Modification, event structure and the word/phrase asymmetry

Modification, Event Structure and the Word/Phrase Asymmetry' Anna-Maria Di Sciullo and Carol L. Tenny Universite du Quebec a MontrM and MIT 1. The Basic Idea We posit that the following difference obtains between morphological (wordinternal) structure and syntactic (phrasal) XP structure: Whereas adjuncts are permitted in both morphological and syntactic structure, complements are only permitted in syntactic structure: (I) xo/xp configurationaI asymmetry The head-complement configuration is not visible, viz., interpretable at the 9-conceptual interface. It is at the XF'-conceptual interface. @i Sciullo 1997) We hrther posit that only DP complements of the matrix verb can have the aspectual hnction of measuring out the event: (2) Measuring-Out Constraint on Direct Internal Argument Direct internal arguments are the only overt arguments which can measure out the event. (Tenny 1994) These two ideas predict certain prohibitions on the kinds of modification that can be introduced by morphological elements. ' This study was supported in part by the Social Sciences and Humanities Research council of Canada graat411-924012 andbytheNational ScienceFoundationgramSBR-9616591. 8 1998 by Anna-Maria Di Scidlo and Carol L. Tenny Pius N.Tamanji and kiyomi Kusumoto (eds.), NELS 28,375-389 Anna-Maria Di Sciullo and Carol L. Tenny 376 2. Spelling Out Some Assumptions 2.1 Event Structure Research in syntax and semantics in recent years has made it clear that natural language grammars do not treat the events that verbs denote as unanalyzable wholes. The meaning of a verb can be analyzed into a structured representation of the event the verb designates. Causation, inchoativity, and stativity are three basic grammatical components of complex events that have been identified in the literature. These three hndamental components participate in structuring complex events into an inner and an outer event, where an inner event is associated with inchoativity and stativity, and an outer event is associated with a causer or a causing event. We assume that these elements of event structure can be represented in some way, and we follow various approaches in the literature, while acknowledging that there are open questions about how this information should be represented. Some representations of event structure from the literature are shown in (3)-(5): He sweeps the floor clean: [ [ He sweeps thefloor ] CAUSE [BECOME [thefloor is clean] ] ] (Dowty 1979,p. 93,example 105.) x wipe the floor clean: x CAUSE [ y BECOME (AT) z] BY [x 'wipe' y] ] Levin and Rapoport 1988, p.2, example 2a. [x CAUSE [ floor BECOME (AT) clean ] BY [x 'wipe'floor] 3 x closes the door: (3e) [ Cul(e) & Agent(e,x) & (3e')[Cul(e') & Theme(e',door) & CAUSE(e,e') & (3s) [Being-closed(s) & Theme(s,door) & Hold@ & BECOMX(e',s)]]]. (Parsons 1990,p. 120, no example number). We assume that representations such as (3)-(5) represent information in Conceptual Structure, of which only a subpart is relevant for syntax. 2.2 Complements and Adjuncts in Syntax Complement and adjuncts are two basic grammatical structures that have been identified in the literature. These two sort of structures participate in the derivation of the linguistic expressions. These structure have been claimed to participate in the derivation of phrases as well as in the derivation of words. For clarity, we assume that these structures are defined as in (6)and (7), following Chomsky (1995)and Collins (1996). In particular, we assume that no new category is Md$cation, Event Structure and the WordPhraseAsymmetry 377 created by adjunction. (6) A is the complement of B, ifF A and B are daughter of a new categoq B. B I\ B A (7) A is an adjunct of B, iff A and a segment of B are daughters. B I\ A B We take such structure to support aspectual information in the course of the derivation of the linguistic expressions and to be visible, viz., interpreted at the conceptual interface. 2.3 2.4 Some Assumptions About the Relation Between Event Structure and Syntax A. We do not assume a perfect one-to-one correspondence between event structure and syntax, but we assume that a certain transparency obtains between predicatdargument relations in event structure and in syntax. We assume the basic predicatdargument relations in event structure are reflected in syntax. B. We assume that scope-taking elements in event structure are adjuncts in morphosyntax. They are not DP complements in syntax. C. We assume that complements in syntax must be arguments in event structure. D. We assume that predicates in event structure are not DPs in morpho-syntax. Organization of the Paper In this paper we examine three types of adverbial modification that can interact with event structure, making a principled distinction between types of event structure modification that can be instanciated in verbal morphology and types of modification that cannot. We compare these with types of adverbial modification that can occur fkeely in syntax, and we argue that the differences between word internal and phrasal modification follows from a basic asymmetry between morphological and syntactic structure. We compare measure modification, which can only occur in syntax, with iterativdrepetitive and bounding modification, which can o w either in syntax or morphology. Finally, we consider the implications of this distinction for the interhx of morphosyntax with the Conceptual System. Anna-Maria Di SciulIo and Carol L.Tenny 378 3. Iterative Modifcation 3.1 WbatItIs Iterative modification indicates that the went represented by the verb, or some part of it, is to be understood as having multiple occurrences. We use the term iterative here in a non-technical sense, as a cover term for a range of such phenomena. Iterative modification may be accomplished through the syntax as well as through verbal morphology. In English, it is supported by adverbial phrases such as again and verbal prefixes such as re-. (8) John reopened the door. John opened the door again. The conceptual structure of expressions including iterative modification can be represented as in (9) where the iterative operator again ranges over the predicatdargument structure: rvm,”(BECOME [^(open (d))1)1)1)1 (9) PAST (^[again ( ^ W (CAUSE w o n Stechow 1995) 3.2 Iterative Modifiers Are Not Complements Iterative modifiers do not contribute to argument satisfaction. Sice complements must be arguments, iterative modifiers are not complements. (10) a. b. c. d. *John smashed again. / *John resmashed. John smashed the vase (again). John dresses (hkily). Jones worded the letter (sloppily). Iterative modifiers are scope-taking elements. Arguments are not scope-taking. Since complements must be arguments, iterative modifiers are not complements. 3.3 Iterative Modification in Syntax and Morphology Iterative modification is observed in several languages in syntax and in morphobgy, as illustrated below with examples from Germanic, Romance and Slavic I Moalification, Event Structure and the WorMhrase Ammetry 379 languages. English (13) a. John rewired the house. b. John Wired the house again. Italian (14) a. Gianni ha riaperto la finestra. 'Gianni reopened the window.' b. Gianni ha aperto la iinestra di nuovo. 'Gianniopened the window again.' Polish (15 ) a. Odmalowalismy sciany domu na bialo. 'We repainted the walls of the house white.' b. Malowalismy sciany domu na bialo mow. 'We painted the walls of the house white again.' The meaning of iterative modification is induced through morphology as well as through syntax. Romance and Slavic, like Germanic offer the option of either prefixes or adverbial phrases. 3.4 Iterative Prefues Iterative modification does not require complementation in XP structure. It also does not require complementation in XO structure, as the iterative prefix is adjoined outside of a complete event projection, as depicted in (16) for change of state and change of place verbs. Anna-Maria Di Sciullo and Carol L.Tenny a-/en- z This is motivated by the fact that the iterative prefix precedes internal spatial prefixes. As noted in Di Sciullo and a p p l e (1994), in French verbal morphology and in Di Sciullo (1997) for Italian, the iterative prefix must precede other prefixes, including spatial prefixes, such as a- and g&-. This hear order restriction amongst prefixes also holds for the languages considered here, including English. (17) a. 11 a r&mbeW*emrebelli la pike. He reembellished/*emrebeUishedthe room. b. 11a rhbouteiUd*emrebouteiUe le vin He reembottled/*emrebottledthe wine. Furthermore, the iterative prefix generally combines with accomplishments, that is with complete events, as illustrated here again with French and English: (18) a. b. c. d. ?11 resait I'italien. ?He reknows Italian. ?I1 est rearrive en retard./?He rearrived late. ?I1 a reconduit une auto./*He redrove a car. II a reconstruit une structure./He reconstructed a structure. (state) (achievement) (activity) (accomplishment) @i Sciullo, 199759) The referent of a verbal projection including an iterative prefix will be interpreted as having the same extension as in the unprefixed verbal projection. This hrther indicates that the iterative prefix may not affect the internal constituency of an event, and thus is not dependent on complementation. (19) a. John rebdt a house. (same house) b. John built a house again. (same or different house) Thus, iterative modification is available in both words and phrases, as it does not crucially require complementation. The differences in interpretation between verbal projection including iterative prefixes on the one hand and iterative adverbials on the other Modification, Event Struciure and the WordPhrae Asymmetry 38 1 follow from the relative scope of the modification. While iterative prefixes may only modify complete precedent events, iterative adverbials may also affect parts of events. 4. Bounding Modification 4.1 What It Is Whereas iterative modification involves some kind of scope-taking over parts of the event structure, bounding modification involves adding an element - namely a temporal terminus associated with a final state - to the event structure. This may be accomplished in either of two ways: by adding a temporal endpoint (20)-(21), or by adding an endstate predicate (22)-(23). Adding a temporal e+int: (20) [x 'drive' car ] ---> [x CAUSE [car BECOME (AT) New York]] (21) a. Miguel drove the car for an hour/ *in an hour. b. Miguel drove the car to New York *for an hour/ in an hour Adding an e d t a t e prealcate: --> [x CAUSE [ floor BECOME clean ] BY [x 'wipe' floor]] (22) [x 'wipe' floor] (23) a. Ari wiped the floor for an hour/ *in an hour. b. Ari wiped the floor clean *for an hour/ in an hour. 4.2 Bounding By Adding A Temporal Endpoint 4.2.1 A Temporal Endpoint Is Not A Complement Of The Matrix Verb A temporal endpoint is an argument of an embedded (prepositional) predicate, rather than an argument of the matrix verb (2.3.A). We see this illustrated in the representation in (2), where the temporal endpoint is an argument of the prepositional element, rather than the matrix verb. 4.2.2 Bounding By Adding A Temporal Endpoint, In Syntax This can be done by adding a goal PP to the VP: English (24) a. Miguel drove the car for an hour/ *in an hour. b. Miguel drove the car to New York *for an hour/ in an hour. Anna-Maria Di Sciullo and Carol L. Tenny 382 Dutch (25) a. Anneke heeft jaredang I *binnen een week bij de opera gedanst. A has years-long I within a week with the opera danced. "Anneke danced with the opera for years I *within a week." b. Anneke was bmen een minuut I *urenlang van het podium gedanst. A. was within one minute I hours-long of the stage off danced. "Anneke danced off the stage within one minute I *for hours." (Van Hout 1996, p. 85, # 4) 4.2.3 Bounding by Adding a Temporal Endpoint, in Morphology This can be done by adding a spatial aflix to the verb, in Romance, as discussed in Di Sciullo (1997), as well as in the Slavic and the Germanic languages, as illustrated below. French (26) a. Rose a porte les fleurs *en cinq minuted pendant cinq minutes. 'Rosecarried the Bowers *in five minuted for five minutes.' b. Rose a apporte les 5eurs en cinq minuted*pendant cinq minutes. 'Rose brought the Bowers in five minutes/*for five minutes.' Italian (27) a. Julio ha corso *in cinque minutVper cinque minuti. 'Julio ran *in five minutedfor five minutes.' b. Julio e accorso in cinque minuti/*per cinque minuti 'Julio 5ed in five minutes/*for five minutes.' Polish (28) a. x przeglosowac swoj plan za piec minut/*w pieciu minutach 'x outvote their plan in five minuted*for five minutes.' b. x niedocenilismyjego zdolnosci za piec minut/*w pieciu minutach 'x underestimates his capacities in five minuted*for five minutes.' English (29) (30) a. Rebecca ran for ten minuted *in ten minutes, to exercise her weak leg. b. Rebecca outran Luke *for five minuted in five minutes, in the last race. a. Karen stayed with her cousins for three dayd *in three days. b. Karen overstayed her welcome at her cousins *for three dayd in three days. Modification, Event Siructure and the WoraPhrase Asymmetry 383 (3 1) a. Mihaela cooked her Christmas turkey for four hours/ in four hours. b. h4ihaela undercooked her Christmas turkey *for only four hours/ in only four hours. 4.3 Bounding by Adding an Endstate Predicate 4.3.1 An Endstate Predicate Is Not a DP Complement An endstate predicate is a predicate, not an argument. Therefore it is not a DP complement (2.3.C), as we see illustrated in (22). 4.3.2 Bounding by Adding an Endstate Predicate, in Syntax This can be done by adding a secondary predicate to the VP: EngIish (32) a. Ari wiped the floor for an hour/ *in an hour. b. Ari wiped the floor clean *for an hour/ in an hour. 4.3.3 Bounding by Adding an Endstate Predicate, in Morphology This can be done by adding an endstate predicate to the verb: Dutch (33) De diktator heeft de gevangene urenlangj *binnen een uur gemarteld. the dictator has the prisoner hours-long / within one hour tortured "The dictator tortured the prisoner for hours / *within one hour." (34) De diktator heeft de gevangene binnen een uur / *urenlang doodgemarteld the dictator has the prisoner within an hour / hours-long dead-tortured "The dictator tortured the prisoner to death in an hour / *for hours." (Van Hout 1996; p. 103, #31) 5. Measure Modification 5.1 A Measure Argument The semantic effect of bounding can also be introduced syntactically through a DP complement. If the DP that is the direct argument of the verb receives a MEASURE aspectual role, it provides a scale or path along which the event is measured out over time, so that the change of state or location it undergoes during the event marks the temporal terminus of the event (Tenny 1994). This path or measure is associated only with 384 Anna-Maria Di Sciullo and Carol L. Tenny particular verbs: verbs involving changes of state, incremental themes, or motion to a god. (35) Mihaela consumed her Christmas turkey *for four hours/ in four hours. We can see the role the complement DP is playing in bounding the event by substituting a mass noun for the direct argument of the verb. It is well-known that the bounding effect is lost under these conditions (Dowty 1979, Verkuyl 1993 and 1972, Hinrichs 1985). (36) Mihaela consumed beer for four hours/ *in four hours. The bounding of the event that is effected through the complement of the verb in syntactic structure cannot be paralleled by morphological means. Morphological compounds, like those below, although they involve the apparent composing of a verb with its complement, do not have the bounding effect of complements in the syntax: (37) Stew-mixing for hours/ *in an hour is hn. (compare Jill mixed the stew for hours/ in an hour.) Door-closing for hours/ *in five seconds is a bad idea. (compare Jack closed the door for hours/ in five seconds.) Glass-fihg for a minutel *in a minute is a silly way to make a living. (compare Rachel filled the glass *for a minutel in a minute.) 5.1 Measure Modification: What It Is Measure modification is modification of the scale or path along which the event is measured out over time (Tenny 1994). This type of modification differs from iterative and bounding modification in modifying the path or measure which is composed through the interaction of the verb and its DP complement together. Measure modification is modification of a temporal terminus already existing in the event structure, by means of modwng the path the event travels towards its terminus. (38) Tiggy closed the door partway. (39) (3e) [ Cul(e) & Agent(e,x) & (3e')[Cul(e') & Theme(e',door) & CAUSE(e,e') 62 (3s) [Being-partway(closed)(s) & Theme(s,door) Br Hold@)& BECOhE(e',s)]l]. (Parsons1990, p. 122.) 5.2 Measure Modifiers in Syntax Measuring modification is instantiated by a class of degree adverbs in English, which may be preverbal (40) or postverbal (41): Modijication, Event Structure and the WordPhrmeAsymmeOy (40) Tiggy partly closed the door. Diane halffilled the glass. The cook completely mixed the stew. (41) Tiggy closed the door partway. Diane filled the glass halfway. The cook mixed the stew completely. 5.3 Measure Modifiers in Morphology 385 Measure modifiers are prohibited in morphology. In this section we examine prefixes which should be potential measure modifiers, as we show thay do not yield a measuring interpretation. We also example deverbal compounds and show that the measure element of meaning present in the verb is lost in the deverbal compound. 5.3.1 Prefues Prefixes which are andidates for a measuring interpretation do not in fact provide that interpretation. We consider English -under. Measure adverbs apply only to the specific set of verbs (and verbal expressions) that have a measure or path in their meaning (Tenny 1997). These include close the door and walk to New York (43); they do not include appreciate her professor (44). (43) a. Jill closed the door parhvay. b. James walked halfway to New York. (44) *Jill appreciated her professor halfway. English -&does not co-occur with close the door and walk to New York (45). It does, however, co-occur with appreciate her professor (46). (45) a. *Jill underclosed the door. b. *James underwalked to New York. (46) Jill underappreciated her professor. Similar co-occurrence restrictions for French sow-: (47) a. James a ferme la fenetre a moiti6. 'James closed the window partway.' b. *James a sousferme la fenare. 'James underclosed the window.' (48) a. James a marche a moitie jusqu'a Paris. 'James walked halfway to Paris.' Anna-Maria Di Sciullo and Carol L. Tenny 386 I b. *James a sousmarchejusqu'a Paris. 'James underwalked to Paris (49) I a. *James a esthe Dr. No. a moitie 'James estimated Dr. No. halfway.' b. James a sousesthe Dr. No. 'James underestimated Dr. No. 5.3.2 Deverbal Compounds Deverbal compounds cannot be modified by measure adverbs. This shows that the measure element of meaning present in the verb is no longer present in the deverbal compound. *Stew-mixing completely is fun. (compare Stew-mixing sloppily is fun.) *Door-closingpartway is a bad idea. (compare Door-closing silently is the rule in this monastery.) *Glass-filling halfway is a hard way to make a living. (compare Glass-filling noisily goes on here every night behind the bar.) Conclusion. Further Discussion Overview The chart below summarizes the facts presented in this paper. Function syntax Morphology Iterative modification Yes Yes Bounding by adding a temporal endpoint Yes Yes Bounding by adding an endstate predicate Yes Yes Bounding through a measuring argument Yes no Measure modification Yes no Morphology Given Di Sciullo and Williams (1987) definition of the head of the word and Kayne's (1994) Linear Corresponden& Axiom the fact that measuring modification requires a complement and that such modification cannot be obtained within the word Modification, Event Structure and the WordPhrase Aynmelry 387 brings hrther support to the view that complements cannot be licensed within the word. In fact, aspectual modification in derivational morphology is only possible to the left: (54) a. b. c. d. to re-overprotect to over-overprotect *to reprotect-over *to overprotect-over 6.3 Interface with tbe Conceptual System It is the nature of the bare output conditions, based on interpretation under asymmetry, that restricts the options, rather than the presence of hnctional categories, such as AGRo, in the derivations (Roeper, 1987; Borer, 1995). The view that aspect is l i e d to AGR-o (Roeper, 1987) does not cover the range of aspectual modification types, including iterative, bounding and measuring modification, as they are supported by other sorts of categories than objects. In fact, adverbial categories as well as resultatives, either adjectival or prepositional play a role is aspectual modification, in addition to objects. AGR-o does not subsume this set of categories, which are not equally specified for case features. AGR [Event-marker] I \ (55) AGR-o I VP I\ [+case V DP +event I I \ +telic] D NF' eat theapples <Likewise, ASP nodes are also partial descriptions for aspectual modification. They fail to predict the basic wordphrase asymmetry with respect to aspectual modification, as our configurationaltheory predicts. Iv (56) I \ ASPP [+event] I \ V -ion Spec ASP' I\ destruc ASP VP I \ V DP N I\ <- I the city Anna-Maria Di Sciullo and Carol L. Tenny 388 In our theory the bare output conditions impose a strong requirement on the form of words and phrases. The two sorts of grammatical objects must be codgurationdy distinct at the interface in order to be properly interpreted by the performance systems. One part of the distinctiveness lies on the visibility of the head-complement configuration in phrases and its non-visibility in words. This asymmetry makes the correct predictions with respect of aspectual modification. In turn it brings into question the licensing of the head-complement configuration in syntactic words such as deverbal compounds, as the putative complement, does not qualie as a syntactic complement with respect to the standard tests. (57) Ce vin est un vrai coupe-la-soif 'This wine is a real thirst-quencher.' b. *La soif a 6te coupee par ce vin. 'Thirst was cut by this Wine.' c. *Ce [coupe-Ia-soS] est efficace lorsqu'ellei est persistante. 'This thirst-quencher is good when it is strong.' a. The difference in style of semantic composition between the iterative and bounding modification on the one hand and the measuring modification on the other hand represents a crucial divide in the grammar, with implications for the morphology-syntax conceptual interface. References Borer, H. 1966. Passive without theta-grids. Ms.W S S . Chomsky, N. 1995. The Minimalist Program. Cambridge, MA.,: MIT Press. Collins, C. 1997. Local Economy. Cambridge, MA.,: MIT Press. Di Sciullo, A.M. 1996. Modularity and X%P Asymmetry. Linguistic Analysis, 26.1, pp. 3-28. Di Sciullo, A.M. 1997. Prefixed Verbs and Adjunct Identification. In A.M. Di Sciullo (ed.) Projections and Interface Conditions. New York Oxford University Press. pp. 5274. Di Sciullo, A.M. & E. Klipple, 1994. Modifying Affixes. WECOL 6. University of Seattle at Washington. Di Sciullo, A.M. & E. Williams, 1987. On the Definition of Word. Cambridge, MA.,: MIT Press. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge, MA.,: MIT Press. Hinrichs, E. 1985. A Compositional Semantics for Aktionsarten and NP Reference in English, Ph.D. dissertation, Ohio State University. Levin, B. & T. Rapoport, 1988. Lexical Subordination. Paper read at the Chicago Linguistic Society. Parsons, T. 1990. Events in the Semantics of English: A Study of Subatomic Semantics. Cambridge, Mass.: MIT Press. Roeper, T. 1997. Events and non-events in derivational morphology. m a s s . Modijkation, Event Structure and the WoraThrme Asymmeby 389 Tenny, C. 1977. Core events and adverbial modification.Paper presented at Workshop on Events and Grammatical Objects, LSA Summer Institute, Cornell University. Tenny, C. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht: Kluwer Academic Publishers. Tenny, C. and F.Heny 1993. Core event structure and the scope of aspect. Ms. University of Pittsburgh. Van Hout, A. 1996. Event semantics of verb kame alternations. A case study of Dutch and its acquisition. Tilburg Dissertation in Language Series. Verkuyl, H. 1972. On the Compositional Nature of Aspect. Dordrecht: Kluwer Academic Publishers. Verkuyl, H. 1993. A theory of aspectuality. Cambridge: Cambridge University Press. Von Stechow, A. 1995. Lexical decomposition in syntax. In Lexical knowledge in the organization of grammar. In U. Egli, P. Pause, C. Schwarze, A. Von Stechow and G. Wienold (eds). Amsterdam: John Benjamins. Departement de Linguistique Universite du Quebec li Montrd Case Postale 8888, succursale Centre Ville Montrd (Quebec) Canada H3C 3P8 di [email protected] Department of Linguistics Massachusetts Institute of Technology E39-245 Cambridge, MA., 02139 [email protected]