Re- prefixation and Talmy’s Parameter

Re- prefixation and Talmy’s parameter Oana Săvescu Ciucivara and Jim Wood New York University 1. Introduction Talmy (1985) proposed that English and the Romance languages differ in that the latter generally express "path" in verb roots, whereas the former expresses "path" in satellite morphemes such as prepositions. English put is a suppletive form expressing a “nondirectional ‘putting’ notion,” its directional component being “determined completely by the particular Path particle and/or preposition present” (Talmy 1985:71). By contrast, Romance languages generally incorporate a directional notion directly into put verbs, but are unable to encode “manner” and direction at the same time. While the Romanian (1) is unambiguously locative, its English counterpart can be interpreted as directional. (1) Sticla va pluti sub pod. bottle.the will float under bridge ‘The bottle will float AT/*TO a location under the bridge.’ In this paper, we present novel data from Romanian re- prefixation, arguing that it provides support for a structural difference between certain English and Romanian constructions with put and related verbs: (2) A repus radioul în funcţiune. has reput radio.the into function ‘He put the radio into use again.’ (3) Am (*re)pus cartea pe raft. have reput book.the on shelf ‘I put the book back on the shelf.’ ∗ Unless otherwise noted, all the Romanian data come from the first author’s native speaker intuitions in consultation with other speakers, and the Icelandic data come from the second author’s fieldwork with Icelandic speakers in New York City. Thanks to our informants for their time and energy. In addition, we would like to thank Alec Marantz and the participants of NELS 41 and LSA 2011 for helpful comments. Part of this work was partially supported by the strategic grant POSDRU/89/1.5/S/62259, Project “Applied social, human and political sciences” co-financed by the European Social Fund within the Sectorial Operational Program Human Resources Development 2007-2013, obtained by the first author. © 2010 by Oana Săvescu Ciucivara and Jim Wood Emily Elfner and Martin Walkow (eds.), NELS 36: 2–13 Re- prefixation and Talmy’s parameter (4) He (*re)put the radio into use. (5) He (*re)put the book on the shelf. We show that the contrasts between Romanian (2)-(3), on the one hand, and between Romanian (2) and English (4) necessitate a syntactic, non-lexical approach to both re- prefixation and to Talmy’s distinction between Romance and English "path" incorporation. 2. Re- Prefixation in English/Icelandic vs. French/Italian Romance languages such as French and Italian have been shown to have a re- morpheme whose properties differ from English re- and its Icelandic counterpart endur- in certain dramatic ways.1 This will show that the similarities between English and Romanian recannot be a matter of historical accident of a cognate morpheme: Icelandic endur- is not a cognate and behaves like English/Romanian, and French/Italian re-/ri- are cognates but behave differently. First, English re- and Icelandic endur- cannot attach to psych-predicates (6-7), while Italian ri- can (8) (Cardinaletti 2003:8). (6) * This reamazes/reshocks me every time. (7) * Þessi mynd hefur endurhneykslað marga. this movie.NOM has reshocked many.ACC (8) Questo mi ristupisce ogni volta. this me re-amazes each time ‘This amazes me again each time.’ Second, in English and Icelandic, re-/endur- cannot attach to auxiliaries (9-10), whereas this is possible in French (11). (9) a. b. * John has rebeen seen. * John would rebe right. (10) a. * Myndin hefur endurverið sýnd. film.the has rebeen shown * María myndi endurhafa rétt fyrir sér. Mary would rehave right for REFL . DAT b. (11) a. b. Jean a re été vu. Jean has re been seen Marie aurait re eu raison. Marie would re have reason ‘Marie would have been right again.’ Third, in English and Icelandic, ‘put’-type verbs, which take two obligatory internal arguments, do not allow re-/endur- (12-13), whereas Italian ‘put’-type verbs do (14). 1 In this paper, all French data on re- comes from Sportiche (2008), and all Italian data on ri- comes from Cardinaletti (2003). Oana Săvescu Ciucivara and Jim Wood (12) * John will reput the book there. (13) * María ætlar að endurláta bókina þar. Mary intends to reput book.the there (14) Gianni ha rimesso la palla sullo scaffale. Gianni has reput the ball on.the shelf This is related to a more general constraint on re-/endur-, namely that it appears to obey the “Sole Complement Generalization” (Levin and Rappaport 1986; Marantz 2009; see also Horn 1980). In English (15) and Icelandic (16), it attaches only to verbs which take one and only one obligatory internal argument. It thus cannot attach to unergative activity verbs. Italian again seems to differ in this respect, to judge by (17), though the data in Cardinaletti (2003) does not provide an obvious minimal pair. (15) * John has resnored. (16) * Jón hefur endurhrotið. John has resnored Mi piace ritradurre. to.me pleases re.translate ‘I like to translate again (something).’ (17) The requirement is that re-/endur- attaches to a verb with an underlying direct object—it need not be a surface direct object, so unaccusatives qualify. (18) a. b. The owner has reopened the door. The door has reopened. (19) a. Eigandinn hefur enduropnað hurðina. owner.the has reopened door.the ‘The owner has reopened the door.’ Hurðin hefur enduropnast. door.the has reopened ‘The door has reopened.’ b. However, as pointed out in Marantz (2009) for English, not just any direct object will do; the direct object must undergo a change of state. This holds for Icelandic as well (21). (20) * Mary retickled me. (21) * María endurkitlaði mig. Mary retickled me In sum, English re- differs from its etymologically related counterparts in French and Italian in a number of ways, and we have touched only on a few of them. Icelandic endur-, which is not etymologically related, behaves more like English in its restrictions. The properties characterizing Icelandic/English re-/endur- are summarized in (22). Re- prefixation and Talmy’s parameter (22) Re-/Endur- requires a verb which: a. Takes an underlying direct object which undergoes a change of state. b. Does not take more than one obligatory internal argument. Moreover, unlike French and Italian, this cannot be a psychological predicate, and re-/endur- strictly attaches to the lexical verb, and not to any auxiliaries. This does not exhaust the differences across these languages, but it suffices to show that the properties of re-like morphemes, when they exhibit English-like behavior, do not come for free for being a prefix with a certain meaning, and cannot receive an etymological explanation. 3. Re- prefixation in Romanian For the most part, Romanian re- prefixation seems to behave like English re-, and not like French and Italian. It allows a direct object undergoing a change of state (23a), it disallows unergative activities (23b) and ‘put’-type verbs with two obligatory internal arguments (23c). (23) a. b. c. Am redeschis uşa. have.1 SG reopened door ‘I have reopened the door.’ * Am resforăit. have.1 SG resnored ??/* Am repus cartea pe raft. have.1 SG reput book.the on shelf INTENDED : ‘I put the book back on the shelf.’ In addition, it cannot attach to psych-verbs or auxiliaries, or to verbs for which the direct object undergoes no change of state. However, there is a coherent class of exceptions where Romanian differs from English. The sentences in (25) appear to violate the sole complement generalization. (24) a. b. I (*re)put the radio to use. I (*re)put it on sale. (25) a. Am repus radioul în funcţiune. have.1 SG reput radio.the in function ‘I made the radio work again.’ L – am repus în v“anzare. CL .3. ACC . M have.1 SG reput in sale ‘I made it available for sale again.’ b. Using make instead of put does not help the English examples either. (26) a. b. I (*re)made the radio work. I (*re)made it available for sale. Oana Săvescu Ciucivara and Jim Wood The data in (25) are apparently a counter-example to the generalization that re- cannot attach to a verb which has more than one obligatory internal argument (or, put differently, one that obligatorily takes a small clause complement). This raises at least two questions. First, why does Romanian re- attach to ‘put’ in some cases (25) but not others (23c), and second, why does English differ from Romanian in not allowing examples like (24) or (26)? We propose the following empirical generalization regarding when Romanian allows re- on verbs with two internal arguments. (27) Romanian re- is allowed on verbs with two obligatory internal arguments if: a. The verb in question is a “light” verb. b. The relation between the two internal arguments is such that one undergoes a change of state and the other specifies the nature of that state. The clause in (27b) refers to the relation between the two internal arguments. In (23c), there is a direct relation between two entities—the relation between the book and the shelf involves two entities in a locative relation, where one specifies the location of the other. In (25), on the other hand, there is no “function” or “sale” which bears a relation with respect to the other argument; rather, they specify the end point of a change of state that the other argument has undergone. That this is the important difference is suggested by the characterization of re-type morphemes above, namely the requirement that their object undergo a change of state. Further support for the hypothesis that this is the important generalization comes from a related difference with ‘give’-type verbs; the latter seem to be subject to the same generalization (and are similarly distinct from English ‘give’-type verbs in this respect). (28) a. b. He (*re)gave the girl her freedom. I (*re)gave him the book. (29) a. I-a b. redat fetei libertatea. CL -has.3 SG regiven girl. DAT freedom.the ‘He gave the girl back her freedom.’ I-am (*re)dat cartea. CL -have.1 SG (*re)given freedom.the INTENDED : ‘I gave him the book back.’ In (29b), the relation between the two internal arguments is one of possession; the book is an entity in a possessive/possessed relation with the other internal argument. In (29a), on the other hand, libertatea ‘the freedom’ is not in an entity in a possessive relation with the girl; rather, the girl undergoes a change of state (she becomes ‘free’), the end state of which is specified by the nominal libertatea. Similar behavior is exhibited by the verb ‘take out’ in Romanian. In (30b) the relation between the two arguments is one involving change of location, and re- prefixation is disallowed. In (30a), however, re- is acceptable, as the book undergoes a change of state, ‘circulation’ being the end-state of the change. English re- is impossible in both cases (31). Re- prefixation and Talmy’s parameter (30) a. b. (31) 4. a. b. Au rescos cartea lui Păunescu din circulaţie. have.3 PL retaken.out book of Păunescu from circulation ‘They have taken Păunescu’s book back out of circulation.’ Am (*re)scos cartea din ghizdan. have.1 SG (*re)taken.out book.the from backback INTENDED : ‘I took the book (back) out of the backpack.’ They have (*re)taken his book out of circulation. I have (*re)taken the book out of the backpack. Analysis Since the behavior of English and Romanian re- is otherwise so similar, and since, as we have seen, this does not come for free for being a re-like morpheme, the challenge is to make sense of how a child could acquire the subtle differences noted above with ‘put’like verbs. That is, what independent difference(s) between English and Romanian could independently yield the difference between English and Romanian re-, such that Romanian makes the distinctions in (27) where English does not seem to? We would like to propose that the relevant difference is in the functional structure of the vP, and that it is the same difference which underlies another difference between the two languages, namely, Romanian’s behavior as a “path-incorporating” language as opposed to English’s behavior as a “manner-incorporating” language. This is evidenced by the contrast between the two languages with respect to how sentences such as (1) can be interpreted.2 Most recent formal approaches to this distinction which usually involve, in one form or another, the claim that Romanian-type languages express a “path” component on the verb (Fábregas 2007; Gehrke 2008; Mateu and Rigau 2010), whereas English-type languages involve a “path” component related to extended projection of PPs (Koopman 2000, Svenonius 2003, 2007, 2008, den Dikken 2010 and other papers in Cinque and Rizzi 2010). With respect to the extended projection of PPs, Svenonius (2008:81) writes: “This is an intense area of ongoing research, but as of this moment it does seem plausible that the difference that Talmy observed between Spanish on the one hand and English on the other could be stated in terms of a functional head [‘Path’ (OS&JW)] that Spanish lacks and English has.” Following the tradition that attempts to capture cross-linguistic differences in terms of properties of functional elements rather than lexical elements (Borer 1984, Chomsky 1995:6, Collins 2005:117, and Kayne 2005:278, 2006), we propose that the difference amounts to where “path” is expressed syntactically—on a variety of p in English, and on a variety of v in Romanian. This is compatible with the view that linguistic features 2 We are aware that the difference between the two kinds of language is not completely rigid or straightforwardly applicable to all words/sentences of a language. This fact further suggests a functional difference in the vP domain, which is the “first phase” and thus susceptible to idiosyncratic differences in the relations between functional and lexical elements (Arad 2003; Marantz 2009). However, there is still a difference between the two languages, and we do not think this is just a difference of “quantity” or “frequency” of the relevant construction types; but since we cannot address all the relevant data here, we leave for future work how our approach would make finer-grained distinctions within languages. Oana Săvescu Ciucivara and Jim Wood are universal, but differ in the way they ‘bundle together’ in languages (Sigurðsson 2004, Zanuttini 2010). The Talmian difference is not the presence or lack of “path” per se, but rather in how “path” relates to other functional structure. For present purposes, we adopt the view that categorial functional heads are subtyped (cf. Folli and Harley 2007) with privative subfeatures, e.g. vCAUS , vDO , etc. (cf. Adger and Svenonius 2011). We propose that PATH can be a subfeature of v (vPATH ) or of p (pPATH ), and that this is subject to parametric variation.3 In the next subsection, we show first how re- can be ruled out with change-of-location uses of ‘put’ in both English and Romanian. In the following subsection, we show how re- can be allowed for change-of-state uses of ‘put’ in Romanian but not English. 4.1 Change of Location To address the question of why re- is disallowed with change-of-location uses of ‘put’ in both English and Romanian, we adopt the assumption that prepositions themselves are not inherently ‘locational’. Rather, they are functional items which can mediate various syntactic relations (such as the introduction of arguments into a structure). Locative uses of prepositions result from an additional morpheme representing “place” (cf. den Dikken 2010, a.o.). Here, we represent this as a flavor of p (pPLACE ), which introduces the ‘Figure’ in its specifier. It is analogous to the Voice/v head that introduces verbal external arguments.4 pPLACE mediates a locative relation between the two entities. (32) pP p� DP the radio pPLACE Pi on pPLACE PP P ti DP the shelf In English locative path expressions, pPATH merges on top of this pP, followed by little v (33a). In Romanian locative path expressions, vPATH merges (33b).5 Given these structures, we can rule out re- with change-of-location small clauses in the following way. In agreement with Marantz (2009), re- selects syntactically for a DP and semantically for a change of state. More specifically, it forces the DP it attaches to to be interpreted as the end state of a change of state event, adding the presupposition that the DP has been in that 3 One might pursue related versions of this hypothesis, e.g. where PATH ‘incorporates’ syntactically in one language but not another, or where PATH ‘bundles’ with v in one language but p in another. 4 We assume that P raises to p PLACE to license the latter, as in den Dikken (2010). Similarly, in English, we assume that pPLACE +P raises to pPATH for the same reason. 5 We assume that the ‘figure’ argument in SpecpP raises to SpecvP for licensing, after which v raises to Voice, as in Chomsky’s (2008) ‘raising to object’ analysis. Re- prefixation and Talmy’s parameter end state before. Because of this, re- cannot attach to the DP in SpecpP, because it would force that DP to be interpreted as a stative eventuality. But this is inconsistent with the semantic predication introduced by pPLACE , which is a relation between an entity and (the set of locations occupied by) another entity. This holds in both Romanian and English. In sum, re- prefixation is ruled out with concrete ‘put’ verbs on the basis of the properties of their complement: if the DPs in the complement need to be interpreted as entities (in this case due to pPLACE ), re- will not be possible. (33) a. English vP vCAUS ‘put’ pP pPATH p j PLACE Pi on pPLACE pP pPATH (to) p� DP the radio pPLACE tj PP P ti DP the shelf b. Romanian vP pP vPATH ‘put’ DP the radio p� pPLACE Pi pPLACE on 4.2 PP P ti DP the shelf Change of State We now turn to the question of why English and Romanian differ, where only Romanian allows re- on ‘put’ verbs when they express a change of state. The explanation in the previous subsection does not extend to ungrammatical cases in English such as (4) (He (*re)put the radio into use). In (4), there is no location, and no entity ‘use’ to which the Oana Săvescu Ciucivara and Jim Wood radio is related. Following the kind of reasoning in den Dikken (2010), we might say that pPLACE is absent in such non-locative uses. Since the radio undergoes a change of state, it can then be interpreted as the end-state of the radio; no functional head relating an entity to the PP is necessary. (34) PP DP P� the radio P in DP use A DP in SpecPP will be interpreted as a set of states, as will its sister P� . These will compose by predicate modification to yield their conjunction: the set of ‘in use’ states which are the end-states of the radio. In fact, we propose that without an entity-introducing head, the DP in SpecPP must be interpreted as a set of states in order to combine semantically with P� . Change-of-state and change-of-location expressions have been argued to share a common formal basis, namely a set of points along some scale (Snyder 2001; Gehrke 2008). We propose that PATH underlies both. In the absence of PLACE, some scale other than a set of contiguous points in space results. Consider the English and Romanian structures in (35). In neither language is there a problem in combining at the PP level; due re- (as well as the lack of a functional head like pPLACE ), the radio must denote a set of states, and it does. Composing this with PATH should not be a problem either; PATH should take a point and return a set of (contiguous) points for which that point is an endpoint. The important difference between the English and Romanian structures, we propose, is that by expressing PATH on v, Romanian allows re- to be in a more local relation to the v complex to which it eventually prefixes. In English, the pPATH head intervenes. (35) a. English vP pP vCAUS ‘put’ pPATH P pPATH in (to) PP (*re-) DP the radio P� P ti DP use Re- prefixation and Talmy’s parameter b. Romanian vP PP vPATH ‘put’ re- P� DP the radio P in DP function To implement this result, we might, for example, invoke the ‘Head-Movement Constraint’, if head-movement is the right way to derive the final position of re- (cf. Harley 2009). Alternatively, it may be that the number of intervening brackets, or PF adjacency, makes a difference at the morphological level. For present purposes, what is important is that the structural difference has nothing to do with re-, but rather with the independent Talmian difference between the two languages (along with the more specific claim about how nonlocative PPs are built). 5. Conclusion We have presented novel data from Romanian re- prefixation, arguing that it provides support for a structural difference between certain English and Romanian constructions with ‘put’-verbs. Romanian re- can attach to verbs taking two obligatory arguments, but only if they express a change of state—otherwise, it behaves like English. In both languages, reis disallowed with concrete uses of ‘put’ because re- requires a DP which denotes an end state, and this is incompatible with locational uses of ‘put’. The fact that re- combines with ‘put’ in Romanian but not English change-of-state constructions is related to a structural difference in how the two languages encode PATH. That this is not an isolated fact about ‘put’-verbs is supported by the examples in (29) and (30) involving ‘give’ and ‘take out’. References Adger, David, and Peter Svenonius. 2011. Features in minimalist syntax. In The Oxford Handbook of Linguistic Minimalism, ed. Cedric Boeckx, 27–51. Oxford: Oxford University Press. Borer, Hagit. 1984. Parametric syntax. Dordrecht: Foris. Cardinaletti, Anna. 2003. On the Italian repetitive prefix ri-: Incorporation vs. cliticization. University of Venice Working Papers in Linguistics 13:7–29. Chomsky, Noam. 1995. The Minimalist Program. Malden, MA: MIT Press. Chomsky, Noam. 2008. On Phases. In Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud, ed. Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta, 133– 166. Cambridge, MA: MIT Press. Oana Săvescu Ciucivara and Jim Wood Cinque, Guglielmo, and Luigi Rizzi, ed. 2010. Mapping Spatial PPs: The Cartography of Syntactic Structures Volume 6. Oxford: Oxford University Press. Collins, Chris. 2005. A smuggling approach to the Passive in English. Syntax 8:81–120. den Dikken, Marcel. 2010. On the functional structure of locative and directional PPs. In Mapping Spatial PPs: The Cartography of Syntactic Structures Volume 6, ed. Guglielmo Cinque and Luigi Rizzi, 74–126. Oxford: Oxford University Press. Fábregas, Antonio. 2007. The Exhaustive Lexicalisation Principle. Nordlyd 34:165–199. Folli, Raffaella, and Heidi Harley. 2004. Flavors of v: Consuming Results in Italian and English. In Aspectual Inquiries, ed. Roumyana Slabakova and Paula Kempchinsky, 95–120. Dordrecht: Kluwer. Gehrke, Berit. 2008. Ps in Motion: On the Semantics and Syntax of P Elements and Motion Events. Netherlands Graduate School of Linguistics: LOT Dissertation Series. Harley, Heidi. 2004. Merge, conflation, and head movement: The First Sister Principle revisited. North East Linguistics Society 34:239–54. Horn, Laurence R. 1980. Affixation and the Unaccusative Hypothesis. Chicago Linguistics Society 16:134–146. Kayne, Richard S. 2005. Some Notes on Comparative Syntax With Special Reference to English and French. In Movement and Silence, ed. Richard S. Kayne, 277–333. Oxford: Oxford University Press. Kayne, Richard S. 2006. On Parameters and on Principles of Pronunciation. In Organizing Grammar: Linguistic Studies in Honor of Henk van Riemsdijk, ed. Hans Broekhuis, Norbert Corver, Riny Huybregts, Ursula Kleinhenz & Jan Koster, 289–299. Berlin: Mouton de Gruyter. Koopman, Hilda J. 2000. Prepositions, postpositions, circumpositions and particles: The structure of Dutch PPs. In The syntax of specifiers and heads, ed. Hilda J. Koopman, 204–260. London: Routledge. Levin, Beth, and Malka Rappaport. 1986. The formation of adjectival passives. Linguistic Inquiry 17:623–661. Marantz, Alec. 2009. Resultatives and re-resultatives: Direct objects may construct events by themselves. Paper presented at the PLC. Mateu, Jaume, and Gemma Rigau. 2010. Verb-particle constructions in Romance: A lexical-syntactic account. Probus 22:241–269. Sigurðsson, Halldór Ármann. 2004. Meaningful silence, meaningless sounds. Linguistic Variation Yearbook 4:235–259. Snyder, William. 2001. On the nature of syntactic variation: Evidence from complex predicates and complex word-formation. Language 77:324–342. Sportiche, Dominique. 2008. Re re again. To appear in a Festschrift for Guglielmo Cinque (OUP). Available on lingBuzz/000770. Svenonius, Peter. 2003. Limits on P: filling in holes vs. falling in holes. Nordlyd 31:431– 445. Svenonius, Peter. 2007. Adpositions, particles and the arguments they introduce. In Argument Structure, ed. Eric Reuland, Tanmoy Bhattacharya, and Giorgos Spathas, 63–103. Philadelphia: John Benjamins. Re- prefixation and Talmy’s parameter Svenonius, Peter. 2008. Projections of P. In Syntax and Semantics of Spatial P, ed. Anna Asbury, Jakub Dotlacil, Berit Gehrke, and Rick Nouwen, 63-84. Philadelphia: John Benjamins. Talmy, Leonard. 1985. Lexicalization patterns: semantic structure in lexical forms. In Language Typology and Syntactic Description, ed. Timothy Shopen, 57–149. Cambridge: Cambridge University Press. Zanuttini, Raffaella. 2010. A Syntactic Analysis of Interpretive Restrictions on Imperative, Promissive, and Exhortative Subjects. Manuscript, Yale University. New York University Department of Linguistics 10 Washington Place New York, NY 10003 [email protected] [email protected]