Degree Phrase Inversion in the Scope of Negation Author(s): Marianne L. Borroff Reviewed work(s): Source: Linguistic Inquiry, Vol.

37, No. 3 (Summer, 2006), pp. 514-521 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/4179379 . Accessed: 30/07/2012 14:11
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.

http://www.jstor.org

514
DEGREE PHRASE INVERSION IN THE SCOPE OF NEGATION

SQUIBS AND DISCUSSION

1 Introduction

Marianne L. Borroff Stony Brook University

In English,degree-modifiedadjectivestypicallyfollow the determiner. In constituent and echo questions, however, adjectives modified by the wh-degree word how must precede the article (la-d). (1) a. b. c. d. [How good a student]is John? *[A how good student]is John? John is [HOW GOOD a student]? *John is [a HOW GOOD student]?

Similarly, in equative comparatives,the comparativedegree word as must precede the adjective. (2) a. Kevin is [as good a student]as Elaine. b. *Kevin is [an as good student]as Elaine. Degree inversions also occur in certaindialects in the context of negation.' I will referto this type of inversion,which has not received as negativedegree inversion(NDI).2 much discussionin the literature, (3) a. John is not [a very good student]. b. John is not [very good a student]. In contrast to (1) and (2), both the normal ordering, (3a), and the inverted form (3b), are acceptable. That is, the inversion in (3b) is optional.These facts raise the questionof how the inversion apparently in (3b) is licensed. In section 2, I discuss the licensing environmentsof NDI, noting that they pattem similarly to the licensing environmentsof negative polarity items (NPIs). Drawing on the parallelismbetween NDI and NPIs, I proposein section 3 thatNDI licensing is a subcaseof negative polaritylicensing, with the landingsite of the degreephrasemovement headed by an NPI.

Thanks are due to John Bailyn, Marcel den Dikken, Dan Finer, Richard Larson, and Peter Ludlow for their input during the development of earlier versions of this squib. I am additionallygratefulto the audiences at the 2000 LSA annualmeeting and the CUNY Syntax Supperfor their thoughtfulcomments and suggestions. Thanks also to two anonymousLI reviewers for their insightful comments. 1Use of the forms shown in (3) cuts across dialect and registerlines on the basis of age and geographicallocation. However, the degree inversion in (3b) is generallypartof a more informalregister,with some speakersjudging it ungrammatical (though I have observed many of the same speakersusing the constructionin casual speech). 2 NDI involves movement of DegPs headed by very, real, that, and too, thoughan anonymousLI reviewer notes thatforms with very and real may be most widely accepted. This is perhapsdue to the fact that that and too both participatein inversions in positive environmentsas well, though NDI of that and too results in a change of meaning (e.g., that is not necessarily deictic in John's not that good a student).

SQUIBS

AND

DISCUSSION

515

2 The Licensing Environment of Negative Degree Inversion As notedabove, NDI is optional(4a-b). However,it requiresthe presence of a negative element (4c). (4) a. John is not [a very good student]. b. John is not [very good a student]. c. *John is [very good a student]. As (5a-c) indicate, respectively, the relevant negative element may be not, a negative adverbial,or a negative verb. (5) a. Mugsy Boags wasn't/*was[very tall a basketballplayer]. b. She never/*alwayswas [very good a dancer]. c. I doubt/*thinkthat this will be [real interestinga class]. The licensing environmentsfor NDI displayedin (5) recall those observedwith NPIs. Like instancesof NDI, NPIs are licensed only in an "appropriately negative environment"(6).3 (6) a. Daisy isn't/*is anyone I know. b. She never/*alwayswas any good a dancer. c. I doubt/*thinkyou can do anythingabout poverty. Nevertheless, the environmentsfor NDI and NPIs are not identical. Giannakidou(1997, 1998, 1999) characterizesnegative polarity licensing as sensitivity to nonveridicality;in other words, NPIs are An operapossible only in environmentsthatexhibit nonveridicality.4 tor is defined to be nonveridicalif it does not entail the truthof its complementand is veridical otherwise.5 (7) An operatorOpis veridicaljust in case Opp ->p is logically valid. Otherwise, Op is nonveridical. While NDI seems to patternwith NPIs with respectto sensitivity to nonveridicality,it is immediatelyapparentthat the licensing environments for the two elements diverge; NDI is not licensed in all
3 Giannakidou(1997) notes an additionalconstructionin which a DegP is licensed only in the presence of negation. (i) shows that all that Adj is not possible in positive sentences, and (ii) that all that Adj constructionscan also participatein NDI. ((iib) is due to an anonymousLI reviewer.) (i) He wasn't/*was all that intelligent. (ii) a. He wasn't/*was all that intelligent a man. b. John didn't make/*madeall that good (of) an impression. Giannakidou explains the distribution of all thatAdj by proposingthat it is an NPI. 4 In this squib, I differ from Giannakidou (1998, 1999) in using the term negative polarity item (NPI) where she would use the term affectivepolarity item (API), in which choice I believe I am following commonusage. Giannakidou uses API to refer to all elements sensitive to nonveridicality,while she limits NPI to referringto elements licensed by antiveridicality. SThis definitionof nonveridicality is a simplifiedversionof the definition given in Giannakidou1999. I use this particularformulationfor simplicity's sake.

516

SQUIBS

AND

DISCUSSION

nonveridicalenvironments.For example, while NPIs are licensed in the antecedentof a conditional,the scope of before, and the restrictive term of a universalquantifier(8), NDI is not (9). (8) a. If you were anyone famous, who would you be? b. Before anyone arrives at my party, I will have cleaned the house. c. Everyone who is anyone will be at my party. (9) a. *If Johnwere [verydiligent a student],he wouldn'tparty all the time. b. *Before he became [very good a student], John was a very bad student. c. *Everyonewho is [very good a student]should pass the final. While the environmentsin (4)-(6) and (8)-(9) sharethe characteristicof nonveridicality, they differ in thatthe formerare antiveridical as well; beyond simply failing to entail the truthof their complement, the operatorsin (4)-(6) entail the falsity of their complement. The definition of antiveridicalityis given in (10). (10) An operatorOp is antiveridicaljust in case Op p is logically valid.
-

The licensing conditionon NDI is such thatit is grammatical only undernegationand without,both of which are antiveridicaloperators. Unlike any-NPIs, instancesof NDI cannot be licensed by the broader group of nonveridicaloperatorsincluding the antecedentof a conditional, the scope of nonveridicalbefore, and the restrictiveterm of a universalquantifier,as exhibited by the ungrammaticality of (9a-c). The distribution of elementsthattakepartin NDI in Englishis reminiscent of the distribution of anotherset of elements that are sensitive to antiveridicality,namely, minimizers in Greek.6Giannakidou(1998) discusses the distributionof these minimizers, noting that they are limited to appearingin the scope of negationand in 'without'-clauses. LEKSI oli mera. (11) a. *(Dhen) ipe not said.3SGword all day 'She/He didn't say a word all day.' b. ... *(xhoris) na pi LEKSI oli mera. without SUBJ say.3SG word all day ... without saying a word all day.' (Giannakidou1998:(120a), (121b))

6 Notethat whilethedistribution of NDIin English of Greek that parallels thedistribution minimizers, of English minimizers is broader. Thedata in(i) and (ii)illustrate justonesuchdivergence; whileEnglish minimizers arelicensed by factiveenvironments, NDIis not. (i) I regret sayinga wordto himaboutit. (ii) *1regret John'sbeingverybada dancer.

SQUIBS

AND

DISCUSSION

517

In additionto being licensed by the overtantiveridicality exemplified by negation and without, NDI shares with Greek minimizersthe characteristic of being properlylicensed by certainindirectlyantiveridical environments. arerhetorical Among these environments questions, which are assumedto be indirectlyantiveridical because they give rise to a negative implicature; positive rhetoricalquestionsentail negative responses.(12a) shows thatminimizersin Greekarelicensed by rhetorical questions,and (12b) shows that NDI is also licensed in this environment.7 (12) a. Pjos dhini DHEKARAja to ti th'apojino? who give.3sG dime for the what FUT.become. lsG 'Who gives a damn about what will happento me?' (Giannakidou1998:(123c)) b. Who says John's very good a student? The datapresentedin this section suggest thata connectionshould be drawn between the licensing of NDI, the licensing of NPIs, and the licensing of minimizersin particular. I will returnto the question of explainingthe distribution of NDI in section 3. One tentativeconclusion to be drawnfrom examiningthe distribution of NDI is that(anti)veridicalitycan license not only lexical elements, but also a construction exhibiting inversion. To complete the descriptionof NDI, I should mentionone additionalcharacteristic. The examplesin (13) differfromnormalinstances of NDI in the presence of the word of, which intervenesbetween the moved degree phrase and the determiner.For all English speakers I have encounteredwho accept NDI, sentenceslike (13a-c) are at least as good as those shown in (5). Additionally, of those who do not accept plain NDI, many allow NDI with of. (13) a. John is not [very good of a student]. b. That's not [thatbig of a deal]. c. Buy a Neon without paying [very big of a price].

7 Anotherindirectlyantiveridicalenvironmentin which NDI is licensed is created by sarcasm in nonnegative environments.For example, I suggest that the apparentyes/no question in (i) becomes indirectlyantiveridicalwith the properintonation(focus and a steep fall-rise contouron IS), in the sense that the speakeris assertingthat the answer can only be negative. In fact, the NDI in (Sc), repeatedas (ii), may be best interpreted as being made possible throughan indirectlyantiveridical environment inducedthroughsarcasticfocus on doubt,ratherthan as being licensed by the negative implicatureinducedby doubt. This possibility is supportedby the fact that NDI is not possible under certainother such verbs, for example, surprise (iii). (i) IS John very good a student? (ii) I DOUBT/*thinkthat this will be [real interestinga class]. (iii) *I'm surprisedthat this is [real interestinga class].

518

SQUIBS

AND

DISCUSSION

3 An Analysis of Negative Degree Inversion To begin to understandthe structureof NDI, we must explain why the DegP appearsin the atypical initial position of the phrase. One possibility is thatit is base-generated there;the surfaceword orderof, for example, very good of a student reflects a structurein which the adjectivephrase [AP good] takes the prepositionalphrase [pp Of [DP a [NP student]]] as its complement. This structurewould make NDI analogous in form to certainother constructionsof the form DegP P DP: for example, [not very good as a student], [not very good for a student], and [not very envious of his brother].These constructions differ from NDI in a few key ways, including that they are felicitous in positive contexts, do not require an overt degree word, and do require an overt preposition.Additionally, while NDI can appearin subjectposition, a canonical position for nominals,other DegP P DP constructionscannot (e.g., Not very good a studentwalkedin vs. *Not very good as a studentwalked in). These facts suggest that while the DegP P DP constructionsare adjectivalin nature,the NDI examples are essentially nominal. This being the case, I propose to take the thatthe pre-determiner approach positionof the DegP in NDI is derived ratherthan base-generated.In what follows, I propose a movement analysis of the surface structureof NDI.8 At first thought,a naturalapproachwould be to view the moved DegP as landing in Spec,DP.9The fact that of can intervenebetween the inverteddegree phraseand the determiner suggests that this is not the case, however. Discussion of relatedphenomenaby Kennedyand Merchant(2000) can help to shed light on this issue. These authors note thatthe moved degreephrasesin questionsandcomparatives may also precede an of element. Considerthe data in (14). (14) a. [How interesting(of) a play] did Brio write? b. [How tall (of) a forward]did the Lakershire? c. Bob didn't write [as detailed (of) a proposal] as Sheila did. d. He took [so big (of) a piece of cake] that he couldn't finish it. (Kennedy and Merchant2000)

8 Troseth2004, which appeared when this squibwas alreadyunderreview, offers a structurefor NDI in which the adjective in NDI is predicative and degree inversion is an instance of predicateinversion. I am unable to address Troseth's proposals here, but I refer readers to her work for an alternative account of NDI. 9 In this section, I assume thatthe pre-determiner DegP in NDI examples has moved therefrom a DP-internal position in an attemptto preserveuniformity of structure between NDI and the noninvertedexamples. See Abney 1987, Delsing 1993, and Lilley 2000 for nonmovementanalyses of pre-determiner degree phrases in English and other Germaniclanguages. See Lechner 1999 for an analysis in which the degree phraseis not adjoinedto NP, but ratheris a complementof DP.

SQUIBS

AND

DISCUSSION

519

Kennedy and Merchant(2000) take the optional presence of of to signal the existence of a functionalphraseFP dominatingDP. They suggest that Spec,FP is the landing site for inverteddegree phrasesin questions and comparatives.The existence of such a phrase was initially proposed by Bennis, Corver, and Den Dikken (1998) as the landing site of moved NPs in nominal predicateinversion constructions. (15) een beer van een vent a bear of a man (Bennis, Corver,and Den Dikken 1998) (Dutch)

The similarity among the NDI, question, and comparativedata suggeststhatthey shouldbe subsumedunderthe same generalanalysis. Thus, the possibility of the presence of of between the moved degree phraseand the determiner indicates that the DegP has moved into the specifierposition of FP in all cases. Given this approach,NDI has the structure shown in (16). On an early minimalistaccountof movement, inversion would be caused by the need to check strong features.10 (16) a. Johnis not [verygood (of) a student]. b. FP

DegP1

F'

very good

DP

of

NP

tj

NP

student Following Kennedy and Merchant(2000), I diverge here from Bennis, Corver, and Den Dikken (1998), who assume that FP is the complement of DP. Supportfor the approachthat FP selects DP as its complement is that the former apparentlyimposes selectional restrictionson the latter;the DP in NDI must always be indefinite."
10 In more recent versions of the theory, different limitationsare placed on the motivationsfor movement:for example, movementis not a requirement (and thus not possible) for featurechecking. If we accept these limitations,the movement exhibitedby NDI may be caused by an EPP featureon the head of FP. 1l An anonymousLI reviewer notes that this observationcan be tracedto Bresnan's(1973) discussionof the syntaxof comparatives. Matushansky (2002)

520

SQUIBS

AND

DISCUSSION

(17) a. John is not [very good a student]. b. *Johnis not [very good the student]. c. *John and Mary are not [very good some/both/twostudents]. In fact, this conditionholds of every constructionin which FP is proposed to play a role; inversion in questions and comparativesis also possible only with a DP headed by a.'2 (18) a. How fast a/*the/*some/*0car(s)] do you want to buy? b. John is [as good a/*the student]as Mary. This section has stressedthe similaritiesin linear orderbetween NDI and other phenomenainvolving movement within the nominal projection.Despite their structuralsimilarities,NDI differs from the others in that it requiresa c-commandingantiveridicallicenser. The next step is to explain why inversionis possible only in this environment. One way of answeringthis question is to posit that the negative licenser acts to makethe presenceof FP possible, andthatFP is impossible outside of negation. In this approach,negation does not license movement;rather,it licenses the landing site for movement.The fact thatmovementgoes handin handwith FP follows from this approach; if FP enters the derivation,its features must be checked. Thus, any derivationin which FP is presentwill necessarilyexhibit DegP-movement. If we claim thatFP is possible only in the presenceof negation, the conditionson its presenceseem to be exactly those on the presence of NPIs. In fact, we saw in section 2 that the environmentsin which inversionis possible are a subset of those in which NPIs are possible; namely, inversion is possible only when minimizers are present. To capturethis intuition,I propose to analyze the head of FP as an NPI itself. We can reflect the fact that NDI is possible only in a subset of the canonicallicensing environmentsfor NPIs by characterizing F? as sensitive to the presence of a c-commandingantiveridicaloperator. Thus, FP in NDI differs from FP in other constructionsbecause only the formerhas featuresthatmakeit sensitiveto licensingby negation.13

also discusses degree movement in detail, as well as the related requirement that movementof the entire DegP to a DP-peripheral position is only possible with an overt indefinite determiner. 12 Anotheroption availableto comparatives is to use a null determiner just in case the indefinitedeterminer a is not possible because of featuremismatch. (i) John and Mary are [as good students]as Kevin and Elaine. However, this option is not available in questions, English nominal predicate inversion, or NDI. (ii) a. ?Johnand Mary want [HOW FAST cars]? b. *those idiots of men c. John and Mary are not very good students. (no NDI reading) 13 An anonymousLI reviewer raises the question of how to explain that the presence of FP, and by extension, DegP-movement,is obligatoryin echo

SQUIBS

AND

DISCUSSION

521

with FP sharesurfacesimilariThe resultis that,while all constructions ties (inversionof DegP, presenceof of), only in NDI is the distribution of FP limited to negative environments. References Abney, Steven. 1987. The English noun phrasein its sententialaspect. Doctoral dissertation,MIT, Cambridge,Mass. Bennis, Hans, NorbertCorver,and Marcelden Dikken. 1998. Predication in nominal phrases. Journal of ComparativeGermanic Linguistics 1:85-117. Bresnan, Joan. 1973. Syntax of the comparativeclause construction in English. LinguisticInquiry4:275-343. Delsing, Lars-Olof. 1993. On attributiveadjectives in Scandinavian and other languages. Studia Linguistica47:105-125. Anastasia. 1997. The landscapeof polarityitems. GronGiannakidou, Dissertationsin Linguistics 18. Groningen: ingen Outstanding Universityof Groningen,Centerfor Languageand Cognition. Giannakidou,Anastasia. 1998. Polarity sensitivity as (non)veridical dependency.Amsterdam:John Benjamins. Giannakidou,Anastasia. 1999. Affective dependencies. Linguistics and Philosophy 22:367-421. compar2000. Attributive Kennedy,Christopher, andJasonMerchant. ative deletion. Natural Language & Linguistic Theory 18: 89-146. Lechner, Winfried. 1999. Comparativesand DP-structure.Doctoral dissertation,University of Massachusetts,Amherst. Lilley, Jason. 2000. The syntax of Germanicpost-adjectivearticles. Paper presented at ComparativeGermanicSyntax Workshop 16, McGill University. Matushansky,Ora. 2002. Movement of degree/degreeof movement. Doctoral dissertation,MIT, Cambridge,Mass. Troseth, Erika.2004. Negative inversion and degree inversion in the workEnglishDP. In Linguisticsin the Big Apple: CUNY/NYU ing papers in linguistics, ed. by Marcel den Dikken, 1-15. Available at web.gc.cuny.edu/dept/lingulliba/papers/troseth _LIBA.pdf.

questions and comparatives.I assume that in these constructionsthe moved DegP itself has a featurethatforces movementto Spec,FP,and thatderivations in which FP is not present will crash because this movement is unavailable. Anotherintriguingpossibility is that the DegP moves to Spec,FP to get to the left edge of an FP-inducedphase, in orderto remain availablefor subsequent checking of its features (e.g., [ + wh] on the part of echo questions).