Gapping and Interpretation
Author(s): Muffy E. A. Siegel
Reviewed work(s):
Source: Linguistic Inquiry, Vol. 15, No. 3 (Summer, 1984), pp. 523-530
Published by: The MIT Press
Stable URL: http://www.jstor.org/stable/4178400 .
Accessed: 17/01/2012 22:15
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 [email protected].
The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.
http://www.jstor.org
SQUIBS
AND DISCUSSION
523
Sag, I. (1983)"On ParasiticGaps," Linguisticsand Philosophy
6, 35-45.
Selkirk,E. (1972)ThePhrase Phonologyof Englishand French,
Doctoral dissertation,MIT, Cambridge,Massachusetts.
Distributed(1981)by the IndianaUniversity Linguistics
Club, Bloomington.
Selkirk,E. (1984)Phonologyand Syntax:TheRelationbetween
Sound and Structure, MIT Press, Cambridge,Massachusetts.
Sells, P. (1983) "Junctureand the Phonologyof AuxiliaryReductionin English," in D. Finer and T. Borowsky, eds.,
Universityof Massachusetts Occasional Papers in Linguistics 8, GraduateLinguistics Students Association,
Amherst.
GAPPING AND INTERPRETATION
Muffy E. A. Siegel,
Temple University
Recently, Gappinghas usually been treated as an interpretive
rule that assigns controllers to or copies materialinto empty
categories, although some linguists have continuedto treat it
as a late syntactic deletion rule.' An interpretiveanalysis of
Gappingand other "deletion under identity" rules means that
"deletionscan be narrowlyrestrictedin sentencegrammar,perhapsjust to deletion of certaingrammaticalformativesand pronouns" (Chomsky (1977, 132)). The interpretiveanalysis of
Gapping, then, has come about largely as a result of the
metatheoreticalaim of constrainingthe various componentsof
grammar.In this squib, I presentmoredirectevidencethatGapping cannot be maintainedas a syntactic deletion rule, but I
also show that there is an unanticipatedmetatheoreticalcost to
treatingit interpretively,because it is sensitive not only to the
logical form of the LF component, but also to the effects of
Affix Hopping,a local rule presumablyamongthe stylisticrules
of the PF componentof the "Y" grammarsoutlinedin Chomsky
and Lasnik (1977) and subsequentwork.
My evidence againstGappingby deletioncomes fromnegative gapped sentences. Although many writers have claimed
This is a greatlyrevised version of a paperpresentedat the 1979
LSA annualwintermeeting.I would like to thankJill Carrier-Duncan,
Jeff Kaplan, Nikki Keach, Tony Kroch, James McCawley, Gary
Milsark,Debbie Nanni, BarbaraPartee, Ken Safir, Justine Stillings,
GregStump,EdwinWilliams,andanonymousLI readersfor theirhelp.
They are, of course, not responsiblefor any errorsI have made, nor
do they necessarilyagree with what I have said.
' See, for instance, Jackendoff(1972), Fiengo (1974), Chomsky
(1976), Williams (1977), Koster (1978), Wasow (1979), and Pesetsky
(1982)for interpretiverules of Gapping,and Hankamer(1973),Stillings
(1975), Kuno (1976), Hankamerand Sag (1976), Neijt (1979), and Sag
(1980)for deletion rules.
524
SQUIBS
AND DISCUSSION
that negative sentences simplycannotbe gapped,2the resulting
strings, such as (1), are not only grammatical,but ambiguous
between a narrowscope negationreading(a conjunctionof negations, as paraphrasedin (2)) and a wide scope negationreading (negationof a conjunction,as paraphrasedin (3)).
(1) Wardcan't eat caviar and Sue, beans.
(2) Wardcan't eat caviar, and Sue can't eat beans.
(3) It is not possible (or desirable)for Wardto eat caviar
and for Sue (simultaneously)to eat (merely)beans.
(1) can have the readingparaphrasedin (2) in a context like,
"Oh, no, I made caviar and beans for dinner,and then I found
out that Ward can't eat caviar and Sue, beans." (1) can have
the readingparaphrasedin (3) if the context is, instead, "Sue
is supposed to be Ward's honoredguest at dinner.Wardcan't
eat caviar and Sue, beans." The narrowscope readingis true
just in case neither person can eat the food named. The wide
scope readingis truejust in case it can't be that they both eat
the foods mentioned.3
Eithera deletion rule of Gappingor a rule of interpretation
can easily account for the narrow scope readingof (1) paraphrased in (2). The former could delete the tense, negation,
modal, and verb of the second conjunctin (2), to producethe
stringin (1). The latter could copy, or choose as controllerin
LF, the tense, negation, modal, and verb of the first conjunct
for the empty categoriesgeneratedin place in the gap in (1), as
in (4):4
(4) Wardcan't eat caviar and Sue AAUx Av beans
2
Ross (1970), Jackendoff(1971), Neijt (1979), and Sag (1980)all
take this position, and Stillings(1975)agrees, thoughwith reservations.
3 It is worth noting that the scope ambiguitydiscussed here can
arise with any sentence operator,not just with negatedmodals.There
are both modalless gapped sentences ((9) and (10) in the text) and affirmativegappedsentences ((i) below) that exhibitthe same ambiguity.
As with examples(1)-(3) in the text, (i) is ambiguousbetweena narrow
anda wide scope readingof the tense andmodal,(ii) has only the narrow
scope reading,and (iii) has only the wide scope reading;it asserts the
possibilityof a conjunction.
(i) Wardcan eat caviar and Sue, beans.
(ii) Wardcan eat caviar and Sue can eat beans.
eats
(iii) Wardcan eat caviar and Sue eat beans.
' Like Williams(1977), I am assumingthe Empty StructuresHypothesisof Wasow(1979);deltasrepresentlexicalcategoriesonly. Since
NOT and Tense are not lexical categories, I will assume that empty
nodes for them are not generated,butthatthey can be copiedor chosen
as controllers(I remainneutralon this point, althoughI cease mentioning both possibilitiesin what follows) by interpretiverules if and
only if the memberof a lexical categorythat they reciprocallyc-commandis.
SQUIBS
AND
525
DISCUSSION
Analyses of the wide scope readingof (1) paraphrasedin
(3), however, are not so straightforward.A stringlike (5) seems
to be the only possible deletion source for the wide scope reading, since it is the only grammaticalsurface structurewith the
appropriatereading.
(5) Wardcan't eat caviar and Sue eat beans.
However, in order for a deletion analysis to work, the tenselessness of the second conjunct in (5) must somehow be accounted for. This would seem to be a hopeless task, though,
for examples (6)-(8) show that tenseless second conjunctsare
not freely generatedin Gappingenvironments,nor can Gapping
always delete just Tense, as Williams(1981)suggests.
(6) *Wardeats caviar and Sue eat beans.
(7) *Oh, no! Chris isn't in the den and the baby be in the
boiler room!5
(8) *Lindahasn'treadherpaperor Richellehave readhers.
Worse for the deletion analysis, (7) and (8) have grammatical
gappedversions with wide scope readingsfor which there is no
reasonable deletion source, given the ungrammaticalityof (7)
and (8):6
(9) Oh, no! Chris isn't in the den and the baby, in the
boiler room!
(10) Linda hasn't read her paper or Richelle, hers.
An interpretiverule of Gappingwould, however, derive the
wide scope readingsof (1), (9), and (10) without the puzzle of
surfacesources with tenseless verbs. If the wide scope readings
are to be accounted for, the rules that establish the scope of
sentence operatorsmust raise operators, on occasion, to produce structureslike (11) in LF.
(11) NOT CAN [s[s Ward[AUXpresent][veat] caviar]and
[s Sue AAux teal
beans]]
A rule of Delta Interpretation,such as that in Fiengo (1974),
could then link the empty elements in the second conjunctwith
the appropriateconstituentsof the first conjunctin (11) to produce wide scope gapped readings.
S I am gratefulto the anonymousreader who broughtto my attention the existence of sentences like (7).
6 Exampleslike (6)-(10) pose a problemfor any late syntacticdeletion analysis, includingthe one in Sag (1980),which nearlyobliterates
the usual distinctionsamong the levels of grammarby allowingrules
like Gappingto be formulatedas syntactic deletion rules that nevertheless have access to informationfromlogicalformfor theirstructural
descriptions.Althoughsuch mixed rules would solve the problemfor
deletion illustratedin (16)-(18), they could do nothingto supplygrammaticaldeletion sources for sentences like (9) and (10).
526
SQUlBS
AND DISCUSSION
It remains, though, not only to discuss at greater length
the operator-raisingrulethatyields structureslike (11), to which
I return,but also to investigatewhetheran interpretiveanalysis
can explain the ungrammaticalityof (6)-(8). That is, can Gapping by Delta Interpretationbe made to treat (12)-(14) differently from the way it treats (11) so as not to predictthat (12)(14) will yield grammaticalsentences with only the Tense
empty, as (11) does?
eat caviarand Sue AAUx eat beans
(12) Ward[present]AUx
be in the den and the baby
(13) NOT [s Chris[present]AUx
AAux be in the boiler room]
have + en read her paper
(14) NOT [s Linda [present]AUx
and Richelle AAux have+ en read hers]
Gappingcan distinguishin the requiredmanner, but only by
adheringto the following constraint:an empty Aux cannot be
interpretivelylinked to Tense in a previous conjunct if Affix
Hopping in the previous conjunctwould move the Tense to a
position not syntacticallyparallelto the second conjunctdelta.
Clearly, the way to build this constraintinto the grammaris to
stipulatethat the results of Affix Hopping must be in place in
S-structure,the inputto LF. Then the requirementof syntactic
parallelismof deltas and their controllers(Fiengo (1974))will
ensurethe desiredresults. WithAffix Hoppingdone in S-structure, as in (15) ratherthan (12), the reason for the deviance of
(6)-(8) and sentences like them becomes clear;there is no syntactically parallelelement in the first conjunctto link with the
delta in the second conjunct:7
(15) Ward [eat+present]v caviar and Sue AAUXeat beans
However, thereare seriousproblemswith performingAffix
Hoppingin S-structure.First, Affix Hoppingis the kindof lowlevel, local housekeeping rule that might best be assigned to
the PF componentof the grammar.More seriously, it has been
assumedthat Tense, as part of Infl, must be a daughterto S in
S-structure,not hoppedandadjoinedto V, so thatit maygovern
and assign nominativeCase to the subjectNP (Chomsky(1982,
52)). Consequently,interpretiveGapping'srequirementthat affixes be hoppedin S-structurecould prove troublesomeenough
to warrantan attemptto returnto Gappingas a late syntactic
deletion rule. However, in addition to the absence of grammatical deletion sources for some gapped sentences, Gapping
exhibits anothercharacteristicindicatingthat it must be an interpretiverule and apply at LF. It is sensitive to logical form.
7 Affix Hoppingmustbe definedon affixesor categoriesof affixes,
not on lexical categorieslike Aux. On the other hand, the empty node
in the second conjunctin (15) must be an Aux node, since deltas represent lexical categoriesonly and Tense, Agr, Infl, etc., are not lexical
categories(see footnote 4). Consequently,we do not expect an empty
Aux node to hop in (15).
SQUIBS
527
AND DISCUSSION
Considerthe following examples:
(16) Tommy wanted to explore the haunted house, and
Susie
wanted4to sleep in the ghost's bedroom.
(17) a. Each of the boys wanted to explore the haunted
house and Susie?*0 anted to sleep in the
ghost's bedroom.
b. Susie wanted to explore the hauntedhouse and
each of the boys
?*0anted to sleep in the
ghost's bedroom.
(18) a. None of the boys wanted to explore the haunted
house and Susie
wanted
to sleep in the
ghost's bedroom.
b. Susie wanted to explore the haunted house and
none of the boys
*0anted
to sleep in the
ghost's bedroom.
The ungappedversions in (16)-(18) are all perfectlyacceptable,
as is the gapped version in (16). The gapped versions in (17),
though, sound decidedly odd, and the gapped versions in (18)
are even worse. A late syntactic deletion rule of Gappingwith
no access to logical form could not avoid producingthe gapped
versions in (17) and (18) if it is to producethe one in (16). Gapping by Delta Interpretation,however, could be constrainedto
prevent the unwantedGapping.Once the scope of quantifiers
has been determined,the gappedversion of (18a),for instance,
would look like (19), ignoringirrelevantfeatures:
(19) (-3x(x is a boy & x wanted to explore the haunted
house)) & (Susie AAUxAV to sleep in the ghost's bedroom)
The condition necessary to block gapped sentences as in (17)
and (18) seems to be that AAUx and Av cannot be interpretively
linkedwith materialif either partyto the linkingis in the scope
of an operator,while the other party is not. Let us call this the
Operator/Verb Interpretation condition (OVI).8
8
OVI apparentlyapplies only in sentence grammar,as distinguishedfromdiscoursegrammarin Williams(1977).It does not operate,
for instance, with "VP deletion" sentences:
(i) None of the boys wanted to explore the hauntedhouse, but
Susie did.
It does, however, apply in the case of conjunctionreduction,if somewhat more weakly than with Gapping:
(ii) *Susie and none of the boys finishedon time.
(iii) ?Susie and all of the boys finishedon time.
528
SQUlBS
AND DISCUSSION
OVI, it turnsout, also figuresin an accountof the behavior
of the originaldata in (1)-(3). There are three ways in which
deltas can satisfy OVI if there is any kind of operatorwith potentialsententialscope in the first conjunct-a modal,negation,
or a quantifier.The first way (Case A) is for a second operator
to be generatedin the second conjunct.This, indeed, seems to
improve sentences like (17) and (18):
(20) All of the boys wantedto explore the hauntedhouse,
all 1
and a few of the girls, to sleep in the ghost's bedsome J
room.9
The second way (Case B) is for the operatorin the first conjunct
to be part of the controller for the A's. In such a case, the
operatorwill be copied inftoor control a A in the second conjunct and thus have scope over the A's there. Case B can obtain
only when the first conjunct operators originatein Aux or V
and have their narrowestscope. For if they originateelsewhere
or are raised in LF, they will not be in an appropriateparallel
position to control the second conjunct's A's (see Fiengo
(1974)). The unavoidableresult of Case B in the data we have
been considering, then, is the narrow scope interpretationof
modals and negationassociated, for instance, with the reading
of (1) paraphrasedin (2). The thirdway that first conjunctoperators can come to have scope over second conjunctA's and
thereby satisfy OVI (Case C) is by having operatorsthat are
interpretedas having scope over both conjunctsat once. This
case is illustratedin (11), which shows the logical form for the
wide scope reading of (1) paraphrasedin (3), and in (13) and
(14).
Such structures as (11), (13), and (14) presumablycome
about as a result of raisingthe operatorsof the first conjunct.
However, if operatorsare adjoinedonly to the beginningof the
first conjunct, as in (21), which is what one might expect for
the scope of sentence operators, OVI will disallow Delta Interpretation,since the first conjunct's Aux and V are in the
scope of operatorsthat do not have scope over the empty categories.
9 A second quantifierintroducedto satisfyOVIseems to be subject
to an interestingseparateconstraint,whichI cannotpursuein this squib.
The second quantifiermust match the first in its monotonicity,as defined in Barwiseand Cooper(1981),that is, roughlyspeaking,its positiveness or negativeness.Contrast(i) with (20).
(i) All of the boys wanted to explore the haunted house and
~none]ate
neither of the girls w
to sleep in the ghost's bed-
few{w
room.
SQUIBS
AND
DISCUSSION
529
(21) [s (NOT CAN [s Wardpresenteat caviar])and [s Sue
AAux Av beans]]
OVI's preventionof Delta Interpretationin a structurelike (21)
predictscorrectly that (1) cannot mean 'For Wardto eat caviar
is not possible and Sue eats beans'. To producethe correctwide
scope readingsof such sentences, then, there must be operator
raising to the higher sentence and subsequentDelta Interpretation, which would produce a structurelike (11) from (21).
This raising to the higher sentence can occur only with
gapped second conjuncts, that is, ones with empty Tense at
least. (2), for instance, with its filled Aux and V, cannot have
the very wide scope reading. Though such a condition on a
special raising rule for a markedconstructionlike Gappingis
easily statable, the reason for it remainsa mystery. However,
such a rule, together with OVI, does predict exactly the ambiguity that we find in sentences like (1). Furthermore,whatever the ultimateform and explanationof the special operator
raisingand OVI, the fact that they must be stated in terms of
logical form constitutes evidence that, although Gappingdepends upon the output of Affix Hopping, it is an interpretive
rule. This means that the role of Tense in the assignmentof
Case, as well as the characteristicsused to distinguishrules to
be assigned to the differentcomponentsof sentence grammar,
may have to be modified, but that deletion under identity, at
least for Gapping,need not be a part of sentence grammarat
all.
References
Barwise, J. and R. Cooper(1981)"GeneralizedQuantifiersand
Natural Language," Linguistics and Philosophy 4, 159-
219.
Chomsky, N. (1976) "Conditionson Rules of Grammar,"Linguistic Analysis 2, 303-351.
Chomsky, N. (1977) "On Wh-Movement,"in P. Culicover,
T. Wasow, and A. Akmajian,eds., Formal Syntax, Academic Press, New York.
Chomsky, N. (1982) Lectures on Government and Binding,
Foris, Dordrecht.
Chomsky,N. and H. Lasnik (1977)"Filtersand Control,"Linguistic Inquiry 8, 425-504.
Fiengo, R. (1974) Semantic Conditions on Surface Structure,
Doctoral dissertation,MIT, Cambridge,Massachusetts.
Hankamer,J. (1973)"UnacceptableAmbiguity,"LinguisticInquiry 4, 17-68.
Hankamer,J. and I. Sag (1976)"Deep and SurfaceAnaphora,"
Linguistic Inquiry 7, 391-428.
Jackendoff, R. S. (1971) "Gappingand Related Rules," Linguistic Inquiry 2, 21-35.
530
SQUIBS
AND DISCUSSION
Jackendoff,R. S. (1972)SemanticInterpretationin Generative
Grammar,MIT Press, Cambridge,Massachusetts.
Koster, J. (1978) Locality Principles in Syntax, Foris, Dordrecht.
Kuno, S. (1976) "Gapping:A FunctionalAnalysis," Linguistic
Inquiry7, 300-318.
Neijt, A. (1979) Gapping:A Contributionto Sentence Grammar, Foris, Dordrecht.
Pesetsky, D. (1982) Paths and Categories, Doctoral dissertation, MIT, Cambridge,Massachusetts.
Ross, J. R. (1970) "Gappingand the Orderof Constituents,"
in M. Bierwischand K. Heidolph,eds., Progress in Linguistics, Mouton, The Hague, 249-259.
Sag, I. (1980)Deletion and Logical Form, GarlandPublishing,
New York.
Stillings, J. T. (1975) "The Formulationof Gappingin English
as Evidence for VariableTypes in Syntactic Transformations," LinguisticAnalysis 1, 247-273.
Wasow, T. (1979)Anaphorain GenerativeGrammar,E. StoryScientia, Ghent.
Williams, E. S. (1977) "Discourse and Logical Form," Linguistic Inquiry8, 101-139.
Williams,E. S. (1981)"On the Notions 'Lexically Related'and
'Head of a Word'," LinguisticInquiry12, 245-274.