The Ambiguity of the English Present Perfect - CiteSeerX

The Ambiguity of the English Present Perfect Author(s): Laura A. Michaelis Source: Journal of Linguistics, Vol. 30, No. 1 (Mar., 1994), pp. 111-157 Published by: Cambridge University Press Stable URL: http://www.jstor.org/stable/4176258 . Accessed: 27/03/2014 22:44 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].

.

Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Linguistics.

http://www.jstor.org

This content downloaded from 128.138.73.68 on Thu, 27 Mar 2014 22:44:11 PM All use subject to JSTOR Terms and Conditions

J. Linguistics 30 (I994), III-I57.

CopyrightC

1994

CambridgeUniversityPress

The ambiguity of the English present perfect' LAURA A. MICHAELIS University of Colorado, Boulder

(Received19 July 1993; revised8 March 1994) This studyexaminesgrammaticaland discourse-pragmatic reflexesof the existential and resultativereadingsof the Englishpresentperfect.I presentboth negativeand positiveargumentsin favorof the claimthat the presentperfectis ambiguous(rather than vague)with respectto thesereadings.In particular,I arguethat the resultative present-perfect representsa formalidiom: a morphosyntacticformcharacterizedby idiosyncraticconstraintsupon grammar,meaningand use. Certainconstraintsupon the resultativepresent-perfect, in particularthat which preventsit from denotinga pragmaticallypresupposedevent proposition,can be MOTIVATED with respectto a discourse-pragmatic oppositioninvolvingthe preterite.However,such constraints cannotbe PREDICTED fromfunctionaloppositionsor any generalsemanticprinciples. Finally,I suggestthat masteryof aspectualgrammarcruciallyentailsknowledgeof such idiomaticform-meaningpairings.

I. INTRODUCTION

Our object of inquiry is the English present perfect (PrP), a verbal periphrasis consisting of the present-tense auxiliary have followed by a past-participial VP (e.g. the Eagle has landed). Semantically, the auxiliary can be regarded as

a sententialoperator(Have)whichscopesa context-freepast-tensesentence. The truth of the resultingpropositionis evaluatedfor the presentinterval. For example,(i) is representedas in (2): (i)

Harryhas finished.

(2)

PRES(W,i)

[Have [AwAiPast(W,j,(Harry finish)]

In (2), the context variables of the past-tense operator have been abstracted,so that the (past)tensein the scope of the perfectoperatorHave is context free: the time i - the referencepoint with respect to which anteriorityis computed- is not anchoredto the time of the utteranceevent. By contrast, the tense operator immediately scoping Have is context

[I] I gratefullyacknowledgethe helpfulsuggestionsand commentsof the followingpeople:

Jean-PierreKoenig, CharlesFillmore,Paul Kay, Knud Lambrecht,Dan Slobin and an anonymousreviewer.None of thesepeoplebearsresponsibility forerrorscontainedherein. III


LAURA A. MICHAELIS

sensitive:it is anchoredto the time of the utteranceevent.I will assumethat the perfectperformsa stativizingfunction: a perfect-formsentenceallows the speakerto encode the occurrenceof an event at a time t-i by asserting that a contingent state obtains at time t. Accordingly,(i) representsa present-tensestativepredication:the state is that of Harry'shavingfinished (Herweg I99 I).

The representationgiven in (2) is interpretedin accordancewith truth conditions requiringthat the argumentproposition be true somewhere within a present-contiguousintervalor 'extendednow': (3) Have (A) is true in M relativeto (w, i) iff thereis a subintervalj of i such that A is true in M relativeto (w,j). (RichardsI982) By contrast,in the definitionof the tenseoperatorPast, the intervalsi and j areinterpretedas instants:the timeof evaluationi does not includethe time j, at which the argumentpropositionholds. Richards'definitionof Past is given in (4): (4) Past (A) is truerelativeto i iff thereis an intervalj earlierthan i such that A is true relativeto j. Under this view, what distinguishesthe PrP from preteriteis that, in the formercase, the argumentpropositionmust be true somewherewithin the presentinterval.In the lattercase, the argumentpropositionmustbe trueat an intervalwhich is prior to the present moment; this intervaldoes not overlapwith or subsumethe presentmoment. The truth-conditionaldefinitionof the PrP given in (3), while a useful point of departure,is inadequate.The definitionsin (3) and (4) do not capturethe meaningdifferenceseen in contrastpairs like (5): (5) (a) I have willed my fortuneto Greenpeace. (b) I willed my fortuneto Greenpeace. There seems to be no reason to presumethat (5a) expressesan event proposition which is true somewhere within an interval whose upper boundaryis now, while (sb) does not. The presumedmeaningof (5a), in which a singleevent of bequestoccurredat a particularpoint in the past, is not compatiblewith expressionslike before,whichdenotea generalpresentcontiguouspast interval.The definitiongivenin (3) does not capturethe fact that sentenceslike (5a)areinterpretedin a way thatmakesthemsynonymous with sentenceslike (sb): a singleevent occurredat some point priorto now. Furthermore,(3) does not enable us to account for the array of PrP interpretationswhich are commonly cited (continuative,resultative,etc.). We will investigatethe proposalthat the PrP is ambiguouswith respectto these interpretations.These readingswill be representedby means of an II2


THE ENGLISH

PRESENT PERFECT

eventuality-basedtemporal logic like that proposed by Parsons (I990), augmentedby 'provisos'representingconventionalimplicata.I will presume that each of these readingsrepresentsa distinctgrammaticalconstruction, i.e. a unique,conventionalizedpairingof form and meaning.The ambiguity at issueis a nonsyntacticambiguity:the distinctreadingscannotbe assigned distinct underlying syntactic structures. Therefore, the PrP form is polysemousin muchthat sameway that wordsmay be polysemous:a single formhas severalrelatedmeanings.Themeaningsto be examinedherearethe three primary readings distinguished by McCawley (197I). They are exemplifiedand paraphrasedin (6): (6) (a) We'vebeensittingin trafficfor an hour. (universal/continuative) A state obtains throughout an interval whose upper boundary is speech time.

(b) We've had this argument before. (existential/experiential) One or more events of a given type are arrayed within a present inclusive time span.

(c) The persons responsible have been fired. (resultative) The result of a past event obtains now.

I will argue that the PrP, althoughformallyparallelto past and future perfects, has semantic, grammaticaland discourse-pragmaticproperties whichdistinguishit from the otherperfectforms.This analysisrequiresthat we reject the compositionalsemanticaccount offeredby Klein (I992), in whichthe meaningsof past, present,and futureperfectsare reducibleto the semanticcontributionof the particularauxiliarytense and the anteriority relationexpressedby the participle.Certaingrammaticalcharacteristicsof the PrP 'cannotbe componentiallyexplainedfrom a semanticpoint of view' (Richards I982: IOI). One such characteristicis the constraint barring 'definite' time adverbs,exemplifiedin (7a): (7) (a) Harryhas joined the navy (*in I960). (b) [It was 1972.] Harryhad joined the navy in

I960.

(8) (a) [Yesterday,the mail arrivedat two.] I had (already)LEFT at two. (b) [Yesterday,the mail arrivedat three.]I had left at TWO. Sentence(7b) demonstratesthat this constraintdoes not characterizethe past perfect.Can one devisea principledaccountof the distinctionbetween (7a) and (7b)? Binnick(I99I) proposesthe followingexplanation:since the referencetime of the PrP is the present,and since referencetime is the time of adverbialreference,the past time referencesupplied by the temporal adverbialin i960 is excluded,becausethis adverbialdoes not describeR. Klein (I992) underminesthe foundationof this argument,by demonstrating that referencetimecannotbe regardedas the sole timeof adverbialreference for the perfect. Klein observes that past-perfectsentenceslike (8a, b) are I 13


LAURA A. MICHAELIS

ambiguousas to whetheran adverbialexpressionmodifiesE or R. In (8a), at two describesR. In (8b), at two modifiesE, while R is threeo'clock.Why then should the formallyparallelPrP forbidmodificationof E? Klein (1992) has sought to account for the anomaly of (7a) by citing a pragmatic restriction, the P(osition)-DefinitenessConstraint: event and referencetimescannotsimultaneouslybe fixedto specificintervalsin a given assertion.Since, in the case of the PrP, R is alwaysfixed via identification with speechtime,adverbialdescriptionof E is necessarilybarred.A problem arises,however,with respectto past-perfectexampleslike (7b). Both E and R are adverbiallyspecified:R is I972 and E is I960 - yet no anomalyresults. Exampleslike (7b)impeachthe validityof the P-DefinitenessConstraint,and substantiatea view like that of Heny and Richards,in whichthe restriction exemplifiedin (7a) is an idiosyncraticfeatureof the PrP. Some grammaticalidiosyncrasiesof this type do not attach to the PrP simpliciter.Certainconstraintsattachto one reading:the resultativereading (6c). These constraintsamountto restrictionsupon the extentto which one can elaborateupon circumstancessurroundingthe past eventdenotedby the VP complement.The restrictionsto be consideredhere are as follows: one cannot use the resultativePrP to (a) provide furtherinformationabout a pragmaticallypresupposedevent, (b) anchorthe event in time by means of a temporaladverb,(c) assertthe occurrenceof an event complexor 'plural event' (Bach I986), (d) specifythe mannerin whichan actionwas performed. I will demonstratethat theseconstraintsdo not attachto the existentialPrP; PrP sentences starred on a resultativereading often have an acceptable existentialreading. The aforementionedrestrictionshave an apparentfunctionalmotivation: the resultative PrP is used to focus upon the presently accessible consequencesof a past event, ratherthan upon the past event per se. The latter function is associatedwith the preterite.The preterite,accordingto Anderson(I98I: 230), is used 'to describea specificpast event for its own sake (the essenceof true narrative)'.This functionis commonlycontrasted with that of the PrP. The PrP is said to indicate'currentrelevance'and to evoke an 'indefinitepast'. A numberof analystshave complainedabout the vaguenessinherentin suchdescriptionsof PrPmeaning(see McCoard1978). These descriptionsare vague in part becausethey presupposethat the PrP has a unitary semantico-pragmaticdefinition. In fact, the role typically assignedto the PrP- that of markingan indefinitepast - conflatestwo quite differentfunctions.The resultativePrP, like the preterite,evokes a small, specificpast interval,i.e. a unique event time. This intervalhappensto be unavailablefor modificationby a temporaladverb(see (7a)), but it can serve as the time frame in which events coded by subsequentpreterite-form assertionsare located.For example,as we will see below,the resultativePrP can be usedin the 'lead sentences'of newsreports,wherelaterpreterite-form assertionsprovidefurtherinformationaboutthe eventexpressedby the PrPI I4


THE ENGLISH

PRESENT PERFECT

form sentence.The existentialPrP, by contrast,evokes not a specificpast intervalbut a present-inclusive time span(generalor restricted)withinwhich one or more events of a certain type are located. Likewise,the 'current relevance'rubriccan be shownto subsumeboth a resultant-stateimplication and a modal notion that has been termed the 'present possibility' requirement. The contrastbetweenthe resultativePrP and preteriteis best understood when examined from a diachronic perspective.Historically,we see the developmentof a discourse-functionalopposition betweenthe two closely alignedforms of past-timereference.In Old and MiddleEnglish,according to Visser(I966), the two forms- resultativePrP and preterite- werelargely interchangeablein both poetryand prose.The diachronicretentionof these two exponents of past-time referencecan be attributedto the ability of speakersto establisha pragmaticcontrastamongsemanticallycommensurate forms(ClarkI987). The diachronicresolutionof 'constructionalsynonymy' will often involve the interactionof semanticand pragmaticfactors:if two distinct constructionsare semanticallyequivalent,these constructionswill come to be seen as functionallyopposed.Theseare casesin which'speakers innovatepragmaticextensionsof grammaticalforms' (Slobin I990: io). In Modern English, the resultativePrP and preteriteparticipatein a discourse-pragmatic opposition; the preteriteis the unmarkedmemberof this opposition.This oppositionarisesfrom the contrastbetweenanaphoric and deicticdeterminationof referencetime.The preteriteis anaphoricin that preterite-formassertionslocatea situationat a 'definite'pastinterval:a time which has eitherbeen previouslyevoked in the discourseor is contextually recoverable.Preterite-formassertionsare said to 'referback' to a linguistic or extralinguistictemporalantecedent.2For example,the sentenceI went swimmingmightreferbackto the temporalantecedentset up by the preteriteform question Whatdidyou do yesterday?In contrastwith the preterite,the resultativePrP expressesdeictictemporalreference.Sentenceslike Tve met someoneelse are used to assertthe existenceof a presentlyaccessibleresult [2] In this discussion,we will not considernarrativewhichentailstemporalprogression.In

such narratives,the referencetime of a preterite-form assertionis a time 'just after' the referencetime of the previouspreterite-form assertion.An exampleof temporaldiscourse is given in (i):

(i)

Margewalkedinto the room. She slippedoff her shoes.

As Partee(2984)observes,the anaphoricanalysisof the preteriteis incompatiblewith the successiveupdatingof the referencetime found in narrativetexts like (i). The anaphoric propertiesof the preteriteemergemost clearlywhenwe look at texts like (ii): (ii) I brokemy finger.Someoneslammedthe closet door on it. In (ii), the referencetime of the second sentencecan be identifiedwith that of the first sentence. The text in (ii) can be said to exemplifyELABORATION MODE, rather than TEMPORAL-PROGRESSION MODE.

I 15


LAURA A. MICHAELIS

(for example,the currentlack of romanticinterestin the addressee);they do not requirereferenceto a previouslyevoked or currentlyaccessiblepast interval.Certaingrammaticalconstraintsupon the resultativePrP, like that related to pragmaticpresupposition,can be attributedto its lack of an anaphoricfunction. I will suggest, however, that the restrictionsupon the resultativePrP constriction(RPC) do not follow automaticallyfrom any general grammatical or functionalprinciples,whetheruniversalor Englishspecific,and that thereforethe resultativePrP qualifiesas a FORMAL IDIoM: a morphosyntactic configurationcharacterizedby otherwise unpredictablegrammaticalconstraintsand 'dedicatedto semanticand pragmaticpurposesnot knowable from form alone' (Fillmoreet al. I988: 505). Idiomaticityis a matter of degree: the propertiesof idiomatic form-meaningpairings are MOTIVATEDto the extentthat theyhavea precedentelsewherein the grammar (see Goldberg1992b). In English,for example,the use of the definitearticle as a degreemarkerin formulaicexpressionslike the betterto eat you withis semanticallymotivatedinsofar as the definitearticleis assignedthis same functionin the so-calledcomparativeconditional(Themore,themerrier).The inheritancerelationswhichlink constructionsare analogousto derivational relationshipsbetween words. Derivationallinks, such as that relatingthe words collate and collateral, might be tenuous, as might the inheritance

relations which link a given constructionto one or more formally and semanticallysimilarconstructions.In such cases, knowledgeof one formmeaningpairingdoes not relievethe learnerof the obligationto learn the relatedpairing. The existentialand continuativereadingsof the PrP closely resemblethe analogousreadingsof the PaP; the correspondingconstructionsare strongly motivated, since adverbial co-occurrence restrictions and interpretive constraintsare largelypredictablefrom the relevantsemantics,sharedby both PaP and PrPforms.In the case of the resultativePrP,however,we find grammaticaland pragmaticconstraintswhich a learnerwould not know simply by knowingthat the perfectform may signal that 'the result of an event continuesto the referencetime'. Therefore,I will presumethat the knowledgethat speakersuse in producingsentenceslike (6c) is represented as a highly idiomaticpairingof form and meaning:the RPC. The remainderof this paperwill be structuredin the followingfashion.In the following section, I will present a discourse-pragmatic analysis of the preterite,which will serve as the basis for an account of the discoursefunctionaloppositionin whichthe RPC and preteriteparticipate.In section 3, I will presentargumentsin favor of the claim that the presentperfectis ambiguousratherthan vaguewith respectto the threerelevantreadings.In section 4, I will providerepresentationsof the three semanticstructuresat issue. In section 5, I will discuss grammaticaland discourse-pragmatic reflexesof the existentialand resultativereadings.In the concludingsection, i i6


THE ENGLISH

PRESENT PERFECT

I will suggestthe consequencesof this study for a conceptionof aspectual grammar. 2. THE PAST TENSE AS ANAPHOR

Partee (1984) has observed that truth-functionalaccounts of preterite meaningcan be dividedinto two generaltypes. In both accounts,the pasttense marker is viewed as an operator Past, which scopes a tenseless proposition.The truth of the resultingpropositionis evaluatedfor speech time t. The firsttype of account,associatedwith Prior(I967), is thatin which a proposition of the form Past (A) evaluates to true iff the tenseless propositionA is true at some time t' earlierthan t. In the second type of account, advocated by Reichenbach (1947), a past-tense sentence is interpretableas true or false only relativeto a given (past) referencetime. Referencetime, accordingto Klein (1992: 535), is 'the time for which, on some occasion, a claim is made'. Partee (I984) observes that under the Prioreanview, the truthof the preterite-formassertiondependson the truth of the base sentenceat soMEtimein the past,whileunderthe Reichenbachian view, the truthof a preterite-formassertiondependson the truthof the base sentenceat THAT time in the past. Most recent formal accounts of tense reference have followed the Reichenbachianview. Reichenbach'saccount appears preferablebecause there is evidencewhich suggests that reference-timespecificationmust be part of the truth conditions of preterite-formsentences.For example,the sentenceI boughtyou a newspaperwill be judgedfalse if the base sentenceis false at the time that the hearerhas in mind (e.g. that morning),despitethe fact that the base sentencemay be true at some other time (e.g. last year). In accordancewith Partee(op. cit.), I will presumethat state predications subsumethe relevantreferencetime. For example,the assertionThesuspect hada beardis made relativeto a specificpast interval,and does not indicate whetherthe state referredto holds now of the individualdenotedby the NP the suspect.In such cases, the past-tenseassertionsignalsthat the speaker wishes to vouch for the presenceof the denotedstate of affairsonly during the relevantpast interval.Whatdo we meanwhenwe referto a past interval as THE RELEVANTINTERVAL? Presumably,this is a specificpast timewhichthe utterer of the preterite-formsentence has in mind, and believes that the hearerhas in mind (or can readilycall to mind). For Hinrichs(I986) and Partee(1984), among others,the requirementof mutual knowledgeof the evoked past intervalis capturedby a representational systemin whichthis intervalis anaphoricallyboundto a specifictime framethat has previouslybeen introducedin the discourse,or is otherwise recoverablefromcontext.Accordingly,some analysts(e.g. Heny 1982) have suggested that the preteriteshould not be representedby an existential quantifierover past times (as Prioreantreatmentswould have it), but by an I 17


LAURA A. MICHAELIS

anaphoricallybound temporal variable. This mode of representationis intendedto capturethe insight that the past tense 'points to a particular (context-determined) [temporal]location' (Cooper I986: 237). The analogy between this form of temporal reference and nominal anaphorais tenuous,since,as Partee(I984: 275) admits,'[i]ntuitively,noun phrases "refer" and tenses do not'. It stretchescredibilitysomewhat to suggestthat a tenserefersto a timein the sameway that a noun phraserefers to an individual.Nevertheless,in eventuality-based systemsof representation like that of Parsons (I990), times are existentiallybound variables,and therebyqualifyas individuals,just as episodes(events)qualifyas individuals. Therefore,we can speakof the past-tenseoperatoras evokingtwo temporal individuals:the time of the utterance(representedas a context-sensitive variable),and the past time of occurrence.We can also speakof the time of occurrenceas LOCATABLEWITHIN A HISTORY, since a historyincorporatesa time line: an orderedset of (temporal)individuals.A historyis definedas a mental record of events locatable along a linear pathway of temporal intervalsleadingto the presentmoment.Accordingto the Hinrichs-Partee model, the interpreterof a past-tenseassertionmust locate the occurrence time within a history,i.e. an orderedset of times (and attendantsituations) whose upperboundaryis the presentmoment. As in cases of nominal anaphora,the temporalantecedentmay be an elementof the linguisticcontextor an elementof the extralinguisticcontext. Linguistic antecedentsinclude frame adverbialslike yesterday, reference timesevokedby previouspast-tenseassertionsand boundtemporalvariables evokedby the subordinateclausesof certainhabitualsentences.Examplesof these three types of antecedentsare given in (ga, b): (9) (a) Yesterday,I got a strangephone call. (b) Harrythrewa party. He got completelysloshed. (c) Wheneverhe touchedthe door knob, he got a shock. In (9a), the preterite-formassertionevokesa past time identifiedwith that denoted by the adverb yesterday.Here, yesterdayhas a frame-adverbial reading:the time at whichthe phone call occurredwas a time withinthe set of times describableas yesterday.In (9b), the time of the second preteriteform assertionis interpretedas the time evoked by the first such assertion: the event of Harry'sbecomingdrunkis located within the intervalduring whichHarrygave the party.In (9c), a somewhatdifferentformof anaphoric past-timereferenceis involved:the past tense predicationappearingin the main clause does not evoke a unique time of occurrence,but a temporal variable, introduced by the subordinate-clausepredication. Here, two temporalvariablesare co-bound:the time of each past door-knob-touching event is the time at which a shock was received. In the case of an extralinguistictemporalantecedent,the time framewith respectto whichthe preterite-formassertionis interpretedmustbe recovered I 18


THE ENGLISH

PRESENT PERFECT

from context. Context in this case is THAT HISTORY TO WHICH THE INTERLOCUTORSHAVE JOINT ACCESS. The historyavailableto the interlocutors includes the immediatepast period representedby the last phase of the discourse. For example, the time of the hearer'smost recent discourse contributionprovides an extralinguisticantecedentfor the preterite-form question in (Io):

(io) I didn't catch the end of your sentence. The 'historical background' may also include events outside of the immediatecontextof the discourse.Theseare eventswhichthe interlocutors can access from a shared history at a time when some event which was previouslyin the offing is known to have occurred.For example, if the speakerand hearershareknowledgethat the hearerwas to requesta raiseat some point during the course of the work day, the speakermay ask the followingquestion,upon the hearer'sreturnfrom work: (iI)

How did you do?

The event whose time of occurrenceprovides the temporal frame of referencefor the preterite-formassertionneed not be recent,althoughthe event must be salient to the degree that the interpretercan recover the occurrencetime.An illustrationof this claimis foundin an anecdote,told by CharlesFillmore(p.c.),the humorouseffectof whicharisesfromthe fact that a preterite-form utterancehas a remote-pastreferencetimethatis nonetheless easily recoverable.Fillmorereportsthat duringa chanceencounteron the streets of London, a fellow linguist invited him to a local fish restaurant famous for its plaice (a type of flounder).Fillmoredeclinedthe invitation but, upon meetingthat same linguistfive years later, greetedhim with the preterite-formquestion How was it? (The addresseereplied, 'Excellent.') Here, the plaice-consumptionevent, althoughcertainlynot recentvis-a-vis the individualhistoriesavailableto each of the two speechparticipants,is salient,becauseno morerecenteventstranspiredwithinthe representationof historyHELD IN COMMONby those interlocutors. In distinguishingbetween temporalantecedentswhich are linguistically expressed and those which can be characterizedas 'belonging to the discoursecontext', it is useful to invoke Lambrecht'sdistinctionbetween identifiabilityand activation status- a distinction used to describe the mentalrepresentations of nominallyencodeddiscoursereferents(Lambrecht, I994). I will extend these concepts to the domain of discoursereferents representing'temporalindividuals',i.e. occurrencetimes establishedin the 'historical record' shared by speaker and addressee. According to Lambrecht,the identifiabilityparameter'has to do with the speaker's assessmentof whether or not a discourse representationof a particular referentis alreadystored in the hearer'smind'. The activationparameter concernsthe 'speakersassessmentof the statusof an identifiablereferentas I I9


LAURA A. MICHAELIS

"activated", as merely "accessible"or as "inactive" in the mind of the hearerat the time of the speechact'. An ACTIVECONCEPTiSone which,in the wordsof Chafe(I987: 22), iS 'currentlylit up, a conceptin the person'sfocus of consciousnessat a particularmoment'. An entitywhichis identifiableis one the interlocutorscan distinguishfrom other entities with which it shares properties criterial for category membership.A formal correlateof identifiabilityis DEFINITENESS,in those languages which have grammaticalizeddefinitenessdistinctions.In such languages,the definitearticletypicallyindicatesthat the noun to which it attaches refers to an entity which is contextually the uniquely salient exemplarof the categoryin question.,For example,in the sentenceI picked up thepackage,the definiteNP thepackagerefersto an entity for which a sharedrepresentationexists in the minds of speakerand addressee.In the case of 'temporal reference',as mentioned,identifiabilityarises from the ability of speakerand hearerto select the evoked past intervalfrom a time line which forms the basis for a representationof history shared by the interlocutors.The interpretermust be able to LOCATE the relevantinterval within that history. Identifiablereferents,accordingto Lambrecht,may be in any of three activationstates: active, accessibleor inactive.Whilean activeconceptis a focus of the interlocutors'consciousness,an accessibleconceptis eitherin the hearer'speripheralconsciousness(as construedby the speaker)or is a salient member of a semantic frame that has been invoked in the discourse.A referentmay be said to be in the hearer'speripheralconsciousnessif, for example,it is a salientpartof the text-externalworld.Deicticexpressionslike thosepicturesmay be said to referto an accessiblereferentin sentenceslike Thosepicturessureare ugly,used to describesome pictureson the wall of the addressee'soffice (where speaker and addresseeare both present in the office). An accessible referent,according to Lambrecht,is more readily broughtto mind by the interpreterthan one whichis inactive.With respect to nominalreference,Lambrechtobservesthat 'the activestatusof a referent is formally expressed via lack of accentuation and typically (but not necessarily)via pronominal coding of the correspondinglinguistic expression'. Inactive or accessible status is accordinglyconveyed by accentuationof the referentialexpressionandfull lexicalcoding(the distinction betweeninactiveand accessiblestatus is not linguisticallyexpressed). In the case of temporalanaphora,we can say that a past periodis active whenit has alreadybeeninvokedin the discourse,and accessiblewhenit has not been explicitlyinvokedbut is salientin the extralinguisticcontext. For example, a preterite-formassertioncan be said to invoke an active past intervalif the past intervalto which the assertionis relativizedhas already been referredto by a previouspast-tensepredication(cf. (9b)). A past time of occurrenceis accessibleif it is contextuallyrecoverable.The temporal individualevoked is in the situationalcontext, and is thereforemore easily 120


THE ENGLISH

PRESENT

PERFECT

conjuredup in the addressee'smind than one which is entirelyinactive.As we observedabove, the situationalcontextmay be the immediatecontextof the discourse(e.g. the time of the last discoursecontributionin (io)), or it may be some salient portion of the history which leads up to the present discourse,e.g. in (i I), the time duringthe work day at which the addressee requesteda pay raise. Whether the interval evoked is characterizableas active or as merely accessiblein a discourse,it representsa closelycircumscribedinterval.If the relevantpast intervalwere not circumscribed,one could not locate it at a particularpoint in a history.In the exampleswhichwe have encounteredso far, the preteriteevokes a past interval that is POINTLIKE. For example, sentence(io) evokesthe point in time at whichthe addressee'slast discourse contributionwas made. Intuitively,it is odd to referto the referencetime as a point, sincewe know that the addressee'sutterancemust have takentime, and that thereforethe referencetime evokedby (io) qualifiesas an interval. However, as Herweg argues (I99I: 982), one cannot distinguishintervals from moments(i.e. degenerateintervals)withoutconsideringthe time units relevantto the cognizer(s)(see also Talmy I988): Since on the conceptual level we deal with mental representationsof time,... viewing a period of time as pointlike means that its internal structure is cognitively neglected as a matter of the granularityof perspectivetaken by the subject.Thus, we allow that one and the same temporalentity be representedas a pointlikeor complextime depending on the situation. In the context of this analysis, 'situation' is to be construed as the particulartime line invoked in the mental representationof a history- a representationwhich the speakerpresumesis shared by the hearer. Our analysis of preteriteis based on a model in which this representationof historyevokesa timeline whose fundamentalunit of temporalmeasurement can be characterizedas a moment.Thismodelrequiresthat the past moment evokedby the preteriteis (a) IDENTIFIABLE(i.e. locatableat a particularpoint along the time line) and (b) ACTIVE or ACCESSIBLEat the time at which the preterite-formsentenceis uttered. The analysisjust offeredis somewhatoversimplified,since the preterite may evoke a past period that has internal structure,i.e. is divisible into subperiods. In the following sentences, the past period refereed to is necessarilyinterpretedas an intervalor 'complex time': (I2)

(a) Did Karla ever call you? (b) GoldmanSachs interviewedBrucetwice.

In (i2a), the adverbever denotesa rangeof timesthat are potentialtimes at which Karla'scalling took place. Accordingto W. Ladusaw(p.c.), ever (likeits negativecounterpartnever) can be usedto preemptthe inferencethat I21


LAURA A. MICHAELIS

a past predicationrefersto a 'small' (i.e. momentaneous)period:(I 2a) refers to a past interval,e.g. a day. In (i2b), the presenceof the frequencyadverb twicerequiresthat the past-tenseassertionbe takenas evokinga periodlong enough to include two interviewingevents. In both (I2a) and (12b), the preterite-formassertionevokes an identifiablepast time that is eitheractive or accessiblein the discoursecontext.Sentence(12b) mightbe usedto answer the question What happened last week? Sentence(i2a) might be used to inquireabout eventsthat occurredon the addressee'sbirthday.The birthday mightbe an accessibletime framein the discourse,i.e. one that has not been explicitlyinvoked,but is presumedsalientto the hearer.In (I2), unlikethe previousexamples,the 'definite'time invokedby the preteriteis a complex time,whichincludestwo or more(potentialor actual)eventtimes.However, the sentencesin (I2) are like the previousexamplesin that the past time evokedcan be characterizedas havinga linguisticor extralinguistictemporal antecedent.Therefore,the time framereferredto by the preterite,whethera pointlikeor complextime, is a salientlyboundedintervalthat can be placed at a particularlocationon a time line (i.e. is identifiable)and is eitheractive or accessiblein the discoursecontext. The circumscribedcharacterof the past time frameinvokedby preterite emergesclearlywhenone contraststhe preteritewith the existentialPrP.The existentialPrPis exemplifiedin (6b). Considerthe contrastpairgivenin (I 3): (a) I went to Paris. (b) I've been to Paris. Sentence (i 3b) illustrates an idiomatic grammaticalproperty of the existentialperfect:the expressionbe to can be used to mean 'visit'; be to is (I3)

ruled out for past-tense assertions (*I was to Paris). Sentence (13a) evokes a

specific,circumscribedpast interval.This sentencemay be a responseto a questionconcerningactivitiesthat the speakerengagedin duringthe past summer.This sentencewould be anomalousas a discourse-initialassertion. In order to qualify as a felicitousdiscourse-initialutterance,(I 3a) would requirea past-timetemporaladverblike in 1992. A frameadverbialof this typewould'activate'a particularpast timewithina representationof history which the speaker,by uttering(I3a), signal an intentionof constructing. By contrast,sentence(13b) does not evoke an identifiablepast interval. The sentencemay be used to denote any numberof visits to Paris by the speaker.Therefore,the questionHow manytimes?would be an appropriate responseto (i 3b), but it would not ordinarilybe an appropriateresponseto (I 3a). Whileeach visit to Parisnecessarilyhas a past time of occurrence,the interpretationof (13b), unlikethat of (i3a), does not requirethe interpreter to invoke a particularpast time of occurrence.Instead,the interpreterneed only envisiona generaltime span,whoseupperboundaryis the presenttime, withinwhichthe eventor eventsdenotedin questiontook place.The interval itself may be denoted by a time-spanadverblike before.Alternatively,the 122


THE ENGLISH

PRESENT PERFECT

lower boundary of that interval may be denoted by a since-adverbial expression (e.g. since the war ended).

The distinctionbetweenthe two typesof past-timereferenceexemplifiedin becomesclearerwhen we look at the interactionof these predication types with frequency adverbials. Notice that if we add the frequency expression three times to each of these sentences,the circumscribedand specificcharacterof the intervalevokedin (i3a) is unchanged.Thatis, (I3a) still refersto a definitepast period(e.g. last year),althoughthis periodis not a pointliketime but a complextime- an intervalcontainingseveralvisits to Paris. In the case of (I 3b), however,the times at which visits to Paristook place arenot placedwithinany temporalboundaries;the eventtimeor times are simply located prior to speech time. An exampleanalogous to (I3) is found below: (I 3)

(I4)

(a) Did Karla ever call you? (= (12a)) (b) Has Karla ever called you?

In (14b),everrefersto an intervalwhicheffectivelylacksa lowerboundary. The speaker'sinquirycan be said to concernthe historyof the addressee's friendshipwith Karla. In (04a), by contrast,the speaker'sinquiryconcerns only a limitedperiodpriorto speechtime. Whilethe most likelyresponseto (14a) would be a simple affirmationor denial, the most likely responseto (14b) would include a frequencyexpressionlike several times. This is a reflectionof the fact that preteriteform sentencesare ordinarilytaken to refer to unique past events- since the referencetimes involved are closely - while sentenceslike (I4b) evoke any numberof instancesof circumscribed the event type denoted,since the past periodinvokedis a broadexpanseof time. Anotherdistinctionbetweenthe preteriteand the existentialPrP involves a feature which I will treat as a conventionalimplicature:the presentpossibilityconstraint.Accordingto McCawley(I97i), among others, the existential PrP requires that the event or episode denoted by the VP complementbe capableof occurringat the presenttime. Examples(I5a, b) demonstratethat the present-possibility constraintdoes not characterizethe preterite: (15) (a) I went to a Neil Young concert. (b) I've been to a Neil Young concert(before). In sentence(i 5a), the indefiniteNP a Neil Youngconcertrefersto a specific performance,whichtook place at a definitepast interval.In sentence(Isb), however,the NP a Neil Youngconcertrefersnot to a specificperformance but to a type of performance(one given by Neil Young). Sentence(I5b) assertsthat withinan intervalupperboundedby the present,therewereone or more instancesof the event denotedby the base sentenceI go to a Neil Young concert. This sentence conventionally implicates that the event 123


LAURA A. MICHAELIS

denotedcould recurat the presenttime.Therefore,Neil Youngmustbe alive at present, capable of giving performances,etc. No such implicationis attachedto (I 5b): this assertionis felicitousif Neil Young is deceased,if he has retiredfrom touring,etc. Thereare (at least)two classesof exceptionsto the analysisof the preterite offeredabove. In the firstclass, describedby Heny (1982) and Partee(I984), amongothers,the past time of occurrenceinvokedby the preteriteis neither active nor (necessarily)identifiable.Considerthe followingexamples: (i6) A: How did Cicerodie? B: He was executedby MarcusAntonius. Shakespearesaid, 'In many'slooks the false heart'shistoryis writ.' In these examples, the preterite-formassertions(e.g. he was executed, Shakespearesaid) do not refer to a period which is under discussion or contextuallysalient. For example,as shown, B's assertionin (i6) need not (17)

answer a question like: Whathappenedto Cicero in 43 BC? Further, the event

time in question need not be locatable by speakerand/or addresseeat a particularpoint in a representationof history(in these examples,a remotepast history).The discussantsin (i6) need not know that the event referred to (Cicero'sdeath)occurredin 43 B.C. Similarly,the uttererof (I7) need not know even the approximateperiod in which Shakespeareexpressedthe quoted sentiment.In (i6) and (I7), the location of the referencetime is not relevantto the concernsof the speakerand/or hearer.In such cases, the meaning of the past tense is appropriatelyrepresentedby the Priorean model: an event occurredat some time prior to now. Partee (I984: 296), makes a similarobservationwith respectto the sentence WhokilledJulius Caesar?: She points out that, in interpreting this sentence

[thehearerdoes]not have to know whenit happenedto know who did it, given that it could only have happenedonce if it happenedat all. In [this] case, the referencetime could potentiallybe the whole of the past. Of course,when a sentencehas a referencetime equatedwith 'the whole of the past', the sentencein essencelacks a referencetime. In anotherclass of exceptionsto the analysisof the preteriteofferedabove, the preteritenot only lacks an activeand identifiableevent time, but also appearsto evoke a deicticallydeterminedreferencetime. In these cases, the preteriteservesa communicativefunction like that of the RPC. Consider the following examples: (i8) I alreadytold you: I'm not interested! (i9)

Pat NIXONdied.

In (i8), the preterite-formassertionapparentlyrefersto a presentstate of affairs: the addressee'sknowledge of the speaker'slack of interest; the sentenceimplies that this state of affairsis the result of a past event: the I24


THE ENGLISH

PRESENT PERFECT

speaker'sreportinghis or her lack of interestto the addressee.Here, the adverbalready,whichordinarilyaccompaniesstatepredications,anchorsthe referencetime of the assertionto the present.The usage exemplifiedhereis, as frequentlynoted, mainly confined to Americandialects of English, in whichthe preteriteappearsto be expandingits semanticrangeat the expense of the RPC. Sentence (i9) is ambiguous,in that it can be taken as an argument-focusresponse to the question Who died yesterday?or as a 'sentence-focusresponse' to a question like What happenedin the news yesterday? (see Lambrecht,I994). In the context which I have in mind, however,(I9) is a sentence-focusassertion,and no past intervalhas been invokedin the discoursecontext.Accordingly(I9) may be a discourse-initial utterance,in whichcase the speakeris simplyreportinga momentousevent to the addressee.On this reading,sentence(i9) has a functionlike that of the 'hot news' PrP.In sum, neither(i 8) or (I9) invokesa past timethat is active or identifiablein the mindsof the interlocutorsat the time of the utterance. In light of exampleslike (i6) and (i9), I conclude, in accordancewith Richards (I982: 134), that

any accountof the apparent'referentiality'of somecasesof the past tense whichmakesit logicallynecessaryfor the well-formednessof a past-tense sentence that there be ... reference to some specific [i.e. contextdetermined] time ... cannot be correct.

Instead,I will presumethat the ' anaphoric'use of the past tenseis one of the communicativefunctionsof the past tense,but not its only function.Let us propose that there is a feature [? anaphoric]which attaches to all exponentsof past-timereferencein a givenlanguage.A formwhichexpresses a past-tenserelation will be regardedas [+ anaphoric]iff it requiresthe interpreterto ' anchor'the eventexpressedby the basesentenceto a ' definite' past interval,i.e. an intervalwhich is both active and identifiable.I will presume that the English past tense is unmarkedwith respect to the anaphoricity feature: it is capable of expressing both anaphoric and nonanaphoricpast-timereference.Below,I will arguethat the resultativePrP participatesin a markednessopposition with the preterite,in which the resultativePrP is markedas [-anaphoric]. We will now proceed to examine the distinct readings of the PrP, exemplifiedin (6), after which we will look at those propertiesof grammar and use which are unique to the resultativereading.The next section will provideevidencethat the PrP is ambiguousratherthan vague with respect to the relevantreadings. 3. VAGUENESS VERSUS AMBIGUITY

Since McCawley'sI97I paper, 'Tense and time referencein English', in whichtemporallogic was usedto describedistinctreadingsof the PrP,many analystsinvestigatingthe semanticsof the PrPhave beenconcernedwith the 125


LAURA A. MICHAELIS

followingquestion:arethe readingsat issuesimplyusesinferredin particular communicativecontexts, or are they distinct meanings? Among those analystswho have approachedthis question,most, includingBauer(1970), Brinton (I988), Dinsmore

(i99i),

Fenn (1987), Klein

(I992)

and McCoard

(1978),have rejectedMcCawley'sclaimthat the understandingsexemplified in (6) should be assigned distinct underlying semantic representations. Instead, these authors have proposed that the distinct understandingsin questionarecomputedby interpretersinvokingfeaturesof the linguisticand extralinguisticcontextin orderto decodean otherwisevaguepredication.3A fundamentaldifficultywith this kind of accountemergeswhen we consider the unreliabilityof contextualcues said to signal a resultativeas against existentialunderstanding.Klein, for example,assumesthat the PrPis vague with respectto both the frequencyof the denotedevent and the distanceof event time from speechtime. Accordingly,he argues [t]hefactthatboth distanceandfrequencyof TSit [eventtime]areleft open gives rise to differentreadingsof the perfect- experiential[existential], resultativeperfect,and others. But these readingsdo not arise from an inherent ambiguity of the perfect. They stem from contextual information... (p. 539) Presumably,the frequencyvariablein Klein's formulationrelatesto the potential for an existentialunderstanding:existentialperfects commonly referto iteratedevents.The interpretermight thereforebe said to compute an existentialunderstandingupon encounteringa PrP sentencecontaininga frequencyexpressionlike twice.For example,the sentenceHarryhas visited Clevelandtwiceis necessarilyassignedan existentialmeaning.The existential understanding,however, is potentiallyavailableeven when the PrP-form predicationdoes not establishthe existenceof multipleeventsof a givenkind. Considerexample(6b), repeatedhere for convenience: (6) (b) We've had this argumentbefore. This exampledemonstratesthat an existentialPrP may denote only one instanceof the eventin question.The interpreterof (6b) can understandthis sentenceas an existentialPrP,even thoughhe or she may have in mind only one instancein which the argumenttook place. Under Klein'saccount,the possibilityof a resultativeunderstandingwill typicallydependupon the speaker'sinferringthat the event denotedby the VP complementis recent.4Result states may be of short duration; if a [3] Not all recentanalystswho havelookedcloselyat the meaningof the PrPhavefocusedon the distinctinterpretations citedin (6). Some,like Inoue (1979), have simplyequatedthe semanticsof the PrP with one of its readings.Inoue'sanalysisof the PrP appearsto be basedon the existentialreadingalone.Parsons(I990), by contrast,concernshimselfonly with the resultativereading. [4] Kleinactuallyarguesonlythatthe Comrie's(I976) 'perfectof recentpast' (= McCawley's 'hot news' perfect)is invokedwhen event time 'immediatelyprecedes'reference/speech 126


THE ENGLISH

PRESENT PERFECT

temporaryresultis to remainin force at speechtime, the causalevent must be recent. However, immediacyof the event to speech time is neither a necessarynor a sufficientcondition for the invocation of the resultative understanding.The argumentthatrecencyis a sufficientconditioncommonly proceedsfrom exampleslike (20): (20)

Have you seen my slippers?

To a question of this sort, the argumentgoes, only an uncooperative hearer would respond: Yes, about a month ago. That is, the questioner presumablyintends a resultativeunderstandingof (20): she is inquiring about a past sighting of the slipperswhich currentlyaffects the hearer's ability to locate the slippers.The questionerdoes not intend an existential understanding,i.e. an interpretationwhich might cause the hearer to mention one or more slipper sightings in the distant past, having no importancefor the presentstate of things.The contextualeffectsobservable here are, however,reducibleto relevance-basedimplicature,togetherwith inferencerelatedto the temporarynature of a given slipperlocation. The interpreterof the PrP-formquestion(20) will recognizethat the only relevant responseis one whichconcernsthe questioner'scurrentquandry:the needto find her slippers.This response entails a resultativeinterpretationof the question.Furthermore,since the conversantsknow that slippersand other such objectstypicallyremainin a givenlocationonly briefly,they also know that the hearer'sknowledgeof the presentlocationof the slipperswill depend upon the hearer'shaving sighted them recently.Therefore,the resultative reading of a PrP-formutterancedoes not depend upon the interpreter's awarenessthat the eventdenotedis recent.Rather,recencyof the eventwith respectto speechtime is intrinsicto the resultativeinterpretationIN CERTAIN SCENARIOS, like that describedas the context for (20). Recencyis not a necessaryconditionfor the adductionof the resultative interpretation.Considersentence(2I): (2 I)

It seems Grandpahas cut me out of the will.

Sentence(2 I) can be construedresultativelyeven in a contextin whichthe disinheritingevent occurredmany yearspriorto the reportingof it, so long as certain effects of that event (penury,etc.) are in force at speech time. Sentence(2I) shows that the causalevent denotedby the RPC need not be recent. Perhapsrecognizingthe recencyof the VP-complementdenotatumvis-avis speechtimeis not an essentialcomponentof the resultativeunderstanding, Klein goes on to arguethat in fact the resultativeunderstandingariseswhen 'contextualinformation... tells us ... that the consequences[of the event]are time. However, as I have argued elsewhere (Michaelis regarded as a subvariety of resultative PrP.

1993),

the hot-news PrP can be

127


LAURA A. MICHAELIS

still to be felt' (p. 539). A similar argument is made by Dinsmore (I99I).

While the existence of present consequencesis certainly intrinsic to the resultativeunderstanding,adductionof the resultativeunderstandingdoes not alwaysdependupon the availabilityof the relevantresultsat speechtime. As McCawley(I98I) has observed,the resultativereadingmay be operative undercircumstancesin whichthe referencetime of the PrP-formpredication is decoupled from speech time. For example, the PrP has a nondeictic (variable)readingin contextslike (22): (22)

Whenever I've seen Madge, she has just had a fight with her boyfriend.

In the main clause of (22), a PrP-form assertion has a resultative understanding.This understandingis present despite the fact that the resultantstate at issue (Madge is agitated or upset) does not necessarily obtain at speechtime. Rather,the resultantstate in questionobtainsat all (past) times characterizedby the speaker'ssightingof Madge. In Partee's terms(I 99 I), the sentenceinvolvesco-bindingof a timevariableexpressedby the subordinateclause: all times at whichthe speakersees Madgeare times at whichan eventof Madge'sfightingwith herboyfriendhas occurred.Here, the referencetime of the PrP, ordinarilyequated with the present, is a variabletime, rangingover valueswithin a present-contiguouspast period. As I will arguebelow, the semanticrepresentationof the resultativePrP includesa conventionalimplicatum:the resultantstate of the eventdenoted obtainsat speechtime.Exampleslike (22) demonstratethat speechtimeneed not provide the temporalframe of referencefor which the resultant-state implicatumis evaluated.Exampleslike (22) are not problematicif one views the resultativeunderstandingas a distinctreadingon the semanticlevel.Such examplessimply show that the resultativePrP has a propertyin common with many other deictic expressions:its 'contextualvariable'need not be anchoredby the speech scene: instead, it can be equatedwith a value or values outsidethe speechcontext (see Partee(I99 I) and Fillmore(197 I) for a discussionof examplesin whichthe contextsensitivityof deicticexpressions is 'closed off'). Exampleslike (22) are difficultto reconcilewith an account like Klein's,in whichinterpreterscomputea resultativeunderstandingon the basis of their knowledgethat a resultantstate of the kind at issue currently exists. Sentence(22) demonstratesthat the currentpresenceof the relevant resultantstate is not a necessaryprerequisitefor evocationof the resultative understanding. Thereis anotherdifficultywhichweakensanalyseslike Klein's:the distinct PrP readingshave distinct grammaticalreflexes.In general, contextually computed meanings do not have grammaticalramifications,while conventionalmeaningsmay. Zwicky& Sadock(I975) arguethat when distinct grammaticalfeatures(of a sufficientlyidiosyncratictype) attach to distinct understandingsof a given construction,the constructionin question is I28


THE ENGLISH

PRESENT PERFECT

ambiguousrather than vague with respect to those understandings.One exampleexaminedby Zwicky& Sadockinvolvessubordinatewh-clausesin sentences like (23), which is ambiguous between headless-relativeand indirect-questionreadings: (23)

I asked what (the hell) she had asked.

Zwicky& Sadockpoint out that the idiomaticexpressionthe hell, placed after the wh-complementizer/relativizer, ensures that (23) receives an indirect-questionreadingratherthan a headless-relative reading.Therefore, they argue, the constructionexemplifiedin (23) is ambiguousratherthan vaguewith respectto the two understandings,sincean idiosyncratic(i.e. non semanticallymotivated)grammaticalfeaturewill generallyattachonly to a conventional(i.e. non context-dependent)understanding. Vaguenessanalyseslike Klein's fail to countenancegrammaticalfacts of the sort to be exploredhere.It is difficultto imaginehow one mightreconcile the claim that the PrP is semanticallyunambiguouswith the fact that, for example,the existentialunderstandingis compatiblewith mannermodification while the resultativeunderstandingis not. Consider the following example: (24)

(a) Our committeechair has (??angrily)tenderedhis resignation. (b) Our committeechairhas angrilytenderedhis resignationevery time we have asked him to take a controversialstand on something.

These examples indicate that the existential-resultative distinctionis a conventionalone, since it has a grammaticalreflex (i.e. a co-occurrence restriction)which is not predictablefrom the two meaningsinvolved.The argumenthereis similarto Zwicky& Sadock'sclaimabout(23).Admittedly, however,the fact that the distinctperfectunderstandingshave distinctcooccurrenceconstraintsdoes not necessarilyobviatea vaguenessanalysis.The PrP readingsin (6) can be regardedas productsof the interactionbetweena vague PrP meaning(perhapsincluding'currentrelevance')and the lexical specificationswhich'fill in' the construction.Underthis analysis,the distinct readingsof the PrP arise via constructionalaccommodation.Accordingto Bauer (1970) and Brinton (I988), these lexical fillers include adverbial meaningsand the Aktionsartpropertiesof the VP complement.Accordingto Brinton (p. 45), 'Consideration of the interaction of the perfect with Aktionsartleadsto a rejectionof the idea that thereare different"types" of perfects.'Brinton'sargumentis based on Bauer'saccount of contextualmodulationeffects,which is summarizedin (25): (25)

(a) If the complementis atelic and the verb is accompaniedby a time-span or durational adverb (since three,for two hours, etc.),

the perfectis continuative. I29 5

LIN 30


LAURA A. MICHAELIS

(b) In the absence of such adverbials,and occasionally in the presence of a frequency adverb like twice, the perfect is existential. (c) If the complementis telic, the perfectis resultative. Despite the attractive simplicity of this solution, the 'algorithm' summarizedin (25) is inadequatein many respects.In general,it is difficult to maintainan accommodation-stylevaguenessanalysisin light of the fact that the PrP constructionis characterizedby TOKEN AMBIGUITY. Let us say that a constructionexhibits token ambiguityif a set of understandings attributableto a grammaticaltemplateoutsideof any lexicalcontextis also availablewhen that templatehas been 'filled in' by lexicalmaterial.As an exampleof a constructionexhibitingtoken ambiguity,let us take the wayconstruction, described by Jackendoff (I990) and Goldberg (1992b). Jackendoffpoints out that sentences like the following are ambiguous between'causative'and 'manner'interpretations: (26) Marvinjoked his way into the meeting. This sentence can be taken to mean either that (a) Marvin convinced peopleto let him into the meetingby joking with them, or (b) Marvinmade jokes as he walked into the meeting. In (26), we see that the wayconstruction,consistingof a verb, a possessiveNP headed by way and a directionalexpression,is token ambiguous,sincethe lexicalfillers(the head verb and directional expression) do not resolve the manner-causative ambiguity associated with the constructional 'skeleton'. By contrast, constructionswhose variegatedinterpretationsderivefrom the modulating effects of linguistic context do not exhibit token ambiguity. Thus, for example,while the English ditransitiveconstructionhas a large array of interpretations(see Goldberg1992a), a giveninstanceof that constructionis unambiguous.The constructioncan be said to expressa vague meaning:it linksthe directobjectfunctionto the role of POTENTIALRECIPIENT.Wherethe headverbis a transferverblikegive,the constructionencodesactualtransfer (Harrygave Margethe book).Wherethe head verbis nontransferverb (e.g. promise) the construction encodes intended transfer (Harry promised Marge

thecar).The ditransitivevalence-construction cannotbe saidto exhibittoken ambiguity, as either the effected- or intended-transferreading is unequivocallyassociatedwith a sentenceinstantiatingthat argument-structure construction;the lexicalverb determinesthe relevantreading.If, however, the relevant linguistic context fails to resolve the vagueness otherwise associatedwith the semanticsof the construction,as in (26), then we have evidence that an accommodation-based(equivalently,integration-based) analysisis not appropriate.In what follows, we will examineevidencethat the conditions set out in (25) leave room for interpretiveindeterminacy vis-'a-visPrP tokens, and that thereforethe PrP is token ambiguous. 130


THE ENGLISH

PRESENT PERFECT

Firstly,the combinationof durationaladverbialand imperfectiveaspect does not entaila continuativeunderstandingof the PrP-formexpression.As Mittwoch(I988) observes,sentenceslike (27) are ambiguouswith respectto continuativeand existentialinterpretations: (27)

Harryhas been in Bali for two days. continuative:Harry'spresencein Bali obtainsfor all timeswithina present-inclusivetime span whose lower bound is two days ago. existential:Therewereone or more visits to Bali by Harrywithina present-inclusivetime span; each of these visits lasted two days.

Sentence (27) demonstrates that the fulfillment of condition (25a) does not

renderthe PrP-formsentenceunambiguous.The factorsdescribedin (25a) are thereforenot sufficientto impose the continuativereading.However, these factors are necessaryconditions: the continuativePrP indicatesthe cessation at speech time of a bounded state. Therefore,the complement necessarily denotes an imperfective situation- one which is typically boundedby means of a durationalor time-spanadverb. Secondly,condition(25b) is neithera necessarynor sufficientconditionfor evocation of the existentialreading.This readingis commonlyascribedto perfectsentencescontainingTELICcomplement-VPs.Considersentence(28): (28)

I've cleanedthe whole house (before).

As Mittwoch (i988) shows, the existential interpretationitself, when adduced, imposes an episodic reading on a stative VP-complement:the sentenceHarryhas beenin Bali is interpretedas meaning'Harryhas visited Bali'. Therefore,the existentialreading,far fromrequiringthe presenceof an atelicor stativeVP-complement,is capableof imposingan eventiveconstrual upon an otherwiseimperfectivecomplementdenotatum.This fact suggests that the existential PrP is appropriatelyregarded as a grammatical construction,since it is apparentthat the lexicalverbACCOMMODATESto the semanticstructureassociatedwith the morphosyntactictemplate. Examples like (29) further demonstrate that condition (25c) is not sufficientto determinea resultativeinterpretation.Sentence(29) has a telic complement-VPand is neverthelessunderstoodas an existentialPrP. Is the presenceof a telic participialcomplementa NECESSARYcharacteristicof PrP sentenceshaving the resultativereading?Sentenceslike (29) can receivea resultativeinterpretation,despitethe lack of a telic VP-complement: (29)

I've knocked.

Sentence(29) mightbe usedby one partyguestto another,as the two wait on the frontporchof theirhost'shome- the hearerhasjust arrived,whilethe speaker has been waiting for several minutes. Given the interlocutors' knowledgeof hailingconventions,the sentenceis used to conveya resultof the followingsort: we shouldnow expectour host to be receivingus. There '3' 5-2


LAURA A. MICHAELIS

may be otherrelevantresults:the hearerneed not knockherself.Notice that (29) in this contextis not an existentialperfect:(29) does not welcometimespan adverbialslike BEFOREor frequencyadverbialslike once. The indeterminateand context-dependentnature of the resultantstate entailedby sentenceslike (29) has led McCoard(1978), among others, to reject the idea that a resultant-stateimplicationis part of the message conventionallyassociatedwith the PrP form. How can a form be said to express the existence of a present result if the form does not necessarily providea clue as to what that presentresultis? Those who pursuethis line of argumentationoverlookthe followingfact: the consequencesof an event reportedby meansof the (resultative)PrP are necessarilycomputedrelative to the setting in which the report is made. Accordingly,a resultativePrP sentence containing a telic participialVP may also be characterizedas evoking an 'indeterminateresult'. Considerthe followingexample: (30)

OK. I've washed your car.

While sentence(30) entails the presenceof a clean car at speechtime, it might also be used to negotiatefurtherconsequencesof that resultantstate (thehearermustnow pay the speaker,etc.).5Exampleslike (30)leadus to the conclusion that the resultant state entailed by the RPC is contextually determined(see also Fenn 1987; ParsonsI990); the constructionalsemantics specifyonly that soMEresultantstate obtains.Complement-verb telicitywill occasionallyenable the interpreterto adduce the relevantresult, but, as shownby (29), thatcue is not necessarilyin evidence.In suchcases,as we will see in the next section, the result in force at speech time is simply that situationwhich is potentiallysignificantto the interlocutorsin theirjoint determinationof immediategoals. I will arguethat wherea presentresultant state has no role in determiningan imminentcourse of action, that state is a poor candidatefor presentationvia the RPC. Analyseslike Brinton'sdo not thereforesucceedin reducingPrPreadings to the interactionof constructionaland lexicalsemantics.Theydo, however, underscorethe fact that the mere presence of co-occurrencerestrictions uniquelyassociatedwith one or the otherof the readingsis a poor diagnostic for ambiguity.As Zwicky& Sadock (I975) argue, ambiguityclaims based upon this type of diagnostic(whichthey referto as the ' addedmaterial'test) fail when the distinct co-occurrencepossibilities are semanticallynonarbitrary,and the lexical materialin questionactuallyinducesratherthan reflectsthe distinctunderstandings at issue.Thepresentanalysisdoes not rest upon this unreliable ambiguity test. First, the diagnostics based upon pragmaticpresuppositionand eventsequencingdo not makereferenceto cooccurrencerestrictions.Second,whilemannerand time adverbsdo represent [51Slobin

(I990) describes the role of such negotiation contexts in the acquisition of the resultative PrP by children.

132


THE ENGLISH

PRESENT PERFECT

'added material',it is not immediatelyobvious how the presenceof one or the other of theseelementswould inducean existentialas againstresultative reading. As I will demonstrate,the past-time adverbs accepted by the existential (e.g. in June) are themselvesambiguous between definite and indefinite readings. Since these adverbs are necessarily interpreted as indefinitesin the contextof the existentialPrP,it appearsthat the existential readingdeterminesthe readingof the adverb,ratherthan vice versa. Additionally,as McCawley(I97I) points out, the claim of ambiguityis bolsteredin this case by the identitytest:6 (3I)

Harry and Marge have been fired.

Sentence(3I) cannot be used to assertboth that Harrywas firedat some point withinhis employmenthistory(existentialreading)and that Margeis currentlyout of work as a resultof having been fired(resultativereading). Anothertest whichsupportsthe ambiguityclaiminvolvesthe cross-linguistic potentialfor the distinctreadingsto be manifestedas distinctlexicalitems (cf. Zwicky& Sadock I975). For example,Mandarinformallydifferentiates the resultativeandexistentialreadings:the resultativereadingis expressedby the coverble, the existentialreadingby the coverbguo (see Comrie1976; Li et al. 1982). 4.

SEMANTIC

STRUCTURES

4.1 . Time-spanperfects

The continuativeand existentialreadings of the PrP have the following commonsemanticproperty:both locate an episode(an eventor statephase) with respectto a time span which includesthe present.In the case of the existentialPrP, one or more events of a given type are located WITHINthis time span; in the case of the continuativePrP, a state phase occupies the entire time span. Features of adverbialco-occurrencereflect the shared semanticstructure.For example,both readingsare compatiblewith sinceadverbialexpressions: (32) (a) Harryhas been in Bali since Saturday. (b) Harryhas been in Bali twice since Saturday. In both (32a) and (32b), the since-adverbialexpressiondenotesthe lower boundaryof a time span whose upper boundaryis speechtime. Mittwoch [6] Dinsmore(i99i) notices exampleslike the following,in which the identitytest yields a differentresultthan it does in (30): (i) Harry has lived in France intermittentlysince the war and so has Jane, continuously.In this example,the conjunctionof continuativeand existential understandings does not resultin zeugma.Onewonders,however,whetherthe presenceof the adverbialspecifications(e.g. intermittently)licenseswhatwould otherwisebe a 'crossedreading'. I33


LAURA A. MICHAELIS

(I988: 207ff.) notices the following distinctionbetween 'universally'and

'existentially' interpreted since-adverbialexpression: in the existential readingof the sentenceSam has been in Boston sinceTuesday,'Tuesdayis excludedfrom the rangeof possibleintervalsof Sam'sbeing in Boston that are covered by the sentence'. That is, episodes fall between the temporal poles representedby upper- and lower-boundingtemporalspecifications; statesextendthroughthosepoles (andperhapsbeyondthem).I will presume, in accordance with Mittwoch, that the since-adverbialconstruction is polysemous,havingboth universaland existentialreadings.These readings are relatedby virtueof the fact that both expressa time span whose lower and upper boundaries are anchored by event and reference times, respectively. Durationaladverbs(e.g. for two hours)overlap functionallywith sinceadverbialexpressions:durationaladverbsmay also specifythe length of a state phase which obtains through to the present time. In this capacity, durational adverbs co-occur with the continuative PrP. Consider the followingexamples: (33) (a) Myron has been upset for an hour. (b) Myron has been upset since three. As shown in (33), durational adverbs differ from since-adverbial expressionsin that the formerdenotean expanseof time occupiedby a state phase,while the latterdenoteonly the lowerboundaryof that expanse- the upperboundarybeingfixedby referencetime (i.e. speechtime in the case of (33b)). Therefore,(33b) is synonymouswith (33a) only if the former is utteredat four o'clock. Since-adverbial expressionslike that in (33b)require that an unspecifiedupperboundarybe fixedby a referencetimedistinctfrom the time of inceptionof the denoted state phase (this referencetime is the presentin the case of the PrP).Durationaladverbs,whichreferto a temporal expanse ratherthan a temporalboundary,do not have this requirement. Therefore,durationaladverbsare compatiblenot only with the continuative PrP but also the preterite: (34) Myron sat on the porch for an hour. Durationaladverbsare not polysemousin the way that since-adverbials are. Whiledurationaladverbsspecifythe lengthof time occupiedby a state phase, they do not have an 'existentialreading',in which some numberof events are located within the period denoted. Therefore,sentenceslike the followingare anomalous: (35) *Larryhas visited us twice for the last threeyears. In (35), the three-yearperiodmust be denotedby an in-headedexpression of temporalextent. Whilefor-headed durationaladverbsdo not have an existentialreadinglike the comparablereadingof since-adverbials, they can I34


THE ENGLISH

PRESENT PERFECT

co-occur with the existential PrP. As we saw in the previous section, sentences like (27) are ambiguous between continuative and existential readings.Sentence(27) is repeatedbelow, along with the relevantreadings: (27)

Harryhas been in Bali for two days. continuative:Harry'spresencein Bali obtainsfor all timeswithina present-inclusivetime span whose lower bound is two days ago. existential:Therewereone or morevisits to Bali by Harrywithina present-inclusivetime span; each of these visits lasted two days.

Dowty (I979), Richards

(I982)

and Mittwoch (I988) argue that the

ambiguityexemplifiedin (27)involvesthe relativescopesof Have(theperfect operator) and the durational adverb. Richards represents this scope ambiguityin the followingway: (36) (a) existential:PRES (W,i) [Have [for two days (Harrybe in Bali)]] (b) continuative:PRES ((W,i) [fortwo days[Have(Harrybe in Bali)]] Evidencefor this scope ambiguityis providedby the fact that when the durationaladverbis preposed,(27) has only the continuativereading(For two days, Harry has been in Bali). The preposing is here said to reflect the

wide scope of the durationalwith respectto the perfectoperator.However, Heny (I982) arguesthat the scopeassignmentgivenin (36b)is not plausible. His argumentis basedon the truthconditionswhichhe assignsto the perfect, whichrequirethat the base sentencebe trueat a NONFINAL subintervalof the interval for which the perfect sentence itself is evaluated. If the overall intervalis a present-inclusivepast period, the base sentence(Harrybe- in Bali) must be true at some time priorto now. In the case of (35b), 'now' is a two-yearinterval,andeachsubintervalof thatperiodis an evaluationtime. As a consequence,at eachsubintervalof the two-yearperiod,the proposition Have (Harrybe- in Bali) must be true at some nonfinalsubinterval. Heny points out, however,that thereare moments(singletons)amongthe subintervalsof the intervaldenotedby a durationaladverb,and a singleton cannotbe said to have a nonfinalsubinterval.Mittwocharguesthat one can overcomethis objectionby droppingHeny's 'nonfinal' conditionfrom the truth conditions for the perfect. Nevertheless, there remains a cogent argumentagainst the scoping given in (36b): durationallybound statepropositions (in this case, for two days [Have [Harry be in Bali]]) cannot, as

required here, be evaluated for the present moment. Bounded-state propositions like Moe be- in the basement for an hour are like event

predicationsin that they lack the subintervalproperty:no subintervalof Harry'sbeing in the basementfor ten minutesis an instanceof that whole episode. Since speechtime is conceivedof as an instantby convention,and since events(includingstate-phaseevents)are not instantiatedat any single momentof the intervalin whichtheyoccur,to assertthe existenceof an event is to reportits culmination.Therefore,in English,neithereventpredications '35


LAURA A. MICHAELIS

nor bounded-statepredicationsare amenableto present-tensereporting,e.g. *Harry is illfor two days.

The scope assignmentin (36b) requiresthat the interpreterevaluatethe truth of a tenseless state-phaseproposition(Have [Harrybe in Bali]) for speechtime. However,as we saw above, speechtime is a moment,and an event proposition cannot be said to be true at a single moment alone. Therefore,(36b) is ill formed.It should be noted that Mittwoch'sanalysis does not fall prey to the objectionjust made. Mittwochdoes maintainthat the continuativeinvolvesa durationaladverbhavingwide scopewith respect to the perfect operator; however, she gives truth conditions for the continuative perfect that would require that (27), on the continuative reading,be evaluatedat the end of Harry'sstay in Bali. This endpointis the presenttime. Giventhis type of analysis,however,I see no needto recognize a wide scopingof the durationalvis-a-visthe perfectoperator.Instead,the continuativecan be assignedthe samescopingas the existential,in whichthe durationadverbhas narrowscope with respectto the perfectoperator.The ambiguity exemplifiedin (27) will then reside in the distinct semantic representationsaccordedthe two perfecttypes: the existentialindicatesthe existenceof one or more events (e.g. state-phaseevents) within a presentcontiguoustime span;the continuativeindicatesthe cessationat speechtime of a phase of a state. A logical representationfor the continuativePrP, loosely based upon Parsons'(I990) formalism,is given in (37)V: (37) 3!e: [State-phase (t)]' (e) 3!t': Culminate (e, t') & 't' is immediately

prior to now'8 This representationstates that thereis a uniqueevent which representsa state phase, that this state holds for an interval t, and that the event [7] The featuresof Parsons'analysisthat are preservedare: quantificationover eventsand timesanduse of the operatorCulminate to assignan eventan endpoint.It shouldbe noted thatParsonsdoes not use the operatorCulminate to describethe perfect,sincefor himthe currentpresenceof a resultstate(denotedby the specification'e's result-stateholdsnow') entailsculminationof the event.SinceI hold thatnot all perfectreadingsinvolvea present resultstate,I cannotrelyupon the presenceof a presentresult-statespecificationto secure culminationof the eventin question. [8] Notice that this formulationdoes not requirethe presenceof an adverbialphraseheaded by since orfor. While state-phasepredicationsare usuallyaccompaniedby an adverbial specifyingthe durationof the phase, they need not be. Herweg(iggib) points to the possibilityof 'implicitbounding'in exampleslike (i): (ii) Sue was upsetwith me twicelast week. Thissentencedenotestwo phasesof the 'upsetstate', as indicatedby the presenceof a cardinal count adverbial(twice),otherwisecompatibleonly with eventpredications(cf. Herwegiggia, b). By the sametoken, continuativeperfectsentencesmay denotea statephasein the absence of a durationaladverb.Note B's responsein the followingexample: (ii) A: Why haven'tyou been around? B: I've been ill. I36


THE ENGLISH

PRESENT PERFECT

culminatesat t'. The state-phasesituationtype is equivalentto the POFECTIVE The statementin singlequotes is meant to representa conventionalimplicature:the time at which the state phase ends (i.e., the time at which the corresponding'pofective event' culminates)is a time 'just before' speechtime. Notice that (37) presupposes the definitionof STATEPHASE:'a period t is a phase of state S just in case S holds at t and t is not properlyincludedin a period t' at which S holds as well' (Herweg I99I: 992). This definitionis problematic,since state-phase predicatesaccompaniedby a durationalexpressionare upwardcompatible.

EVENT TYPE recognizedby Herweg(I99I).

For example, the sentence Harry was sick for at least a week implies that the

period duringwhich Harrywas ill in fact exceededa week'stime. Herweg'sdefinitionof state phase entailsthat the state cannot obtain for timesotherthan those for whichit is assertedto obtain.However,we know that speakersmay highlight phases of a state that are in fact properly includedwithin some largerperiod duringwhich that state holds as well. Accordingly,let us adopt the followingreviseddefinitionof STATEPHASE:'a statephaseis a periodduringwhicha stateholds, whereany periodproperly includingthat periodis outsidethe speaker'simmediatefocus of attention'. The reviseddefinitiongives us a betterresultwhen we attemptto describe sentenceslike the following: (38) Paul's been living in Bouldersince at least I989. In the case of (38), we must account for the speaker'sintuitionthat the state of Paul's living in Boulderwill probablycontinuepast now, and may have obtainedpriorto i989. Given the reviseddefinitionof state phase, we can say that a personchoosingto utter (38) is directinghis or her attention to a singleperiodat whicha stateholds,whiledisregardingtimesoutsidethat period at which the state may hold as well. While(37) presupposesthe definitionof statephase,the definitionof state phase in turn presupposesthe definitionof state, which in most formal modelsinvolvesuniversalquantification:'if S is a state whichholds at time t, all subintervalsof t are also times at whichS holds as well'. Accordingly, McCawley(197I, I98i), refers to the continuativePrP as the 'universal perfect'. The continuativePrP is said to indicate that all times within a present-inclusiveinterval are times at which the denotatum of the VP complementholds. McCawley'sdefinitionprovidesa clear explanationfor the fact that the existentialand continuativePrPs are synonymousunder negation.Considersentence(39): (39) Nobody has told me the truth. Sentence(39) has the followingequivalentreadings:(a) universal:for all timeswithinsome present-inclusive time span,the base sentencenobodytellme the truthholds, and (b) existential:thereare no times within a present'37


LAURA A. MICHAELIS

inclusivetime-spansuch that the base sentencesomeonetell me the truth holds. I will representthe semanticstructureof the existentialas in (40): (40) 3e: Event (e) 3t: t < now Culminate(e, t) & 'the event type is one

which is replicableat the presentmoment' In (40), the existentialquantifierbindingevent and time variablesallows for multipleinstantiationsof a given event. As a consequence,this formula does not evoke a uniquetime of occurrence.The temporalvariableis simply restrictedto times prior to now, although a lower-boundaryspecifierlike sincenoonmay furtherrestrictthe range.All eventsmustinstantiatean event typewhichfulfilsthe 'present-possibility' requirement.McCawley(I98I: 82) describesthe requirementas follows: 'the speakerand addressee'sshared knowledgedoes not rule out the continuedoccurrenceof eventsof the kind in question'. This requirementis representedhere by the materialin single quotes; it is viewedas a conventionalimplicature.McCawleydemonstrates the existenceof this constraintby means of exampleslike (4I): (4I) (a) Have you seen the Monet exhibition?

(b) Did you see the Monet exhibition? McCawleyobservesthat (4Ia) is appropriateonly given a situation in which an event of exhibition-visitingby the addresseeis still possible.That is, the exhibitionis still open and the addresseeis capableof viewingit before it closes. If the exhibitionis closed or the addresseeis not in a positionto see it, only the preterite-formquestion(4ib) is appropriate. The acceptabilityof a given instance of the existentialPrP will depend upon the extentto whichinterlocutorscan construethe eventdenotedas one which can still occur. Often, this construalrequiressome work. McCawley (I97I) points out that the widely discussed sentence (42) is a priori anomalouson an existentialreading: (42) Einstein has visited Princeton.

Sentence(42) is odd, since Einstein is deceasedand therebyno longer capable of visiting any institution. If, however, one construes the event categoryat issue as containingeventsof 'Princeton-visitingby a luminary', the sentenceis acceptable.This readingis aided by the presenceof narrowfocus accent on the subject: (43) How can you say that Princetonis a culturalbackwater?]EINSTEIN has visited Princeton. The narrow-focusaccent evokes the existenceof an open proposition'x has visited Princeton', such that Einstein, as well as living individuals representingcurrentpotentialvisitors, are argumentsof this propositional function.This then illustratesthat the formal objectwhich correspondsto 138


THE ENGLISH

PRESENT PERFECT

the eventtype can be eithera full proposition(the addresseevisitsthe Monet exhibit)or a propositionalfunctioncontainingan argumentvariablewhose rangeis restrictedto certainentities(e.g. famous academics).Belowwe will encounterevidencewhich suggeststhat the present-possibilityconstraintis reflectedin co-occurrencepossibilitiesfor temporal adverbs- possibilities which characterizethe existentialas againstresultativePrP. We noticed above that the existentialPrP co-occurswith cardinalcount adverbslike twice,which specifythe numberof instantiationsof the event type denoted by the participialVP. Another class of frequencyadverbs welcomed by the existentialPrP are ordinal count adverbs.Considerthe followingsentences: (44) (a) This is the first time that Mel has ever eaten sushi. (b) This is the second time that Mel has (*ever)eaten sushi. Sentence(44a) specifiesthat therewas one instanceof Mel's eating sushi withina present-inclusive time span.The adverbeverindicatesthat the time span lacks a lower boundary- an unlimitedarray of past times is under consideration. Sentence (44b) shows that, in a PrP context, ever is incompatiblewith ordinals greater than one. Why should this be? W. Ladusaw(pers.comm.)suggeststhat the time-spanat issuein sentenceslike (44b) has an implicit lower boundary:the time of the FIRST sushi-eating event. Therefore,(44b) does not referto an unlimitedrange of past times, and everis not acceptable. 4.2. The resultativeperfect

McCawley(I98I) and Mittwoch(I988) have suggestedthat the PrPhas two basic readings: continuative and existential. While Mittwoch does not considerthe possibilityof a resultativereading,this possibilityis explicitly rejectedby McCawley.Abandoninghis earlierclaim that the PrP is threeways ambiguous, McCawley argues that the resultative understanding 'should be treated as an existential present-perfectaccompaniedby an implicature(whetherconversationalor conventionalI do not know)that the event type that is referredto would normallyresultin the presentstate of affairsthat the speakerconveys is the case' (I98I: 84). Under this view, a sentencelike (45) is merelyan existentialPrPaccompaniedby an implicatum to the effectthat a past eventof Harry'smovingawaywouldordinarilyresult in his currentabsencefrom the neighborhood: (45) Harryhas moved out of the neighborhood. McCawleyreducesthe resultativeto an existentialreadingbecause the resultativereadingfails to pass the logical 'contradictorytest'. That is, the negativeversion of a PrP sentencebearingthe resultativereadingdoes not express the contradictoryof the affirmativeversion. McCawley (197I) '39


LAURA A. MICHAELIS

definedthe resultativereadingas indicatingthat the 'direct effect of a past event still continues' (i98I: 8I). Given this analysis of the resultative reading,we arriveat an odd resultconcerningnegativeresultativesentences like (46): (46) I haven'tbrokenmy arm. McCawleyagues that under his earlieranalysis of the resultative,(46) could be used to assertthat an arm once brokenis now healed(i.e. that the direct effect of a past event does NOT continue). This is not, however, a possible reading of (46). Sentence(46) simply indicatesthat there was no event of arm breaking within the relevant interval; it is therefore an existentialPrP.McCawley'sargumentto the contrarynotwithstanding,I will give the resultativeand existentialreadingsdistinctsemanticrepresentations. In what follows I will show that McCawley'sobservationabout (46) can be explainedwithoutdiscardingthe assumptionthat the resultativereadingis a distinctreadingon the semanticlevel. I will representthe resultativereading as follows: (47) ]!e: Event(e) 3!t: t < now Culminate(e, t) & 'e's resultsstate holds now'. In (47), an eventis an existentiallyboundvariabledescribedby a predicate indicatingthe event type. As indicatedby the ]! scoping events and times, there is only one event-timepairingprior to now. The operatorCulminate pairs the event with its time of culmination.The materialenclosedin single quotesrepresentsa conventionalimplicature;as such,it cannotbe cancelled. Considerthe contrastin (48a, b): (48) (a) I put your shoes in the closet, but they'renot therenow. (b) I have put your shoes in the closet, *but they'renot therenow. Both sentencesassert that the speakerplaced the shoes in a particular location,and implicatethat the shoesnow residein that location.In the case of (48a), however, the resultant-stateimplicationis akin to a generalized quantity-basedimplicature:the assertionis relevantand/or informativeonly insofar as the event describedhas some presentconsequences.While the resultant-stateimplication attached to the preterite-formassertion is DEFEASIBLE(LevinsonI983), that attachedto a present-perfect formassertion is not. The existentialand continuativePrPs can have resultant-stateimplications like that linked to the preteritein (49a): (49) (a) I've read De Oratore threetimes (so I can explainit to you). (b) I've been ill (so I haven'tgotten aroundto it). The resultant-stateimplicationsattached to (49a, b) are evoked by a hearerattemptingto discernthe relevanceof the PrP-formassertion.The 140


THE ENGLISH

PRESENT PERFECT

sentencesin (49) can be regardedas instancesin whichtwo PrPreadingsare mutually compatible (see Norvig I988). For example, (49a) has both existential and resultativereadings: three reading events have occurred and, as a consequence,the readernow has knowledgeof the text in question. As noted earlier, the result state whose existence is conventionally implicatedby the RPCis not necessarilydeterminedby the linguisticcontext. The presentresultwhichthe speakerwishesto call attentionto is not always akinto the outcomecodedby a telicVP-complement.Sentence(29), repeated below, illustratesthat resultantstates can be computedfor PrP sentences which contain atelic VP-complements: (29)

I have knocked(so someone should be coming,etc.).

In general,the interpreterof an RPCconstructmustcomputethe relevant resulton the basis of extralinguisticand linguisticcues (e.g. the outcome,if any, entailedby the Aktionsartof the participialcomplement)Onlylexically encodedend-stateswill be subjectto the defeasibilitytest applied(48). McCawley'scontradictorytest demonstratesthat when one negates a resultativePrP,one is negatingonly the existentialassertionthatthe denoted event occurred.This fact, however,need not be takenas indicatingthat the resultativereadingis reducibleto the existentialreading.One can instead presumethat the resultativeinterpretationattachesto affirmativesentences only. This constrainthas an obvious semanticmotivation.In negatingthe sentence I have broken my arm, one necessarilydenies the existential assertion. This in turn removes the possibility that the resultant-state implicatumcan be satisfied;the resultantstatecould not possiblyobtain,as the requisitecausal event did not occur. The resultant-stateimplicatumincorporatesa pragmaticvariable:it can be viewedas a directiveto the interpreterto findin the context,linguisticor extralinguistic,a methodof relatingthe past event denotedto some feature of the presentsituationcharacterizable as a consequenceof that event.As Li et al (I98i) observe with respectto the Mandarinperfect markerle, the presentationof a present result often representsa demand for action (includinga verbal response)from the hearer.Consider,for example,the following sentence: (50) My car has been stolen.

In (50), the absenceof the car is evoked as potentialcatalystfor future action (e.g. phoning the police). The resultantstate is that situationwhich determineswhat happens next (Slobin I990). That is, the resultantstate inferredby the heareron the basis of the PrP-formassertionis not only the currentabsenceof the car, but also the fact that somethingmust be done about the absenceof the car. Sincesome resultantstatesrepresentsituations requiringimmediateresolution,the felicityof the resultativePrP will often dependupon recencyof the participial-VPdenotatumvis-'a-visspeechtime. '4'


LAURA A. MICHAELIS

Notice that (5) wouldnot be an appropriateresponseto a questionerseeking to know, a week or so afterthe theft, why the speakeris drivinga rentalcar. The resultantstate arisingfrom an eventmay last indefinitely,but represent a temporarysituationfrom the point of view of 'crisis management':the resultantstate (qua actionablesituation)will cease once the interlocutors have determined a strategy for reckoning with the consequent crisis, althoughthe situationwhichprecipitatedthat crisis(e.g. the absenceof the car) may not cease. 5. GRAMMATICAL AMBIGUITY

REFLEXES OF EXISTENTIAL-RESULTATIVE

As mentioned,the existentialand resultativePrP differin that the RPC (a) cannotreferto a pragmaticallypresupposedeventin, e.g. a contentquestion or cleft (b) rejectstemporalmodificationof event time (c) rejectsparticipial complementsdenotingevent complexes,and (d) does not welcomemanner modificationof the VP-complement.Letus examineeachof theseconstraints in turn. 5. I. Pragmatic presupposition According to Lambrecht (I99I:

I), the pragmatically presupposed com-

ponent of an assertion is '[t]he set of propositionslexico-grammatically evokedin a sentencewhichthe speakerassumesthe heareralreadyknowsor believesor is ready to take for grantedat the time of the utterance'.For example, in a sentence like It was Harry who borrowed the rake, an event

proposition Someoneborrowedthe rake is pragmaticallypresupposed.A convenientshorthandto be employedhere allowsus to the referto the cleft sentence as pragmatically presupposing an EVENT,rather than an EVENT PROPOSITION.

In their anaphoriccapacity,past-tensesentencesserveto elaborateupon circumstancessurroundinga pragmaticallypresupposedevent,i.e. an event previouslyassertedrelativeto the referencetime in question.An exampleof the 'elaborationmode' is providedby the secondand thirdsentencesof the narrative in (5I):

(5I) Haywardpolicehavearrestedthe primesuspectin last week'sstring of laundromat robberies. Two off-duty officers confronted the suspectas he left a local 7-I I. A back-upunit was calledin to assist in the arrest. Following Kamp & Rohrer (I983: 26I), we can observe that the second

and thirdsentencesin (5I) 'are naturallyunderstoodas constitutiveof, and thus as temporallyincludedin, the event introducedby' the first sentence. The first sentence, a resultativePrP, is used to inform hearers of the 142


THE ENGLISH

PRESENT PERFECT

occurrenceof the arrest.Subsequentsentences,pragmaticallypresupposing that event,providefurtherdetailsabout it. The resultativePrP cannothave other than an event-reportingfunction in narrativeslike (5I). Notice the oddity of (52) (where # indicatesanomalyon the RPC reading): Haywardpolicehave arrestedthe primesuspectin last week'sstring of laundromatrobberies.#Two off-dutyofficershave confronted the suspectas he left a local 7-I I. # A back-upunit has been called in to assist in the arrest.

(52)

The anomalyof (52) can be attributedto the restrictionstatedin (53): (53) The RPC cannot be used to furtherdescribe('elaborateupon') a pragmaticallypresupposedevent. Both Comrie(1976) and Dinsmore(I98I) haveproposedversionsof (53), although both authors fail to recognize, as we will recognize,that this constraintattachesto the RPC ratherthan to the PrP per se. One can see furthermanifestationsof (53) in (54) and (55): (54) I can't come tonight. I've brokenmy ankle in a skiingaccident. (55) A: My God! Look at that cast! B: # I've brokenmy ankle in a skiing accident. In (54), the speakeris both establishingthe occurrenceof an event (of anklebreaking)and simultaneouslyprovidingsome informationabout that event(it occurredduringskiing).In (55),by contrast,speakerB is responding to an utterancewhichlicensesthe assumptionthat the recentoccurrenceof a fracture is knowledge common to the interlocutors.The response, therefore, provides further information about that pragmaticallypresupposedevent. As such, it cannot appearin PrP form. There are certain grammaticalconstructionsdedicatedto providingor requestingadditionalinformationabout a pragmaticallypresupposedstate of affairs.Amongtheseconstructionsareclefts,as noted,and wh-questions.9 [9] The RPC is not necessarily incompatible with wh-questions requesting information about circumstances surrounding the event denoted by the VP-complement. In particular, we find that who-questions like the following are acceptable: (i)

Who has made this terrible mess?

Here, the wh-question is used to request the identity of the agent responsible for the past action denoted, presumably presupposing the open proposition x made this terrible mess. Exceptions like (i) may be attributable to a general property of who-questions: such questions may require that the addressee accommodate to the relevant presupposition, rather than necessarily exploiting shared knowledge of that presuppositional material. Lambrecht (994), for example, cites cases like the following: (ii)

Who wants a cookie?

The speaker of (ii) cannot be said to be accessing the presupposition that someone in fact wants a cookie; the question in some sense simply induces the hearer or hearers to behave 'as if' this

'43


LAURA A. MICHAELIS

As predicted,neitherconstructiontype readilyacceptsthe RPC, as shownin (56) and (57): (56) Don't thank ME. # It's HARRY who's selectedthe wine. (57) #Where have you found my watch? With specificclasses of predicates,nominaldefinitenessand pronominal anaphora will evoke a pragmaticallypresupposedevent; the RPC will accordinglybe proscribed.As Dinsmorepoints out (I98I) one such class of predicatesare verbs of creation.When the direct object denotes a unique createditem,a PrPdenotingan eventof creationhas a resultativeratherthan existentialinterpretation,since a given act of creationis not replicable.An NP invokingan existentialpresuppositionwill, when servingas the objectof a verb of creation,requirepragmaticpresuppositionof the creationevent. Notice the contrastbetween(58) and (59): (58) Look! Myron'spainteda little picture. (59) 4Myron's paintedthe little picture/it. In (58), the existenceof the pictureis not presupposed,and thereforethe paintingevent is not presupposed.The PrP can thereforebe used to denote the creationevent.In (59), by contrast,the occurrenceof the paintingevent is establishedas knownto the interlocutorsvia the existentialpresupposition linked to the definiteor pronominalNP. Sentence(59) has the effect of providingfurtherinformationabout a pragmaticallypresupposedevent;the sentencetells us the identityof the painter.The PrPis therebyunavailablein (59).

Similar effects are observable among verbs of transfer, where a linguisticallyuninstantiatedrecipientrole is filledby the speaker.Whenthe themeargumentof sucha verbis coded by a definiteor pronominalNP, this signalsthat the codedentityis IDENTIFIABLE.Recallthat an identifiableentity is 'one for whicha sharedrepresentationalreadyexistsin the speaker'sand hearer'smind at the time of the utterance'(Lambrecht,1994). With respect to an itemtransferredtowardthe interlocutors,identifiabilitystemsfromthe availabilityof that item to the recipients,i.e. its placementat the deictic center.Availabilityin turnentailsthat the occurrenceof the transfereventis pragmaticallypresupposed.Use of the PrP is accordinglyruledout in such contexts.. Consider (6o) and (6i): propositionweresharedknowledge.Similarly,in the case of (i), the knowledgethat someone made this particularterriblemess (and that thereis in fact such a mess)cannotbe said to be sharedby speakerand addressee.The uttererof (i) is in fact directingthe attentionof the addresseeto a messthat has not previouslycome underdiscussion.In casesin whichthe event propositionis necessarilysharedknowledge,questionsaboutagentidentitycannotbe expressed by meansof the RPC: (iii) *Who has done your hair?

I44


THE ENGLISH

PRESENT PERFECT

(6o) How touching.The Millershave sent a fruitcake. (6i)

#The Millershave sent the fruitcake/it.

In (6i), as against (60), the identifiable nature of the theme argument (the

fruitcake) evokes a mutual understandingthat a transfer event- one responsiblefor the accessibilityof the theme- has occurred.'0The PrP is ruledout, owing to the fact that (6i) providesadditionalinformationabout a pragmaticallypresupposedtransferevent: it identifiesthe senders. The restrictiongiven in (53) does not rule out sentenceslike (62). This sentenceis contrastedwith sentence(57), repeatedhere as (63): (62) Wherehave you hiddenmy watch? (63) #Where have you found my watch?(= (57)) As we saw, (63) is anomalousbecauseit requestsadditionalinformation about a pragmaticallypresupposedfindingevent.Why is the PrPacceptable in (62)? In uttering(62), the speakeris seekingfurtherinformationaboutthe resultantstate of the hiding event, i.e., wherethe hiddenentity resides.In (63), by contrast, the location query concerns the past event alone: the discoverysite is not construedas the locationwherethe found entityresides at present.Wh-questionslike (62) are acceptablebecausethey are construed as requestsfor informationabout circumstancessurroundingthe resultant state ratherthan the prior event. Example (64) demonstratesthat the constraintgiven in (53) does not characterizethe existentialPrP: (64) Wherehave the police arrestedthe suspect? Sentence(64) is ruledout on a resultativereading;the sentencecannotbe used to inquireabout the place of arrestof a suspectnow in custody.This sentence is, however, renderedacceptablewhen an existentialreading is invoked:the speakerseeksto locate an arrayof eventsin whicha particular suspectwas arrested.Under this reading,an appropriateresponseto (64) might be: He has been arrested in Berkeley, in WalnutCreek, etc. Given that

(64) is nonanomalouswhen construedexistentially,we are led to conclude that the constraintstatedin (53) is associatedexclusivelywith the resultative PrP. This constraintdoes not appearto be an arbitraryone when we assume that the RPC cannot express anaphoric temporal reference.The nonanaphoricnatureof the resultativePrPis shownby the fact that it cannotbe used to invoke an extralinguistictemporal antecedent.In section 2, we [io] The readeris askedto ignorea readingof (6i) in whichthe NP thefruitcakehas a type

reading.Sincethe articlein this casewouldbe a genericarticle,ratherthanone indicating a uniquelyidentifiablereferent,(6i) wouldnot presupposethe transferof the themeto the deicticcenter.Instead,givena type readingof the NP thefruitcake,we wouldbe inclined to see (6i) as an event-reporting sentence.

'45


LAURA A. MICHAELIS

noticed that the preterite- when operatingin an anaphoriccapacity- may evoke an extralinguistictemporalantecedent.For example, the preteriteform assertion in (65) evokes the time of the discourse contribution immediatelyprior: (65) I didn't hear. I had the water running. The extralinguisticantecedent may be the time of an event whose occurrencein the text-externalworld is thought to be salient: (66) Did you see that huge wasp fly by? In (66), the time evoked by the preterite-formquestionis not that of a recent discourse contribution,but rather the time of an event which is presumedsalient within the shared past experienceof the interlocutors. Notice that the RPC is unavailablein the contextsdescribed: (67) (a) #1Ihaven'theard.I had the waterrunning. (b) #Have you seen that huge wasp fly by? Both (67a) and (67b) have potential nonresultativeinterpretations.In (67a),the PrP-formassertionis likelyto be interpretedas a continuativePrP: the addressee'slast discoursecontributionis framedas a state phase whose durationis includedwithinthe periodduringwhich the waterwas running. In this case, the PrP-formassertioncan be paraphrasedin the followingway: I haven't heard what you've been saying for the past few minutes. That is, all

times at whichthe addresseewas speakingare timesat whichthe waterwas running. This interpretationis unavailable in a context in which the addressee'sinaudiblecontributionwas, say, a monosyllabicresponseto a yes-no questionpreviouslyposed by the speaker. In (67b),the PrP-formassertionhas a possibleexistentialreading,in which the speakerpresupposesthat the wasp is a regularvisitor, and that one or more sightingshave alreadyoccurred.In this case, the relevantreadingis evokedby the presenceof yet or before.Theseexamples,alongwith example (64), demonstratethat (53) does not characterizethe PrP per se. This constraintis a parochialrestrictionassociatedwith the RPC. Although(53) has a discourse-pragmatic motivation,viz. the anaphoricity contrastbetweenthe RPC and preterite,(53) wouldnot be predictablesolely from knowledgeof this contrast.Sentenceslike (62) show that referenceto a pragmaticallypresupposedevent is in fact permittedin contentquestions whichseek furtherinformationabout presentresultantstates.Furthermore, knowledgeof the anaphoricitycontrastwouldnot enablethe learnerto infer apparentlyinexplicableexceptionsto constraint(53). Oneexceptioninvolves why-questions.Considerthe contrastbetween(68) and (69) (noticedby C. Fillmore,p.c.): (68) Why have you signedyour name in red ink? (69) ?Whathave you signedyour name in red ink for? I46


THE ENGLISH

PRESENT PERFECT

It is difficultto understand,a priori, why the question in (68) should welcomethe RPC, whilethe apparentlysynonymousquestion(69) does not. constraint(53) has exceptionswhichmust be learned;the particularsof (53) cannot thereforebe predictedfrom the anaphoricityrestrictionupon the RPC exemplifiedin (67). Nevertheless,the fact that the RPC is markedas [-anaphoric] provides a reasonable'after the fact' explanationfor the existenceof (53): although not all anaphoricpast-timereferenceinvolves event elaboration,all instances of 'event elaboration' involve anaphoric past-timereference. 5.2.

Time adverbs

Whilesentenceslike (7Oa)areunacceptable,sentenceslike (7ob)arepossible: (70) (a)

*Harry has visited Rome in 1970.

(b) Harryhas arrivedon Tuesday(before). Heny (I982) argues that the type of adverb exemplifiedin (7ob) 'has instead of a specificreferentialsignificance,which picks out some definite time in the past, a quantificationalinterpretation'(p. I5i). Heny's analysis of (7ob)is somewhatelliptical,but we can presumethat by 'quantificational interpretation'he meansan interpretationthatis coherentwiththe semantics of the existentialPrP- existentialquantificationover events and associated timeswithina present-inclusivetime span. As Klein points out (I992: 549), adverbshavingthis interpretationdo not 'fix a singletime span' in sentences like (7ob). Called 'indefinite time adverbials'11by Heny, these adverbs denote a calendricaltime that is not unique,but recursat regularintervals: June, three o'clock, Winter. The adverbial refers not to a token of the

calendartime but to the type.Thus,the adverbialsof (7ob)mightbe realized as nonspecific indefinite NPs - on a Tuesday, etc.

'Cyclic' timeadverbialscontributeto the characterization of an eventtype that can recurat present.For example,presumingthat Harryis still alive, etc., the event of Harry's arrivingon Tuesday can be duplicatedon the specifiedday of the week in which (7ob) is uttered.Since speechtime is the time for which one presumesthat replicationof the vent is possible,speech time must be an intervalthat is largeenoughto accommodateany token of the cyclic time specified.In (7ob), for example,this intervalis equatedwith the currentweek; the time at which Harry'sarrivalcan be repeatedis the [ii]

The term 'indefinite time adverbial' is somewhat misleading, in that the adverbial class discussed here does not include a number of other adverbs commonly regarded as 'indefinite': recently, before, in the past year. These adverbs, as McCawley (I97I) argues, serve to restrict the range of the existentially bound time variable in the case of the existential perfect. Further restriction may be provided by a cyclic adverb. Hence, in Your TV show has gone on on Tuesdayfor the past year, the latter adverb confines times within the range to those within the present year; the cyclic adverb further restricts the airing times to those belonging to the set of 'Tuesdays'.

'47


LAURA A. MICHAELIS

Tuesday of this week. Following Parsons (I990), we may representcyclic adverbswithinthe logicalrepresentation(40) by includinga clausespecifying that all times bound by the existential quantifier belong to a set of 'Tuesdays', 'Noons', etc. Adverbialslike on Tuesday,which necessarilyhave a cyclic or indefinite readingwhen combinedwith the PrP,'2are ambiguousin othercontexts.In (7I), for example, the adverb may be interpretedas coding a definite or indefinitetime: (7I)

The healthinspectorcomes in on Tuesday.

The present-tensepredicationin (7I) is ambiguousbetween a futuratepresentinterpretationand a habitualinterpretation.On the futurate-present reading,the adverbialon Tuesdayrefersto a definiteintervallocatedin the future.On the habitualreading,on Tuesdayrefersto a cyclictime, tokens of whichrecurregularly.In sentenceslike (7ob), on Tuesdayis unambiguousonly the cyclic interpretationcohereswith the semanticsof the existential PrP. Therefore,the mere presenceof a past-timeadverblike on Tuesday cannot be regardedas inducing or creating an existentialreading of an otherwisevaguePrPconstruct.Suchadverbsmerelyhavethe potentialfor an indefinitereading- a potentialexploitedby the interpreterin reconcilingthe semanticcontributionsof PrP constructionand time adverb.In the case of the RPC, no such reconciliationis possible- a time adverblike at noon can have neithera definitenor an indefiniteinterpretation.An indefinitereading is not possible becausethe resultativedenotes a unique,nonreplicablepast event. A definitereadingis proscribed,owing to a constraintwhichwe saw earlier to be idiosyncratic:the time of the event in question cannot be specified.Thus, sentence(72) has an existentialreading,but not a resultative reading: (72)

Harryhas walkedthe dog at noon.

Sentence(72) can be used to assert that there have been one or more noontime dog-walking events by Harry within some undefinedinterval contiguouswith the present.On this reading,the sentenceconventionally [I2]

Klein (I992: 549), points out an instance in which an adverbial having the potential for a cyclic interpretation fails to have this interpretation when combined with the perfect. Consider (i) and (ii): (i) Chris has been in New York at Christmas (before). (ii) *At Christmas, Chris has been in New York.

(Klein incorrectly stars (i), taking at Christmas as necessarily definite.) The anomaly of (ii) can be attributed to the wide scope of the time adverbial signaled by its preposed position. The scope in question can be represented in the following fashion: at Christmas(Pres(Perj(Chris be in New York))). The operator Perf derives a state proposition, which, as shown, must be evaluated for the present moment. This present evaluation time clashes, however, with the past time reference coded by at Christmas. Since this adverbial is necessarily within the scope of Perf in (i), that sentence is readily interpretable as an existential perfect containing a cyclic past-time adverb.

148


THE ENGLISH

PRESENT PERFECT

implicatesthat this event can still occur. Sentence(72) cannot, however,be used at two or three in the afternoonto assert the presentexistenceof a walked dog, panting or enervatedas a result of having been exercisedby Harryat noon that day. The RPC-basedconstraintagainstspecificationof eventtime is amenable to a post hoc semantico-pragmaticexplanation. Since the RPC is [-anaphoric] it cannot be used to 'corefer'with an adverbialantecedent'; the RPC cannotusurpthe anaphoricfunctionof the preterite.However,the constraintbarringa resultativereadingof (72) appearsto be an idiosyncratic one. As noticed by Binnick(i99i) and Comrie(1976), among others, this constraintis uniqueto the EnglishRPC. Comrieobserves(p. 54): It is not clearthat the mutualexclusivenessof the perfectand specification of the time of a situationis a necessarystate of affairsin a language.In Spanish,for instance,wherethe Perfectdoes have specificallyperfect[sc. resultative]meaning,it is still possible to specifyexactly the time of the past situation,as in me ha levantadoa las cinco 'I have gotten up at five o'clock' (in replyto a questionwhy I am looking so tired). While learnersmight deduce a discourse-functionalfoundation for the time-specification constraint,that constraintis not a necessaryor predictable concomitantof resultative-PrPsemantics. 5.3. Event serialization

The RPC cannot be used to assert the occurrenceof an event sequence. Considersentences(73a, b): (73) (a) (b) (c) (d)

I have cleanedthe house and fed the dog. I have (??now)cleanedthe house and then fed the dog. Have ([I clean the house]and then [I feed the dog]) [Have(I clean the house)]and [have(I feed the dog)]

Sentence(73a) is ambiguousbetweenexistentialand resultativereadings. The former reading is brought out by the presenceof before; the latter readingis evoked by the presenceof now."3Owing to the presenceof the conjunctionand then,sentence(73b) has only the existentialreading,which is incompatiblewith nowin this context. The conjunctionand thenmust be interpreted as conjoining the predications denoted by the participial complements: I clean- the house and I feed- the dog. The interpretation in

question is representedin (73c). Here, the past event denoted by the conjoined participlesrepresentsan event sequence.Accordingly,sentence [I

3] The use of now as a method of imposing a resultative interpretation of the PrP is somewhat questionable, as now is also compatible with an existential interpretation of the PrP: Now rve been to Paris twice. The reading of now intended in (72b) is that in which now indicates that the consequences of event sequence denoted obtain at speech time.

'49


LAURA A. MICHAELIS

(73b) cannot be used to assertthe presenceof a clean house and fed dog at speechtime. By contrast,(73a) can have a resultativeinterpretationof this sort. This interpretationrequirescoordinationof the type representedin (73d). Under a resultativeinterpretation,(73a) codes two distinctassertions about presentstates of affairs. One may attempt to account for the constraintobservablein (73b) by arguingthat the referencetime of the PrP is the present,and that therefore the PrPis incompatiblewith the advancementof the referencetime required in the description of an event sequence (Partee I984; Dowty I986). The

validityof this type of accountis underminedby the fact that, as shown in (74) and (75), and PrPunderan existentialreadingcan be used to referto an event sequence: (74) Have you ever washedyour car and then had it rain? (75) Harry has often had a few too many and then regrettedit in the morning.

Sentences(74) and (75) are interpretedas asserting(or questioning)the existenceof one or more instancesof an event complexwithin a given time span. Thus, the existentialperfectcan be used to assertthe past occurrence of an event complex.The resultativeinterpretationis compatibleonly with the assertionof a simplexpast event. This constraintis difficultto justifyin termsof the resultativesemanticsrepresentedin (47); thereis no compelling reasonto assumethat an event complexcannotyield the requisiteresultant state. The constraint barringparticipialcomplementsrepresentingevent complexesthereforeappearsto be an idiosyncraticfeatureof the RPC. 5.4. Manner adverbs

Following Ernst (I987: 79), I will describemanneradverbsby means of a predicate-modification rule, describedin (76): (76) For any adverbmodifyinga predicatex, thereis an entity g which is a propertyof/aspect of/'something about' the eventualityof xing (by the subject)such that ADV (g). Ernst points out that a number of adverbs, like appropriately,may functionas eithersentenceor predicatemodifiers:we findbothAppropriately, Carol handledJay's lawsuit and Carol handledJay's lawsuit appropriately.In

the formercase, a contextualnormis computedfor possibleeventualities.In the latter case, a contextual norm is computed for possible manners of performingthe action in question. Certain manner adverbs encoding rapidity, like quickly, present an apparentproblemfor this scheme.Whilequicklycan referto the mannerin whichthe agentexecutedan action(as in She spokeveryquickly),it can also referto the speedwith which an event culminatedfollowinganotherevent. 150


THE ENGLISH

PRESENT PERFECT

In the lattercase, VP-initialpositionis preferred:WhenMarge'sunexpected guests arrived, Irving quickly made a quiche (versus ?made a quiche quickly).

Here,we neednot inferthat makingof the quicheinvolvedrapidmovements on Irving'spart,etc.; we simplyadducethattherewas a shorter-than-average time lag betweenthe event of the guests'arrivaland the point at whichthe makingof the quicheculminated.In suchcases,quicklyappearssynonymous with immediately.

Therefore,adverbs like quickly do not yield unequivocalresults with respect to the interactionof RPC and mannermodification.Notice that resultativesentenceslike the followingare ruled out: (77)

*The committee has quickly rejected my proposal.

If, however, quickly in (77) is taken to be synonymous with immediately,

then the anomalyof (77) can be attributedto the fact that evocationof an anterior reference-point(i.e. an event prior to the rejection event) is incompatiblewiththe resultativereadingof the PrP.As we sawin section5.2, denotatumcannotbe specified. the occurrencetime of the complement-verb This meansthat this eventcannotbe temporallylocatedby meansof a pasttimeadverblike at noonor by meansof 'indirect'temporalspecification:visa-visan anteriorpast referencepoint.The adverbquickly,on the immediately reading,necessarilyevokes a point with respectto whichthe encodedaction representsa rapid subsequent development.This point is 'virtually' a referencetime for the subsequentevent. Accordingly,sentenceslike the followingare anomalous: (78) *1 sent in an excellent proposal and those cretins have quickly rejectedit. Here, the time of bringingin the proposalis the anteriorreferencepoint with respectto which rejectionrepresentsa rapideventuality.The anterior event representsa temporalanchorfor the eventdenotedby the PrP.Notice that in the absenceof the adverbquickly,whichestablishesa temporallink between the sending and rejection events, the sentence is acceptable. Existential PrPs, which are not constrained by the time-specification constraint,acceptmannermodificationof the immediatelytype: (79) Wheneverhe has broughtin somethingunusual,the committeehas quicklyrejectedhis proposal. In (79), the referencepoint is the event of bringingin somethingunusual. The reference-pointevent may be contextuallyevoked,in sentenceslike the following: Harvard has typically rejected me immediately.This sentence may

be utteredin a situationin whichthe heareris awareof a set of time points at which the speaker submittedan applicationto Harvard;the speaker assertsthe rejectionoccursimmediatelyaftereachof the contextuallyevoked submissionevents. '5'


LAURA A. MICHAELIS

It is not clear, however, that quickly(on the immediatelyreading) is necessarilyincompatiblewith the RPC. The anteriorreferencepoint (with respectto whichrapidityof an eventuationis computed)may not explicitly evoke a past time. In such cases, quicklyis welcomedby the RPC. Consider the followingexample: (8o) The king of pop has quickly become the king of psychobabble. (Austin, American Statesman, 25 February

I993)

This sentence is a reference to Michael Jackson's recent televised revelationsof childhood emotional abuse. It evokes an anteriorreference point with respectto which the eventuationof Jackson'scurrentglossolalic state is a rapid development.This anteriorreferencepoint is the time at which Jacksonwas declared'king of pop' (at an MTV awardsbanquet). However, this past referencepoint is not invoked by a preterite-form predication,as in (78). Instead, the interpretermust reconstructthe past referencepoint on the basis of extralinguisticknowledge or pragmatic accommodation:therewas a certaintime at whichMichaelJacksonbecame the king of pop. Thatis, the interpretermustevokea coronationeventon the basis of the use of the title thekingof pop to describethe subjectdenotatum. That such an inferencingstrategyexistsis shownby the relativeanomalyof sentenceslike the following: (8i) ?TheDuchess of York has quicklybecomea figureof ridicule. Here, the NP TheDuchessof Yorkdoes not evoke a salientpast reference point with respectto whichthe Duchess'presentfallenconditionrepresents a rapiddevelopment. Given the possibility that an adverb of the quicklyclass may, when receivinga readingakin to that of immediately,welcomethe RPC, we have reason to disregardsuch adverbswhen we examinethe interactionof the RPC with adverbial manner-modification.Therefore, we will confine ourselvesto data like the following: (82) The presidenthas (??angrily)called a halt to the pressconference. (83) JudgeWapnerhas (??loudly)overruledthe defense'sobjection. As shown in (82) and (83), the RPC does not acceptthe manneradverbs angrilyand loudly.Notice, however,that (84) and (85) are acceptable: (84) WheneverMr Hume has questionedhim, the presidenthas angrily called a halt to the pressconference. (85) Wheneverthe defensehas made that objection,JudgeWapnerhas loudly overruledit. Sentences(84) and (85) are interpretableonly as existentialPrPs.That is, the sentencesassertthat there are a numberof eventualitesof a given type 152


THE ENGLISH PRESENT PERFECT

within a present-contiguoustime span, and that these eventualitiescan be characterizedas actions havinga particularproperty. Adverbs consistentlyrejectedby the RPC are those which Ernst calls 'pure' manneradverbs: a lexicosemanticclass comprisingadverbswhich 'representa quality specificallylinked to a certain type of predicate(e.g. thoseinvolvingsound,movement,etc.)' (ErnstI987: 78). Theseadverbs,e.g. loudly,arepredicatemodifiersalone.Otherclassesof adverbs(e.g. evaluative and agent-orientedadverbs) can serve as both predicate and sentence modifiers. Such adverbs co-occur felicitously with the RPC only when servingas sentencemodifiers,as shown in (86) and (87): (86) Stupidly,Bill has respondedto some hecklers. (87) Bill has respondedto some hecklers(??stupidly). In (86), the evaluation of stupidity attaches to the event of Bill's responding rather than to some property of his response, as in (87). Presumably,acceptanceof sentence-adverb modificationby the RPC can be attributedto the following fact: in cases like (86), the event-descriptor stupidlyevokes the presentconsequencesof the event denoted.In (86), one understandsthat Bill's respondingto the hecklerscan be characterizedas stupidonly insofaras thereare undesirableconsequencesof that event (Bill sacrificeshis dignity,etc.).As a predicatemodifier,stupidlydoes not facilitate inferencerelatedto the presentresult:one cannotjudge a prioriwhetherthe consequencesof Bill's having respondedin an obtuse mannerare desirable or undesirable.Becausepredicatemodifiersdescribethe event per se, they appearat odds with the communicativepurposeof the RPC: to depictthe presentconsequencesof a past event. Manneradverbsare, however,readily accommodatedby the existentialPrP, which simplyassertsthe existenceof one or more events of a given type; a manneradverbwill contributeto the identificationof this type. The constraintbarringmanner adverbshas some semantico-pragmatic basis: the RPC focusesupon the consequencesof an action,ratherthan the mannerin whichan actionwas performed.However,the constraintbarring mannermodificationrepresentsan idiosyncraticcharacteristicof the RPC, since it does not follow directlyfrom the relevantsemantics.There is no reasonin principle,it seemsthat one cannotreportupon both the mannerof an action and the currentlyaccessibleconsequencesof that action. The constraintbarringmanneradverbsattachesvia conventionto the RPC; it is not otherwiseinferrable. 6. CONCLUSION The EnglishPrP does not representa unitaryaspectualconstruction,but a complexof suchconstructions.Whilethe PrPencodeswhat has beentermed the 'currentrelevance'of a past event,currentrelevancemustbe subdivided '53


LAURA A. MICHAELIS

into continuative,resultativeand existentialvarieties: in the case of the continuativePrP, the culminationof a state phase immediatelypriorto the currentmoment;in the case of the RPC, the currentpresenceof a resultant state; in the case of the existentialPrP, the currentreplicabilityof an event. These three varietiesof currentrelevanceare regardedas distinct perfect readings,describedby the logical representations(36), (39) and (46). In this paper, I have proposed that these readings are manifestedas distinctgrammaticalconstructions.Evidencefor thisproposalis providedby grammaticalrestrictionsunique to expressions encoding the resultative reading(39).TheRPC,as envisioned,willserveas thelocusof allidiosyncratic restrictionsdiscussedhere: those pertainingto pragmaticpresuppositionof E, temporalspecificationof E, 'pluralization'of E and mannermodification of E. I have suggested that constraints related to temporal-adverb modification and pragmatic presupposition can be attributed to the discourse-pragmatic oppositionbetweenRPC and preterite- an opposition which involves the feature [? anaphoric]. I have also argued that all constrainsdescribedherecan be regardedas instancesof a generalrestriction wherebythe RPC fails to unify with constructionswhich imbue the event denotedby the VP complementwith an unduedegreeof saliencevis-a-visits currentlyaccessibleconsequences. The availabilityof discourse-pragmatic andsemanticmodesof explanation for the observedconstraintsshouldnot, however,be taken as evidencethat these constraintsare predictableor 'follow from' the semanticsa priori. Thereis no necessaryincompatibilitybetweenassertingthe currentexistence of a resultantstate and elaboratingupon circumstancessurroundingthe causal event. Furthermore,as we saw in section 5.I, the RPC does not necessarily bar anaphoric past-time reference; there are exceptions to constraint(53). I presumethereforethat the constraintsdiscussedin section 5 must simply be learnedalong with the RPC. This is not to say that the speaker fails to recognize that these constraints have a semantic and discourse-pragmaticbasis. In all likelihood, however, this recognition representsafter-the-factinferencing,in the sense of Goldberg (1992b): abductivereasoningappliedby speakersattemptingto 'make sense' of the inputforms.The availabilityof a semanticmotivation,as well as knowledge of a systemof discourse-functional oppositions,will serveas a mnemonicaid for the learnerattemptingto masterthose constraintsupon grammarand usage associatedwith the RPC. This case studysuggeststhat thereis no necessarydistinctionbetweenthe so-called literal meaning of a constructand the conditions governingthe felicitioususe of that form in discourse.Here, we have examinedtwo such conditionsassociatedwith the RPC. First, the RPC evokes a form of pasttime referencethat is nonanaphoric.One manifestationof this restrictionis the constraintstatedin (53): the RPC cannot be used to 'elaborateupon' a pragmaticallypresupposedevent proposition.Second,the RPC is typically '54


THE ENGLISH

PRESENT PERFECT

used only in thosecontextsin whichthe resultantstatedenotedis relevantfor the joint determinationof immediategoals. An appropriateaccountingof theseand otheruse conditionsassociatedwith the RPCrequiresan approach in which,as Heny puts it (I982: 154), 'pragmaticconsiderationscan interact freely with the semantics'- that is, a conception of grammarin which a grammaticalconstructionis a complexof syntactic,semanticand pragmatic features. The generaltheoreticalframeworkwhichI have invokedhereis in certain respectscomparableto that describedby Hornstein(I990). Like Hornstein, I have presumedthat the interpretationof some set of temporalexpressions is mediatedby grammar,wherethe autonomousnatureof that grammatical componentarisesfrom the fact that 'its ... principlesof organizationare not definablein termsof semanticnotionsrelatingto the temporalinterpretation of a sentence'(HornsteinI990: 5).14 I differ from Hornstein, however, not only in invoking a broader conception of this grammar,in which semantic and discourse-pragmatic constraintsare 'grammatical'featureson a par with formalconstrains,but also in invokinga nonuniversalincarnationof that grammar,whoseminimal symbolic units are language-specifictemporal-aspectualconstructions (Michaelis I993). With respectto the PrP in particular,my pursuitof the constructionalapproach does not representa significantdeparturefrom Hornstein'sview,sincehe himselfadmits(p. I I4) that 'not everyidiosyncrasy of the present perfect should follow from universalgrammar'and that '[p]eculiaritiesthat can be determinedfrom the behavior of the present perfect in well-formed simple sentences need not be accounted for by principlesof UG'. In the presentstudy, this constellationof idiosyncratic featuresattachesto a pragmaticallyspecializedunit of meaningand form, the RPC. I see such language-particularconstructions not as mere 'taxonomic artifacts'- without relevancefor the descriptionof linguistic competence (see Chomsky 1992) - but as grammaticalunits whose acquisition is intrinsic to the learner's mastery of a system of temporal reference. The present approach, which targets both (a) construction-specific constraintson grammar,interpretationand use and (b) 'ecologicallybased' use conditions, provides for both broad-basedand fine-grainedaspectual analysis.This approachenablesthe analyst to acknowledgethat aspectual meaning is expressedby means of highly idiosyncraticforms, while not neglectinga 'macrocosmic'approach,in whichthe meaningsand functions

[14]

The formal system which Hornstein describes as the 'syntax of tense' does not represent the formal realization of the tense operators themselves. The primitives at issue are the R-, E- and S-points of Reichenbach's system of tense representation, and the 'syntactic constraints' advanced pertain to restrictions upon the linear ordering of these points, e.g. the manner in which the linear representations of main and subordinate clauses 'line up'.

I55


LAURA A. MICHAELIS

of aspectual construction are determinedwithin a system of universal semanticcontrastsandlanguage-particular discourse-functional oppositions. REFERENCES Anderson,L. (I 98 I). The'perfect'as a universalandas a language-specific category.In Hopper, P. (ed.) Tenseaspect.Amsterdam:Benjamins.227-264. Bach, E. (I986). The algebraof events.Linguistics and Philosophy 9. 5-I6. Bauer,G. (1970). The English'perfect'reconsidered.Journal of Linguistics 6. I89-I99. Binnick,R. (i99i). Time and the verb. Oxford:OxfordUniversityPress. Brinton, L. (I988). The development of English aspectual systems. Cambridge:Cambridge UniversityPress. Chafe,W. (I987). Cognitiveconstraintson informationflow.In Tomlin,R. (ed.) Coherence and groundingin discourse.Amsterdam:Benjamins.2I-52. Chomsky,N. (1992). A minimalistprogramfor linguistictheory.MIT OccasionalPapersin Linguistics I.

Clark, E. (I987). The principle of contrast: a constraint on language acquisition. In MacWhinney, B. (ed.) Mechanisms of language acquisition. Hillsdale, NJ: Lawrence Erlbaum

Associates.1-33. Cooper, R. (I986). Tense and discourse location in situation semantics.Linguisticsand Philosophy 9: I7-36.

Comrie,B. (1976). Aspect.Cambridge:CambridgeUniversityPress. semanticsor pragmatics?Linguistics and Philosophy 9. 37-6I. Dinsmore,J. (i98i). Tensechoiceand time specificationin English.Linguistics19. 474-494. Dinsmore, J. (I99I). Partitioned representations: a study in mental representation.Dordrecht:

Kluwer. Dowty, D. (I979). Word meaning and Montague grammar. Dordrecht:Reidel. Dowty, D. (I986). The effects of aspectualclass on the temporalstructureof discourse: Ernst,T. (I987). Whymannermodificationis exceptional.In Aske, J., Beery,N., Michaelis,L. & Filip, H. (eds.) Proceedings of the ThirteenthAnnual Meeting of the Berkeley Linguistics

Society.Berkeley:BerkeleyLinguisticsSociety,Inc. 77-87. Fenn, P. (1987). A semantic andpragmatic examination of the Englishperfect. Tuibingen:Gunter

Narr Verlag. Fillmore,C., Kay, P. & O'Connor,M. C. (I988). Regularityand idiomaticityin grammatical constrictions:the case of Let alone.Language64. 63-76. Goldberg,A. (I992a). On the inherentsemanticsof argumentstructure:the case of the English ditransitive. Cognitive Linguistics 3. 37-74.

Goldberg, A. (I992b). Argument structure constructions. PhD dissertation,University of California,Berkeley. Heny, F. (I982). Tense,aspectand time adverbials.II. Linguistics and Philosophy 5. 109-154. Herweg,M. (I99I). Perfectiveand imperfectiveaspect and the theory of events and states. Linguistics 29. 969-IOIO.

Hinrichs,E. (I986). Temporalanaphorain discoursesof English.LinguisticsandPhilosophy9. 63-82. Hornstein, N. (I990). As time goes by: tense and Universal Grammar. Cambridge, MA: MIT

Press. Inoue, K. (1979). An analysisof the Englishpresentperfect.Linguistics I7. 561-589. Jackendoff,R. (I990). Semantic structures. Cambridge,MA: MIT Press. Kamp,H. & Rohrer,C. (I983). Tensein texts.In Bauerle,R., Egli,U. & von Stechow,A. (eds.) Meaning, use and interpretationof language. Berlin: Walter de Gruyter. 250-269.

Klein, W. (I992). The present-perfect puzzle.Language68. 525-552. Lambrecht,K. (I 99I). Sentential-focus structuresas grammatical constructions.Paperpresented at the I99I WinterMeetingof the LSA. Lambrecht,K. (1992). Thepost-focalcomme-Nconstructionin spokenFrench.Ms., University of Texas,Austin. Lambrecht,K. (1994). Information structure and sentence form. Cambridge:Cambridge UniversityPress(in press). 156


THE ENGLISH

PRESENT PERFECT

Levinson, S. (I983). Pragmatics. Cambridge: Cambridge University Press. Li, C., Thompson, S. & Thompson, R. M. (I982). The discourse motivation for perfect aspect: the Mandarin particle le. In Hopper, P. (ed.) Tense aspect. Amsterdam: Benjamins. 19-44. McCawley, J. (I97i). Tense and time reference in English. In Fillmore, C. & Langendoen, D. T. (eds.) Studies in linguistic semantics. New York: Holt Rinehart. 97-II3. McCawley, J. (I98I). Notes on the English perfect. Australian Journal of Linguistics I. 8 -9o. McCoard, R. (1978). The English perfect: tense choice and pragmatic inferences. Amsterdam: North-Holland Publishing Company. Toward a grammar of aspect: the case of the English perfect construction. Michaelis, L. (I993). PhD dissertation, University of California, Berkeley. Mittwoch, A. (I988). Aspects of English aspect: on the interaction of perfect, progressive and durational phrases. Linguistics and Philosophy II. 203-254. Norvig, P. (I988). Interpretation under ambiguity. In Hall, K. (ed.) Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, Inc. I88-201. Parsons, T. (i99o). Events in the semantics of English. Cambridge, MA: MIT Press. Partee, B. (I984). Nominal and temporal anaphora. Linguistics and Philosophy 7. 243-286. Prior, A. (I967). Past, present andfuture. Oxford: Oxford University Press. Reichenbach, H. (I947). Elements of symbolic logic. New York: Macmillan. Richards, B. (I982). Tense, aspect and time adverbials. I. Linguistics and Philosophy 5. 59-I07. Slobin, D. (I990). Talking perfectly: discourse-pragmatic origins of the present perfect. Ms., University of California, Berkeley. Talmy, L. (I988). Cognition and grammar. In Rudzka-Ostyn, B. (ed.) Topics in cognitive grammar. Amsterdam: Benjamins. I65-205. Visser, F. (I966). An historical syntax of the English language (part 2). Leiden: E. J. Brill. Zwicky, A. & Sadock, J. (I975). Ambiguity tests and how to fail them. In Kimball, P. (ed.) Syntax and semantics, Vol. 4. New York: Academic Press. I-36. Author's address: Department of Linguistics, University of Colorado, Boulder, Colorado 80309, USA. E-mail: [email protected]

157
