seem to be rather complex forchildren below six, the children were apparentlynot forced to perform above their level. The relia-bility of the tests was increased because thechildren’s results were compared with the per-formance of adults on the same tests.

3. Analysis: specificity and discourserelatedness

3.1. SpecificitySchaeffer obtained two major results. First, in

Dutch two-year-olds, direct object scrambling doesnot take place in about 70% of the obligatorycontexts. Secondly, Italian two-year-olds leave64% of the direct object clitics empty. Schaeffer’sanalysis of these results is based on the followingassumptions about adult language:

– Phrase structure contains a Discourse Phraseand a Specificity Phrase: [DiscP [SpecP [NegP[AgroP [VP]]]]].– Scrambled direct objects (higher than negation)move because of the feature [+specific] to thespecifier of Specificity Phrase. The feature [+spe-cific] is the syntactic expression of referentiality.


– Scrambled direct objects move further to thespecifier of Discourse Phrase to be related to thediscourse.– Direct object clitics are inherently specific andare generated in the head of Specificity Phrase.

In Schaeffer’s analysis, there are two importantdifferences between child language and adultlanguage. In the first place, specificity is optional-ly marked in child language. This explains utter-ances as the following:

(1)Bert gaat niet de peer opeten (C. 2;9)Bert goes not the pear up-eat‘Bert is not going to eat the pear.’Schaeffer p. 57, ex. (1.e)

(2)Ernie gaat niet twee bomen uitknippen (M. 2;4)Ernie goes not two trees out-cut‘Ernie is not going to cut out two trees.’Schaeffer p. 59. ex. (2.a)

Without focus on any constituent, sentence (1) isexcluded in adult Dutch, since it contains a defi-nite, hence [+specific] unscrambled direct object.Schaeffer argues that in this child utterance, thedirect object is unmarked for specificity, despiteits definite determiner. Similarly, the indefiniteunscrambled direct object in (2) would be un-marked for specificity, even if it has a referentialinterpretation.

In the second place, Schaeffer argues that thechild is able to interpret DPs “directly” instead ofby movement to SpecDiscP. This explains why thedirect objects in (1)–(2) do not move to SpecDiscPeither.

I would like to discuss what kind of develop-ment is predicted by this analysis of (1)–(2).Schaeffer explains the development in syntax by adevelopment in pragmatics, which causes dis-course relatedness, and hence specificity to be-come obligatorily marked. One may wonderwhether this excludes language-specific develop-ment or not. Note that the child’s grammar differsfrom the adult’s in the following ways.

In Schaeffer’s analysis, a DP in adult lan-guage that is referential, semantically, always hasthe feature [+specific]. For the child, however, thisdoes not hold. The child may express referentiali-ty with a [+specific] D and then scramble thedirect object DP. S/he also has the option not toexpress it at all (as in (1)–(2) and then leave thedirect object below negation. This means that inchild language the interface between semanticsand syntax must be different. This is the firstdifference. This difference between child andadult language has impact on development.During the development toward the adult state,the child must acquire the obligation of express-ing referentiality syntactically. This seems to bepossible by acquiring language-specific features,such as those that are related to scrambling. Thechild has to acquire the ways in which referential-ity is expressed in syntax (e.g. scrambling) and inmorphology (e.g. determiners). Summing up, theacquisition facts cannot be explained by develop-ment in pragmatics only. There must be language-specific development with respect to specificity.

The second difference is related to the statusof Specificity Phrase in the grammar of the two-year old. If the specificity feature is not marked,as would be the case in (1–2), the feature on thehead of Specificity Phrase cannot be checked. As aresult, Specificity Phrase cannot be projected,since if it were projected, the derivation wouldcrash (Chomsky 1993). The fact that the childdoes not project Specifity Phrase in context re-quiring this projection in adult language pleadsagainst Full Competence. Yet, there are still the30% utterances in which scrambling does takeplace. In the representation of these utterances,Specificity Phrase must be present, if they areanalyzed as containing the feature [+specific].Schaeffer uses the optional presence of scram-bling as an argument for Full Competence. I willreview this argument in section 4.

Another question the optionality of specificity

raises is how it affects the form of DPs in the child’soutput. Specificity is expressed on the D-head of areferential nominal expression. The overt definite(or the specific indefinite) determiner is taken to bethe morpho-syntactic expression of specificity (p.21–22). If specificity is optional at an early stage,there are, in principle, two ways in which thisoptionality can affect the child’s output. The childmay either leave the D-head phonologically empty,or have a lexicalized D that is unmarked for specifi-city. Schaeffer’s analysis allows both options. On theone hand, the optional spell-out of specificity isargued to lie at the basis of the optional phonologi-cal realization of direct object clitics by Italian two-year-olds. On the other hand, Dutch childrenphonologically realize definite (and specific indefi-nite) determiners that are argued to be unmarkedfor specificity. Since direct object clitics and definitedeterminers are both inherently [+specific] in adultlanguage, it is not immediately clear why the childwould treat them differently. Why do direct objectclitics have to be phonologically empty when theyare unmarked for specificity, whereas this does nothold for definite determiners? An explanation ofthese differences would strengthen the analysisconsiderably.

3.2. Discourse relatednessSchaeffer’s data contain several interesting

facts that support the claim that children’s dis-course differs from adult’s discourse. The fact thatthere are differences confirms earlier findings.Children differ from adults in that they usepronouns without first establishing a referent forthat pronoun by using a DP (e.g. Karmiloff-Smith1979). Further, they use definite DPs instead ofindefinite DPs, which is confirmed by Schaeffer’sresults (Fig. VI, p.73).

Schaeffer finds another difference. Her Ital-ian and Dutch subjects below six produce muchmore full direct objects than adults:

Schaeffer p. 62, Figure IIIItalian children

overt clitic omitted clitic full direct object

2 022% (22) 64% (63) 14% (14)3 062% (179) 15% (43) 23% (68)4 089% (237) 00% (0) 11% (28)5 091% (227) 00% (0) 09% (23)adults 100% (439) 00% (0) 00% (0)

Schaeffer p. 64, Figure IVDutch children

full object dem. pron. strong pers. pron. clitic

2 52% 26% 6% 16%3 15% 26% 1% 58%4 07% 07% 0% 86%

Schaeffer explains the Dutch data by the assump-tion that the direct object clitic remains emptybecause of the absence of the feature [+specific],so that the direct object must be realized as a fullDP. Yet, the similarities between Dutch andItalian may also point to a more general tendency.

These results show exactly the opposite of the‘out of the blue’ use of pronouns known from theliterature. Here, children use full DPs instead ofpronouns, although the referent has been madeclear in the previous discourse. Apparently, thechild can both use a pronoun for a DP (the ‘out ofthe blue use’) and a DP for a pronoun. It may bethe case that this is related to the fact that the‘out of the blue’ use of pronouns has been found insubjects, whereas Figure III describes the use ofpronouns and DPs in direct objects. The informa-tional status of subjects is different from that ofdirect objects, which may have an impact onacquisition.

The differences between child and adultdiscourse cannot be explained completely bymaking the assumption that the child ignores thediscourse and does not distinguish between dis-course-related and non-discourse-related. MacNa-mara (1982) shows that children distinguishdefinite from indefinite DPs in reception beforethe age of two. Therefore, children are not com-pletely ignorant of the discourse.

As Schaeffer notes herself, this issue deservesmore research. Further inquiries concerning

discourse in children may also shed more light onthe development of focus. Focus effects fall outsidethe scope of this book, but they interact withscrambling in adult Dutch, and may have aninfluence on acquisition, too.

4. The Full Competence HypothesisIn this section, I would like to comment on

the following question. Does the optional presenceof functional elements in the child’s productionprovide evidence for the Full Competence Hypoth-esis? The reliability of the answer to this questiondepends on the reliability of the methods withwhich the optional presence of functional ele-ments is assessed. At least three variables shouldbe taken into consideration: age, variation be-tween individuals, and variation between differ-ent types of construction.

With respect to age, it is important to ac-knowledge that direct objects are used by manychildren below age 2. Schaeffer’s youngest Dutchsubject is aged 2;4. Further research with youngerchildren may give different results, comparable tothose of Hoekstra and Jordens (1994), who findthat object scrambling is initially absent in childDutch.

Variation between individuals may alsoinfluence the results. Because of this variation, itis rather problematic to assess optionality incross-sectional research. One of the possibilitiesthat I can think of is that among the two-year-oldsthere is a group of children showing no directobject scrambling, a group that has optionalscrambling, and a group with obligatory scram-bling. If one takes all results together, these threetypes of individuals are all considered as one type,showing optional scrambling. To exclude thispossibility, a characterization of individualsshould be made.

Further, not all functional elements appear atthe same time. Some constructions are early,others are late. At early stages, some construc-tions are present in the child’s production, othersare absent. Therefore, the presence of one con-struction does not provide a sufficient argumentfor Full Competence, as opposed to Partial Com-petence.

Summing up, further research with youngerchildren, taking into account their individualdifferences would shed more light on the amountof functional structure available. Ideally, theresults on direct object scrambling and objectclitic placement are compared with results onother constructions.

5. ConclusionIn the literature on first language acquisition,

there has been relatively little attention for thedevelopment of the direct object, whereas thedevelopment of the subject has been documented ingreat detail. Therefore, Schaeffer’s dissertation fillsa gap in our knowledge about the acquisition ofsyntax. Since the topic of research is relativelyunexplored in general, the analysis still leavesseveral questions unanswered, as I have shownabove. In my view, the strength of the book lies inthe intermodular approach. In this respect, I hope itwill inspire further research along the same lines.

ReferencesBarbier, I. (1993) ‘An experimental study of scrambling

and object shifting in the acquisition of Dutch.’ Paperpresented at GALA, University of Durham.

Chomsky, N. (1993). ‘A minimalist program for linguistictheory.’ In K. Hale and S. Keyser (eds.) The View fromBuilding 20. Cambridge, Mass.: MIT Press.

Hoekstra, T. and P. Jordens (1994) ‘From adjunct to head.’In T. Hoekstra and B. Schwartz (eds.) LanguageAcquisition Studies in Generative Grammar. Amster-dam, John Benjamins.

Karmiloff-Smith, A. (1979) A Functional Approach to ChildLanguage: a study of determiners and reference.Cambridge, Cambridge University Press.

MacNamara, J. (1982) Names for Things: a study ofhuman learning. Cambridge, Mass., MIT Press.

Schaeffer, J. (1995) ‘On the acquisition of scrambling inDutch.’ In D. MacLaughlin and S. McEwen (eds.)Proceedings of the 19th Annual Boston UniversityConference on Language Development 2, 521–532.Somerville, Mass., Cascadilla Press.

Book reviews Glot International, Volume 3, Issue 5, May 1998 16

1. IntroductionGrounded Phonology is basically a theory

that defines the class of permissible featurecombinations and a theory of rule structure basedempirically on a deep analysis of ATR systems.Although published in 1994, the manuscript hasbeen circulating since 1992 (a more archaic ver-sion since 1986) and it has much influencedphonological research since then.

The theory has three basic components: atheory of underspecification, which is (correctly, Ithink) not based on the segment; a set of condi-tions on permissible structures which is cruciallyrelated to phonetic interpretation (conditionsmust be “phonetically grounded”); and a theory ofrule structure which restricts the class of possiblephonological rules to a set of parametric combina-tions. The theory of feature combination definesthe general mechanisms that determine bothlexical and derived feature combinations; thissubtheory is presented in the first of the threebasic chapters of the book, chapter 2, “Combinato-rial Specification”. It also defines the way featurecombination relates to their physical interpreta-tion: representations are impossible or permissi-ble, and in the second case, less or more markeddepending on the strength of their phonetic plau-sibility (chapter 3, “Grounding Conditions”). Thethird main chapter (chapter 4, “ParametricRules”) introduces a set of parameters whichdetermine a more restricted class of permissiblerules. These three parts correspond to specificsubtheories organized as independents modules ofphonological theory.

There is a general “Introduction” (1–41), anda final chapter (chapter 5, “Conclusions”) in whichsome of the shortcomings of the theory are dis-cussed and the integration of the theory in anarchaic version of optimality theory is explored.

2. Segments and pathsThe “Introduction” (chapter 1) stresses the

modular character of the theory presented, andsets up some preliminary theoretical tenets onwhich subsequent developments are based.

Archangeli and Pulleyblank assume a fea-ture-geometric approach to feature organization,in particular one in which well-formed nodes mustdominate feature content. There are two basicnotions, F-element and path. An F-element is afeature (binary or monovalent) or any other node.The notion path is presented in this introductionand further developed in chapter 2. A path is asubtree of a prosodic and feature-geometricalstructure. It is formed by a set of nodes (features,prosodic categories, or other nodes) linked byassociation lines such that only one token of eachnode type is present. In (1) there are two mainpaths, marked with a continuous line and abroken line.

Some lower elements (e.g. +HI) are language

particularly assigned to an anchor, which isalways a node dominating them in the same path,e.g. an ATR node can have a mora or a syllablehead as an anchor. When structural changes inthe structure motivated by rules or conventionsassign some F-element [= feature, node] α to someanchor β a path from α to β is created, i.e. associa-tion lines with the necessary intermediate nodesare generated (Node Generation convention).

A set of conditions, path conditions (discussedin chapter 2), define path well-formedness. Pathconditions are conditions on feature/node combi-nations: if a condition disallows two F-elements,they are incompatible in the same path, but theycan be present in a representation if they are notwithin the same path, e.g. if one is floating. Noticealso that conditions hold at the underlying levelsand at any other level (this follows from the Well-formedness Principle (2)):

(2)Well-formedness Principle: Representations [nota bene, at alllevels — JM] and relations between representations [i.e., rules— JM] are well formed.

(3)Locality Condition: Phonological relations respect Adjacencyand Precedence.Adjacency: α is structurally adjacent to β iff:a. at least one of the two is unassociated, both are on the sametier, and no element intervenes between the two on that tier;or,b. both α and β are associated to the same anchor tier and noanchor intervenes on that tier between the anchors to which aand b are associated.

(4)Precedence Principle: Precedence relations cannot becontradictory.

(5)Anchor Hypothesis: Anchor paths establish core-tier ordering.

Adjacency (the definition is almost identical toMyer’s (1987)) has important consequences forrule application. In connection with the OCP, itdefines, for instance, floating elements as adja-cent, hence subject to it. It also defines adjacencyin representations with three elements on a tier:in (6) α and β are not adjacent in the “twin peak”configuration (6a). On the other hand, the gappedconfiguration (6b) is ill-formed, being disallowedby the Locality Condition and the PrecedencePrinciple.

(6)a. twin peak b. gapped

µ1 µ2 µ3 µ1 µ2 µ3

α β α

In (6a) there is an anchor (µ2) intervening be-tween µ1 and µ3 to which α and β are associated,therefore by (3b) α and β are not adjacent, henceno OCP effect. (6b) is ruled out because α is in apath with µ1, hence since µ1 precedes µ2, α alsoprecedes µ2. The same applies to α and µ3, exceptthat in this case µ2 precedes µ3, hence µ2 precedesα. The result is that α both precedes and is pre-ceded by µ2, in contradiction with the PrecedencePrinciple.

Principle (4) might in fact not be needed.Since structures must be well-formed (2) andprecedence must be respected (3), any contradic-tion in terms of precedence is a violation of well-formedness. No formal definition of the Anchor is

given. From the discussion it seems to follow thatit states the following: a) If there is a path pcontaining α and its anchor µ, then α does notprecede µ, and µ does not precede α ; b) If thereare paths p1, p2 containing α1 and its anchor µ1, α2and its anchor µ2, respectively, then if µ1 precedesµ2, α1 precedes α2, and vice versa. (Notice thatprecedence is irreflexive.)

3. What are permissible featurecombinations?Combinatorial Specification, developed in

Chapter 2, differs from both contrastive andradical underspecification. The primitives of thetheory are F-elements, i.e. features like [+ATR], ornodes like Place and association status (linked/free); in the cases analyzed in the book F-ele-ments are vowel features. F-elements can bemissing (underspecification), linked, or free.Representations are sets of F-elements related byassociation (by association lines in the usualnotation) supplemented with prosodic structure.Segment is a derived notion, and a segment is apath (the converse is not true).

Specific combinations of elements in a pathcan be banned universally or language particular-ly. This corresponds approximately to the notion“possible segments in language X.” Thus thenumber of possible segments depends on thenumber of F-elements and on the combinatorialrestrictions, which are expressed by (path) condi-tions. Consider a language with a vowel systemdefined by the features +HI, +LO, –BK. Since anyof them can be unspecified underlyingly there are23 = 8 combinations, namely those shown in (7a).The conditions in (7b–d) eliminate three items,those represented by ‘*’.

(7) a.* * u i * a o e

+HI +HI +HI +HI+LO +LO +LO +LO–BK –BK –BK –BKPath Conditions: (Universal) b. If +HI then not +LO

c. If +LO then not +HI(Particular) d. If +LO then not –BK

If we eliminate condition (7d), then there will betwo low vowels, e.g. [a] and [æ]. If all three condi-tions apply, then we get (7a), a regular five vowelsystem. The more elements a language allows, themore complex the segment system will be. Forvowels, in addition to ±HI, ±LO, ±BK, there canalso be ±RD and ±ATR.

There is another way to restrict the set ofavailable “segments.” Consider a a language withthe same features but with different featurevalues, namely –HI, +LO, +BK. We get (8).

(8)a1 a2 e u a3 a4 u i

–HI –HI –HI –HI+LO +LO +LO +LO+BK +BK +BK +BK

We now have four combinations containing +LOwhich are not excluded by any universal conditionlike (7b,c). An eight system like (8) constitutes apossible case, but it is surely very marked. Ingeneral, given two representations, like a3 = [+LO,+BK] and a4 = [+LO], there are three possibleoutcomes. First they may surface as phoneticallydifferent vowels. Or they can have an identicalphonetic representation, in which case there isabsolute neutralization and the system is marked:there must be strong evidence for positing thesetwo different underlying representations —evidence is of course based on their differentphonological behavior (this is the case of Ainuwhich has two vowels like a3 and a4). The thirdpossibility is that the more complex vowel isexcluded. This exclusion is achieved through thecombined action of two principles, Representation-al simplicity and Recoverability. The former statesthat the complexity of a representation is deter-mined by the number of terminals plus thenumber of association relations; the second onestates that “Phonological representations andphonetic content are related.” They are interpret-

BASE RULES PHONETICALLY

by Joan MascaróReview of GROUNDED PHONOLOGYby Diana Archangeli and Douglas Pulleyblank

(1)Root Root

Place Place

Dorsal Labial

+HI +RD


ed in the sense that complexity should be minimalwhile allowing phonetic content to be assignedunambiguously. The consequence is that a lan-guage can have underlying paths like a3 and a4,but only if there is clear phonological evidencethat both a-vowels have a role in the language;otherwise simplicity will rule out a4, the morecomplex one.

Thus, combinatorial specification differs fromradical underspecification because it merelyprefers underlying redundant information to beabsent, but does not disallow it (if present, redun-dant information corresponds to a marked situa-tion, which shows up in the phonology).Underspecified structure may be specified in thecourse of the derivation, by F-insertion or spread-ing, or by redundancy rules. In p. 110–111 Arch-angeli and Pulleyblank argue that redundancyrules introducing an F value do not have to pre-cede necessarily the first rule that mentions thisvalue.

Archangeli and Pulleyblank consider the sizeof vocalic inventories predicted by combinatorialspecification. The size depends on the number ofelements, n, and on their linked or free status. Ifall combinations are allowed for n linked F-elements (specified features), and since anyelement can be underspecified in a representa-tion, there are 2n possible combinations; if the F-elements are n free elements, then the number ofpossible paths is of course n+1. Two examples ofsuch cases (all elements linked underlyingly, allelements free) are studied, the vowel systems ofBarrow Inupiaq and Tiv. Barrow Inupiaq hasthree linked F-elements, [+HI], [+RD], [+LO]which define the set {a, u, i1, i2}, i2 being the total-ly unspecified V. Tiv defines with the three free F-elements [–BK], [+RD], [+LO] the system {i, e, a,u, o, f}, f representing a low rounded V. In Tiv,however, the total number of vowels is 2n–2=6,since two combinations, those which are [+HI,+LO] are universally disallowed.

Thus, the class of possible representationscan vary only along the following parameters: the(sub)set of F-elements, their underlying or (only)derived structures, their combinatorial restric-tions, and their linked/free character. In otherwords, the class of representations of a givenlanguage is defined by the class of F-elementsactive in its phonology, the conditions that limitthe class of combinations of these elements, by thepossibility that they appear underlyingly or onlyin derived representations, and their free vs.linked character. This yields quite (but not ex-ceedingly) large figures. For vowels, I get 167systems, without considering the effect of lan-guage particular conditions which surely willproduce a much bigger figure (notice, though, thatthese figures represent often systems that are“segmentally” identical but structurally different:they impose different phonologies, which meansthat the system of phonological processes is con-siderably restricted). Archangeli and Pulleyblankmake an approximate estimate of the number ofphonological systems generated by their theorytaking 20 linked F-elements, which gives220=1,048,576 possible combinations. The obviousneed to constrain the combinatorial system leadsto the theory of grounding.

The statements that establish the conditionsof well-formedness of F-element combinations, thepath conditions, are formulated in the shorthandnotation F1/F2, interpreted as ‘if F1 then F2’, or‘if F1 then not F2’. Thus the HI/ATR Conditionmight be interpreted as ‘if [+HI] then [+ATR]’.Here an improvement of the notation to make theshorthand F1/F2 more transparent would havebeen helpful. In many cases the + or – value forthe feature is not obvious, nor is the negative orpositive value of the consequent. A more transpar-ent shorthand notation might have been a betteralternative, e.g. expressions like HI/ATR,–ATR/–HI, DORS/*LAB, where a ‘+’ value is assumedunless ‘–’ is mentioned, and ‘*’ stands for thenegator.

3. Sympathetic and antagonistic featurecombinations: groundingChapter 3 develops the theory of grounding,

that is phonetic grounding of path conditions.Combinatorial specification defines, for any givenlanguage, a set of underlying paths, (“segments”),through the choice of features and conditions.This causes two problems. On the one hand thereis overgeneration of phonological inventories, asmentioned above; on the other hand, systems are“flat” in the sense that no difference in natural-ness or marking is established. The obvioussolution is to add mechanisms that introducemarking; highly marked objects are impossible(hence the class of objects is restricted), possibleobjects are differentiated in terms of marking.There are three basic types of marking specifica-tions: marking of F-elements (features, nodes),marking of rules (discussed in the followingchapter), and marking of path conditions. Mark-ing of F-elements is not investigated in the book;only ATR is discussed in this respect (some specu-lations about high appear in p. 466 fn. 66, andsome references to low and round on p. 376). Theformat of markedness statements is as in (9);no-tice that if a feature is monovalent, only (9a)applies. (10) gives a specific example.

(9) Format for Markedness statementsa. F tends to be used [X-amount] actively.b. If F is used actively, its active value tends [Y-amount] to

be [αF]; the passive value of F tends [1–Y-amount] to be[αF].

(10) [ATR] Markedness statementa. [ATR] tends to be used <0.15> actively.b. If used actively, its active value of [ATR] tends <0.6> to be

[+ATR]; the passive value of [ATR] tends <0.4> to be[−ATR].

(10a) states that most languages do not use thefeature [ATR]. The [X-amount] specificationindicates the probability (between (0 and 1) that alanguage might use [ATR]; since Archangeli andPulleyblank do not give specific values, I haveintroduced a value only for illustration. (10b)gives the probability that a language that uses[ATR] actively, will use [+ATR] or [–ATR] as theactive feature.

Chapter 3 is devoted to the third type ofmarkedness statements, marking of path condi-tions. Path conditions, now grounded conditionsare assigned, like F-elements, what we mightterm a “grounding value.” This value establishesan ordering of conditions in terms of their relativedegree of phonetic grounding or phonetic plausi-bility. More specifically, Archangeli and Pulley-blank propose the following principles:

(11)I Path conditions invoked by languages must be

phonetically motivated.II The stronger the motivation for a path condition Φ,

a. the greater the likelihood of invoking Φ,b. the greater the likelihood of assigning a wide scope

to Φ within a grammar, and vice versa.

To illustrate, a condition like if [nasal] then[+voice] is phonetically grounded (and a conditionlike if [nasal] then [–voice] is not) because velicopening reduces pressure in the glottal zone thusfacilitating vocal cord vibration; in addition voice-lessness in nasals reduces perceptibility. It followsthat the first condition is grounded. Consider forinstance the conditions in (12).

(12)a. If [+high] then [–low] maximally sympatheticb. I f [+high] then [+ATR] sympathetic, nonantagonisticc. If [+high] then[–ATR] antagonistic, nonsympatheticd. If [+high] then [+low] maximally antagonistic

Sympathetic relations are an instance of “en-hancement” in the sense of Stevens, Keyser, andKawasaki (1986): in (12) tongue body loweringand tongue root advancement enhance tonguebody raising. (12a), which is maximally sympa-thetic, yields totally unmarked, hence phonetical-ly necessary representations, while (12d) yieldsmaximally marked, phonetically impossible repre-sentations. (12b), which is phonetically grounded,

must not but can be a (grounded) condition of anatural language. It will yield more or lessmarked structures according to the extent towhich is sympathetic, i.e. phonetically grounded;(12c), on the other hand, is not phoneticallygrounded, it expresses a phonetically implausiblerepresentation. The result is that, given (I), both(12c, d) are prohibited, they cannot be part of theset of grounded conditions.

The phonetic articulatory basis of conditionsthat relate to tongue body height and tongue rootposition is examined (of course the basis can alsobe perceptual). The degree of strength in phoneticgrounding is not discussed in detail in the book(provisionally, four degrees of grounding are used,very strong, strong, medium, week). The assump-tion is that values can vary from minimallygrounded to maximally grounded.

Grounded conditions like those in (11) take astheir domain a) the well-formedness of underlyingand derived representations, b) specific ruleswhich are positively specified for one or moregrounded Grounded conditions.

While systems like {i, e, a, o, u}, {i, 6, a, f, u}are possible and common, {I, e, !, o, ~}, {i, 6, !, f,u} (! = [–lo, –ATR]), are extremely rare or notfound. This follows from the existence of thestrong grounded conditions that establish thathigh vowels are [+ATR], that [–ATR] vowels arenonhigh, that low vowels are [–ATR], and that[+ATR] vowels are nonlow.

As for rules, since grounding determines well-formedness and well-formedness holds also ofderived structures, a rule that creates an un-grounded structure is proscribed. An examplewould be a rule spreading [–ATR] on high vowelsbut not on mid and low vowels. Specific conditionscan hold of particular rules: a rule inserting[−ATR] only on low vowels is specifically subject toIf [–ATR] then [+LO].

4. Rules are parametric combinationsThe form of rules is addressed in chapter 4.

Archangeli and Pulleyblank indicate, quite cor-rectly, that there “have been few attempts at aconstrained theory of graphically representedrules” (285). By ‘graphic rules’ or ‘rule pictures’they mean standard notations using autoseg-ments, association relations depicted as lines, thecommon stricken line (=|=) for delinking and thediscontinuous association line (-----) for spread-ing, together with more or less mixed SPE nota-tion. It should be stressed though, that theproblem with graphic rules doesn’t lie in thegraphic character (a two dimensional representa-tion), but in the fact that they are notationallyvague and unconstrained.

“Parametric” rules can insert F-elements,delete them, and spread or delink them. The classof rules includes rules associating underlyinglyunlinked (“free”) elements, since as Archangeliand Pulleyblank argue, there are no universalassociation conventions, and their effect is takenup by individual particular rules. Parametric rueshave the following format (12):

(12)a. Argument Default Default

(some F-element)b. Parameters Function INSERT DELETE

c. Type PATH F-ELEMENT

d. Direction LEFT TO RIGHT RIGHT TO LEFT

e. Iteration ITERATIVE NONITERATIVE

f. Structure A-Structure NONE FREE

g. requirements T-Structure FREE NONE

h. Other A-Conditions (grounded conditions/context)k. requirements T-Conditions (grounded conditions/context)

The action of a rule is thus limited to insertion ordeletion (b) of an association line (spreading anddelinking, respectively), or of an F-element (c);this action can proceed iteratively in one or theother direction (d), and might require that theargument, the F-element, or the target be free (f,g) — otherwise it can be free or linked — and thatthe F-element or the target meet a set of variedconditions. This set of “other requirements” en-larges potentially the class of possible rules, sinceArchangeli and Pulleyblank do not restrict the


structure of contexts. In fact it can be considereda kind of “segmental residue “ (in the sense ofPoser (1982)), that will be hopefully incorporatedinto a sufficiently constrained theory by futureresearch.

As an illustration, consider the process rulethat inserts [+ATR] leftwards in Menomini (13).The argument, [+ATR], is subject to a contextualrequirement, namely the presence to its right of a[+ATR] and nonlow (it is nonlow because ofATR/LO = If [+ATR] then [–LO], which forces the[–LO] value). It is also subject to a requirement onthe target, namely that it obey also ATR/LO,hence it must be nonlow, and that it be long(bimoraic: µµ).

(13) Menomini [+ATR] SpreadArgument +ATR

Default NondefaultParameters Function INSERT

Type F-ELEMENT

Direction RIGHT TO LEFT

IterationStructure A-Structure NONE

requirements T-Structure FREE

Other A-Conditions ___<+ATR & ATR/LO>requirements T-Conditions ATR/LO, µµ

There is an important part of the book that I havenot reviewed, namely the detailed analysis ofmany phonological processes (normally ATRinsertion and spreading) on which the theoryrests. These analysis include 16 different lan-guages families, among them Niger-Congo (7languages) and Bantu (4 languages). Furtherresearch on ATR systems, and vowel systems ofthese languages, will be heavily influenced byGrounded phonology.

As Archangeli and Pulleyblank themselvespoint out, in particular at the beginning of Chap-ter 5, “Conclusion”, there are many open ques-tions left: the analysis is restricted to F-elements(especially to vowels), to the feature ATR, and toinsertion mechanisms; it does not determine howrules interact, and the sample of languages islimited. I think that this last mea culpa is totallyunjustified. Not many works can offer an analysiswhich is supported by a similarly broad andvaried empirical base. The shortcomings, or thepotential problems of the Grounded phonologytheory lie rather in the restriction of the evidenceto ATR insertion and spreading phenomena. Itwould come as no surprise if some of the theoreti-cal proposals put forward in the book would runinto difficulties when other features, other proc-esses, consonantal systems, syllable structure,stress, prosodic structure, etc. were considered.

In any case I think that, even if theoreticalchanges of various sorts have changed manyaspects of phonological theory since Groundedphonology’s elaboration, most of the proposals inthe book are sufficiently motivated and insightfulenough to remain valid when adequately incorpo-rated in current and future theoretical models.

ReferencesArchangeli, D. and D. Pulleyblank (1994).

Grounded Phonology. Cambridge, Mass.: MITPress.

Myers, S. (1987). Tone and the structure of wordsin Shona. Ph.D. dissertation, University ofMassachusetts, Amherst.

Poser, W. (1982). Phonological Representationsand Action-at-a-Distance. In H. van der Hulstand N. Smith (eds.) The Structure of Phono-logical Representations (Part II), Dordrecht:Foris.

Stevens, K.N., S.J. Keyser, and H. Kawasaki(1986). Towards a phonetic and phonologicaltheory of redundant features. In J.S. Perkelland D.H. Klatt, eds., Symposium on Invari-ance and Variability of Speech Processes,426−449. Hillsdale, N.J.: Lawrence ErlbaumAssociates.

GLOW 1978–1998

In this issue we devote some extra attention toGLOW. The GLOW (“Generative Linguistics inthe Old World”) Organization was founded 20years ago by Jan Koster, Henk van Riemsdijk andJean-Roger Vergnaud, “to further the study ofGenerative Grammar in Europe.”

Making sure that the GLOW colloquium isheld somewhere in Europe every year is just oneof the activities of the organization. Among otherthings, GLOW has also been actively involved inco-organizing summer schools (like the ones inSalzburg and Girona). The GLOW satellite confer-ence in Hyderabad, India, earlier this year, is thestart of yet another, Asian, branch of activities;the next Asian GLOW is planned for next year inJapan. (GLOW’s website is: <http://cwis.kub.nl/~fdl/research/gm/glow.)

In this issue of Glot International, we have areport of the GLOW in Hyderabad as well as areport of the 1998 GLOW colloquium (ordinair),which was hosted by Tilburg University, the homeinstitution of Henk van Riemsdijk, one of GLOW’sfounders. What follows is a short interview withhim.

“We’re a great field,but it is small and inconsiderable danger”A short interview withHenk van Riemsdijkby Lisa Cheng and RintSybesma

It looks like a pretty goodGLOW this year, but the number of people is notoverwhelming. There don’t seem to be many gradu-ate students either. In the past, especially as agraduate student in Europe, you needed a damngood reason for not going to GLOW. Now if youdon’t go to GLOW, you don’t go to GLOW. Clearlythe field is in disarray, it is in bad shape.

I don’t think you are right. There are simply moreconferences than in the past. Looking at GLOWthis year it is rather gratifying to see how youngthe public is. There are young generations, andthey do tend to come. But they come selectively.GLOW still is an important conference, it simplyis not exclusive anymore.

So the field is in good shape?

No, we are not in good shape and it sometimesfrightens me. Sometimes I am worried about thecontinuity of the field.

Intellectually it is in pretty decent shape. Ihave personal little qualms about whether mini-malism is on the right track, that sort of little

things. Listening to the talks here, I think thatthe level is quite high and there is lots of interest-ing stuff going on. So I am excited about that.

What I mean is: Is it in good shape from theperspective of logistics, the infrastructure, thepower structure? How easy would it be for theopposition or the “bad” world out there to get ridof us? From this perspective, things don’t lookthat good.

The “bad” world out there — you mean the “other”linguists?

Yes. Linguistics at large. The linguistic establish-ment. Looking back to the initial stages of GLOW,20 years ago, I had something like an idea thatthat was the start and we would slowly conquersome power basis in Europe, not to become theonly game in Europe, but at least one of the majorones. I think many generativists, especially youngones, tend to be quite mistaken about the extentto which that has been successful. The reason isthat we have made some progress, we have cent-ers where we attract students. These students geta good training, they go to conferences, the wholeworld in which they move is a generative world.But they often don’t realize that it is an extremelysmall world. Take a country like Germany. It isnot a poor country, they’ve got a lot of universities.But if you want to pin point the centers of genera-tive grammar, depending on your generosity, youend up with maybe half of a dozen.

Chomsky would kill me, if he heard this (andyou’re gonna write it up so he will kill me), but Iactually think that generative linguists have beendoing too little to entrench themselves in whatyou could call “the linguistic establishment”. I amlooking at this from the European perspective,where distinctions and diversities are perhapsbigger than what you would find in the UnitedStates. But, take Germany again, it is people indiscourse analysis and pragmatics and so on thatare in power.

Occasionally you see what that means. Someyears ago we had this big typological projectcalled EUROTYP, funded by the European Sci-ence Foundation. The persons who were asked toorganize it were Simon Dik and Ekkehard König.In the original plans that they had, they had ninesub- compartments in that project. They agreedthat one of those should be on generative gram-mar. Their actual plan was: we do complementa-tion, word order, themes like that, they went up to8, then they said: now we have to do somethingwith generative grammar. Let’s call the ninthproject the Principles and Parameters Approach.Then they called and asked me whether I woulddo it. I said no. It is ridiculous to have a Principlesand Parameters Approach next to a theme likeword order. I said: you can give me a topic and wework on it from our perspective just like the

CONFERENCESGLOW Special:

ReportsGLOW Hyderabad, India, January 20–23by Georges TsoulasGLOW Tilburg, April 15–18, including the workshopsby Ileana Paul, Marc van Oostendorp, and Jan-Wouter ZwartInterview with Henk van Riemsdijkby Lisa Cheng and Rint Sybesma

ConSOLE 6, December 15–17, University of Lisbon(by João Costa)

Conference reports Glot International, Volume 3, Issue 5, May 1998 19

typologists, the pragmaticists etc. work on theirtopics from their perspective. That is how itworked out, we did a group on Clitics.

This is just an illustration of how we areperceived by a big international organization likethe European Science Foundation in combinationwith the linguistic field at large.

Is there anything we can do about it?

One thing that we can do, which we certainlydon’t do enough, is step outside the trodden pathsof the generativism. I feel embarrassed pursuingthis matter now, because we were talking aboutGLOW, which is the exact opposite. But evenhere, for the workshop we decided to invite Gre-ville Corbett, who, for all intents and purposes, isnot a generativist. He is a very good morphologist.He is here and he seems to be enjoying himself.

Personally, instead of going to conferenceslike NELS these days, I go to non-generativeconferences in places like Prague. It has nothingto do with preaching the gospel, but it is to showthese people that you can interact with them in asignificant manner. I think that it has to be done.Sooner or later, we will have to be there, on a parwith the rest.

So you think that there is not enough interactionbetween the different types of linguists? Is it be-cause we have the wrong attitude? Too aloof?

It comes from both sides. It definitely comes fromboth sides. For example, in March, I went toPrague for a conference called Bridges and inter-faces. In Prague you still have these relics of thePrague school. I am not the youngest anymoremyself, but these people are a generation aboveme. They are 60ish, 70ish, and they want toperpetuate the Prague school. But the Pragueschool hasn’t led to any significant insights for 30years now. So that it makes no sense and you cansee that the younger generation would actuallylike to do other things, like generative grammar,or formal semantics. They would love to be part ofan international enterprise, which they don’t getto be because their superiors tend to be ratherfierce and control the jobs. So they have to doPrague school. So what you see is that the young-er generation give talks in which they sort ofdisguise what they really want to do under a lot ofPrague school terminology. And they are helped ifthere is some diversity in their conferences so youget to talk to them, and they get to talk to you.And all of a sudden you have some interaction.

Another area where that is true is typology.There are many people out there doing typology.In our tradition, lots of people also have big trapsopen about lots of typological generalizations. Butnobody knows what the others are doing, becausethey don’t go to each other’s conferences.

There are lots of sub-disciplines that we don’thave a quarrel with. We don’t interact. We don’tmeet those people at conferences, but they arethere. The thing is, they constitute a certainestablishment.

A number of years ago, I was spending asummer in Vienna. Vienna is sort of the center ofthe Societas Linguistica Europaea, our Europeansister organization. They publish the Folia Lin-guistica. If you ask me, it is piss poor in content,but it is run by some of the real Establishment ofLinguistics in Europe. I actually explored with thepeople there the possibility of joining up in oneway or another. They could certainly do with somerejuvenation at the content level and we could dowith some of their clout in adminstrative circles.Not too surprisingly, it didn’t work out. But maybesome grains from that discussion are still around,maybe in a few years, we should try again.

There is a lot of linguistic activity. GLOW, as aconference, may have lost its focal point function,but the field seems pretty lively. There are manyconferences and workshops, some on very specifictopics. Or is there a downside to that too? Thateven generativists don’t talk to eachother anymore?

There are a lot of specialized conferences and ofcourse, most of that is good. Where I am a littlehesitant is that there are these ad hoc workshopson very specific topics — interesting and adequatetopics, there is nothing wrong with that — butsetting them up, getting the money, organizingthem, getting people to come, and so on, consumesa lot of time and energy. I feel that too much timeand energy goes into that type of activity, as ifpeople really lack the imagination to think ofother things that they might put their energy intofrom which the field may benefit equally well oreven more. If anybody that wants some sugges-tions as to what else they could do, there are lotsof things to suggest. I was talking about theestablishment angle. Certainly people couldinvest more in that. Get ourselves entrenched inthe echelons that matter.

Another thing is related to the fact thatlinguistics, not only generative linguistics, butlinguistics at large is a poor field. And it is arelatively small field. The consequence of that isthat the infrastructure of the field is quite bad.We have journals and stuff like that. But that isabout it. In my university, economics is dominantso I know that for most economics journals youcan get the contents on the screen of your compu-ter any time. These are projects sponsored byKluwer, Elsevier, the big companies. But forlinguistics? Zero. Well, there’s Glot International,it’s a nice exception, but it is you guys who did it.Not Kluwer or CUP or one of the other big guys.Okay, there is some movement. I understand thatMIT Press is going to make Linguistic Inquiryelectronically available. And there is this abstractproject which is done by Blackwell, that too is astep in the right direction.

What I’m trying to say is that, on the whole,we are badly organized. The number of publica-tions is overwhelming but, clearly, Kluwer is notgoing to think of doing anything big for us them-selves. So here is something that people couldtake the initiative for. Instead of organizinganother ad hoc workshop, someone might actuallytake this particular thing on and start askingaround, pushing, bitching with the publishers,stuff like that. That is just one example. I thinkthat there is a lot more that could be done.

And GLOW is going to do it?

Sometimes I do this Gedanken-experiment: sup-pose GLOW hadn’t been created 20 years ago,would the situation now be such that peoplewould spontaneously create it? I find that ex-tremely hard to answer, partly also becausegenerally speaking I’m in favor of anarchisticinterpretations of how science progresses. On theother hand, I also think that people shape eventsto a certain extent. If you have a field full ofMachers, the field is different from a field that isfull of chaotic, unorganized people.

We have inherited a great field, we have beenreally lucky. The field is, to put it very modestly,interesting and exciting enough for us to want tocontinue at least at the present level (maybe notat a much higher level because we cannot findemployment for the people anyway). I have al-ways found it a privilege to be part of this enter-prise. But I think that people underestimate thedanger in which we are in that it might get lost orcollapse somehow. GLOW can play a role in this,but it can’t do it on its own.

GLOW’s role ought to be: looking after thewelfare of the field, particularly in Europe, and toa certain extent elsewhere in the world. I thinkthat’s a good purpose. It is a purpose which is asacutely necessary now as it has ever been before.

GLOW Reports

Extraordinary GLOW 1998Central Institute for English and Foreign Lan-guages,Hyderabad, India, January 20–23by Georges Tsoulas

Usually, GLOW conferences are accompanied by anumber, this one isn’t, usually GLOW conferencesare held in Europe, this one wasn’t. The Hydera-bad GLOW, was an “Extraordinary” venue, had itbeen numbered it would have been GLOW 21, butas Henk van Riemsdijk carefully explained in hisinaugural address GLOW counting is a trickybusiness (GLOW’s 20th birthday will be celebrat-ed during the Tilburg meeting). The idea of anAsian GLOW, or a GLOW in “the very old world”as K.A. Jayaseelan, the GLOW convenor, put it, ora GLOW in “the really old world” according toChomsky’s message that was read during theopening session, has a history of some four years.One can only applaud to the initiative and offerthe highest praise to those involved in the organi-sation to make it possible. With an unfailingsense of dedication they did an enormous amountof work. This GLOW was a truly “Extraordinary”conference in all the senses of the word.

The conference started on the 19th of Janu-ary with a lengthy inaugural session where the80-odd participants had the chance to hear thehistory of the organisation of the conference by K.A. Jayaseelan, and the history of the GLOWorganisation by one of its founding members,Henk van Riemsdijk.

In a conference report, it is somewhat unusu-al to focus on the inaugural addresses — and withgood reason as they are usually innocuous anddevoid of interest. In this case we hope to beexcused for staying a bit longer on it. One of theissues that was addressed by K. A Jayaseelan andH. van Riemsdijk was the future of the AsianGLOW. Undoubtedly, one needs to recognise theindisputable, and undisputed, linguistic richnessof the Asian continent, the number of activetheoretical linguists but also the sometimesdifficult economic conditions and the rather wideeconomic gap separating countries. The latter factas well as the geographical hugeness of the conti-nent were given quite some attention in theinaugural addresses. they seemed to be presentedas obstacles to be overcome if the enterprise of anannual GLOW event in Asia were to succeed (theprecise example given by H. van Riemsdijk was ofa GLOW venue in Japan or Korea and the diffi-culties that linguists from say India would have toface in meeting the travel expenses etc...). Howev-er, given the success of the present venue it seemsthat these issues should not be considered insu-perable problems, although at some point or otherone would have to address them. Perhaps this isnot the place for suggestions but on one of theabove counts, how does one address the “problem”that Asia is about 4 times the size of Europe? Or,on the financial side, everyone who has talked tolocal GLOW organisers knows that they had toshow extreme ingenuity and inventiveness inorder to meet the costs. The same would have tohappen in Asia. Despite these questions, it wouldbe a great pity, we think, if the Asian GLOW wasa one off venue. Fortunately, there seems to be anoffer for the organisation of the next GLOW inAsia from Nanzan University in Japan. One canonly wish that the above questions will findappropriate answers and that the second AsianGLOW will soon be announced.

The actual conference started on the 20th ofJanuary with a talk by Andrew Simpson on“VP-Final Modals and Pied Piping in East andSouth East Asia”. In this talk Simpson discussedthe distribution of certain modal verbs in a varie-ty of East and South East Asian languages suchas Thai, Cambodian, Vietnamese, Cantonese etc.concentrating mainly on contemporary Thai. Hismain focus was the distribution of the modal verbdai in that language. This verb has a curiousdistribution: all modal verbs in Thai occur in


normal pre-VP positions with the only exceptionof dai which occurs postverbally. Simpson arguedthat this raises important problems for Cinque’sversion of the Universal Base Hypothesis ofclause structure. Technical points aside, hisanalysis involved raising of a whole part of theclause from a fully regular base structure inaccord with the Universal Base Hypothesis.Finally he suggested that an extensive process ofBorrowing had taken place, as this paradigm,although very marked, regularly occurred in thelanguages of the region in which the constructionwas at various stages of development and reanal-ysis.

George Tsoulas and Kook-Hee Gill’s talk,“Multiple Subjects and Multiple Specifiers: ANecessary Connection?”, focused on the multipleNominative construction in Korean and put intoquestion the widely held assumption that Koreanand Japanese pattern in the same way withrespect to their phrase structure properties ac-counting for these constructions. Their evidencefrom scope and binding facts showed that, at leastfor the cases they examined the C-commandrelations between the two subjects must be as-sumed to be different in the two languages. Theyfinally gave a negative answer to the question inthe title of the talk, as the Korean data pointed toan analysis which made no use of the possibilityof projecting multiple specifiers.

In his presentation entitled “Aspect andEvent Structure in Vedic”, Paul Kiparsky ad-dressed the issue of the proper analysis of thepast tenses in Vedic and the problems it raises forthe “two dimensional” theory of tense and aspectas inspired by Reichenbach’s classic work. Heargued that the set of primitives provided for inthis theory is not only not sufficient to provide aunitary representation for the aorist and theperfect in Vedic, but the primitives do not evenallow one to distinguish between the two tenses.As a remedy he proposed two refinements to theReichenbachian approach: first, the specification,as part of the representation of certain temporaland aspectual categories, of the mapping of thepredicate’s event structure into the parametersthat define temporal relations; the second refine-ment is simply the blocking effect from morpho-logical theory. The default readings of the tensesare defeated by morphologically specified tenses.Finally Kiparsky argued that the distinctionsmorphologically manifested in Sanskrit are cov-ertly present in the syntax of English.

Two talks in the conference addressed ques-tions arising from the analysis of Kannada. Jef-frey Lidz, in his talk “Causation and Reflexivityin Kannada: Evidence for Post-Syntactic Morphol-ogy” concentrated on a valency puzzle in Kanna-da, and argued that it provided evidence for amodel of grammar in which syntactic structure fora verb is generated independently of its lexicalmeaning, the grammaticality of the sentencedepending on the satisfaction of some correspond-ence conditions between the syntax and thelexicon. The puzzle comes from the distribution ofthe causative and reflexive morphemes. In Kan-nada, lexically causative verbs cannot take anexternal causative morpheme whereas verbs thatare non-causative in the lexicon can. However,both types of verbs have an anticausative variantmarked with the reflexive morpheme. Lidz wasled to propose that the reflexive has no semanticcontent in itself: it is simply the spell out of ahead in a particular configuration. The secondtalk on Kannada was by R. Amritavalli: “Kanna-da Clause Structure”. Amritavalli argued thatTense and Aspect in Kannada are distinguishedfunctionally rather than categorially, finiteness inKannada being located in the Agr, Neg, or Modalheads whose aspectual complements may beinterpreted as Tense depending on the context.

Tara Mohanan, in her paper “MorphemeInternal LF Representations”, (read by K.P Mo-hanan), argued that the GB model of the gram-mar should be modified in order to permit accessto morpheme internal semantic representations.The alteration of the model she suggested

amounts to the recognition of a lexical LF mediat-ing between LCSs and Argument structure on theone hand and phrasal LF on the other. She moti-vated her lexical decomposition account by avariety of constructions in English (among others,scope of almost, word internal negation etc., andtheir counterparts in Malayalam and Mandarin).

Still on the Malayalam front K. A. Jayasee-lan presented a fascinating set of data and anequally fascinating account of the relation be-tween question formation and quantifier forma-tion, in his talk entitled “Questions Quantifiersand Zero Derivation in Malayalam”.

The invited speaker of the conference wasMamoru Saito who gave a lecture in which hepresented joint work with Hiroto Hoshi on theJapanese light verb construction. Their analysiswas complex and intriguing and it is not possibleto do justice to it in the limited space alotted tome here. The main thrust of their argument wasthat the light verb construction in Japanese isbest analysed as an instance of LF incorporation.They used Grimshaw & Mester’s (1988) analysisas guide for their account of the Japanese con-struction. The LF incorporation takes place fortheta role assignment where theta roles areconsidered formal features on the assigners andmore generally they are associated with categorialfeatures. The more radical conclusion that Saitoand Hoshi drew from their analysis is that theyconstrued it as evidence against the existence ofA-chains and A-traces which they indeed arguedshould be dispensed with. Saito and Hoshi arguedthat the LF-incorporation analysis presentedevidence for the minimalist model. However, onewonders how this evidence should be interpretedwhen taken in conjunction with their conjectureconcerning the elimination of A-chains. Moreprecisely one wonders what should happen to theheads of A-chains once the latter have been elimi-nated. The options, of course, are initial mergingat their spell-out position or traceless movement.It seems rather pointless to have the latter optionbut the consequence of the first one is clearly thatit lends support to a representational rather thana purely derivational model, or perhaps a move-mentless mono-stratal model. These issues wereleft open but they will certainly be exciting futureepisodes of the theory Mamoru Saito outlined.

Four talks addressed issues in optimalitytheory: the three phonology talks “Paradigmaticinfixation in Miskito” by Vivian Lin, “AbkhazMabkhaz: m-reduplication in Abkhaz, weightlesssyllables, and base reduplicant correspondence”by Benjamin Bruening and “Conflict Resolutionin Optimality Theory: The Case of The Anti-Faithfullness Requirement in ReduplicativeMorphology” by K.G. Vijayakrishnan and finallyK.P. Mohanan’s talk “Crosslinguistic Variabilityof Laws and Their Interaction: Parameter Settingor Constraint Parametrisation?” who focused onissues of reflexive binding in Malayalam.

A talk focusing on non-Asian languages wasHenk van Riemsdijk’s “On Light Nouns” whopresented an analysis of the partitive constructionin continental Germanic, based on the line ofthinking that assumes functional categories to benot independently specified bundles of featuresbut rather defined in terms of the categorialfeatures on their lexical hosts.

Other presentations in the main conferenceincluded “Ability Modals and their Actuality Entail-ments” by Rajesh Bhatt, “Only in Hindi” by Shra-van Vasishth, “Ergativity and Ergative Splits inAsian Languages” by Gloria Gocchi, “NO Raising”by Masanori Nakamura, “Compounding in Hindiand Bangla” by Rajendra Singh and ProbalDasgupta, “Light Verbs, Empty Prepositions andZero Derivation in Malayalam” and “Rightward Ho!A critical Discussion of Antisymmetry, RightwardMovement, and the Syntax/Pragmatics Interface”by Steven Schaufele.

The theme of the GLOW workshop, whichwas jointly sponsored by CIEFL and the Depart-ment of Linguistics of the University of Trondhe-im, was “Verb Typology of African and AsianLanguages”. There were seven regular papers and

one invited speaker, Chris Collins (who spoke on“Pluractional Verbs in Hoa”). The other presenta-tions included “A typology of serial verb construc-tions in Dagaare and Cantonese” by K.K. Lukeand Adams Bodomo, “Japanese Adjunct ‘Control’verbs” by Hironobu Hosoi, “In Support of Agr as aFunctional Category” by K.V. Subbarao, “LexicalLogical Form and Verb Typology: A Case Studyfrom Malayalam” by Tara and K. P Mohanan,“The structure of Maithili verb Morphology” byYogendra P. Yadava, “A Typology of Counterfactu-al Marking in Modern Indo-Aryan Languages” byRajesh Bhatt and “The Four Copulas in Odia” byBibhuti Bhusan Mahaparta.

Trying to present a full report on such aconference in such a limited space is obviously anear-impossible task. I hope that I have managedto convey the deep sense of satisfaction at alllevels, intellectually and in personal interactionalike, felt by all the participants of this confer-ence. There is truly a lot which the organisersshould be commended for. It was a pity that mostof the European and American linguists in theinitial program failed to turn up, which not onlycaused a lot of last minute alterations in theprogramme but some resentment as well (thoughby no means with our gentle hosts), as one cameto realise that the support given to this enterpriseby many European and American researchersfailed to materialise in the end. (This does notapply to everyone and the importance of thesupport by the GLOW organization must not beunderestimated.)

Finally let me just reiterate everyone’s wishthat the Asian GLOW continues after this year. Itprovided a prestigious and geographically accessi-ble forum for discussion and exchange of ideas forAsian linguists — among themselves and withthose who had come from other places.

GLOW 21, Tilburg, 15–17 April 1998GLOW Workshops, Tilburg, 18 April 1998by Ileana Paul, Marc van Oostendorp, andJan-Wouter Zwart

GLOW ColloquiumThe following papers were presented at the 21thGLOW Colloquium, organized at the University ofTilburg, April 15–17, 1998, featuring ‘features’ asthe colloquium topic.

Paolo Acquaviva (University of Venice,‘Uniform Lexicalization: Deriving Spell-Outwithout [±strong] Features’) addressed the issueof how to describe word order variation withoutusing the notion of [±strong] features. Adoptingthe morphology-after-syntax approach, he pro-posed that a Principle of Lexicalization (‘if afeature bundle F under the head X° is matched bya unit of autonomous morphology, then F is lexi-calized under X°’, where a unit of autonomousmorphology is any category by reference to whichmorphological generalizations can be made)accounts for which position in a phrase structuretree gets spelled out.

Artemis Alexiadou and Elena Anag-nostopoulou (ZAS Berlin and ZAS Berlin/Uni-versity of Tilburg, ‘Restrictions on ArgumentPlacement: Adjacency on Case-Assignment and anAmbiguity with Attract-F’) aimed at establishingthe following generalization: Subject-inversionwhere the subject is VP-internal is not possiblewhen there is an object inside the VP, unless thisobject is a PP. Evidence for this generalizationwas taken from the syntactic conditions on theavailability of VP-internal subjects across lan-guages. The proposed generalization was shownto reduce to a constraint against covert Case-driven movement out of VP of both the subjectand the object.

Maya Arad (UCL/University of Geneva,‘Stativity, Eventiveness, Agentivity: Object Expe-riencers Revisited’) presented a highly enlighten-ing analysis of object experiencer verbs likefrighten, amuse, concern, delight, worry, etc. Sheshowed that these verbs may have three readings(agentive, eventive, and stative), and that only thestative reading is associated with ‘psych proper-


ties’. Moreover, any construction can be interpret-ed as a psych-verb construction, once certainconditions are met (having to do with the animacyof the experiencer and the incapability of physicalaction of the subject, as in an idea hit me). Aradshowed that what is characteristic of the psych-interpretation of verbs is that the perception ofthe stimulus and the resulting mental state arenot ordered in sequence but in parallel. Givencertain assumptions about the mapping of thetime path onto phrase structure, it follows that inpsych-constructions both the stimulus and theexperiencer are generated in specifier positions of(shell structured) VPs (rather than vP). Theanalysis argues against earlier inversion accountsof Object Experiencer verbs.

Jonathan David Bobaljik (Harvard/McGill,‘The Autonomy of Syntax and the Typology ofMovement’) discussed consequences of the shift inexplaining word order variation from differenttiming of movements (the overt vs. covert distinc-tion) to different spell-outs of copies. He wasespecially interested in cases where the highercopy was irrelevant to both spell-out (PF) andinterpretation (LF), mentioning as candidatesexpletive association and quantifier float.

Richard Campbell (Oakland University,‘Case, pro, and definiteness effects’) claimed thatCase features are not features of noun phrasesbut of pronominals. Common noun phrases cancontain pro, whose Case features must be identi-fied (hence the Case Filter). Nominals not as-signed Case contain a non-pronominal emptycategory different from pro, and must be indefi-nite existentials.

Rose-Marie Déchaine and Mireille Trem-blay (University of British Columbia/MIT andQueen’s, ‘On Category Features’) presented auniversally acclaimed paper on the conceptualnecessities of lexical and functional categories,showing how certain functional categories requirecertain lexical categories in their complement, butnot the other way around. Using data from a widevariety of languages, they reached nontrivialconclusions—for instance, that Aspect requires V,but T does not (it does not have V-features). Theevidence leading to this conclusion is that stativepredicates can be any category, but eventivepredicates (eventivity being the result of aspectshift) are always spelled out as V. The set ofconceptually necessary categories proposed in-cludes D (which has N-features), N, T, Aspect(which has V-features), V, and P. Lexical categoryfeatures such as [±N,±V] are epiphenomenal.These features are not needed to define catego-ries, which are basically labels for syntacticobjects existing only in the linguist’s eye. There isno inherent distinction between lexical and func-tional categories. Both types of categories areconfigurationally defined. Functional categories inaddition have semantic features which must bechecked. In the lexical domain, Merge is notdriven by valency, but by predication. The authorsconclude that their approach favors a representa-tional architecture of grammar.

Hamida Demirdache and Myriam Uribe-Etxebarria (University of British Columbia andUniversity of the Basque Country/Basque Centerfor Language Research, ‘On the Projection ofTemporal Structure’) proposed that the grammarof Tense and Aspect be reduced to the same set ofuniversal substantive and structural primitives.Tense and Aspect are defined as dyadic predicatesof spatiotemporal ordering. Utterance Time,Assertion Time, and Event Time are representedas specifiers of the heads Tense, Aspect, and V,respectively, of which the first two can be instanti-ated as expressing spatiotemporal ordering rela-tions (inclusion, subsequence, precedence). Thesystem was demonstrated in a range of types oftenses and aspects from various languages.

Anna-Maria Di Sciullo (UQAM, ‘Featuresand Asymmetrical Relations in MorphologicalObjects’) focused on the features computed underthe word-level, and defended the hypothesis thattheir associations in ‘morphological objects’ andtheir visibility at the interfaces is based on the

satisfaction of specific local asymmetrical rela-tions (the complement, adjunct, and specifierrelation). Morphological objects are binarybranching adjunct-head configurations, and theonly features supported by morphological objectsare the ones associated with the adjunct relation:the iterative feature and the delimiting feature.

Marcel den Dikken and Rint Sybesma(University of Tilburg and University of Leiden,‘Take serials light up the middle’) took up theanalysis of transitive constructions in Chomsky(1995, chapter 4) involving a layered structure ofa light vP and a lexical VP and applied it to takeserial verb constructions and middle construc-tions. Middle constructions were argued to lackthe vP, in contrast with take serial verb construc-tions, where v is occupied by take. They showedthat the lower verb in serial verb constructionslooks more like a middle verb than like a genuinetransitive verb.

Gorka Elordieta (UCLA/Basque Center ofLanguage Research, ‘The phonological import ofsyntactic features’) claimed that prosodic phonolo-gy needs to be supplemented with another set ofsyntactic domains for phonological processes. Twoadjacent syntactic heads which stand in a feature-checking relation can form such a domain. Thedata on which Elordieta based his claims werevowel assimilation in Basque and liaison in aninformal register of French. In the talk he showedhow standard assumptions of prosodic phonologyare insufficient: the phenomena take place be-tween a lexical head and certain functional heads,but not others. According to Elordieta the reasonbehind the asymmetry is that there is no syntacticfeature checking going on in the latter cases,whereas in the former, there is.

Ricardo Etxepare (Basque Center of Lan-guage Research, ‘Paratactic Dependents andCovert Merge’) discussed embedded clauses withroot properties, arguing that these involve mergerin the covert syntax of an abstract performativeelement.

Maria Rita Manzini and Leonardo M.Savoia (University College London/University ofFlorence and University of Florence, ‘ObjectInflection’) presented and analyzed a wide varietyof data from Italian dialects involving objectcliticization. They argued that clitics are orderedin an aspectual and an inflectional string. Order-ing parameters do not reflect the underlying orderof the string, but only the circumstance thatclitics can in general be realized either in theiraspectual or in their inflectional positions.

Ana Maria Martins (University of Lisbon,‘On the need of underspecified features in syntax.Polarity as a case study’) was concerned withmotivating in syntactic terms a typology of polari-ty items well established in the semantic litera-ture, by adopting a theory of features integratingthe notion of underspecification. The strong/weakdistinction familiar from the semantics literatureis described in terms of specified, underspecified,and variable-underspecified negative features (cf.Rooryck in Probus 1994).

Gereon Müller (University of Stuttgart,‘Feature Strengthening’) discussed the problemthat remnant topicalization in German maycreate unbound traces, but not unbound interme-diate traces. This problem was discussed in thecontext of the more general question of how todescribe successive cyclic wh-movement in termsof feature checking (it involves feature strenght-ening of the wh-features of the embedded CP,forcing extraposition).

Léa Nash and Alain Rouveret (Universityof Paris VIII, ‘Feature Fission and the Syntax ofArgument DPs and Clitics’) applied their theory offeature checking through feature fission (creationof a higher functional projection for an additionalfeature checking operation) to the licensing ofarguments either in or outside VP — in particu-lar, argument clitics. The analysis takes argumentfeatures to be formal features, and adopts Alexia-dou & Anagnostopoulou’s (above) conjecture thatthe internal and external argument of the verbcannot both be realized inside VP at Spell-Out.

Jairo Nunes and Eduardo Raposo (Uni-camp and UCSB, ‘Portuguese Inflected Infiniti-vals and Configurations for Feature Checking’)discussed the phenomenon of Portuguese, wherethe appearance of inflection on infinitives isconditioned by the presence of an adjective withagreement features in the infinitive’s complement.They argued that this supports a description ofagreement patterns without making use of Agree-ment Phrases, and discussed consequences forchecking theory.

Marc van Oostendorp (University of Lei-den, ‘Phonological Feature Domains and theContent of Epenthetic Vowels’) presented argu-ments that we need to take into account themorphological affiliation of phonological segmentsin order to account for root-controlled vowelharmony in Turkish and for the exceptionalbehavior of epenthetic vowels in both Turkish andIcelandic.

Javier Ormazabal and Juan Romero(University of the Basque Country/Basque Centerfor Language Research and Instituto Universitar-io Ortega y Gasset, ‘Attract-F: A Case AgainstCase’) proposed to replace Case features (whichare present for theory-internal reasons only) byanimacy features (which are interpretable) andused a discussion of the properties of double objectconstructions to show that the feature relevant inobject movement is animacy rather than Case. Inconjunction with the Minimal Link Condition,this explains the ordering restrictions on clitics inRomance (the Person-Case constraint of Bonet1991).

Keren Rice and Trisha Causley (Universi-ty of Toronto, ‘Asymmetries in Featural Marked-ness: Place of Articulation’) argued that there aretwo problems connected to the markedness ofplace features. On the one hand, there is a greatdeal of cross-linguistic consistency as to whatpatterns as unmarked phonologically. But on theother hand, it is not always the same feature thatpatterns as unmarked cross-linguistically. Riceand Causley solved this by presenting a theory ofplace both for the consonantal and for the vocalicdomain that involved different types of unmark-edness. For instance, central vowels and velarconsonants are represented by an empty placenode. This representation is quite literally un-marked in the sense that it involves a minimalstructure. For this reason central vowels andvelar consonants count as unmarked in thesesegments. On the other hand, certain languagesdo not allow these segments, because they disal-low completely empty place nodes. Rice andCausley presented their analysis in terms ofOptimality Theory but it is reasonably clear thattheir approach could be transferred without toomany problems to other theories of phonology.This does not mean that their proposal will beuncontroversial because it involves a few repre-sentational oddities that may not be acceptable toall scholars: e.g. the features [Labial] and [Dorsal]are subsumed under a class node [Peripheral].

Elizabeth Ritter and Heidi Harley (Uni-versity of Calgary and University of Pennsylva-nia, ‘Sorting out you, me, and the rest of theworld: A feature-geometric analysis of person andnumber’) presented a feature geometry of pronom-inal elements and agreement features involvingthe notions Participant, Addressee, and Speaker(for person features), and Individuation, Group,Minimal, and Augmented (for number features).The geometric structure of these morphosyntacticfeatures is provided by Universal Grammar. Thepresentation involved an analysis of pronominalparadigms from a variety of languages.

GLOW ’98 Syntax Workshop ‘AgreementSystems’The workshop began with Rita Manzini andLeonardo Savoia’s paper ‘Subject Verb andExpletive Associate Agreement.’ Noting parallelsbetween expletive-argument constructions andclitics, they proposed an aspectually-based ac-count of agreement. Thematic roles are aspectualfeatures that are generated in functional catego-


ries and must move to a DP position to checkthese features.

Øystein Vangsnes (‘The role of genderagreement in Scandinavian: a study of possessiveconstructions in four Swedish dialects’) pointedout that only in those dialects that do not markagreement within the possessed NP can aprenominal possessor and a definite article cooc-cur. He argued that this restriction is due to thepossibility of checking gender agreement onlyonce within DP.

Drawing on a wide range of languages, invit-ed speaker Greville Corbett noted the frequentfeature mismatches and other irregularities inagreement systems (‘Agreement systems: featuremismatches and the pervasiveness of agreementchoices’).

Maria Teresa Guasti and Luigi Rizzi(‘Nonagreeing do in child English and eht expres-sion of Agr’) focused on data from first languageacquisition to show that tense and agreement aredistinct syntactic positions. Although agreementis optional at a certain stage in English childgrammar, the optionality disappears in interroga-tive contexts. They argued for a principle and aparameter that connect overt movement andmorphological agreement. Finally, they addressedthe structural nature of uninflected clauses inearly English.

Henry Davis discussed ‘Person splits, phi-features and temporal architecture.’ He showedthat unrelated languages manifest a distinctionbetween 1/2 subjects and 3 subjects and relatedthese differences to structural hierarchy. Moreo-ver, he argued that person features are connectedwith certain temporal projections in a clause andreduced the EPP to the necessity for T to berelated to phi-features.

Presenting data from Zuni, a linguistic iso-late, Lynn Nichols (‘Agreement and locality inlexical structure’) showed that what might appearto be an irregular agreement pattern is in fact aresult of complex predicate formation via P incor-poration. Agreement is sensitive to argumentstructure and is disrupted by lexical derivation ina well-defined class of verbs.

Invited speaker Maria Bittner addressedthe well-known link between agreement anddefiniteness (‘On agreement and definiteness’).She argued for a loss of functional categories toaccount for the gradation in semantic individua-tion that arises with such processes as antipas-sive, noun incorporation, “stripping” and lack ofperson agreement marking. Lexical categories areassumed to be basic predicates, while functionalcategories introduce existential presupposition byadding a pro.

GLOW ’98 Phonology Workshop ‘Phonologi-cal Opacity’The emphasis of the GLOW colloquium sessionstraditionally is on syntax rather than phonology.This year’s conference featured two lectures onphonological structure and one on the interfacebetween phonology and syntax, reported on above.All three gave arguments in favor of an enrichedtheory of phonological representations ratherthan concentrating research on the issue of rulesversus constraints.

The phonology workshop on opacity on theother hand went straight into the heart of thediscussion between proponents of bistratal Opti-mality Theory and the defendants of polystratalgenerative phonology based on extrinsicallyordered rules. Since the two extremes were bothrepresented, the discussion was quite lively. AsMorris Halle pointed out during this day, thephenomenon of opacity was one of the drivingforces for setting up the mechanism of extrinsicrule ordering. On the other hand, the existence ofthese phenomena pose serious problems for Opti-mality Theory, because their analysis appears tocrucially involve some additional levels of repre-sentation beyond the input and the output. Sever-al ways of dealing with these phenomena havebeen proposed within the OT literature, and itwas one of the purposes of this day to evaluate

those proposals.The contribution of David Odden was called

‘Cyclicity and Counterfeeding in Kimatuumbi’,but Odden was quick to point out that the paperdidn’t deal exclusively with either cyclicity, coun-terfeeding or Kimatuumbi. His data were drawnfrom several (Bantu) languages and he showedthat an OT analysis of these facts would involve aquite heavy machinery, including SympathyTheory, two-level constraints and output-outputrelations. Neither of these three ‘supplements’ toOT on its own is sufficient to handle all the rele-vant facts. An analysis of Bantu in terms of ex-trinsically ordered rules is less problematic, butMr. Odden seemed hesitant to draw the conclu-sion out of this that such an analysis would there-fore be preferable. From an empirical point ofview, both theories seem to be equally equipped tohandle the relevant facts.

Marc van Oostendorp in his talk ‘Non-derivational opacity in allomorph selection’claimed that there are certain types of opacitywhich can be handled in a sympathy approach butnot in a rule-based approach. There are cases inDutch where a form behaves phonologically as ifit has chosen the unmarked allomorph, eventhough on the surface the form displays themarked allomorph of a given affix.

Ronald Sprouse unfortunately was not ableto be physically present at the workshop. Hispaper was read by Orhan Orgun. Mr. Sprousespaper ‘Enriched Input Sets as a source of opacityin Optimality’ presented a tristratal version ofOptimality Theory, in which we have two passesof generation. The first generator function is onlyable to add material to the input, not to deleteanything (this function is therefore comparable tothe original definition of Gen in Prince and Smo-lensky 1993). The second function can apply allsorts of operations to the output of the first Gen.Furthermore there are sets of faithfulness con-straints relating input, enriched input and out-put. Mr. Sprouse demonstrated that thisextension of Optimality Theory (which is morerestricted than Sympathy Theory) can handle thedata of Turkish epenthesis and deletion and thecomplicated facts of Yowlumne (= Yawelmani).

Orhan Orguns own paper ‘PhonologicalOpacity and Synchronically Arbitrary Alterna-tions: What Two-Level Phonology can contributeto OT’ also proposed to extend the Gen function ofOptimality Theory, but in a completely differentmanner. Mr. Orgun tried to give an articulated,language-specific implementation of the genera-tor, formulated in terms of the two-level theory ofphonology that Mr. Orgun developed in his thesis,based on the work of Koskenniemi. Mr. Orgunproposed that the generator function of a givenlanguage may only generate those candidates thatconform to this two-level system. He claimed thatthis is only a minimal extension to the theory (wemay need language-specific rules anyway) andthat the two-level system is rich enough to ac-count for complex cases of opacity such as theones in Hebrew, Icelandic, Yowlumne andKashaya. The question then arises what thedivision of labor is between the two-level genera-tor and the evaluator function; Mr. Orgun left thisquestion open to further research.

The talk by Ania Lubowicz was called‘Derived Environment Effects in OT’. Ms. Lubow-icz argued that Derived Environments are bydefinition in most cases also instances of opacity:certain segments do not undergo a given phono-logical process, because they do not occur in aderived context. Ms. Lubowicz analysed this asthe result of constraint conjunction; a faithfulnessconstraint is conjoined with a well-formednessconstraint in the domain of a segment. If thefaithfulness constraint is violated in a given(derived) context, we are not allowed to violate thewell-formedness constraint at the same time.Therefore, the segments in derived environmentsare likely to undergo a certain change in order tobecome more well-formed. Yet segments in under-ived environments do not have to violate the samefaithfulness constraint, and therefore it becomes

irrelevant wether or not they violate the well-formedness constraint: they can stay the way theyare. Ms. Lubowicz showed that this analysisworks both for phonologically and for morphologi-cally derived environment effects in languageslike Polish and Slovak.

Bill Idsardi studied some formal properties of‘Opacity, sympathy and derivations’. The emphasiswas on the treatment of opacity within sympathytheory. Taking the analysis of Tiberian Hebrewvowel epenthesis and glottal stop deletion, andIcelandic j-deletion and vowel epenthesis as astarting point, Mr. Idsardi showed that currentSympathy accounts are less parsimonious (thereare more constraints/rules at stake), more abstract(the analysis involves a larger numer of relevantbut non-surfacing representations), and moredelicate (changing the grammar slightly may havedramatic consequences for candidates which seemunaffected by the basic analysis) than the availablerule-based accounts. Furthermore, SympathyTheory also is not more restricted than rule-basedtheory according to Mr. Idsardi: also instances ofthe so-called Duke-of-York Gambit may be analyzedusing sympathy. The latter point was also madeearler during the day by David Odden.

The last talk of the workshop was MorrisHalle’s ‘English Stress 1968–1998’. In this talk,Mr. Halle gave an overview of the developments ofthe theory of English stress since the appearanceof The Sound Pattern of English. The main devel-opment is an enlarged insight into the representa-tion of phonological structure. The idea ofgrouping segments into syllables, projectingsyllables onto an independent stress plane andenriching that plane with a limited number ofboundary symbols, allows us to formulate therelevant rules in a way that is maximally simple.The theory thus established is supplemented witha small number of exception markings. Mr. Hallefurthermore adopted Burzio’s proposal that cer-tain word-final ‘super-light’ syllables such as -ureand -y do not get projected to the stress plane atall. This further simplified the rule-based accountof English word-level stress.

ConSOLE 6, December 15–17, Univer-sity of Lisbonby João Costa

Last year’s issue of ConSOLE (Conference of theStudent Organization of Linguistics in Europe)was hosted by the Department of General Lin-guistics of the University of Lisbon.

As usual, the event was a good opportunityfor graduate students to exchange ideas abouttheir ongoing work, and to divulge in their re-sults. There were 24 presentations, which enabledsome discussion on a wide set of topics. In thisreport, I present a very brief summary of thesubject of each talk, just to make you anxious forthe proceedings!

Olaf Koeneman (OTS/Utrecht) opened theconference presenting the results of joint workwith Ad Neeleman (‘Transitive Expletive Con-structions’). Developing proposals by Williams(1980) on predication and a non-universal hypoth-esis on clause structure, they claimed that Transi-tive Expletive Constructions are only possible inlanguages that project CP. This projection createsa position for expletives to be inserted which doesnot violate unicity of predication, since only inthis case is the expletive outside VP’s predication-al domain. Covering a wide range of Germaniclanguages, they were able to derive Vikner’s(1995) generalization that Transitive ExpletiveConstructions are dependent on V2.

Elizabeth McCoy (University of York,‘Heads’) proposed an analysis for the followingword-order variation fact in Irish non-finiteclauses. In Irish, non-finite clauses are SOV.However, in Southern Irish, there may be only oneovert argument before V. If the two are overt, theemergent word order is SVO. Her explanation wasbased on a dual analysis for the particle a in Irish:in Northern Irish, it heads AgrO, while in South-ern Irish, it is a verbal particle. Assuming that at


least one out of two adjacent strong functionalheads must be visible, she derives the impossibili-ty of having more than one argument in SouthernIrish: it is impossible, since a is not a functionalhead. This paper raised some interesting issuesconcerning the meaning of feature strength.

Anikó Lipták (HIL/Leiden University,‘Focus and ECM’) argued that Hungarian displaysa Long Focus Movement construction, which israther unexpected, since the moved constituentagrees with and gets Case from the matrix verb.Lipták followed Rooryck’s (1997) analysis of ECM,according to which the embedded AgrSP raises toSpec,CP, from where the moved constituent mayraise to Spec,AgrOP of the matrix verb and to thefocus projection. Under this analysis, no impropermovement arises, and the case and agreementfacts are explained.

Luciana Storto (MIT, ‘Karitiana: A VerbSecond Language From Amazonia’) describedseveral properties from Karitiana, a languagespoken in Amazonia, arguing that this language isV2. She claimed that, based on the Karitianadata, V2 might be interpreted as V-movement to Ior Focus, but not necessarily to C.

Juliette Waals (OTS/Utrecht University,‘Syllable Structure and Consonant Duration inDutch’) presented the results of three experimentson the temporal structure of consonant clusters inDutch. The conclusions were that consonantduration decreases as sonority increases. Theresults also confirmed the Metrical SegmentDuration Hypothesis, according to which segmentduration reflects syllable structure.

Marina Vigário (University of Minho, ‘Onthe Prosodic Status of Stresless Functional Wordsin European Portuguese’) argued that stresslessfunctional words in European Portuguese areprosodic clitics, attached postlexically to theProsodic Word. Her arguments were based onasymmetries of behavior between functionalwords and other words with respect to severalprosodic and phonological phenomena. Sheshowed that stressless functional words have abehavior which in general does not respect con-straints applying to Prosodic Words.

Laurel LaPorte-Grimes (University ofConnecticut, ‘Infl as Agr or T, but not both’) pre-sented a new approach to the V-movement con-trast between English and French, extending andmodifying Lasnik’s (1995) morphological analysis.Grimes’ analysis was based on a modification ofthe assumptions concerning the featural specifica-tion of the different verbal forms. It also succeed-ed in explaining the fact that in what aboutanswers bare verbs appear inflected in English,but in the infinitive in French.

Anabela Gonçalves (University of Lisbon,‘On Restructuring Constructions in EuropeanPortuguese’) discussed clitic climbing and restruc-turing in European Portuguese. According to herproposal, the difference between restructuringand non-restructuring verbs has to do with theactivity vs. inertia of the embedded functionalprojections.

Rajesh Bhatt (UPenn and MIT, ‘ActualityEntailments of Ability Modals’) opened the secondday of the conference, describing the behavior ofability modals with respect to actuality entail-ments. He showed that ability modals may beambiguous between past episodic and past genericreadings, which raises problems for the tradition-al LF for these modals. He then proposed ananalysis in terms of a decomposition of abilitymodals into a generic operator and a stage-levelpredicate ABLEslp.

‘Verbal Small Clauses’ was the topic of JoanRafel’s paper (University of Girona). He arguedthat small clauses may be headed by verbalcategories. He then applied such a structure tothe analysis of three constructions: pseudo-rela-tives in French, naked infinitives in English andthe Prepositional Infinitival Construction inEuropean Portuguese.

Susanne Glück and Roland Pfau (GoetheUniversity, ‘On Classifying Classification as aClass of Inflection in German Sign Language’)

looked at classifiers in German Sign Language.They compared the traditional incorporationanalysis with their proposal that classifiers areinflection markers. They argued for this analysison the basis of agreement patterns and pro-licensing in left-dislocation constructions.

Maria João Freitas and Matilde Miguel(University of Lisbon, ‘Prosodic and SyntacticInteraction: the Acquisition of NP FunctionalProjections in European Portuguese’) investigatedthe distribution of word-final fricatives in chil-dren’s first productions, and argued that codasmay emerge earlier than expected, if they corre-spond to morpho-syntactic information beingmastered by the children since very early. Theanalysis proposed for fricatives in coda was alsoproductive for the explanation of emergence ofsyllabic material occupying determiner positions.

Mikael Vinka (McGill University, ‘Nonpredi-cative Particles in Swedish’) investigated thebehavior of non-predicative particles in Swedishand argued that, in order to explain the interac-tion between these particles and NP-movement,they should be analyzed as maximal projections,which are able to undergo XP-movement to theSpecifier of AspP, and, from this position, blockNP-movement.

Hironobu Hosoi (McGill Universiy, ‘TheLexical Extension Control Verb Analysis’) arguedthat the Japanese V-V Compound “Control” Verbis not a control verb. He proposed instead thatcontrol verbs are formed under lexical extension.This proposal enabled him to explain some differ-ences between this type of verb and regular Eng-lish control verbs.

Viola Miglio (University of Maryland ‘AFormal Approach to Dialectology: Vowel Reduc-tion in Romance’) showed that Optimality Theorymay account for dialectal differences by means ofminor re-rankings of constraints. The empiricaldomain of her analysis was Vowel Reduction inRomance.

Susan Garrett (UPenn, Positional Faithful-ness and Truncation in Child Speech: What GetsKept and Why?’) looked at child Spanish trunca-tions, and argued that truncation is not arbitrary.Children systematically keep the stressed sylla-ble/foot, which was explained in terms of prosodicfaithfulness to the relevant prosodic domains.

‘The acquisition of expletive definite articlesin Modern Greek’ was the subject of TheodoreMarinis’ talk (University of Potsdam). He showedthat determiners are acquired in accordance withtheir semantic content: the more contentful theyare, the earlier they are acquired. The earlyacquisition of expletive determiners followed froma theory of triggers, according to which contrast-ing environments are relevant for determiningthe order of acquisition.

Adolfo Ausin (University of Connecticut,‘English Verbal Morphology and Feature Move-ment’) extended Lasnik’s analysis of Englishverbal morphology, defending the F-movementapproach. He claimed that the difference betweenauxiliaries and the other verbs is that only theformer are opaque to F-movement. Hence, onlythe former must move as categories.

Laura Siegel (UPenn, ‘Gerundive Nominalsand the Role of Aspect’) described some asym-metries between two types of gerundive nominals(POSS-ing and PRO-ing). She claimed that thesetwo types of gerundive nominals may not receivea unified analysis, since they are not able to referto the same type of events. She explained theseasymmetries, connecting them to the role playedby progressive aspect.

Dalina Kallulli (University of Durham, ‘TheCommon Basis of Clitic Doubling and Scram-bling’) claimed that clitic doubling and scramblinghave a common basis: defocusing. Her analysiswas based on a set of similarities between thesetwo phenomena in Germanic and Greek, Albanianand Rumanian.

José Luis Méndez (Universidad Autónoma deMadrid) proposed ‘A morphosyntactic analysis fornominal compounds in Spanish,’ based on thesmall clause analysis for possessive and attribu-

tive relationships.Esterella de Roo (HIL/Leiden University,

‘Nominal Projection in a case of AgrammaticAphasia in Dutch’) investigated the functionalstructure in spontaneous speech of a Dutchagrammatic patient, in order to determine wheth-er functional structure is completely absent orjust unrealized. Her results were related to stud-ies on extended projections and to underspecifica-tion of functional heads.

Marie Claude Boivin (MIT, ‘En-cliticization,Raising to Subject, and Case’) made a new propos-al for en-cliticization, explaining the asymmetrybetween quantitative and genitive en when ex-tracted from derived subjects. Her analysis wasbased on Case. She claimed that the absence of anovert nominal (quantitative en) makes the raisedNP unable to check Case, rendering raising toSpec,TP illicit.

Danny Fox (MIT, ‘Economy, Asymmetries,and the Ellipsis Scope Generalization’) revised hisown work on the scope of ellipsis, arguing on thebasis of down-stressing data that the Strict CycleCondition is not necessary to explain restrictionson scope. He now used accommodation (as resultof inferences) as a new tool to explain the relevantdata.

The invited speaker of the conference wasEduardo Raposo (University of California,Santa Barbara) who proposed an analysis ofclitics in Portuguese, arguing for Postal’s ideathat clitics are determiners.

Next year’s ConSOLE will be in Bergen,Norway, organized by the new board: MichaelRedford, Tina Cambier-Langeveld and AnikóLipták.

CONFERENCE CALLSSociety for Pidgin and Creole LanguagesLos Angeles, USA (in conjunction with LSA)8—9 January, 1999

Abstracts are invited for a 20-minute paper onphonology, morphology, syntax, semantics, lexicon,social aspects of language, history of the discipline orany pertinent issue involving pidgin and creolelanguages.

Deadline: 31 July, 1998

Info: [email protected] (Armin Schwegler); http://www.ling.su.se/Creole/calendar/

ConSOLE 7University of Bergen, Norway9—11 December, 1998

Papers are solicited from students in the field ofgenerative linguistics, more specifically in, but notlimited to, phonology, morphology, semantics, signlanguage, language acquisition and syntax.

Deadline: 1 August, 1998

Info: [email protected]; http://www.ling.uib.no/ConSOLE-7

14th Comparative Germanic Syntax WorkshopUniversity of Lund, Sweden8—9 January, 1999

Abstracts are invited for 30 minute papers on com-parative Germanic syntax.

Deadline: 1 August, 1998

Info: [email protected] (ChristerPlatzack)

International Conference of the LinguisticSociety of BelgiumUniversity of Antwerp, Belgium

11—12 December, 1998

Topic: Modal Verbs in Germanic and Romance Lan-guages.

Deadline: 15 August, 1998.

Info: [email protected] (Patrick Dendale);[email protected] (Johan van der Auwera)

Goodies Glot International, Volume 3, Issue 5, May 1998 24

WinSAL-V, which stands for Windows SpeechAudio Lab with Video option, is a tool for digitalspeech processing and acoustic analysis. TheWinSAL program provides a wide variety ofanalysis methods which cover all of the basicspeech analysis techniques. The video option alsoallows the program to load and display video files,which is a feature that is not found in most otherspeech analysis programs. The CD-ROM versionof the program comes with a database of soundfiles and a manual with an informative speechanalysis tutorial. The basic WinSAL package costs399 DM (about $220), the CD-ROM version is agreat value for only 50 DM (about $28) more. Thevideo option costs another 50 DM. The WinSAL-Vpackage on CD-ROM is a solid value for thoseinterested in doing basic speech analysis, espe-cially if you are interested in audiovisual signals,as it provides many of the same functions assoftware packages that cost $1,000 or more.(WinSAL is an especially good deal for students,who receive a 300 DM discount.)

WinSAL requires a PC compatible computerwith a 486 or faster processor, 8 MB of memory,Microsoft Windows 95 or NT, and a 16-bit soundcard. The program occupies about 10 MB of diskspace. Audio and video is handled through win-dows, so no special setup or equipment should berequired. A Pentium processor, 16 MB of memory,and a video card which can display 16-bit color isrecommended. The software was tested on a 486DX2 50 MHz processor, 24 MB of memory, Win-dows 95, with an 8X CD-ROM drive and a SoundBlaster AWE 32 sound card. The program workedperfectly well and performed all of its functionswith adequate speed on this machine, however,those interested in doing a lot of spectrographicanalysis or working with signals of sentencelength or longer will want to use a Pentium ma-chine.

WinSAL is a speech signal analysis anddigital signal processing program capable ofloading, saving, and recording WAV format soundfiles. With the video option, WinSAL-V can alsoload AVI video files. Audio and video are playedback together, allowing some capability for audio-visual speech analysis. WinSAL-V cannot recordvideo files, but such software is available fromMedia Enterprise for a price (1,999 DM, about$1,100). The WinSAL interface provides a varietyof control options, including pull-down menus, abutton bar, and some functions available througha context sensitive right mouse button menu.There are some keyboard shortcuts, but only forthe set of functions accessible through the rightmouse button menu. A click of the left mousebutton plays back a small portion of any signal inthe area in which it was clicked. While this fea-ture is nice, it has apparently made the usualclick-and-drag highlighting unusable. To select aportion of a signal, you have to press the shift keyand click at the same time, then drag. The coordi-nation required is rather precise, and in manycases it took me several attempts to get it to work.Once a portion of the signal is selected, it can beplayed, saved as a separate file, or expanded to fillthe entire analysis window using a special menuwhich pops up after highlighting. However, afterchoosing from the menu, the highlight is removed,so it is very difficult to play an arbitrary portion ofa signal over and over again without zooming inon it. To zoom out requires an equally unintuitive

Ctrl+click combination. In all windows, the pro-gram displays the coordinates of the mouse cursorin the relevant units, which I find to be a veryhandy feature for measurement. Since this is awindows program, the display of the windows canbe adjusted, minimized, and maximized. Onesmall annoyance is that once a window is maxi-mized, it cannot be minimized in the usual fash-ion (there is no restore command in the windowmenu and the buttons in the upper right corner ofthe window disappear, only leaving options likecascade, tile, or close all windows).

Signal analysis options are divided into threegroups: long time, short time, and spectrogram.The long time analysis options include a standardwaveform, energy display, and fundamentalfrequency display. The energy and fundamentalfrequency displays include a small waveformwindow which is time aligned with the display.The energy display shows RMS amplitude as adefault. The energy window can be configuredusing a parameters dialog box available from thecontext sensitive menu. The window size overwhich each segment of the measure is computedand other options for measuring energy (averageand max/min over the window) can be selected.Similar control over parameters is available forall of the analysis functions, making the programfairly configurable. However, for the benefit of thenovice user, it would be nice if there was a way toreset the parameters to the defaults withoutreinstalling the program from scratch. The lack ofa default discourages exploring the options. Thefundamental frequency display computes anestimate of the fundamental frequency using theaverage mean distance function (AMDF, which isvery similar to an autocorrelation). The maximumand minimum allowable frequency for the pitchtrack is adjustable, a very important feature forproper analysis.

The short time functions are linear predictivecoding (LPC), fast Fourier transform (FFT),Cepstrum, AMDF, and autocorrelation. Eachshort time window is accompanied by a timealigned waveform which has red cursors markingthe segment of the signal which is displayed inthe analysis window. The relative display of thewaveform and analysis window is fixed and can-not be configured (though the window itself can,of course, be resized). The size of the analysiswindow is adjustable for all of the short timefunctions, so the standard trade off of frequencyresolution versus time resolution is available.Unfortunately, the frequency range over whichthese functions operate cannot be parameterized,so displays of the most relevant frequency rangescan only be created by digitizing at a lower sam-pling rate. The LPC order can be parameterized,another crucial feature. The Cepstrum, AMDF,and autocorrelation operate just like the LPC andFFT, displaying values just over the analysiswindow. The program cannot compute thesefunctions for intervals over the entire signal,which makes these displays of limited value forthe researcher.

The values of the short time displays (andlong time displays) can be saved into an ASCIItext file to be used for subsequent processing.Overall, I believe the short time displays will be ofmost value to the student learning about digitalsignal processing. Since the analysis tasks cannotbe automated by macros or special functions, they

are useful analysis tools for small projects orthose requiring careful hand measuring only. Thedisplays can also be printed or saved as BMPformat graphics files. I tested the printing of anLPC window on a 300dpi HP Laser Jet IIIP andfound the print of the waveform to be of goodquality and the overall display to be satisfactory.The program can also save a screen shot to a BMPfile, so more intricate displays can be created andeither printed out, inserted in another document,or modified with a graphics program.

The final analysis option is the spectrogram.WinSAL uses color spectrograms, and this optionis unfortunately not configurable. Not surprising-ly, the spectrogram was by far the slowest analy-sis tool, but its speed was reasonable on mymachine even for short phrases. The frequencyrange of the spectrogram can be adjusted. Thespectrogram can be printed out or saved as agraphics file. The program uses a color spectro-gram for the printout as well, which I suspectwould look very nice on a color printer. A printedspectrogram on my laser printer came out aswhite on black, and was very dark and not verylegible to me. To make black on white output Iwould recommend saving the spectrogram as aBMP file, converting it to a grayscale image in agraphics program, and then reversing the polarityof the scale.

The handbook that comes with the CD-ROMversion has a short manual on the use of theprogram, though all of this information is alsocontained in the on-line help for the program. Thehandbook also contains a basic introduction tosignal processing which I found a very helpful anduseful tutorial. The tutorial is full of tips andtechniques for signal analysis and also containsan explicit example stepping through some of thecommon functions and telling you what youshould see for a sample sound file included on theCD. The CD contains a catalog of sound and videofiles for both English and German. All of thesegments are covered for each language. Conso-nants are given in word initial position if possible.This is the same catalog of sounds in the Speech-lab/Sprachlabor program (reviewed in GlotInternational 3,4; a demonstration version ofSpeechlab is included on the CD). The CD in-cludes a few videos and sound files not found inSpeechlab. There are videos of laryngeal function,which are a little hard to appreciate without away to view the video frame by frame, and somevideos of an infant vocalizing. The sound filesinclude some isolated vowels, two formant synthe-sized vowels, and sentence length recordings ofread speech in English and German. There arealso longer passages in German read by twofemale talkers.

Overall, the WinSAL package is a good value,especially the CD-ROM version with handbook.The video catalog on the CD-ROM can only befully appreciated with the video option, though Ithink the best use of the video is for instructionalrather than analysis purposes as there are novideo controls. The ability to step through thevideo frame by frame with buttons or the key-board would enhance this portion of the programquite a bit. This software has valuable featuresfor students and researchers who need to do asmall to moderate amount of speech analysis. Iwouldn’t recommend this program as the primaryspeech analysis tool in a speech lab, as it does nothave the full functionality and programmability ofmore expensive programs designed for speechprofessionals. The lack of macros and automatedprocessing tools make this package inappropriatefor large research projects. In addition, the varie-ty of minor annoyances like the strange form ofhighlighting and color only spectrograms mightactually interfere with efficient analysis andgenerating printed results on a daily basis. Thissoftware will be an outstanding value if thesecomplaints are addressed in future versions.WinSAL would make an good supplementalspeech analysis package for any lab, used forinstructional purposes and smaller projects, ascopies could be purchased for several machines ata reasonable cost.

SIGNAL ANALYSIS AND DIGITAL SIGNALS

by Stefan FrischReview of WinSAL-Vby Ingolf Franke

Extra Glot International, Volume 3, Issue 5, May 1998 25

THE NUMBER OF DEATHA linguistic mystery in eight installmentsby Chris Sidney Tappan

Chapter 4Towards some hypotheses

Paul had just had a flat tyre, that was all.It did not take him much time to convince James

that her idea that Bill Fyffe had been killed by mistakeand that he — Paul — had been their real target wasabsurd. “I’m a linguist and I’m a nice guy — who wouldwant to kill me?!” he told her.

He was right, of course.So they went and had lunch together, as previously

planned, in the restaurant called “The Drinking Don-key”. It used to have a different name, “The HungryHorse”, and in those days it was their favorite hang-out.After lunch they would go to Paul’s office, where Fyffe’sbody had been found, to see whether anything wasmissing or something else had been disturbed. So theyhad thought it a nice idea to do lunch together. They hadnot had lunch together on a weekday for a very longtime. Even the restaurants had changed names.

After getting some beer, they placed their order(two simple sandwiches with lettuce) and looked at eachother, both thinking that they should do this more often.

“So what did you find out this morning?” Paulasked.

“I talked to several people, but there was not muchof interest,” James said. It was clear that she wasdisappointed. “I mean, most just confirmed what wealready know, that everybody hated Bill and for goodreasons too.” She opened her notebook and leafedthrough it, not looking for anything in particular, justgoing through it and see what was the thing she wasmost eager to talk with Paul about.

The waiter brought the food and asked themwhether they wanted another beer. They did.

James said: “I got the list of everybody who was inthe building the night of the murder, that is to say, allthe people who were officially there, who had picked upa key, of course. There are only seven. Well, eight,including you. ‘A quiet night,’ the janitor-in-chief saidwhen he gave me the list. When I told him that my ideaof a quiet night did not include a murder, he said that adead person is always more quiet than one who is notdead, which he himself thought was very funny. I amthinking of getting the smart ass detained for 24 hours,or maybe 48, just to irk him. What do you think?”

Paul smiled at her. Wasn’t she wonderful?“Anyway,” James continued. “As to those seven

people, I talked to four of them this morning. The otherthree will come in this afternoon. The four I talked tocame early, left early, did not see or hear anything. Twoof them could confirm each other’s presence, they hadbumped into each other in the restroom, but they hadbeen on a different floor than where Bill’s office is, sothey could hardly be expected to be able to provide anyinteresting information. We’ll check their alibis. One ofthe others was a Martian and he at least could tell mesomething useful. He said he had actually been there tosee Bill. Everybody apparently knows that Bill is in theoffice every night. His name is Nichael. He said he hadbeen there with Stephem. I will see Stephem thisafternoon. They were in the office right next to Bill’s. Hesaid there was nobody there. Bill never showed up andthey left at about nine.”

“Ah, Nichael,” Paul said. “I know him, he is one ofmy informants. He is interesting because his phonologyis all screwed up. Did you notice?”

“I had a hard time understanding him, yes. Why dothese guys have such funny names, anyway?” Jamesasked, putting a piece of sandwich in her mouth. “Imean, like Johm, our neighbor, why are they calledJohm instead of just John?” she added, barely under-standable.

“They have some funny phonological rules, that’sall,” Paul answered, looking around to see whether hecould find the waiter with their beers. “Rule number 1 ofthe Martian phonology of English is: turn N into M atthe end of a word or M into N at the beginning of aword. Rule number 2 is: Rule 1 applies only once. AndRule number 3 is: It only applies to names of people.The interesting thing about Nichael’s speech is that heviolates rules 2 and 3, though not all the time, of course(that would be too easy). So he does not limit rule

number 1 to names of people but to all unique objects.So he says sum instead of sun, Nars instead of Mars.And he says noom instead of moon because he violatesrule number 2 too. How many nooms does Nars have, doyou know? I neam, how unique is a noom?”

James looked at him. She had no idea what he wastalking about. So she simply ignored him, saying: “As Isaid, I haven’t gotten much further. I mean, apart fromthe guys who were in the building that evening, I alsotalked to many other people. The only interesting but atthe same time ridiculous thing I heard was that some-body, a Martian actually, threatened to kill our Mr. Billthe day of his death. And you know why? Because Billhad told the Martian that there was not enough paper inthe printer! Apparently, the Martian was printingsomething in the printing room, and Bill had been theretoo and in going out he had made that comment aboutthere being not enough paper. Someone — I think she isin your department, let’s see (leafing through her notes)Esperanza Holyfield — overheard this while enteringthe printing room, and she found this Martian there,fuming with anger, apparently because of this commentabout the paper, mumbling that he would kill him for it.But, I mean, killing somebody because he tells you thereis not enough paper in the printer — I mean, how muchmore farfetched is this whole story going to be?”

Much to James’ surprise, Paul did not burst out inlaughter, but instead, he reacted very seriously. “AMartian threatened to kill Bill?” he asked in whatseemed sincere astonishment. “That is remarkable. Heis always so very nice to them. Who was it? Do you knowhis name?”

James looked at her notes again. “I think his namewas Nartin or something.” All of a sudden, she wasmuch more alert too. As a consequence, her appetite wasgone, and she felt very thirsty. And then the waiter —another one actually — brought their beers. Thanks, shenodded to the waiter, and she took an eager bite.

“Nartin?” Paul was saying. “Hmm, that is interest-ing. Nartin had applied for a job and he didn’t get it. Billwas on the committee of course, but it is well-knownthat Bill was very much in favor of hiring Nartin. Heeven said it on TV, the other day, remember?, that morejobs should go to Martians?”

James did not remember but she believed him.Paul was going to say something, but then their ownwaiter came with two more beers. They said theyalready had their beers, thank you, and wanted tocontinue their conversation, but the waiter would not letthem. He stood at their table and said angrily: “No, youdid not get your beers. These are your beers, the oneson your table are not yours!” And he got even moreangry when he discovered that they had already beensipping from the beers the other waiter had put on theirtable. “Now they can no longer be served to the peoplewho ordered them!” he exclaimed. “You should havewaited for me!”

“You should have waited on us!” Paul declaredwith much too much drama. And the joke was notappreciated anyway.

“Exactly!” the waiter said triumphantly. “I’m gladyou see my point.” Then stood there silently.

“Can’t you bring the beers on your tray to whoeverthe beers we are drinking were for?”

“They won’t accept them. They did not order these.They ordered those!”

Paul was stunned. “What kind of place is this?” heasked, looking at James, but obviously not expecting ananswer from her. “I mean” — turning to the waiter —“How are we supposed to know that these beers were notfor us! We order them, a waiter brings them —”

“Exactly!” the waiter interrupted him victoriouslyat the top of his voice. “A waiter brings them, not thewaiter. Not the same one you placed the order with!”And he pointed at the text in bold red print on themenu, right under the name of the restaurant. It read: Ifyou order a beer, make sure you get it! When Paul andJames looked up, the waiter had stalked off, leaving thebeers on their table.

“Arrest the smart ass!” Paul said to his wife. But

the police detective was flabbergasted and did not move.

After lunch they went to Paul’s office. The technical staffof the police had done their work: finger prints, photos,everything. Now, all James wanted to know from Paulwas whether anything was missing from the office.

Paul looked around. “No,” he declared. “Not as faras I can see now.”

They looked around some more, but Paul did notsee anything unusual.

“So he was lying here?” Paul asked, pointing at thechalk marks on the floor.

“Yes,” James said.“Can I work in the office now?” He said. “I don’t

know whether I’d feel like it, of course. It remains astrange idea that somebody was killed right here.”

“Yeah.” James was silent for a while. Then she said:“Yes you can work here. We are finished with this room.I’ll go up and talk to some more people. I have a lot ofappointments lined up.”

They looked at each other. Let’s try to get homeearly today, they thought.

The conversation between Nartin and Bill in the print-ing room kept bothering Paul. What had actually beensaid? While pondering over this question, he sat downbehind his computer and turned it on. But before thestarting up had been completed, Paul had already leftthe office. He had decided to pay a visit EsperanzaHolyfield.

She was in. She was also some sort of a linguist,although it had never really become clear to Paul whataspect of language she was interested in let alone thathe had any idea as to what aspect of language she wasstudying.

“Hi, Esperanza.”“Hi, Paul, come in.”“Thanks.”“How are you doing? It must be quite a shock to

find out that somebody got killed in your office. Andsomebody you know too. I’m glad he did not get killedhere.”

Why would he? Paul thought. That is actually aninteresting question, he thought. Bill Fyffe had been inhis — Paul’s — office for a purpose. Presumably, he wasthere for something he could not have done in somebodyelse’s office, like Esperanza’s. Unless he was going fromoffice to office and had already been here. Forget thisline of thought, Paul told himself.

“Is your wife making any progress?” Esperanzaasked.

“Not really,” he said. “But you told her somethingthat interests me. About the printing room incident withNartin so to speak. Could you tell me exactly whathappened and do you think you can tell me everybody’sexact words?”

“Strange, isn’t it? I was really shocked. I had neverheard Nartin say anything like that.” Esperanza stilllooked shocked. “But their exact words...”

“Try it. I have the feeling that it is important,” Paulencouraged her.

“Well, I was going to print something, I don’t evenknow anymore what it was. So I walk up there and Billand Nartin are in that room. Nartin is printing some-thing, and they are talking about the reasons whyNartin did not get that job in the department. I hadforgotten something so I went back downstairs to get it.Coming back, I bumped into Bill in the doorway. He wasjust coming out. And I heard him say something like: ‘Ithink you don’t have enough paper!’ He says this in ajoking manner. You know how he is, I mean was, alwaysso much nicer to them than to us.”

“Were those his exact words?” Paul asked her,ignoring her last comment and before she could continueher story.

“Something like that. ‘I think you don’t haveenough paper!’”

“And then what happened?”“So I go in and I was shocked to see that Nartin

was angry as hell. I mean, I had not expected thisbecause Bill had made that comment jokingly, and whenI had been there earlier, the atmosphere had been fine,they were sitting there like friends, no animosity. Butnow, Nartin was fuming. ‘Did you hear what he said?!’he asked me. I said, You mean that comment about thepaper? ‘Yes!’ he says. ‘I could kill him for it!’ he said. ‘Icould kill him for it,’ that were his exact words. I couldhardly believe what I heard. Especially from Nartin.And I mean, it was only about the paper in the printer!”

“You said that they were talking about the reasonswhy Nartin did not get the job when you were up thereearlier, right?”

“Right.”“What did they say? What were they talking

about?”“I don’t know, I wasn’t there for a long time, I

walked in, and walked out again, right away.”

Interview Glot International, Volume 3, Issue 5, May 1998 26

tional Grammar, the multiple notions of a subjectthat you get by looking at multiple strata likePerlmutter has done, and how that has turned upfacts of great interest.

So frameworks are only useful as a heuristic? Tofind more facts?

Well I love facts. Newmeyer has described meand my generative semantics buddies as datafetishists, and I’m quite happy to be so described.You can really get high on data.

But of course, let me stress that we datafetishists are able to get such gigantic highs out ofour experiences with data partly because of theinteractions between theory and data. So, yes, tomy mind, the main value of theories in any sub-ject is how they lead people to the discovery andappreciation of facts, how they help you findthings that were there that you didn’t knowabout, or make you appreciate the significance ofthings that you knew were there but that you didnot know the significance of. In the preface of mysyntax textbook I mention black holes. It takessome awfully specific theories of light and thedistribution of matter, and so forth, before it ispossible for astronomers to say, “By God, I’ve got ablack hole here.” We’ve got the same sort of thingin linguistics. The sort of approach to syntax thatI take, taking constituent structure really serious-ly, as well as a very specific version of the CyclicPrinciple, has given me all sorts of neat stuff, likeexplanations of why quantifier float and there-insertion and so forth disambiguate quantifierscope. The way that I set it up, that follows fromstrict Cyclicity. Of course, in the process of comingto these neat facts, I discover holes in my theoryand deficiencies, which I then alter. Theories leadto facts, which lead to better theories which leadthe data fetishist, if he is lucky, to a yet greaterhigh!

When it comes to theories, I also keep in mindthe very important point that Lakatos, the philos-opher of science, made, that when people talkabout the success of a theory, it really isn’t atheory they’re talking about. They’re talkingabout some community in which research is doneover decades, maybe even centuries, where theassociated theory will often undergo enormouschanges. The continuity that people perceive, isthere in virtue of a research program, which doesnot change much over the course of the decades oreven the centuries. The people working withinthat research program keep altering their theo-ries in what look to them like small ways. Butchanges in a theory are never inherently large orsmall — they’re only large or small from a per-spective such as a research program.

So you got into linguistics for the facts?

Originally, I was in mathematics, but for noparticularly good reason — I really hadn’t thoughtmuch about what I was getting myself into. I wastaking some language courses and enjoyed themand at the same time I got more and more turned

off by mathematics. Eventually I thumbedthrough the University of Chicago time scheduleand saw that there was something listed as lin-guistics. So I sat in on a linguistics course taughtby Eric Hamp, which I greatly enjoyed. Then I gota mathematics scholarship to study in Germany,at the University of Münster. But instead of doingvery much mathematics, I took all sorts of lan-guage courses, including a Dutch course. Duringthat year, I got more and more turned off bymathematics. After I got back to Chicago, I want-ed to take a language course just for the fun of it.Japanese was offered at a convenient time, so Itook it and I fell in love with the language rightaway. I also started looking around in the libraryfor linguistic books. I came upon Syntactic Struc-tures and it really turned me on. Not long afterthat I saw the announcement for the new linguis-tics graduate program that they were starting atMIT. I applied, got accepted, went there, became alinguist and I have been enjoying life much moreever since then.

How come?

Linguistics is fun!

Why?

Languages are weird and wonderful things. Aslong as you are perceptive enough there is plentyto keep you happy and busy. In that respect, Ithink linguistics is way to hell more interestingthan mathematics. There’s a wonderful quotationfrom Bertolt Brecht at the beginning of Feyer-abend’s Against Method: Ordnung gibt es meistenswo nichts ist. Sie ist eine Mangelerscheinung.“There is order mainly where there is nothing. It’sa phenomenon of absence.” And, well, Ordnunggibt es in der Mathematik.

Not in linguistics?

There is this principle, associated with Ebeling, tothe effect that phonemic analysis is most effectiveon the languages that one knows less about. It’sprobably true for any linguistic analysis. This isan illustration of the Brecht quotation. This iswhy linguistics is so interesting.

In addition, there are tie-ins between linguis-tics and virtually everything. That’s another thingthat really turns me on about linguistics — it’s anexcuse for me to be a perpetual student. Sincelinguistics ties in with pretty much everything, Ican, with a straight face, sit in on philosophycourses and claim that is to my professionalbenefit.

Back to your surfing the frameworks...

The sort of moving around that I do is a lot likethe shopping tour I just described. What I actual-ly do in a lot of cases is what I call “to charitablymisrepresent” what is done in other frameworks.Let me explain what I mean by “charitable mis-representation”. In a paper I did recently, I talkquite a bit about this really nice paper that Caro-lyn Heycock had in Linguistic Inquiry about two

years ago that presents some very interestingfacts. Heycock explains why certain interrogativesentences with anaphoric relations are ungram-matical, while others are grammatical, in terms ofhow they differ in what she calls LF. She says“LF” but I can charitably misrepresent her ashaving talked about “logical form”. Logical form,of course, isn’t LF because if you think of logicalform literally, it’s something that ties in to logic: itwould involve scopes being represented explicitly.The term “LF” has an etymological relationship tological form, but not much of a relationship be-yond that. In LF — so-called — you don’t haveenough nodes of the right categories to representall the scopes that you get in a sentence withmultiple quantified expressions. This results fromthe demand within that framework that syntacticrepresentations be homologous, that whateveryou’ve got in one level have a counterpart in otherlevels. This limits how many sentence nodes youhave to serve as scopes. Anyway, I read CarolynHeycock’s paper and I say: Great, she’s got awonderful idea and the difference that she’s got inwhat she calls LF, is in fact the difference inhonest-to-God logical form. In that paper, shetalks about a difference between wh-expressionsthat involve predicate elements like how proud ofBarbara versus things that involve non-predicateelements like how many accusations againstJohn. She comes up with the, I think, absolutelycorrect conclusion that the predicate expressionshave to be inside their clauses in logical structure(here I charitably misrepresent her) versus theother ones which, in logical structure, could beoutside the clause. Putting that into my frame-work, where you’ve got deep structures that are atleast approximate semantic structures, andmaking heavy use of the really powerful version ofthe Cyclic Principle that I have, I can then saythat the derivations that you get with theserespective deep structures give me a violation ofone of the anaphora constraints in the one case,and no violation in the other case. And I’m veryhappy: I got an explanation of something interest-ing, and I got it through the charitable misrepre-sentation of something done in anotherframework.

So when something new comes up like OptimalityTheory, do you plunge into it and see how you cancharitably misrepresent it?

I generally don’t “plunge”. But I come into contactwith things done in various approaches, like bythumbing through journals and by going to confer-ences and staying in the same room because it isn’tworth the trouble to get up and move to anotherroom. (It’s like watching television.) I get exposed toall sorts of things, a whole lot of which have nothingparticular to recommend them, but quite a few ofwhich have interesting stuff. And if a certain ap-proach seems to be leading somewhere, in that itrecognizes or discovers interesting facts or that itappreciates the significance of things that have notbeen previously appreciated, then I’ll look at morestuff done in that framework.

The first big dose of Optimality that I hadthat really impressed me was a paper that DougPulleyblank gave a couple of years ago aboutvowel assimilation in Yoruba. He gave neat expla-nations for why the assimilations go in one direc-tion in one set of words and in another directionin another set of words.

With regard to Optimality in syntax, I’veheard some talks and I’ve read miscellaneouspapers that have at least caused me once in awhile to think in, let’s say, Optimality theoreticterms about things in syntax. For example, Isubmitted a paper a few months ago to a Fest-schrift for a person whose name I guess I’m sup-posed to still keep in confidence. In this paper, Igot into some odd differences between declarativesentences and questions in English with regard towhen you’ve got an auxiliary verb and a negative,whether the auxiliary verb and the negative staytogether or not. So take, for example, Can’t you seewhat he’s doing? Perfectly normal question. Can

“WHAT’S RIGHT ABOUT X-BAR SYNTAXIS THE X AND THE BAR. WHAT’SWRONG IS EVERYTHING ELSE!”

by Lisa Cheng & Rint SybesmaAn interview with James McCawley

Continued from page 1


you not see what he’s doing? Sort of okay, but thepreferred form would be the first one where you’vedragged the negation along in the inversion. Nowtake questions corresponding to You really can’tsee what he’s doing: Can you really not see whathe’s doing? That’s perfect, way to hell better than:Can’t you really see what he’s doing? Here, I think,you’ve got something that you can describe interms of ranked constraints, in the sense thatthere’s a tendency — you can call it a constraint ifyou want — to have surface scope relations matchlogical scope relations. The obligatory inversionyou’ve got in English questions would then forceyou in that particular case to have some sort ofdiscrepancy between logical constituent structureand surface constituent structure. We’re talkingabout the interrogative counterpart of somethingwhere you’ve got really above not and not abovecan and you’ve got to do inversion. That’s gonnagive you a surface structure where at least thecan is higher than the really. But if you strand thenegation, then you can at least minimize theamount of discrepancy that you’ve got betweensurface c-command relations with three elementsand the corresponding logical relations. So maybeI benefitted to some extent from having beenexposed to some Optimality stuff.

Any comments on what is going on in the field inother respects?

I am quite annoyed at a lot of stuff done in syntaxrecently, especially this absolutely screwy concep-tion of syntactic categories that now prevails inthe kinds of syntax coming out of MIT and itsvarious branch offices. Let me give you a little bitof background. In the 1980s I would say, “What’sright about X-bar syntax is the X and the bar.What’s wrong is everything else”. When I said“what’s right is the X and the bar”, I was referringto the X as it was used in the 70’s and early 80’s,when it was still a part of speech. Since the mid80’s you’ve had all sorts of stuff other than partsof speech being used as the X of X-bar categories,as in this thing NegP that now turns up all overthe place. Well, the negative elements in thevarious languages of the world belong to particu-lar parts of speech. There are negative adverbs,negative adjectives, negative prepositions, nega-tive conjunctions, etc., and they share the syntaxof those parts of speech, in addition to whateverproperties they have in virtue of being negative.Calling something a NegP obscures the syntacticproperties that those words have got in virtue ofbeing adverbs, adjectives, or whatever. In addi-tion, in the case of negative adverbs, it misrepre-sents their syntactic status. Negative adverbs,like adverbs in general, are modifiers and are notthe heads of larger things. Actually, one of thethings that annoys me about syntactic categoriesas they’re treated in real recent MITish stuff isthat it’s really become hard for MITish people tosay “modifier” anymore. I mean, all sorts of thingsthat to me are obvious modifiers now get repre-sented as heads of things that they aren’t headsof. This particular conception of categories hasbecome undeservedly popular and is yieldingseriously screwed up analyses.

But in principle, according to what you saidearlier, even that should still lead to something.

That’s true. Even holding all sorts of crazy ideaslike this conception of syntactic categories thatI’ve been complaining about hasn’t prevented themore astute practitioners from coming up withreally neat stuff.

Do you think that people should not specialize insyntax or phonology or so and keep working in asmany fields as possible?

It certainly is worth the linguist’s while to stay intouch with all the things that have contact withthe field he wants to specialize in. It may some-times give you a slightly different perspective, aperspective that may help you see things. There’s

nothing wrong with specialization — it’s actuallyhighly desirable. But specialization shouldn’tmean that you close yourself off completely fromthe rest of the world. It should be concentratingon a particular field, keeping in mind that it is toyour advantage to be aware of things that arerelevant to that field, but aren’t strictly speakingwithin it.

But whether people should or should notspecialize... Different people get different degreesof pleasure out of different things. My advice is:get your jollies however you can. I can tell youwhat sorts of things I have gotten fun out of, butI’m sure there’d be all sorts of fields that otherpeople have gotten lots of fun out of. Good forthem!

And for you, what is the thing that turns you onthe most?

I guess a sort of broadly conceived syntax, where“broadly conceived” means I really don’t particu-larly give a hoot about demarcating it from se-mantics, pragmatics, discourse structure, andwhat have you. I do syntax, but I pay attention toall these things, and I do not particularly carewhether what I’m doing is what every self-pro-claimed syntactician would call syntax.

Are there any properties which you think arecrucial for being a good linguist?

I don’t know. There may be some characteristicswhich are advantageous. Others probably don’tmatter. Or at least they don’t hurt. I mentionedmy background in mathematics to you. I often tellmy students: the advantage to a linguist of abackground in mathematics is like a vaccination.It gives you immunity against horrible diseasesthat you might otherwise catch, but it doesn’thave any positive value in itself.

Somebody described you as probably one of themost well-travelled linguists in the world. Whatwas your worst conference?

One memorably bad conference was on my firsttrip to Japan in 1965. It was a conference called,“Second World Congress of Phoneticians,” organ-ized by the Phonetic Society of Japan. It was aweird conference. First of all, it was an interna-tional conference, but most of the Japanese partic-ipants gave their papers in Japanese. The secondthing was, a huge proportion of the papers had todo with a recent kidnapping case. There were taperecordings of the kidnapper’s telephone requestsfor ransom and there were papers by phoneticianswho had been doing phonetic analyses of theserecordings, trying to identify the person from thevoice. All the papers ended with the phoneticianssaying that the culprit is yet to be apprehended.Another thing that was memorable was that thepre-prints for the conference were being distribut-ed in plastic envelopes that had on them thewords: “Second World Congless of Phoneticians.”So that was a memorable, reasonably bad, confer-ence.

And since 1965 there haven’t been any bad ones?That must say something about the field!

Give me some time, I’m sure I can think of one ortwo more!

Any other odd place you did linguistics?

I once did a question and answer session for theInternational Institute of Dravidian Linguistics,in Trivandrum, Kerala, India. A transcript of thething was published in their newsletter. I’ve listedit in my publications as “freely translated intoIndian English.”

One other interesting experience that I hadwas a three week course at Guangzhou Instituteof Foreign Languages in 1987. While I was talk-ing with some of the faculty there, including aGerman instructor, I mentioned that I had given

Holland Academic Graphics

Prices for individuals:NLG 30 (ex. 6% VAT where applicable, ex. P&P, only when ordered

directly from the publisher)

Prices for libraries and institutes:NLG 40 (ex. 6% VAT where applicable, ex. P&P)

Postage and handlingNetherlands NLG 10

Europe NLG 12.50

Rest of the world NLG 17.50

P.O. BoxThe Hague

The Netherlands

53292

2505 AG

e-mail: mail hag.nlhttp: www.hag.nlfax: +

@

31 70 4480177

[ 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 ]

Recent dissertations from UiL OTS

Advertisement

Licensing and interpretation of invertedsubjects in Italian

Genitive Case and genitive constructions

First steps in wh-movement

Manuela Pinto

Petra de Wit

Jacqueline van Kampen

Licensing and interpretation of inverted subjects in Italian

Genitive Case and genitive constructions

First steps in wh-movement

wh

do wh

aims atidentifying the structural and interpretive primitivesinvolved in the phenomenon of subject inversion inItalian and at translating them into formal terms assuggested in Chomsky (1995) and Reinhart (1995,1996). The modular account Pinto proposes views thephenomenon of subject inversion as the result of theinteraction of the principles of the computationalsystem with the syntactic properties of the subject andwith conditions of derivational economy and interfaceeconomy.

In analysing Italian subject inversion and invertedsubjects, special attention is paid to focus contexts andoften unrealized loco-temporal arguments.

The major aim of this study is not only to offer anexhaustive and explanatory analysis of inversion inItalian, but also to achieve this without burdeningsyntax with unneccessary or stipulative devices. Al-though the empirical domain is restricted to Italian, theanalysis proposed here will hopefully form the basis fora crosslinguistic approach to inversion in Romance.

1997. 261pp. ISBN 9054340649. Paperback.

explores the syntax ofarguments and a variety of possessive constructions innominal clause structures in languages such as Dutch,English, German and Russian. Couched within aminimalist framework, De Wit presents a coherenttheory of how possessives and arguments are licensed inthe functional layers of DP and she explains the role ofderivational affixes, case markers and prepositions inthis process. The analysis is extended to examples ofpredication in DP as in binominal noun phrases. Inaddition, it takes into consideration the historicaldevelopment of the morphological genitive and itsrelation to current genitivelike constructions.

1997. 223pp. ISBN 90 5434 066 5. Paperback.

In the acquisition of language isconsidered from a generative point of view. Theemperical basis is a longitudinal corpus of two Dutchchildren (aged 1;9-10;0 and 1;7-7;0). The main topic isthe acquisition of V2nd and -movement.

Van Kampen presents analyses for: (a) twowordutterances, (b) -insertion, (c) -drop/topic drop, (d)left branch violations, (e) spell-out of intermediate Cpositions.

A central question is how multilevel representationsare learnable by means of an input-driven procedure.

1997. 219pp. ISBN 90 5651 028 2. Paperback.


lectures in German at various places. Then thisGerman instructor started pleading with me togive a lecture in German. I said, what the hell,how can I turn down a request to give a lecture inGerman in Guangzhou China?! So I worked upthis lecture titled: “Kritische Bemerkungen zuChomskys neuster Syntaxtheorie”. It was a spe-cial experience.

It must take some extra time to prepare a talk inGerman.

I’ve done a whole lot of lecturing inGerman. It takes me a bit of extrawork. I’ll have to dig up a dictionaryand look up unfamiliar vocabulary.But the last two weeks I’ve beengiving lectures in German at awhole bunch of universities. I stud-ied in Germany a long time ago, sothat’s no big deal. I’ve done that lotsof times.

I find it really interesting andenlightening to work up lectures inother languages. (I’ve also givenlectures in Spanish a couple oftimes.) In a lot of cases, it alerts meto terminological gaps in the otherlanguage. I really had quite a hardtime working up a lecture on thelogic of conditional sentences that Igave in German in Hamburg a fewdays ago. German linguists that Iconsulted asking for terminologicalsuggestions told me that this is atopic which, when Germans writeabout it, they generally write inEnglish. Or if they write it in Ger-man, they publish it as workingpapers that you couldn’t find in Chicago.

You are also quite famous for your knowledge ofall the best restaurants all over the world inwhatever cuisine.

You are exaggerating. I try to acquire such knowl-edge simply because having such knowledge isvery useful for someone who likes good food.

There’s a story of somebody organizing a confer-ence in New York, who asked you to write theconference restaurant guide, although you were infar away Chicago.

That’s a bit of an exaggeration too. Somebodyasked me for suggestions and I gave them some.

Still, the New Yorkers organizing that conferenceapparently trusted your knowledge of the restau-rant situation in New York better than they didtheir own.

I make a point of exploring New York’s China-towns. There are some really interesting restau-rants that are not that well known to round-eyes,or hairy barbarians, whatever the appropriateterm is.

I remember that even in a place like Madison youknew all the good places.

Madison is a great place to visit. It’s the onlyplace outside of Nepal that I’ve been to that hastwo Nepali restaurants.

When you know where to find the good restau-rants, you supposedly know where to find the badrestaurants as well.

I once ate in a very bad Chinese restaurant inOslo, probably the most memorably bad Chineserestaurant I have ever been to. When I visitedOslo for a few days in 1969, I asked friends ofmine there if they could recommend a Chineserestaurant. You know, good Chinese restaurantsdo turn up in odd places, like Alice Springs, Aus-tralia, for example, to mention one excellent

restaurant. So my friend says: “We think the bestChinese restaurant is so and so in such and suchgatan. Why don’t you go there.” I went to thatplace, I walked in and looked around and, evi-dently, it had, until recently, been a Norwegianrustic restaurant — they had knotty pine pan-eling on the walls and cartwheel chandeliers. Andthen they had Chinese lanterns hanging from thecartwheel chandeliers. A Norwegian waitressappeared and greeted me in Norwegian. Sheseated me and showed me this menu that was

entirely in Norwegian. I ordered a couple of thingsand they were absolutely wretched. The reallymemorable thing, though, was that there werelittle jars on the table containing condiments, andthat the labels on the jars were not in Chinese,but in Japanese, and the contents had no relationto what the labels said.

Did you tell the waitress?

No. I didn’t think it would be worth the trouble.

Are there any good Chinese restaurants in Hol-land that you can recommend?

Let me give you a tip on how to identify goodChinese restaurants in Holland: the good ones arethe ones that have their menus written in Chi-nese and English, but not Dutch. If they have amenu in Dutch, don’t go in!

You must have a fantastic memory. Not onlybecause of all those restaurants but in the questionperiod of one of the talks this morning you quotedShakespeare, Play such and such, Act so and so.Just like that.

I was cheating, actually. I consulted something Ipulled out of my bag — one of my collections ofexamples. Every year I print out this collection oflinguistic examples, that I’ve gathered over thelast year. The general title is “A Linguistic FleaCircus”. I was quoting from my 1995 LinguisticFlea Circus.

Linguistic Flea Circus?

It’s a collection of examples. Data fetishism, I toldyou! I write things down in a notebook that Icarry around with me. Things I hear on television,or, while reading a newspaper, whenever I see aninteresting example, I’ll mark it and tear out thatpage. As you can see, there is a huge number ofSpanish examples. I watch a lot more Spanishtelevision than English language television. Mostof these Spanish examples come out of telenovel-as. What it comes down to is: I collect linguisticexamples.

But data fetishism has paid off in my linguis-tics. I have all these interesting examples, organ-ized in a way that I’m familiar with, so wheneverI need examples, this will help me find them.There’s a paper that I published about six orseven years ago in a Berkeley Linguistic Societyvolume, that was about a really odd phenomenoninvolving coordination. I had found maybe one ortwo examples in a year. It was only after I hadcollected maybe a dozen examples over a period ofseven or eight years that I started to figure out

what was going on. Oneexample of this turned upin an opera review in one ofthe Chicago newspapers. Itwas something like: “TheTemple of Dagon, whoseexterior we see in Act I andinterior in Act III, lookslike a Hollywood movieset.” Now, whose exterior wesee in Act I and interior inAct III? You’ve got a shareddeterminer, whose appliesto both conjuncts. You alsohave gapping, but thepossibility of the gappingthere is contingent uponyour having a shareddeterminer — the Temple ofDagon, whose interior wesee in Act I and whoseexterior in Act III is muchlower in acceptability. AfterI had collected a dozen orso of those examples, therewere enough of them that Idecided this was worthtrying to write a paper

about. So my Flea Circus does pay off!

This Temple of Dagon example you certainlyquoted from the top of your head. So you do have agood memory.

I just happened to hit on the right mental file.

EpilogueAfter we sent the interview to Professor McCawleyfor him to have a look at, one of the comments hesent back to us was the following. It is related toour first question:

I am surprised that in the interview I did notexpress more repugnance at the thought of “every-body working in the same framework, using thesame theory” — that’s as disgusting a prospect aseverybody espousing the same religion or every-body speaking the same language. Fortunately,such a nightmarish situation could arise only outof a political situation that was even more night-marish.

ReferencesChomsky, N.A. (1957). Syntactic Structures. The Hague:

Mouton.Feyerabend, P.K. (1975). Against Method. London: New

Left Books. [3rd edition, 1993.]Heycock, C. (1995). Asymmetries in reconstruction.

Linguistic Inquiry 26, 547–570.Lakatos, I. (1970). The methodology of research pro-

grammes. In Criticism and the growth of knowledgeed. by I. Lakatos & A. Musgrave. Cambridge:Cambridge University Press, pp. 91–196.

McCawley, J.D. (1988). The Syntactic Phenomena ofEnglish. Chicago: University of Chicago Press. [2ndedition, 1998.]

McCawley, J.D. (1993). Gapping with shared determin-ers. Berkeley Linguistics Society 19, 245–53.

McCawley, J.D. (to appear a). Why surface syntacticstructure reflects logical structure as much as itdoes, but only that much.

McCawley, J.D. (to appear b). Some interactions be-tween negation and inversion.

Newmeyer, F.J. (1980). Linguistic Theory in America.Orlando: Academic Press. [2nd edition, 1986.]

IN THIS ISSUEjlawler/glot3-5.pdfState-of-the-Article Glot International, Volume 3, Issue 5, May 1998...

Documents

Transcript of IN THIS ISSUEjlawler/glot3-5.pdfState-of-the-Article Glot International, Volume 3, Issue 5, May 1998...