Conditionals - Web Hosting at UMass Amherst - University of

50 Pages·2004·0.2 MB·English

Conditionals
RajeshBhatt,RoumyanaPancheva
[email protected],[email protected]
1 Overview
This contribution provides a survey of the major issues in the syntax of con-
ditionals. Wepresentandcriticallyevaluatethefindingsintheliteratureper-
taining to conditional structures. We furthermore advance a particular view,
not articulated in its entirety before, namely that conditional clauses (e.g., if-
clauses)areessentiallyfreerelativesofpossibleworlds. Similarlytothemore
familiarinstancesoffreerelativesofindividuals,conditionalclauses(i)likely
involve clause-internal operator-movement to Spec, CP; (ii) they receive the
interpretationofdefinitedescriptions; and(iii)theymayparticipateincorrel-
ative structures as happens in the subcase of conditionals with the proform
then.
Conditionalstructuresinvolveanadverbialclause,oftenreferredtoasthe
CONDITIONALCLAUSE,ANTECEDENTorPROTASIS(theunderlinedconstituent
in (1)), and a main clause, known as the CONSEQUENT or APODOSIS. Con-
ditional structures are interpreted, in general terms, with the proposition ex-
pressedbytheantecedentclausespecifyingthe(modal)circumstancesinwhich
theproposition expressedbythemainclauseistrue. Thus, (1)statesthatthe
possibleworlds/situationsinwhichAndreaarriveslate(thedenotationofthe
conditional clause) are possible worlds/situations in which Clara gets upset
(thedenotationofthemainclause).
(1) IfAndreaarriveslate,Clarawillgetupset.
Conditionalsasin(1)areknownasHYPOTHETICAL CONDITIONALS. They
arethemostcommonkindofconditionalstructuresdiscussedintheliterature,
andconsequently,ourcontributionwillmostlyfocusonthem. Othertypesof
conditionals exist as well, notably RELEVANCE CONDITIONALS, as illustrated
in(2):
(2) Ifyouarethirsty,thereisbeerinthefridge.
InthecaseofRelevanceconditionals,clearlytheantecedentdoesnotspec-
ifythecircumstancesinwhichthepropositionexpressedbytheconsequentis
true, as the latter is, in fact, asserted to be true (in the world of evaluation).
Rather,the possibleworlds/situations in which theproposition expressedby
the antecedent is true, are possible worlds/situations in which it is relevant,
from the perspective of the speech act, that the proposition expressed by the
consequentclauseistrue. ItisasifinaRelevanceconditional,thereisanim-
plicitperformativeclauseembeddingthesurfacemainclause,andthisperfor-
mativeis thetrueconsequent ina(Hypothetical)conditional structure(Ifyou
1
arethirsty,thenitisrelevantforyoutoknowthatthereisbeerinthefridge).1.Here
wewillnotdiscussRelevanceconditionalsinmuchdetail(cf.Iatridou(1991)).2
Conditionals are not unique in their overall structure, rather conditional
clauses belong to a class of adverbial clauses that includes, among others,
clausaladverbialsoftime,cause,andconcession,asillustratedin(3).3
(3) a. IfAndreaarrivedlate,Claramusthavegottenupset.
b. WhenAndreaarrivedlate,Claragotupset.
c. BecauseAndreaarrivedlate,Claragotupset.
d. AlthoughAndreaarrivedontime,Claragotupset.
Like the other clausal adverbials, conditional clauses are typically intro-
ducedby aCP-relatedelement, acomplementizeror anoperatorin Spec,CP
(cf.if,when,because,althoughin(3)above).Andliketheotheradverbialclauses,
conditional clauses may precedeor follow the main clause. Historically, and
typologically, clausal adverbials are related, though of course, in individual
languagestheymayhaveundergonedistinctdevelopmentandasaresultdi-
vergedfromeachother. Forinstance, inEnglish, conditionals allowthepres-
enceofaproforminthemainclause‘linked’totheadverbialclause(i.e.,then),
andconcessivesdotoo(i.e.,still,nevertheless). However,because-clausesdisal-
lowsuchproforms,andwhen-clausesallowthemonlymarginally(cf.(4)).
(4) a. IfAndreaarrivedlate,thenClaramusthavegottenupset.
b. WhenAndreaarrivedlate,(*/??then)Claragotupset(??then).
c. BecauseAndreaarrivedlate,(*forthatreason)Claragotupset(*for
thatreason).
d. AlthoughAndreaarrivedontime, (still/nevertheless)Clara(still/
nevertheless)gotupset.
In this contribution we discuss the internal syntax of conditional clauses
and the possibilities for their attachment in the syntactic structure. We draw
some,thoughnotextensive,parallelswiththeotherkindsofadverbialclauses,
andsuggestadirectionfortheiranalysisinauniformmanner.
Acaveatregardingthescopeofourpresentationisinorder: Hereweonly
discussindetailissueshavingtodowiththesyntaxofconditionals. Someref-
erencetotheirsemanticsismade,whennecessaryfortheexposition, butitis
notputinformalterms,andisnotclaimedtobeinanywayexhaustive.There
1Forthisreason,RelevanceconditionalsareoftencalledSPEECHACTCONDITIONALS
2Iatridou1991discussesanothertypeofconditional,FACTUALCONDITIONAL,whichissome-
whathardertodistinguishfromRelevanceandHypotheticalconditionals. Wenotesomefacts
aboutFactualconditionalsinsection5.
3Someoftheseclausaladverbialsareknownasadverbialsofcontingencyintraditionalgram-
mar(cf. Quirketal.(1985)): conditionals(e.g.,ifq,p),resultatives(e.g.,p,soq),causatives(e.g.,
becauseq,p),exceptives(e.g.,p,except(that)q),purposeclauses(e.g.,p,sothatq),andconcessives
(e.g.,althoughq,p),wherepisthepropositionexpressedbythemainclause,andqistheproposition
expressedbytheadverbialclause.
2
isavastphilosophicalliteratureonthetopicofconditionalreasoningandlogic,
andalsoagrowingnumberofformalsemanticanalysesofconditionalsinnat-
ural language, to which we could not do a proper justice even in references
(still,seeHarperetal.(1981),andJackson(1991)amongmanyothers.)
2 Defining Conditionals
2.1 Whatis aconditional?
Beforewebegin,wehavetoclarifythebasisonwhichwedecidewhetherapar-
ticularsentenceisaconditionalornot. Abovewehavedefinedconditionalsas
structures involving an adverbialclause interpretedas stating the conditions
under which the proposition expressed by main clause is true (or its truth is
relevant,asinthecaseofRelevanceconditionals). Surelythereareothersyn-
tacticwaystoconveyaconditionalmeaning.Thefollowingexamplesillustrate
justsuchcases:
(5) a. Kissmydogandyou’llgetfleas.
b. Foryoutodothatwouldbenice.
Sentence(5a)isinterpretedalongthelinesofIfyoukissmydogyou’llgetfleas;
similarlysentence(5b)receivesaconditionalinterpretationsuchasItwouldbe
niceifyoudothat.CaseslikethesearenotsomepeculiarityofEnglish.Inanum-
ber of languages a structure involving an imperative clause conjoined with a
non-pastindicativeclausereceivesaconditionalinterpretation:theimperative
clauseisinterpretedastheantecedentclauseofaconditional,andtheindica-
tiveclauseinthecoordinationisinterpretedastheconsequent(cf.Clark(1993),
Han(2000),a.o.). Thisstructure-to-meaningmappingappearstobecomposi-
tional, given that the coordinating conjunction or has the semantic import of
the coordinating conjunction and plus negation of the proposition expressed
bytheimperativeclause.Inotherwords,whereas(6a),withandasthecoordi-
nator,isinterpretedasifp,q(wherepandqarethepropositionsdenotedbythe
imperativeandindicativeclauses,respectively),(6b),with orasthecoordina-
tor,isinterprtedasif p,q. Thefactsof(6)arefurthermorecrosslinguistically
attested. :
(6) a. Kissmydogandyou’llgetfleas.
pandq ifp,q
!
b. Kissmydogoryou’llgetfleas.
porq if p,q
! :
Turningto(5b),heretherelevantfactsinderivingtheconditionalinterpre-
tationarethenon-finiteness ofthesententialsubject,andthemoodmorphol-
ogy in the main clause. In many languages a specialized mood, often called
3
conditionalmood,isemployedinsuchcases.4 Again,asinthecaseofthecon-
joinedimperative,thetwoclausesinvolvedin(5b)correspondsystematically
to the antecedent and consequent of a conditional: the non-finite sentential
subject is interpretedas the antecedent, andthe main clause is interpretedas
theconsequent ofaconditional. (SeePesetsky (1991)foradiscussion ofsuch
structures.)
Do the regularities characterizing the structure-to-interpretation mapping
insentencessuchastheonesin(5)andtheircross-linguisticavailabilityjustify
considering these to be conditionals? We believe that comprehensive discus-
sions ofconditionality should includeaninvestigation ofcasessuchasthese.
Theterm CONDITIONAL initsstrictsense, however,isbeingusedintheliter-
ature only to refer to constructions involving an adverbial clause merged to a
particularposition in amain clause. We follow this convention hereanddis-
cussfurtheronlysuchadverbialstructures,partlyforreasonsofspace,partly
because the literature has been almost exclusively devoted to the adverbial
structures,butpartlyalsobecausetheconventionalsplitisolatesawell-defined
classofcaseswhosepropertiescanbefruitfullyexplored. Thefactthatwedo
notmentionanyfurthercasesliketheonesin(5)shouldnotbetakentomean
thattheyshouldbeexcludedfromawiderstudyofconditionalexpressions.
2.2 The markingof conditionals
Languagesuseavarietyofmeanstoindicatethataparticularsyntacticstruc-
ture is a conditional rather than some other construction that involves two
clauses. Without an attempt to give an exhaustive description of the range
of options and typological tendencies, we present in this section some of the
commonstructuralmeansofformingconditionals.
Overtmarkingoftheprotasis(theantecedentoftheconditional)appearsto
bethecommoneststrategy,crosslinguistically(cf.Comrie(1986),Zaefferer(1991)).
Thiscanbedonebyemployingparticularlexicalitems(i.e.,freemorphemes),
inflectionalmorphology,orbypurelysyntacticmeans,e.g.,verb-movement).5
TheEnglishif,theGermanwennandfalls,andtheMandarinChineseruguoex-
emplifythemarkingofanantecedentusinglexicalitems,arguablyfunctional
elementsintheCP-domain-complementizersoroperatorsinSpec,CP.
(7) a. Ifitissunny,Iwillgoforawalk.
b. WennSteffigewinnt,wird gefeiert. (German)
if Steffiwins Impers.Passivecelebrate
‘IfSteffiwins,peoplecelebrate.’
4SeeIatridou(2000)whoarguesthattheconditionalmoodisfurtherdecomposable,andoneof
itsconstituentsisapasttensemorpheme,which,however,isnotinterpretedalongatemporalbut
alongamodaldimension.
5By‘purelysyntactic’wemeantheabsenceofaccompanyinginsertionofaspecializedlexical
itemortheappearanceofspecializedinflection.Ofcourse,syntaxstillunderliestheuseoflexical
itemssuchasthecomplementizerif,orthelicensingofconditionalinflectionontheverb.
4
c. Falls Steffigewinnt,wird gefeiert. (German)
in-caseSteffiwins Impers.Passivecelebrate
‘IncaseSteffiwins,peoplecelebrate.’
d. Ru´guoˇ Zhangsanhe¯ jˇıu, woˇ ma` ta¯ (Mandarin)
if ZhangsandrinkwineI scoldhim
‘IfZhangsandrinkswine,Iwillscoldhim.’
Manylanguagesusetemporalwh-pronouns(e.g.,Germanwenn‘when/if’6)
as conditional markers (cf. Traugott et al. (1986)). Another common lexical
deviceforformingaconditionalareinterrogativecomplementizers/operators,
e.g.,Englishifisalsousedinembeddedyes-noquestions.7
Themarkeron theantecedentdoesnot havetobeasingle lexicalitem. It
canalsobeaphraseasisthecasewiththeEnglishincase,andtheSpanishcon
talqueliterally‘withsuchthat.’
(8) Te perdono con tal que vayas. (Spanish)
To-youforgive-1swithsuchthatgo-Sbjv-2s
‘Iforgiveyouifyougo.’
Languagescanalsomarktheantecedentthroughinflectionalmorphology
ontheverbintheantecedentclause.ExamplesofsuchlanguagesincludeWest
Greenlandic,Turkish,andBasque.
(9) (WestGreenlandic,cf.Fortescue(1984))
pakasa-anna-rukku pissanganar-niru-vuq
surprise-just-2s.3s.CONDbe.exciting-more-3s.INDIC
‘Ifyoujustsurprisehim,itwillbemoreexciting.’
Theinflectionthatservesastheovertmarkeroftheantecedentclauseneed
not be unambiguously conditional (i.e., only realized in conditionals). Some
languagesmarkantecedentclausesbyusingimperativeverbs(intheabsence
ofacoordinatingconjunction). Consider(10),fromJacab(toappear):
6Technically,the‘when/if’ambiguityobtainsonlyinnon-pastclausessincepasttensetemporal
adverbialsrequireadifferenttemporalrelativepronoun,wann.
7Finally, epistemicandoptativemodality,copulas, andtopicmarkersaretypologically very
oftenemployedinconditionals(cf.Traugottetal.(1986));someexamplesaregivenbelow:
(i) (Arrernte,cf.Wierzbicka(1997))
a. Ingwenthepeke kwatyeurnte-me.
tomorrow maybewater fall
‘Perhapsitwillraintomorrow.’
b. Kwatyepeke urnte-me,ayengepetye-tyekenhe.
water maybefall 1sgS come-Vb-Neg
‘IfitraisnIwon’tcome’
5
(10) Znaj jakakoj-nibud’inostrannyjjazyk, rabotal by
know-imper.2sgI some-kind foreign languageworkedwould
perevodcˇikom. (Russian)
translator-instr
’IfIknewsomeforeignlanguage,Iwouldworkasatranslator.’
Morphosyntactically,conditionalslike(10)differfromtrueimperativesin,
at least, the absence of subject-verb agreement and the fact that they can be
formedfromverbsthatdonotnormallyappearintheimperative,suchashap-
penorturnout(seeHacking(1998),Jacab(toappear),fordiscussion). Theim-
perativeverb,whichinimperativescanshownumberdistinction,withforms
for2sgand2pl,canonlybe2sginitsuseinthistypeofconditionals.8
The use of subjunctive morphology is another common formal device in
buildingconditionalantecedents. ConsiderthefollowingexamplesfromRus-
sianinthisrespect(fromHacking1998):
(11) Procˇitala by onaetu stat’ju, onasmogla by
read-sg.femsubjshe this-accarticle-accshe can-sg.femsubj
otvetit’ navasˇ vopros. (Russian)
answer-infto yourquestion
‘Had she read/were she to read the article, she would have been/be
abletoansweryourquestion.’
Interestingly,whenthereisnoconditionalcomplementizer,asintheabove
Russianexamples,theverb,whethermarkedassubjunctiveorimperative,un-
dergoesmovementtoC.I-to-Cmovementisinfactanotherformalmechanism
for forming antecedents of conditionals, often employed by languages in the
absenceofanyotherindicator,lexicalormorphological, ofconditional mark-
ing.
InEnglish, I-to-Cmovementisrestrictedtotheantecedentsofcounterfac-
tual conditionals (cf. Iatridou and Embick 1994), and also some future-less-
vivid conditionals, i.e., conditionals that implicate that if p is the proposition
expressedbytheantecedent, pismorelikelythanp(thetermFUTURE-LESS-
VIVID is introduced in Iatrid:ou 2000). In other languages, inversion is less
restricted and is available in indicative conditionals as well, as the German
examplebelowillustrates.
(12) a. HadIknown,Iwouldnothavegone.
8And,giventheabsenceofsubject-verbagreement,pro-dropinsuchconditionalsisimpossible
inRussian. Allthesefeatures,inadditiontotheadjoinedstatusoftheconditionalclause,distin-
guishaconditionalsuchas(10)fromaconjoinedimperativewithaconditionalmeaning,asin(i),
againfromJacab(toappear).
(i) Udar’ menjai ty ob etompozˇalejesˇ!(Russian)
hit-2sgme andyou-sgaboutthis regret
’Hitme(you-sg)andyouwillbesorryaboutit.’
6
b. Werehetocome,wewouldnotgo.
c. Hast du was, dannbistdu was (German)
Haveyousomethingthen are yousomething
‘Ifyouhavesomething,thenyouaresomething.’
The preceding discussion might suggest that the explicit marking of the
antecedentiscrosslinguisticallyobligatory. Thisisnotthecase. InBengali(cf.
Comrie(1986))andHindi, forexample, itis thepresenceofthe thenwhich is
obligatory,notthepresenceoftheif.
(13) (Hindi,cf.McGregor(1995))
(agar)mehnat karoge to safal hoge
if hard-workdo-Fut.2Plthensuccesfulbe-Fut.2Pl
‘Ifyouworkhard,you’llbesuccessful.’
Themarkeroftheapodosis(themainclause)inBengaliandHindiisclearly
ofpronominalorigin. Comrie(1986)notesthatallinstancesofovertapodosis
markingknowntohiminvolveparticles,oftenofpronominalorigin. Hesug-
geststhatthesemaythereforebeanalyzableasresumptivepronouns.
Finally,thereareconditionalconstructionswherenoovertmarkingofany
sortseemstobenecessary. MandarinChineseallowsforconditionalinterpre-
tationintheabsenceofanyovertmarkingofconditionality, since rugou‘if’is
optional,andsoisthepronominalintheconsequentclausejiu‘then’:
(14) (ru´guoˇ)Zhangsanhe¯ jˇıu, woˇ (j`ıu) ma` ta¯ (Mandarin)
if ZhangsandrinkwineI thenscoldhim
‘IfZhangsandrinkswine,(then)Iwillscoldhim.’
However,Comrie(1986)notesthatintheabsenceofanyovertconditional
marking,asentencelike(14)isambiguousbetweenavarietyofrelationshold-
ingbetweenthetwoclauses(e.g.,if/when/because).9
In summary, conditionals are formed through a variety of means. They
share a basic bi-clausal structure, with the antecedent adjoined to the main
clause, though, as we will see below in the discussion of adnominal condi-
tionals, other adjunction possibilities are also attested. The internal syntax
9SimilarfactsobtaininthecaseoffreeadjunctsinEnglish.Afreeadjunctsuchastheonein(i)
canbeinterpretedasanadjunctofcondition,cause,orconcession,dependinginpartonthetense
ofthematrixclause:
(i) a. Workinghard,Iwillfinishat8.
b. Workinghard,Ifinishedat8.
Stump (1985) proposes that the grammar assigns a conjunctive interpretation to a free ad-
junctstructure(I(will)work(ed)hard&Iwillfinish(ed)at8),withcontextprovidingtheeventual
interpretation.
7
of the antecedent clause involves the CP-domain, where presumably clause-
typingfeaturesarelexicalizedbyspecialcomplementizersortheytriggerverb-
movement. A particularly interesting question arises regardingthe structure
ofconditionals: how, in the absenceof aspecializedmarker, such asacondi-
tional complementizer or conditional inflection, can a clausal adjunct receive
theinterpretationofaconditional? Inotherwords,ruguoinMandarinmaybe
positedto carrytherelevantfeaturesthat asemantics forconditionals would
need,andthatwoulddistinguishanadjunctclauseheadedbyruguofromone
headed by e.g., because. However, given that I-to-C movement in English is
clearlynotlimitedtoconditionals,thatis,itisalsofoundinmatrixquestions,
and in certain focus contexts,10 why is it that an adjunct as in (12a) is inter-
pretedasaconditionalratherthanasabecause-clause? Similarfactsobtainin
otherlanguagesbesidesEnglish,thatemployI-to-Cmovementinconditionals.
3 Structural Issues
WebeginwithadiscussionofsimpleconditionalswithoutthenlikeIfyouopen
the refrigerator, it won’t explode. Once these basic cases have been discussed,
wewilllookatconditionalswiththenandthedifferentstructuralpropertiesof
otherkindsofconditionals.
Abasicobservationisthatif-clausescanbeclause-initialorclause-final.To
thelimitedextenttowhichtheycanbeclause-medial,theymustbesetoffby
parentheticalintonation.
(15) (fromLasersohn(1996))
a. clauseinitial:
Ifyoubotherhimlongenough,Johnwillgiveyoufivedollars.
b. clausefinal:
Johnwillgiveyoufivedollarsifyoubotherhimlongenough.
c. clausemedial:
John,ifyoubotherhimlongenough,willgiveyoufivedollars.
*Johnifyoubotherhimlongenoughwillgiveyoufivedollars.
We will address two questions concerning the if-clause here. First, what
is the structural location of the if-clause with respect to the main clause, and
second,whatistheclause-internalsyntaxoftheif-clause.
10Consider(i):
(i) a. NeverbeforehasDanaseenanythinglikethis.
b. OnlywithRebeccawillPatbehappy.
8
3.1 The Attachmentof theif-clause
Greenberg(1963)statesthefollowinguniversalconcerningthelinearorderbe-
tweentheantecedentandtheconsequentclauseofaconditional.
Universal of Word Order 14: In conditional statements, the con-
ditional clause precedes the conclusion as the normal order in all
languages.
Comrie (1986) claims that while many languages allow for both clause-
initial and clause-final placement of the if-clause, there are also rigidly verb-
finallanguages,wheretheclause-finaloptionisunavailable.11,12
GreenbergandComrie’sobservationsareimportant,butitshouldbenoted
that they are observations about the surface positions of if-clause. Typolog-
ical tendencies are compatible with if-clauses having an origin distinct from
their clause-initial surface position. As we will see soon, there is evidence
that the clause-initial/clause-final difference reflects differencein attachment
height and that at least some clause-initial if-clauses need to be derived via
movementfromclause-finalif-clauses.
3.1.1 If-clausesasAdverbialClauses
Onelogicallypossibleanalysisofconditionalsentencesisthatthetwoclauses
arecoordinatedsyntactically,withiffunctioningasa‘trueconjunction,’tobor-
row a phrase from Jespersen (1956). There have not been explicit syntactic
proposalsthattheantecedentandconsequentofconditionalsentencesareco-
ordinatedclauses, as faras we areaware.13 There are, in fact, important dif-
ferencesbetweenifandacoordinatorlike and. Whereasif-clausescanappear
11Alanguage,inwhichComrieclaimsthisisthecase,isTurkish.Asfaraswehavebeenableto
verifywithnativespeakers,however,allthefollowingwordordersarepossibleinTurkish:
(i) a. Bizegerhava gu¨nesliol-ur-sa disaric¸ik-acag-iz.
weif weathersunny be-pres.-cond.out go-fut.-2pl
‘Ifitissunnywewillgoout’
b. Bizdisaric¸ik-acag-iz egerhava gu¨nesliol-ur-sa
weout go-fut-agr.2nd.plif weathersunny be-pres.-cond.
‘Wewillgooutifitsunny’
c. Egerhava gu¨nesliol-ur-sa bizdisaric¸ik-acag-iz.
if weathersunny be-pres.-cond.weout go-fut.-2pl
‘Ifitissunny,we’llgoout’
12Additionally,GreenbergandComrie’stypologicalclaimmayneedtobequalified.Geis(1985)
citesZwicky(p.c.)fortheclaimthattherearelanguageswhichcannotplaceif-clausesinsentence
initialposition.Wehavenotbeenabletoverifythisclaimforaparticularlanguage.
13Geis(1985)pointsoutthatHeina¨ma¨ki(1974)proposedthattemporalconnectiveslikewhen,
beforeanduntilbeanalyzedascoordinatingconjunctions.Giventhesimilaritiesbetweenwhenand
if,itseemsreasonabletoassumethatHeinamaki’sproposalcouldextendtoif.
9
bothsentence-initiallyandsentence-finally,thesameisnottrueinthecaseof
coordinatedstructuresinvolvingand/but/or.
(16) a. Joewillleaveand/but/orMarywillstay.
b. *And/but/orMarywillstay,Joewillleave.
Furthermoreonlyandeven canmodify if-clausesbut not second conjuncts
incoordinations.
(17) a. Leewillgiveyoufivedollarsonly/even[ifyoubotherhim].
b. Leewill giveyou five dollars(*only/*even) [and/but/orKen will
giveyouten].
Inbeingabletoappearbothsentence-initiallyandfinallyconditionalclauses
arelikeotheradverbialclauses:
(18) a. Iwillleaveatnoon/becauseyouleave.
b. Atnoon/becauseyouleave,Iwillleave.
Clearevidencethatsentence-finalif-clausesareconstituentsoftheVPand
thereforeadverbialscomesfromVPdeletionanddosoanaphora.Themostcon-
servative interpretation of the data below is that the place-holders referback
to constituents, and therefore that the conditional clauses are constituents of
theVP.Hencetheyarealsoadverbials(giventhattheyarenot nominalargu-
ments).
(19) a. IwillleaveifyoudoandJohnwill ,too.
leaveifyoudo
b. IwillleaveifyoudoandJohnwilldosotoo.
Evidencefor the adjuncthood of if-clausesalso comes fromtheir behavior
underit-clefting.if-clausescanbecleftedbutnotoutofawh-island.
(20) (fromCollins(1998))
a. ItisifthestudentfailsthattheteacherwillfiretheTA.
b. ?It is if the student fails that Bill said that the teacher would fire
theTA.
c. *ItisifthestudentfailsthatBillwonderswhytheteacherwillfire
theTA.
Theseverityoftheviolationin(20c)-whatusedtobeanalyzedasanECP
rather than a Subjacency violation - suggests that the if-clause is an adjunct
(giventhatitisclearlynotasubject).
The data involving modification by only and even (cf. 17), and VP ellip-
sis phenomena (cf. 19) provide strong evidence against the view that the an-
tecedent and consequent of conditionals are coordinated. They support the
viewthatif-clausesareadverbials,similarly totemporalphrasesandclauses.
Furthermore,pronominalizationbythensuggeststhatif-clausesareadverbials,
sincetheiranaphoricreflex-then-isanadverb.
10