The Prepositions a, of and by as dummy case-markers : two structural cases. Case as a Solution of Bound Pronouns

THE PREPOSITIONS A, OF AND BY AS DUMMY CASE-MARKERS: TWO STRUCTURAL CASES. CASE AS A SOLUTION OF BOUND PRONOUNS Jelly Julia de Jong University of Groningen & GISSL 1992 In this paper I explain the fact that bound pronouns in Iberian languages are grammatical in oblique case-marking contexts and ungrammatical in structural case positions. When we search for an explanation for this phenomenon relating it to case-theory, we diswver that the dummy case-marking character of the preposition a plays an important role. An analysis of the contexts where we Bnd this preposition gives us insight into the structure of specific and non-specific NPs in these languages and leads us to conclude that there are two types of structural object case (as proposed for Finnish by Belletti (1988). for Turkish by En$ (1991) and De Hoop (1992). and for Inuit by Bittner (1988) and Bok-Bennema (1991)). Furthemrore. I will show that also bound pronouns in snake-sentences and bigger domains in English can be explained in terms of case-theory. The dummy case-marking character of the prepositions of and by and the assumption that Num is a genitive caseassigner plays an important role here. The theoretical framework in which this paper is written is that of the theory of Government and Binding, developed in Chomsky (1981) and subsequent work. 1. Bound Pronouns Are Contrary to the Current Theory This section provides an introduction to the bound pronoun problem and an overview of the contexts where we can find these pronouns. The standard binding conditions A and B as formulated in Chomsky (1981:188), given in (I), are not able to account for the so-called 'snake-sentences', illustrated in (2): Caalan Working Papers in Linguistics (CWPL) 1992: 183-203. Universitat Autbnoma de Barcelona. (1) A. An anaphor has to be bound in its governing category. B. A pronoun has to be free in its goveming category. (2) Johni saw a snake near himi In this case the pronoun hirn is bound in its governing category (GC). This is in conflict with principle B of the standard binding conditions which states that a pronoun has to be free in its GC. We find bound pronouns in snake-sentences in many languages: Jani ziet een slang naast 'mi. 'John sees a snake near him.' Jani seach in slange nEst himi. 'John saw a snake near him.' Jeani voit un serpent $i c6tt de luii. 'John sees a snake near hirn.' Giannii vede un serpente vicino a luii. 'John sees a snake near him.' Ioni a vazut un sarpe linga el,. 'John saw a snake near hirn.' Juani vio una serpiente cerca de tli. 'John saw a snake near him.' En Joani va veure una serp a prop d'elli. 'John saw a snake near him.' O Joaoi viu um gat0 ao pé delei. 'John saw a cat near hirn.' Xani veu unha serpe perto deli. 'John saw a snake near hirn.' (Dutch) (Frisian) (French) (I talian) (Romanian) (Spanish) (fitala (Portuguese) (Galician) Lees and Klima (l%3) proposed a sentential solution for (2): (12) Johni saw [a snake to be near himi] This is what nowadays would be called a small clause solution. Chomsky (1981) briefly considers such solutions in chapter 5 but rejects them as being not generally applicable to all cases. The inadequacy of the smail clause solution to the problem clearly appears in Dutch cases with intransitives, as we can observe in (13): (13) Jan keek om zich heen. 'John looked around him.' If the domain introduced by om were a small clause, we would have the structure in (14): (14) Jani keek [s PROi om zichi heen] The problem is that with such intransitives, there is no other antecedent for PRO than the subject itself. But since zich is also bound by the subject, the reflexive is bound in the local domain, while it should be free.1 In other words, with a structure like (14) one would expect zichzelfrather than zich. Thus, the small clause analysis gives exactly the opposite results from what is needed. This anaiysis must therefore be rejected. Moreover, as empirical investigation shows, several Romance languages allow pronouns to be bound in more local contexts, as we can observe in (15)-(20): (15) Jeani parle de luii. John talks about him (French) As shown by Koster (1985: 148, (19)): a. zich and zichzelfare bound in their minimal full clause; b. zichzelfmust be bound in a small anaphoric domain, zich must be fre. (16) ??~ianni~ parla de luii. John talks about him (allowed in certain contexts) (17) Ioni totdeauna vorbeste despre eli. John always talks about hirn (18) Juani habla de Bli. John talks about him (19) En Joan, parla d'elli. The John talks about hirn (20) O Joaoi fala sempre delei. The John talks always about hirn (15)-(20)'John (always) talks about himself.' (Italian) (Romanian) (Spanish) (catala) (Portuguese) Assuming that lui, el, él, ell and ele are pronouns, the standard binding conditions are not able to account for these locally bound pronouns. The French example (15) is widely known in the literature but the Italian example (16) is marked as ungrammatical in the literature so far. The native speakers I talked to, however, told me that it is much easier to get a bound reading when the sentence is supported by contextual infonnation. In a forthcoming article I will discuss the specific problems with bound pronouns in French and Italian. I will also postpone the discussion of some Romanian and Galician sentences I got via questionnaries. In this article I will discuss bound pronouns in the Iberian languages. Section 2 discusses the possibility of bound pronouns in Frisian and presents the Chain condition that Reinhart and Reuland (1991a, b) developed to explain this phenomenon. We will see that this Chain condition explains why bound pronominals in Iberian2 languages are grammatical in oblique case-marking contexts. In this section I will also show what kind of contexts do not allow bound pronouns. The Ikan languages are the Romance languages spoken in the Iberian Peninsula. 2. A Case-related Solution to the Bound Pronouns The anaphoric system of Frisian presents a serious problem for the standard binding theory too. Like English, Frisian has a two-member system; it has an anaphor himsels, and a pronominal hirn. However, unlike Dutch and English, Frisian has locally bound pronominals. The generalization is that wherever Dutch allows zich (a so-called SEanaphor3), Frisian allows a bound pronominal. In their search for an explanation of the distribution of bound pronouns in Frisian, Reinhart and Reuland (1991a, b) have developed a Generalized Chain condition on Achains which is based on a general notion of Chain links by Chomsky (1986a, b): (21) Generalized Condition on A-chains A maximal A-chain (al, ...., a,,) contains precisely one link -alwhich is fully specified for grammatical features (structural case features and +features). Pronominals are fully specified for phi-features. The referential dependence of anaphors is syntactically reflected in having a paradigm which lacks a distinction in at least one grammatical dimension. Anaphors fai1 to have a full paradigm for singular-plud or gender distinctions. (Reinhart & Reuland (1991b)) In (22) the pronoun se 'her', which bears structural case: produces an ungrammatical sentence but har is grammatical in this bound position. - Reinhart and Reuland (1989) introduce the tems SEand SELF-anaphors. SELF-anaphors are complex anaphors like e.g. zichzelf (Dutch), seg selv (Norwegian), hlmself(English), se stesso (Italian). SE-anaphors are Simplex Expression-anaphors like e.g. zelf(Dutch), seg (Norwegian), self(English), se (Italian). Two pronominals, namely the 3rd person singular feminine and the 3rd person plural (common gender), have two object forms: both have se as well as har (or -in pluralharren). The two foms can be used interchangeably. However. unlike harand harren, se is ungrammatical when locally bound. J. Hoekstra (1991) shows that seis ungrammatical in a number of positions where har and harren are allowed: (a) in the object position of prepositions (which assign oblique case); (b) in the object position of transitive adjectives (which assign oblique case to their objects, cf. Van Riemsdijk (1983)); (c) in the experiencer argument position of psychological verbs (22) a. Jeltsjej skammet h~i. Julia shames her-INH b. *Jeltsjej skammet se,. Julia shames her-STRUCI' 'Julia is ashamed of herself.' (Frisian) (22b) contains a chain that links two elements bearing structural case and thus violates the Chain condition in (21). Reinhart and Reuland argue that (22a) is grammatical because har bears inherent case and that in the entire sentence there is only one element which is specified for structural case: the proper name Jeltsje. We will see that this Chain condition also accounts for the (un)grammaticality of locally bound pronouns in Spanish, Portuguese and Catalan. Consider again examples (18)-(20), repeated here as (23)-(23, and examples (26)-(34): (23) Juani habla de Cli. (24) En Joani parla d'elli. (25) O Joaoi fala sempre delej. (26) Gregoriq 10 ha comprado para Cli. Gregory it has bought for him 'Gregory has bought it for himself.' (27) Se'l prq va comprar per elli. (not: per a ell) itpro buy-PAST-3sg for a him (not: for a him) 'He bought it for himself.' (Spanish) (atala (Portuguese) (Spanish) (as Den Besten (1985) and also Belletti & Rizz. (1988) show, tbis position is assigned oblique case). and (d) in free dative constructions (which are assigned inherent, dative case). Hoekstra comes to the conclusion that (A) 'Se must bear structural case'. (that is, either accusative or nominative since se also is used for 3rd person singular feminine and 3rd person plural (commou gender) subjects). Since he shaws that in all four circumstances just mentioned the pronoun har(ren) is allowed, the~ is another, implicit, conclusion: (B) 'Har(ren) can bear inherent case'. There is a fifth circumstance where seis ungrammatical and har(ren) is graamatical, namely: (e) in a bound position in a reflexive construction. (28) O paii comprou a casa para elq. The father has-bought the house for hirn 'Father has bought the house for himself.' (29) a. *Mario se ama a 81. Marioloves a hirn b. Mario se ama a si mismo. Mario loves a SE self 'Mario loves himself.' (30) a. *En Joan s'estima a ell. The John loves a hirn b. En Joan s'estima a si mateix. The John loves a SE self 'John loves himself.' (3 1) a. *O Joao ama-se a ele. The John loves a hirn b . O Joao ama-se a si pr6prio / a si mesmo. The John loves a SE self 1 a SE self 'John loves himself.' (32) a. "Fernando se lava a 61. Fernando washes a him b. Fernando se lava a sí mismo. Fernando washes a SE self 'Fernando washes.' (33) a. *En Pere es renta a ell. The Peter washes a hirn b. En Pere es renta a si mateix. The Peter washes a SE self 'Peter washes.' - (Portuguese) (Spanish) (Catalan) (Portuguese) (Spanish) (34) a. *O Pedrq lava-se a elei. The Peter washes a him b. O Pedro lava-se a si mesmo I a si pr6prio. The Peter washes a SE self I a SE self 'Peter washes.' (Portuguese) The prepositions in (23)-(28) assign oblique case to the object. This means that these objects do not have structural case. The only element in these sentences which is fully specified for all grammatical features is the subject. This means the objects are able to form a Chain with the subject without violating the Chain Condition (21). But what about (29)-(34)? Why can't we have bound pronouns here? Section 3 concerns the properties of the preposition a that precedes [+animate] and [+specific] direct object pronouns in the Iberian languages. The case-marking character of this preposition and the Chain condition, together, explain why bound pronouns in the Iberian languages are ungrammatical in certain contexts. 3. Preposition a as a Dummy Case-marker At first sight all object pronouns in (23)-(34) are assigned oblique, thus inherent, case by the preceding preposition. But the value of the preposition is not identical in all sentences. In the Iberian languages indirect objects and direct object strong pronouns with the features [+animate] and [+specific] are preceded by the preposition a. For example: (35) Me ha visto a mi. me has3sg seen a me (36) Jo el corregir6 a ell. I him correct-FUT a he (37) O homem burlou-te a ti. The man cheated-3sg-you a you (Spanish) (catalan) (Portuguese) (38) Vexo-o a ell. seelsg him a he (Galician) Jaeggli (1982) points out that the preposition a 'to', when inserted before a direct object strong pronoun with the features [+animate] and [+specific], is a dummy case-marker. He was the first to relate case assignment to the phenomenon of clitic-doubling.5~6 Zubizarreta (1985) shows that a is a dummy preposition functioning as a semanticaily empty case-marker by pointing out that the accusative direct object preceded by a is thematically unrestricted. She argues that an argument that is realized as object of the verb or as subject is semanticaily unrestricted since other roles than Agent may be assigned to the subject position and other roles than Theme may be assigned to the object position. However, the thematic role of an argument Jaeggli uses this to explain that in River Piate Spanish clitic-doubling is only possible with [+animate] objects. Let us consider the circwnstances in which clitic-doubling is dlowed in Spanish: (i) a. Veo a 10s chicos. 'I see the boys.' b. Los vw a ellos. 'I see them.' c. Los vw a 10s chicos. 'I see the boys.' (ii) a. Compro 10s libros. b. *Los compro 10s libros. 'I buy the books.' In (iib) the object libros 'books' is not [+animate] so no a-insertion takes place. In (iib) the ditic absorbs case, this means case cannot be assigned to the object libros. In (ib) the clitic dso absorbs case but insertion of the preposition a takes place because the object is [+animate] and [+specific]. As the preposition a is a dummy casemarker it is able to assign case to the object ellos. This is the reason why (ib) is grammatical and (iib) is ungrammatical. The dummy case-marking character of the preposition a leads Demonte (1987) to suggest a distinction between true PPs and pseudo-PPs (=NPs) in the grammar of Romance languages. In her reasoning the a + NP sequences of (29)-(38) ai^ p~e~do-PPs. (50) b. *This problemi virtually defines a solution of iti. (51) a. Sami doesn't remember the man after himi (at the podium). b. *Sami doesn't remember any brother of himi. (52) a. mich picture near Johni does hej like? b. *Which pictures of Johni did hej like? (53) Pictures of Johni seem to himi to be beautiful. (54) a. Which picture near Johni did hej buy? b. *Which picture of Johni did bei buy? In all these cases the pronoun is bound in its governing category. The judgements on the (b) sentences range from highly dubious to ungrammatical. The (a) sentences are grammatical in general (as we will saw in section 7, the ungrammaticality of (47a) turns out to be a strong argument in favour of my analysis in tems of case and chains). It is not difficult to observe that in most of the (a) sentences the object is preceded by a preposition that assigns oblique case. The objects in the (b) sentences, however, are preceded by the preposition of which assigns structural case. This means that in the (b) sentences we find two elements with structural case in the same chain, and hence these examples are ruled out. In the (a) sentences, however, no violation of the Chain Condition takes place so they are felt to be much better. The idea that in English a genitive in the complement of NP or AP is realized by means of the preposition of which, as a semantically empty case-marker, assigns objective case is put forward by Chomsky (1986a:87, 192-204, footnote 139). More concretely, I will assume that Number has its own functional projection in English (like Ritter (1991a, b) does for Hebrew and Picallo (1991) for Catalan) and that it is Num that assigns genitive case just like Infl assigns nominative case and V assigns accusative case.1OThe preposition of is a dummy element which 'helps' Num to assign genitive: l0 Ritter (1991b) discusses genitive constructions (both construct states and free genitives) in Hebrew. She says that in a construct state the N raises to the head of DP to pick up a phonetically nul1 genitive case-assigning determiner, which she labels Dgen. There are two characteristics that distinguish the free genitive from the construct state: First, the definite article ha may appear in initial position of the free genitive; and second, the free genitive subject bears an overt genitive case-marker Sel. Since a free genitive DP can be headed by the A Num 4p Ab, of x N attaches to Nurn, Num assigns genitive case to the object via the preposition otll Section 7 explores the case-marking properties of the preposition by in English. 7. Preposition by as a Dummy Case-marker It is interesting to observe that (47a), repeated here as (56), is ungrammatical: (56) *Theyi remember a discussion by themi. definite determiner, the surface order cannot be derived by raising N to Dgen in this context. Moreover, the presence of Sel, the ovat case-marker on the subject, suggests that genitive case is assigned by some other means than Dgen in the free genitive construction. Nevertheless (beause of subject-object assymetry) Ritter wants to derive the surface order by moving the head noun. Therefore, she supposes that the free genitive construction is derived by N-raising, but not to the head of DP. According to the Head Movement Constraint, one is forced to postulate another head position intermediate between D and N. Ritter assumes, following Anderson (1982) and others, that there is a distinction between inilectional affixes and derivational affixes. Inilectional affixes are attached to the lexical stem as a consequence of head movement and thus must be heads of syntactic projections. Number is an inilectional affix (in Hebrew and, for our sake, in English) and is thus realized as the head of a syntactic category. Moreover, the position of this category is intermediate between DP and NP. So, the intermediate category is NumP and its head bears the number specification of the noun phrase. l1 It is not probable that these DPs are headed by a phonetically nul1 genitive case-assigna (Dgen) which assigns genitive because they can have a definite determiner in initial position. Bresnan (1982), and later Zubizarreta (1985), claim that in English the preposition by is a semantically empty case-marker which thematically does not restrict the object. As Zubizarreta has pointed out:l2 "In English, the adverbial status of the by-phrase in the passive construction is not so obvious because it has lost its lexical meaning. As (57a, b) illustrate, the by-phrase in the English passive (as well as in other languages such as French, Spanish, and Italian) need not be agentive. This shows that the preposition by in this construction does not restrict the semantic role of its object. The by-phrase is interpreted as having the same semantic role as the lexical external argument of the verb. This is an unusual situation for an adverbial: adverbials are typically semantically restricted phrases. (57) a. The letter was rkived by Mary. b. The house is surrounded by trees. Further proof that by is a dummy preposition is provided by example (5%). Only semantically unrestricted grammatical positions may be predicated of. As noted by Bresnan (1982) and illustrated in (56b) for French, the by-phrase in the passive construction can be predicated of. Since adjectives in French carry overt agreement features, there can be no doubt that solile in (58b) is functioning as a predicative adjective and not as an adverb. (58) a. John said he was passed by Mary in the hall yesterday drunk. b. Ces vers ont Cté réciíés par Marie compl&tement soQle. Those verses were recited by Mary completely drunk.' (Zubizarreta ( 1985254-255)) Since inherent case is related to a specific O-role (cf. Chomsky (1986a); Belletti (1988)) we must come to the conclusion that the object preceded by the preposition by in (56) does not bear inherent case. I assume that by assigns structural case. 12 The numbering of the examples of the quotation has been modified to agree with the numbering of the present paper. Now, let us look at sentence (56) again and see whether we have a solution for its ungrammaticality. We assume that the preposition by is a dummy case-marking preposition whereas the head of the PP in the sentences (46) and (48)-(51) is a 'real' preposition with 'true' semantic value. Moreover, the preposition by assigns structural case to its obejct rather than inherent case. This means that in these sentences we find a chain linking two elements bearing structural case, violating thus the Chain condition (21). 8. Conclusion In this paper I have shown that the Chain condition developed by Reinhart and Reuland does not only account for bound pronouns in Frisian but also for bound pronouns in all domains in Spanish, Catalan and Portuguese and in snake-sentences and bigger domains in English. According to Jaeggli (1982) the difference between specific NPs, which are a-NPs, and nonspecific NPs which are 'bare' NPs is related to case. We saw this is indeed plausible since according to Bittner (1988), Bok-Bennema (1991), Enq (1991) and De Hoop (1992) the type of interpretation an object gets is correlated to the type of case assigned to this object. Following De Hoop (1992) on a link between two types of structural case and different readings on objects, it can be argued that Spanish a is a case-marker inserted at S-structure in order to license strong structural case on [+specific] [+animate] objects. In the case of Spanish, it seems very likely that the other type of objective case which gives rise to a weak reading, is also a structural rather than an inherent case. Furthemore, Jaeggli's hypothesis is also in accordance with the Chain condition developed by Reinhart and Reuland (1991a,b). In fact, Reinhart and Reuland's A-chain condition can be adapted somewhat in the sense that the notion structural case is replaced by strong structural case in order to distinguish the latter type of case from inherent as well as from weak structural case with respect to A-chains (this was independently put forward by De Hoop as well). A version of this paper was presented at the I1 Coloquio de Gramatica Generativa (VitoridGasteiz, April 1992). I would like to thank Carme Picallo for helpful and inspiring comments on my work during the 1992 Girona International Summer School in Linguistics. This research was supported by the Foundation for Linguistic Research, which is funded by the Netherlands organization for research, NWO. References Belletti, A. (1988) The case of unaccusatives', Linguistic Inquiry, 19.1, 1-34. Belletti, A. and L. Rizzi (1988) 'Psych-verbs and O-theoryf, Natural Lunguage and Linguistic Theory, 6,291-352. Besten, H. den (1985) The ergative hypothesis and free word order in Dutch and German' in J. Toman (ed) Studies in German Grammar, Foris, Dordrecht, 23-64. Bresnan, J. (1982) 'Control and Complementation', Linguistic Inquiry, 13.3,343-434. Bittner, M. (1988) Canonical and noncanonical argument expressions, Ph. D. dissertation, University of Texas, Austin. Bok-Bennema, R. (1991) Case and agreement in Inuit, Foris, Dordrecht. Chomsky, N. (1981) Lectures on government and binding, Foris, Dordrecht. Chomsky, N. (1986a) Knowledge of Lunguage: Its nature, origin and use, Praeger. Chomsky, N. (1986b) Barriers, MIT Press, Cambridge. Demonte, V. (1987) 'C-command, Prepositions, and Predication', Linguistic Inquiry, 18.1, 147157. Ens , M. (1991) The Semantics of Specificity', Linguistic Inquiry, 22.1, 1-25. Fodor, J. and I. Sag (1982) 'Referential and Quantificational Indefinites', Linguistics and Philosophy, 5.3, 355-398. Hoekstra, J. (1991) Pronouns and Case. On the distribution of Frisian harren and se 'them' and the relation between abstract case and morphological case, ms. Fryske, Ljouwert. Hoop, H. de (1992) Case Confguration and Noun Phrase Interpretation, Ph. D. dissertation, University of Groningen. Jaeggli, 0. (1982) Topics in Romance Syntax, Fons, Dordrecht. Johnson, K. (1992) The syntax of nominals and binding theory', hand-out AIO-course, June 1992, Amsterdam. Koster, J. (1985) 'Reflexives in Dutch' in Grammatical Representation, Fons, Dordrecht. Lees, R. and E. Klima (1963) 'Rules for English pronominalization', Language, 39.1, 17-28. Picallo, C. (1991) 'Nominals and nominalizations in Catalan', Probus, 3.3, 279-316. Reinhart, T. and E. Reuland (1989) 'Anaphors and logophors: An argument structure perspective' in J. Koster and E. Reuland, Long distance anaphora, Cambridge University Press, Cambridge. Reinhart, T. and E. Reuland (1991a) 'Binding conditions and chains', ms. University of Tel Aviv /Groningen. Reinhart, T. and E. Reuland (1991b) 'Pronouns, simplex anaphors and chains.' The Seventh Workshop on Comparative Germanic Syntax, November 22-24, Stuttgart. Riemsdijk, H. van (1983) The case of German adjectives' in F. Heny and B. Richards Linguistic Categories: Auxiliaries and related puzzles, vol. I: Categories, Reidel, Dordrecht. Ritter, E. (1991a) 'Two functional categories in Noun Phrases: Evidence from modern Hebrew', in S. Rothstein Perspectives onphrme structure: heads and licensing. Syntax and Semantics, 25, Academic Press, New York. Ritter, E. (1991b) 'Evidence for number as a nominal head', GLOW, Leiden. Zubizarreta, M.L. (1985) 'Morphophonology and morphosyntax: Romance causatives', Linguistic Znquiry, 16.2, 247-289. Department of General Linguistics-A.T.W. Oude Kijk. in't Jatshp. 971 2 EK Groningen The Netherlands e-mail: [email protected]