Optionality in the Placement of Asturian Clitics

Optionality in the Placement of Asturian Clitics Guillemo Lorenzo Universidad de Oviedo I discuss in this paper some recent proposais about clitic placement in Western Romances. Focusing on Asturian data, I evaluate them explonng their extension to contexts not considered originaüy by their authors. I conclude that the better way to give a coherent account to all these data is the proposal (due to Rouveret) of a functionai head specifically designed for the placement of clitics, which generates or not depending on the kind of sentences in which clitics appear. This hypothesis fits well with certain instances of optionaiity observed in infinitivai contexts. It is proposed that two altemative derivations are opened in these cases, both licensed on economy grounds. This paper explores a bunch of syntactic topics: the nature of pronominal clitics and the syntax of its placement, the nature of epistemic verbs, the optional character of certain derivations (a non welcome idea within the latest outcomes of Generative Theory), the morphological conditions for long head movement application . . . They are all unified in a try to evaluate two recent hypotheses about clitic placement in Western Romances. I mainly use data from Asturian (an almost forgotten Western Romance). The paper is organized as follows: section 1 presents the basic facts about clitic placement in Asturian, and introduces Rouveret and Uriagereka's hypothesis about them; section 2 offers certain data that do not seem to fit well with any of them; section 3 concludes, finally, that only Rouveret's hypothesis, implemented with some theoretical devices independently needed by the theory, can surmount the conflictive data of the previous section. Catalun Working Papers in Linguistics (CatWPL) 4.1 (1994): 9'3-120 Universitat Autbnoma de Barcelona 1. Clitic Placement in Asturian: An Overview Asturian declarative sentences with the word order that can be considered standard or unmarked (SVO) place obligatorily the clitics after the verb ('enclisis'): (1) a. Xuan IlimpiGy 10s zapatos a Maria. Xuan clean-PAST-3sg-cl (to her) the shoes to Maria 'Xuan cleaned María the shoes.' b. *Xuan y llimpi6 10s zapatos a Maria. Xuan cl (to her) clean-PAST-3sg the shoes to Maria Enclisis is thus the normal placement for Asturian clitics, as in the rest of Western Romances (Portuguese and Galician).' Asturian data is described by Sánchez Vicente and Rubiera Tuya (1985) and D'Andrks (1993). Its clitic paradigm is summarized in bis schema: As can be noticed from the examples in (1). dative clitics can always be doubled in Asturian. Doubling of an accusative clitic, on the other hand, is only possible when the corresponding DP has been focused, as in (6). singular plural 3 lu (ac. masc.) - la (ac. fem.) 10 (ac. neuter) Y (dat.) 1 ms 1 me 2 te 2 VOS 3 10s (ac. masc. neuter) les (ac. fem.) yos (dat.) In certain contexts, however, the situation is exactly the opposite, and 'proclisis' becomes the only option for cl~tic placement. The most important of those contexts are summarized bellow : (i) Sentences introduced by a wh-word (2) a. iC6m0 10s llimpió Xuan? how cl(them) clean-PAST-3sg Xuan 'How did Xuan clean them?' b. *¿Cóm0 llimpiódos Xuan? how clean-PAST-3sg-cl(them) Xuan (3) a. iQu6-y dixo Maria? what (ac)-cl(to him) say-PAST3sg Maria 'What did Maria say to him?' b. *¿Qd dixo-y Maria? what (ac) say-PAST-3sg-cl(to him) Maria (ii) Sentences introduced by a negative word (4) a. Non te 10 dicía por iso. not cl(to you) cl(it) say-PAST-lsg for that 'I did not say it to you for that reason.' b. *Non dicíatelo por iso. not say-PAST-lsg-cl(to you)-cl(it) for that (5) a. Inxamás lu vi delante. never cl(him) see-PASTlsg in front 'I never saw him in front of me.' b. *Inxamás vi-lu delante. never see-PAST-lsg-cl(him) in front (iii) Sentences with focused phrases (6) El padre dió-y algunes perres. the father give-PAST-3sg-cl(to him) some money The father gave him some money.' a. Algunes perres-y les di6 el padre. some money-cl(to him)-cl(them) give-PAST-3sg the father 'Some money was given to him by the father.' b. ?*~l~unes perres dió-y les el padre. some money give-PAST-3sg-cl(to him)-cl(them) the father 1 .I. Some Analysis All along this paper I will assume that Asturian clitics behave as Xos that in a certain point of the derivation adjoin to the higher inflectional head, a position that they reach by a movement operation from an argument position (see section 2.1.1 below). This is represented in (7): I will not deepen here into the question about the trigger of such an operation. I will follow Kayne's (1991) prohibition against cliticisation over traces, and I will consider without further reasoning that it can explain why clitics need to go out of the lexical projection of verbs. I will accept (also as Kayne) that the higher inflectional head is the final target of this movement operation in most Romance languages. What I really want in this paper is to evaluate two recent proposals about the ordering restrictions observed by clitics in Western Romances, that I will try to summarize in a few paragraphs. 1.1.1. Rouveret (1992). The gist of Rouveret's (1992) proposal is the existence in the lexicon of Western Romances of a special functional head, whose mission is to receive clitics from the lexical projection of verbs, where they find no conditions for cliticisation. Rouveret calls W this functional head.2 W, projected higher than AgrS, asks for the satisfaction of certain conditions within its checking domain: namely, the presence of either a verbal and a nominal element. Rouveret says that this property derives from a mixed morphological character of W. The sentential subject, once satisfied the agreement conditions of the AgrS head, moves up to the specifier of WP, checking the nominal feature of W. Up to this point, clitics (already located in W) appear before the verbal complex (which is located in AgrS). But W still waits for the checking of its verbal feature. This circumstance will work as a trigger for the movement of the verbal complex, which adjoins to W and finishes the derivation before clitics. (8) summarizes all these steps: The W is after Wackernagel, who proposed the law that clitics must appear following the first sentential constituent in Indo-European. The extension of this law to Romances Languages in hown as Tobler-Mussafia Law. See about these questions Anderson (1993) and Salvi (1990). NP; A Y -TGAgS C1. AgrS TP A 4'A\ This derivation explains the ordering of clitics within declarative sentences exhibiting an unmarked word order. How can we explain the opposite ordering, observed in the remaining cases? Rouveret explains that the common aspect of all these cases is the presence of another functional head: Comp in sentences introduced by a wh-word, and (I will assume following Laka (1990)) Z in sentences with negative items or focused phrases. All this functional heads manifest as abstract features, which impose a requirement of lexical identification on the representation. Rouveret assumes that such a requirement cannot be satisfied by a head labelled as W. Thus, Grammar avoids its generation in this kind of contexts, and it is the verbal complex (labelled as AgrS) which adjoins to those heads identifying the abstract features. A CI-V ordering is obtained in this way, as represented in (9): 1.1.2. Uriagereka (1993). Uriagereka tries to explain the peculiarities of clitic placement in Western Romances without introducing any distinctive functional category in their lexicons. He maintains that the responsible of this idiosyncrasy is a category which is actually present . in all Romances, though with different morphological specifications in each of them. He calls this category F, and he says that it serves to code syntactically the paint of view of the speaker or the discursive subject. F is, among other things, the position for clitics. Uriagereka proposes that in Western Romances F is a sort of affix that asks for the help of another element in order to acquire phonetic shape. In his opinion, this requirement can be satisfied in two ways: (i) cliticisation under the infiuence of a goveming element, or (ii) incorporation into it of another head. I (ii) is the situation observed in declarative sentences with the unmarked order: clitics are 1 located in the position F and, there being no goveming element, the verbal complex adjoins to them. The resulting derivation is exactly as (8), except for the presence of F instead of W. The remaining cases are explained by the fact that in all of them there is an element which governs the clitics in F; so, altemative (i) is accomplished. It is represented in (10): cliticisation C1 NP AgrS' A AgrS (...) A V-T-AgrS It is worth noticing at this point that Unagereka's proposa1 about clitic placement in Westem Romances makes some wrong predictions. For instance, his hypothesis is incompatible with the verb-subject inversion observed in interrogative contexts in all these languages, as shown in (1 1) with Asturian data: (11) a. ¿dnde trabayaba Xuan el aiiu pasau? where work-PAST-3sg Xuan the year last 'Where does Xuan work last year?' b. *¿onde Xuan trabayaba el añu pasau? where Xuan work-PAST3sg the year last .- If this phenomenon is explained (as usually is) as movement of the verbal complex into Comp, cliticisation might take place over the verbal complex placed in Comp (as in (12)), and not over the wh-word in [Spec, SC]. (12) is, however, an ungrammatical sentence: (12) *i [e qué [comp di6i-Y Xuan ti a Maria]]? what give-PAST-3sg-cl (to her) Xuan to María So, Uriagereka's hypothesis is problematic even in some matrix sentences. We will see in the next section that in no matrix contexts it is absolutely useless. 2. Clitic Placement within Infinitival Constructions Clitics located in embedded infinitival constructions exhibit in most of the cases the possibility of appearing before or after the infinitive, even in sentences in which typical triggers of proclisis are present: (13) a. Ye una tontería non facelo. is a nonsense not do-cl(it) 'It is nonsense not dolng it.' b. Ye una tontería non 10 facer. Is a nonsense not cl(it) do 'It is nonsense not doing it.' (14) a. Preguntó 6nde facelo. ask-PAST-3sg where do-cl(it) 'He asked where to do it.' b. Preguntó ónde 10 facer. ask-PAST-3sg where cl(it) do 'He asked where to do it.' (26) talanta [CP [comp etense>] [Agrsp kgrs facer (1o)l güeyll Notice that in (26) a potential trigger for the clitic anteposition is present. According to the kind of analysis developed in this paper (see 1.1. I), the abstract operator which appears in (26) must be lexically identified: the movement of the verbal complex from AgrS up to Comp could satisfy this requirement. But such an operation would result in an important defect for the representation, because the clitic would be left stranded. However, we can imagine again that the clitic itself adjoins to Comp after the movement of the verbal complex, arising a string with the Cl-V order of (25a). In (27) all the steps of this derivation are represented: With a factive verb as llarnentar a C1-V order is not obtained, because its context lacks a trigger for a derivation as (27). Up to this point we have an partia1 account of the contrast exhibited in (23). Anyway, (23a) is not the only possible ordering of the elements of an infinitival clause in the context of an epistemic verb. (28) is also correct in Asturian: (28) Talanta facelo giiey. wonder-3sg do-cl(it) today 'He is thinking about doing it today.' 2.2.1. Pensar is Not un Epistemic Verb in Asturian. An altemative idea about the Cl-VIVC1 altemation within infinitival clauses introduced by epistemic verbs relies on the possibility of rejecting the presence of the abstract temporal operator in Comp (or in liberating it of the requirement of lexical identification), and on the supposition that we are in front of an instance of 'clitic climbingt.3 In other words, it could be that what is at stake in sequences of the type 'epistemic verb + infinitive' is a structure which allows the movement of the clitics of the infinitive up to the clitic positions of the main verb.4 Following this idea, in a sentence as (23a) (repeated below as (29a)) a clitic which is serving as an argument of the infinitive does not find any obstacle in its way to the main inflection. Once there, it should obey the standard conditions introduced in section 1. Therefore, in an unmarked situation it should appear after the main verb. (29b) represents that: (29) a. Talanta 10 facer giiey. b. [WP [W talant9 [W loiII [A~~sP [A~S ti tjII (. . .) h ti facer güe~lll But it must be noticed that if 10 had really cliticised over the main verb in (29), it might happen that the introduction of any of the standard triggers of proclisis should obligatorily place the clitic before the verb. This prediction is not borne out: (30) *Non 10 talanta facer güey. not cl(it) wonder-3sg do today The conclusion is that the clitic still relies on the infinitive in (29). The curious thing about this conclusion is that epistemic verbs are usually introduced as typical mediators of clitic climbing contexts, as can be seen in (3 1) with the verb pensar ('to ~his possibility was suggested to me by Amaya Mendikoetxea @.c.). See Kayne (1989) for an specification of the characteristics of such a structure. (31) a. Pensaba facelo guey. think-PASTlsg do-cl(it) today b. Pensabalo facer güey. think-PAST-lsg-cl(it) do today c. Non 10 pensaba facer guey. not cl(it) think-PAST-lsg do today5 Pensar is usually offered as a prototypical epistemic verb, and it maybe behaves as such in Portuguese. I believe, however, that pensar does not belong to this kind of verbs neither in Asturian nor in Spanish. This idea seems to be clear in sentences as (32), in whichpensar has a modal value that can be translated as 'to have the intentionlwill to do x'. In Asturian (as in Spanish), pensar is only able to develop an authentic epistemic value in a compositional way, with the participation of a preposition. These two values are clearly distinguished in the Spanish sentences bellow: (32) a. Modal Pensaba acabar10 hoy. think-PAST-lsg finish-cl(it) today 'I have the intention of finishing it today.' b. Epistemic Pensaba en acabar10 hoy. think-PAST-lsg in finish-cl(it) today 'I was imagining a situation in which I finished it and such a situation was a part of tcday.' I arn consciously avoiding to give a literal translation to the sentences in (31). In tbis secalon I arn precisely írying to fix the mect equivalencies to them. The translations in (32) are certainly artificial, but I am only trying to fix and to distinguish the two different values of pensar. What is important in this context is that only a sentence as (32a) allows clitic climbing: (33) a. Lo pensaba acabar hoy. cl(it) think-PAST-lsg finish today 'I have the intention of finishing it today.' b. *Lo pensaba en acabar hoy. cl(it) think-PAST-lsg in finish today Summing up, pensar can be classified within the verbs that Picallo (1990) calls 'semimodals', which sometimes exhibit all the properties of modal verbs (among them, to allow clitic climbing), and sometimes behave as standard verbal heads. My conclusion is that while talantar is a pure epistemic verb,6pensar it is not in its basic meaning. It can only develop an epistemic value in a derivative way. Talantar, as an epistemic verb, subcategorizes CPs, in whose head an abstract temporal operator is located;7pensar, as a modal verb, subcategorizes an infinitival VP (if we follow Kayne (1989:240-241)) or it is generated as an VP adjoined to an infinitival VP (if we follow Picallo (1990:289)). Within any of those theoretical frameworks it is possible to justify the clitic climbing operation, which becomes impossible if we suppose the presence of a CP projection in between, as in my analysis of sentences with epistemic verbs (as talantar). Given this analysis, sentences of this kind share their basic structure with those of (14) and (15) (repeated as (34b) and (34~). with an interrogative operator in Comp. In all the cases, an abstract element in Comp starts a derivational process which concludes with the Cl-V order. The situation is similar to that in (13) (repeated as 6 Talantar is translated by Neira and Piíieiro (1989:571) as 'pensar'. 'cavilar' (i.e. . 'to think', 'to ruminate'). 7~s can be noticed from examples (15) and (17) above, ihese CPs can contain boa declarative or interrogative sentences. (34d)), with an abstract negative operator in X. Therefore, the optionality of clitic placement in the context of an epistemic verb is due to the same reason of the optionality in the context of an interrogative or negative word: (34) a. Talanta facelo giiey. = Talanta 10 facer giiey. wonder-3sg do-cl(it) today wonder-3sg cl(it) do today 'He is thinking about doing it today.' b. Preguntb 6nde facelo. = Pregunt6 ónde 10 facer. ask-PASTSsg where do-cl(it) ask-PAST3sg where cl(it) do 'He asked where to do it.' c. Talantaba c6mu facelo. = Talantaba c6mu 10 facer. wonder-PAST-3sg how do-cl(it) wondered how cl(it) do 'He wondered how to do it.' d. Ye una tontería non facelo. = Ye una tontería non 10 facer. is a nonsense not do-cl(it) is a nonsense not cl(it) do 'It is nonsense not doing it.' It is important to notice that if it is true that all the cases in (34) ask for an unitary explanation, Uriagereka's hypothesis seems to be unoperative in non-matrix contexts. Leaving the question of the optionality aside and focusing only on the versions with proclisis, (34) finds no explanation within his framework, because in such a sentence there is no lexical element whose presence serves to prevent the movement of the verbal complex, and the main verb is unable to govern the clitic, given the presence of a CP projection. 2.3. Optionality (2) Sections 1.1.1 and 2.2 contain a possible explanation for the version with proclisis of the sentence types represented in (34). In all these cases there is an abstract operator in a functional head position which needs to be lexically identified by the infinltival complex. The infinitive moves up to that position, and the clitic is left stranded, against a general condition on representations. In order to repair this situation, the clitic itself adjoins to a higher head, giving rise to the Cl-V order. But this explanation still leaves two questions unresolved: (i) how is the derivation in the altemative version with enclisis?, and (ii) how it is possible that Grammar leaves open two alternative derivations, which seems to be against economy considerations? 2.3.1. Long Head Movement. Thus far I have explained proclisis as an effect of an adjunction operation triggered by the necessity of avoiding clitic stranding. Consequently, the versions with enclisis should be explained by a derivation which does not leave the clitic in such a situation. I will defend in this section that this is what we obtain if the verb avoids adjunction to AgrS in its way to Z or Comp. From a morphological point of view, a derivation like that finds certain motivation in the fact that Asturian infinitives lack agreement features lexically manifested. We can think that Asturian infinitives are able to rise up to Z: or Comp in order to identify their abstract features avoiding the abstract AgrS head. This derivation is represented in (35):8 I understand (following Kayne (1991)) that the clitic is in a legitimate enviroment in (35). assuming that prohibition again cliticisation over traces does not preclude cliticisation over other kinds of empty elements (as an empty agreement feature). from its original position: given the GTC, when it adjoins to X (which contains the verbal complex at this point of the derivation), it transfers to X the property of governing its original government domain, and the V-T complex becomes the governor of its own trace. It is true that (37) seems to contain a violation of the prohibition against clitic stranding (see Kayne (1993:8-12)). However, the derivational step represented in (37) takes place in LF, where such a condition is probably irrelevant. It must be taken into account that the prohibition against clitic stranding is a condition on the phonetic processing of clitics, which loses its relevance once the LFIPF split has taken place. We have, therefore, two alternative derivations, both apparently allowed by Grammar: (38) (i) V-T adjoins to X passing along (Cl) AgrS; and (ii) <AgrS> adjoins to V-T -X in LF. (39) (i) V-T adjoins to AgrS; and (ii) (CI) V-T-AgrS adjoins to X. 2.3.2. Economy of Derivationr. The introduction of economy principies among the mechanisms of Grammar aims to restrict.the number of possible derivations for each representation, favoring the less costly ones attending to certain criteria (see Chomsky (1991, 1993)). Therefore, the conclusion reached in the previous section must be considered problematic from the point of view of the theory of derivations. The two alternative derivations proposed up to here are schematically represented in (40): Both derivations have the same number of steps. From a purely numerical point of view, both are equally optimal. An aspect of those derivations that could favor (38) is that one of its steps takes place at LF. Under a certain interpretation of Chomsky's (1993) Procrastinate Principle, we could conclude that derivation (38) is more economical than derivation (39). However, what Procrastinate actually establishes is that a derivation must refrain their steps as much as possible, in tems of the level at which they apply: LF movement are less costly, I and derivations must choose this option unless the structure contains an element which determines its processing in the Overt Syntax. Attending to this, Procrastinate cannot serve as a referee in our case, because the LF step in (38) does not contrast with an overt equivalent in (39). Actually, both derivations apply their own steps in the optima1 level, considering the conditions that intervene in each case. Therefore, it must happen that the two derivations that I have been introduced are equally economical for the Grammar, which can maintain both. A potential problem for this solution is why it does not apply in the context of finite verbs, where only derivation (39) is possible. But what is at stake here is probably the fact that all the features of the finite inflection are overtly manifested, as is also the case with Portuguese inflected infinitives (see again (36)), which determines a strict application of the HMC.lO l0 Wbich is compatible with Lema and Rivero's (1990:14) idea about the morphological character of this condition. 3. Conclusion In this paper I have dealt with some problems of clitic placement within infinitival contexts in Asturian. I have used them in order to evaluate Rouveret and Uriagereka's hypothesis about clitic placement in Western Romance. My conclusion is that only the former can be extended to contexts not considered originally by any of the authors without introducing any new stipulation in the theoretical frarnework that they use. This paper is part of a research made during my stay at MIT in the academic year 19931994. I want to thank the Department of Linguistics and Philosophy for giving me Visiting Scholar status for a year, and all the people that made me happy all that time, specially Anna Alvarez, Ana Ardid, Víctor Longa, Javier OrmazBbal, Xavier Villalba and Miriam UribeEtxebarría. Very special thanks are due to Amaya Mendikoetxea and Gemma Rigau, for their help and insplration in the development of this paper. I also thank Xose Lluis Garcia Arias for confirming some of my intuitions about the Asturian data, and an anonymous CatWPL reviewer for valuable comments. References Anderson, S.R. (1993) 'Wackernagel's Revenge: Clitics, Morphology, and the Syntax of Second Position', Language 69, -98. Alvarez, R., X.L. Regueira and H. Monteagudo (1992) Gramática Galega, Galaxia, Vigo. Baker, M.C. (1988) Zncorporation. A Theory of Grammatical Function Changing, Chicago University Press, Chicago. Barbosa, P. (1994) 'Clitic Placement in Old Romance and European Portuguese and the Nul1 Subject Pararneter', ms. MIT. Benucci, F. (1992) 'Prepositional Particles and the Portuguese Personal Infinitive', University of Venice Working Papers in Linguistics 92.1.8. Chomsky, N. (1991) 'Some Notes on Economy of Derivation and Representation' in R. Freidin (ed) Principies and Parameters in Comparative Grammar, The MIT Press, Carnbridge, 417-454. Chomsky, N. (1993) 'A Minimalist Program for Linguistic Theory' in K. Hale and S.J. Keyser (eds) The View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger, The MIT Press, Cambridge, 1-52. D'AndrCs, R. (1993) Allugamientu de 10s Pronomes Atonos col Verbu n'Asturianu., Departamentu de Filoloxía Española, UviCu. Kayne, R.S. (1989) 'Null Subjects and Clitic Climbing' in O. Jaeggli and K.J. Safir (eds) The Null Subject Parameter, Kluwer, Dordrecht, 239-261. Kayne, R.S. (1991) 'Romance Clitics, Verb Movement, and PRO', Linguistic Inquiry 22,647686. Kayne, R.S. (1993) The Antisymetry of Syntax', ms. Graduate Center, CUNY. Lema, J. and M.L. Rivero (1990) 'Long Head Movement: ECP vs. HMC', paper presented at the 20 NELS Conference, Pittsburg. Longa, V.M. (1994) The Galician Inflected Infinitive and the Theory of UG'. See this volume. Neira, J. and M.R. Pifieiro (1989) Diccionario de 10s Bables de Asturias, Instttuto de Estudios Asturianes, Oviedo. Picallo, M.C. (1990) 'Modal Verbs in Catalan', Natural Language and Linguistic Theory 8, 2853 12. Raposo, E. (1987) 'Case Theory and Infl-to-Comp: the Inflected Infinitive in European Portuguese', Linguistic Znquiry 18,85109. Rouveret, A. (1992) 'Clitic Placement, Focus and the Wackernagel Position in European Portuguese', paper presented at the ESF Workshop on Clitics, Donostia. Salvi, G. (1990) 'La Sopravivenza della Legge di Wackernagel nei Dialetti Occidentali della Peninsola Iberica', Medioevo Romanzo 17, 177-2.10. Sánchez Vicente, X.X. and C. Rubiera Tuya (1985) 'Enclisis, proclisis y pronomes átonos n'asturianu', Lletres Asturianes 18,71431.