Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 287 Vrinda Subhalaxmi Chidambaram Princeton University Slovenski jezik – Slovene Linguistic Studies 6 (2007): 287–301 Relative and Pseudo-Relative Clauses in Slovene Avtorica raziskuje raznovrstne oziralne stavke v slovenščini s posebnim ozirom na strukturo, ki se tradicionalno obravnava kot izjemna glede na to, da je v njej neskladje med številom v naslonskem zaimku in številom v odnosnici, npr. Najboljši iskalec si, [RC kar smo jih imeli.] 1 Razprava kaže, kako sodobna generativna skladnja lahko reši ta problem, in sicer z dvema možnima rešitvama. The author explores the various types of relative clauses in Slovene, focusing on a construction that has traditionally been assumed to be a special type of relative clause in which there seems to be a mismatch between the number feature of the clitic pronoun and that of the relative head, e.g., Najboljši iskalec si, [RC kar smo jih imeli.] 1 best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd. sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the best seeker that we have had (them).’ The paper illustrates how contemporary generative syntax can handle the problem of this feature mismatch, giving two possible solutions. 1. Introduction2 Chomsky noted in The Minimalist Program (1995) that certain basic syntactic phe- nomena continue to lack an adequate account. Among these, he mentions relative clauses. Many scholars (notably Alexiadou, Bianchi, Chomsky, Kayne, Lasnik, etc.) have investi- gated relative clauses in various languages, but none has offered a universally consistent archetype for their structure. The inevitable result of this theoretical gap is that there is no perfect definition of ‘relative clause’ and no perfect diagnostic that one can apply to determine which constructions can be labelled as relative clauses and which cannot. To date, this issue has not arisen as a major problem in the corpus of generative syntax, as in most of the world’s languages the distinction between relative clauses and other varieties of clauses (i.e. than-clauses selected by a comparative adjective) is unambiguous. Most languages have certain lexemes (often syncretic with wh-words or complementizers) that introduce a relative clause as well as a consistent relationship between the relative clause and the main clause. These two factors make the relative clause in such languages easy to identify and fairly simple to analyze (language specifically). There are, however, certain languages where the distinction between relative clauses and other types of clauses is blurred, bringing to light the crucial question of the UG principles determining the struc- ture of relative clauses. Slovene is one such language. 1 Example taken from Harry Potter in jetnik iz Azkabana. 2 Many thanks to Alja Ferme, Gašper Tkačik, and Sašo Živanovič for their grammaticality judgements and insights on this project. 288 Slovenski jezik – Slovene Linguistic Studies 6 (2007) In this paper, I will identify several types of relative clauses in Slovene, focusing on one particular construction (to be known as the kar-phrase). I will provide two possible analyses of this construction, both based on a generative syntactic model. In one, I will follow the traditional assumption that it is indeed a relative clause. In the second analysis, I will provide a more explanatory model, distinguishing this type of clause from relative clauses and likening it to other types of embedded clauses that also appear cross-linguistically. 2. A Brief Typology of Slovene Relative Clauses The most common relative clause construction in Slovene involves the invariant complementizer ki and a resumptive clitic pronoun bearing the person, case, number and gender of the object, as shown in example (1). The resumptive pronoun appears exclusively when it represents the dative, accusative or genitive object of the embed- ded predicate, and under such conditions, it is obligatory, as illustrated by the devi- ance of sentence (2). (1) Govorim s človekom, [RC ki ga poznaš.] speak-1st.sg.pres. with man-inst.sg.masc that-REL him-3rd.sg.acc.masc.pron know-2nd.sg.pres. ‘I am speaking with the man that you know.’ (2) *Govorim s človekom, [RC ki poznaš.] speak-1st.sg.pres. with man-inst.sg.masc that-REL know-2nd.sg.pres. ‘I am speaking with the man that you know.’ This type of relative clause (hereafter, ki-type relative) is the most frequently used rel- ative construction in Slovene. Relatives clauses, like the ones in Slovene, that include a resumptive pronoun as opposed to a gap are attested in several other languages, such as Brazilian Portuguese, the Venetian dialect of Italian and Bergamesco (Bianchi 2004). Under a Kaynian analysis (Kayne 1994, Bianchi 1999), the relative head is generated inside of the embedded relative clause, and when it moves into the matrix clause, it leaves behind a trace, which in Slovene, Brazilian Portuguese, Venetian Italian and Bergamesco, is overtly realized as a resumptive pronoun. Put differently, the relative head in (1), namely človekom starts out in the embedded clause as the object of the verb poznaš. As the structure builds up by means of the Merge opera- tion, človekom raises into the main clause leaving behind a trace within the embedded clause. This trace is overtly realized as a clitic pronoun which must move (presum- ably as a PF operation) into the second position of the embedded relative clause. The appearance of the overt trace or resumptive pronoun is easily explained in terms of the Case Filter and the Full Interpretation condition (as defined in Chomsky 1995). Poznati is a verb that has two arguments, two theta-roles to assign, and an uninterpre- table accusative case feature to check. In order to circumvent a Case Filter violation and a violation of Full Interpretation, the trace bearing the accusative case checks the accusative case feature on the verb and is realized as the accusative pronominal clitic bearing all of the φ-features of človekom, namely the feature complex: [+masculine +3rd person +animate +singular] Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 289 There are also instances of ki-type relatives where there is no accompanying clitic pronoun. This occurs when the relative head is base-generated as the subject of the relative clause. (3) Punce, [RC ki študirajo lingvistiko ] so lepe girls-nom.pl.fem that-REL study-3rd.pl.pres. linguistics-acc.sg.fem be-3rd.pl.pres pretty-pl.fem. ‘The girls, who study linguistics, are pretty.’/ ‘Girls who study linguistics are pretty.’ There is no reason why one could not analyze this construction just as the other ki- type relatives that contain accusative or dative clitics. The relative head, in this case punce, moves from the subject position of the relative clause into the main clause. The fact that there is no resumptive pronoun left in the subject position of the rela- tive clause can be attributed to Slovene being a pro-drop language, in which subject pronouns are generally covert. The ki-type relative clause, then, is consistent with existing analyses of relative clauses. While ki-type relatives can be used in constructions where the relative head moves from subject, object, or indirect object position within the relative clause into the matrix clause, they cannot be used in cases where the relative head is moved from an adjunct position in the relative clause. That is to say, ki-type relatives appear to be used exclusively for relativizing arguments. If there is an adjunct in the embedded clause that is then relativized, the resulting relative clause will contain a properly declined form of the pronoun kateri, meaning ‘which’ or ‘who’, as shown in (4). This is one type of wh-relative (a relative clause using a pronoun formed from a wh-word) used by Slovene. (4) Tam je punca, [RC s katero igram tenis] there is girl-nom.sg.fem. with-prep whom-inst.sg.fem. play-1st.sg.pres. tennis-acc.sg.masc. ‘There is the girl I play tennis with.’/ ‘There is the girl with whom I play ten- nis.’ Slovene also has a second type of wh-relative. These are most often used in free relatives and correlatives. The morphological form of the relative pronouns used in these clauses is: wh-interrogative + [r], as shown in the table below. Table 1: WH-Relative Pronouns Interrogative Relative pronoun English koga kogar who (acc. or gen.) komu komur whom (dat.) čigav čigar whose kaj kar what (nom. or acc.) česa česar what (gen.) čemu čemur what (dat) 290 Slovenski jezik – Slovene Linguistic Studies 6 (2007) This table lists a number (though not all) of the declined relative pronouns formed from interrogative wh-words. The two examples below illustrate the use of these rela- tive pronouns in a free relative (5) and a correlative (6). (5) Kdor ni z nami je proti nam. whoever-REL.nom. is-not-3rd.sg.pres. with-prep. us-2nd.inst.pl. is-3rd.sg.pres. against-prep. us- 2nd.pl.dat. ‘Whoever is not for us is against us.’ (6) Česar nimam, tega nočem what-REL.gen. not-have-1st.sg.pres. that-gen.sg.masc not-want-1st.sg.pres. ‘What I do not have, I do not want.’/ ‘That which I do not have, I do not want.’ The use of these relative pronouns is more or less restricted to the free relative and correlative constructions3 with the significant exception of kar, which I will address below. 3. The Kar-Phrase A more complex construction in Slovene is a clause that closely resembles the ki-type relative, but that uses the relativizer kar. Unlike in the examples of free rela- tives and correlatives, kar here cannot be a wh-pronoun, as it is indeclinable, like ki. These constructions, like ki-type relatives, require the presence of a clitic pronoun in the embedded clause, although it is often not immediately clear what acts as the pronoun’s antecedent. (7) Najboljši iskalec si, [RC kar smo jih imeli.] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the best seeker that we have had (them).’ (8) *Najboljši iskalec si, [RC kar smo imeli.] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. have-L-par- ticiple ‘You are the best seeker that we have had.’ Just as with ki-type relatives, the clitic pronoun can also represent a dative argument of the embedded predicate, as shown in (9). (9) Najboljši iskalec si, [RC kar smo jim dali] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.pl.dat.pron give-L-participle nagrado. award-Acc.sg.fem. 3 The structures of free relatives and correlatives have been discussed and analyzed by various scholars, but as this is not the main object of my inquiry, I will simply say that such constructions do exist in Slovene and will refrain from providing an analysis for them here. (see Toporišič 2000).  In fact, the pronoun could even be found in the instrumental case in which it has no clitic counterpart, as in ‘On je bil najboljši iskalec, kar sem z njimi kdaj igrala.’ (‘He was the best seeker that I have ever played with (them).’) Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 291 ‘You are the best seeker to whom we have given (them) this award.’ The unusual aspect of this type of clause (hereafter, kar-phrase) that distinguishes it from the similar ki-type relative is the apparent mismatch between the clitic pronoun and its antecedent. The clitic pronouns in the embedded clauses of sentences (7) and (9) are plural, while their antecedents appear in the singular. With ki-type relatives, such a mismatch is inadmissible, as shown in (10). Conversely, with kar-phrases, a matching between the number features of the antecedent and the clitic pronoun yields a deviant sentence, as illustrated in (11). (10) *Najboljši iskalec si, [RC ki smo jih imeli.] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the best seeker that we have had (them).’ (11) *Najboljši iskalec si, [RC kar smo ga imeli.] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.sg.acc.pron have-L-participle ‘You are the best seeker that we have had (him).’ In this paper I will explore the kar-phrase, offering two possible analyses that will account for its unique behavior. The behavior of kar in sentences such as (7) and (9) indicates that it is not a relative wh-pronoun, as it shows neither animacy nor declines as one would expect. Instead it remains in a frozen form, namely the nominative/accusative singular inani- mate. This seems to suggest that ‘kar’ is not a wh-pronoun, but that it is an invariable complementizer heading a complementizer phrase. That having been established, we are immediately faced with a question: is the kar-phrase a relative clause CP compa- rable to the ki-type relative and differing from it only in some minor way or is it an entirely distinct type of CP? 4. If it looks like a relative and walks like a relative... Textbooks and grammars of the Slovene language identify kar as a relativizer (Lenček 1982, Derbyshire 1993, Albretti 1995), which is not an unqualified assump- tion. The structure of kar-phrases is indeed similar to the structure of the high-fre- quency ki-type relatives. The problem that obtains, following this assumption, is how to handle the number feature mismatch. Assuming, as traditional and modern gram- mars have, that the kar-phrase is truly a type of relative clause, one must account for the difference between it and ki-type relatives. It is crucial to note that ki-type relatives can also include a plural resumptive pronoun, but only in cases where the overt relative head is plural, as shown below in example (12). (12) Najboljši iskalec si od vseh iskalcev [RC ki smo best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. of all-gen.pl. seeker-gen.pl. that BE-1st.pl.pres. jih imeli.] 3rd.pl.acc.pron have-L-participle 292 Slovenski jezik – Slovene Linguistic Studies 6 (2007) ‘You are the best seeker of all the seekers that we have had.’ In the matrix clause of sentence (12), there is a prepositional phrase whose object is in the plural. It is semantically a partitive prepositional phrase, identifying the group of entities that the superlative quantifies over (de Hoop 2003). The structure of this phrase is straightforward, if one assumes that the relative clause itself is embedded inside the prepositional phrase. Figure A The resumptive pronoun in Slovene is also a second-position clitic, so it does not appear in the position shown in (fig.A), but in a higher position in the embedded clause (presumably moved there by a PF operation). Aside from this, the structure is straightforward, insofar as the number feature of the resumptive pronoun does match the number feature of the relative head. While this derivation seems natural, it is by no means inarguable and it is, in fact, controversial. It is important to note that the structure in (fig.A) rests on assumptions that devi- ate from the analysis of relative clauses offered by Kayne (1994), subsequently ad- opted and expanded upon by Bianchi (1999), in which the head of the relative clause is merged into the object position of the relative clause and then moves into SpecCP. (13) a. [DP the [CP that John made [DP claim]]] b. [DP the [CP [DP claimj ] that John made tj ]]] (Kayne 1994, Alexiadou et al. 2000) The derivation that Kayne proposes, in fact, gives rise to a number of questions con- cerning Case assignment, even in simple relative clauses, such as (1) in Slovene. (1) Govorim s človekom, [RC ki ga poznaš.] speak-1st.sg.pres. with man-inst.sg.masc that-REL him-3rd.sg.acc.masc.pron know-2nd.sg.pres. ‘I am speaking with the man that you know.’ The head of the relative clause is assigned instrumental case by the preposition ‘s’ in the matrix clause. Under Kayne’s head-raising analysis, it is unclear how the Case checking properties shown here would obtain (i.e. does the preposition check its Case across a CP barrier?). Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 293 The kar-type relatives present further problems for any analysis. Assuming that it is structurally no different from a ki-type relative clause, one would have to posit the existence of a covert plural head that would give rise to the plurality of the resump- tive clitic pronoun. The structure, then, would be somewhat similar to what is shown in example (12), (fig.A), namely (14), where the prepositional phrase need not (but certainly could) be pronounced. (14) Najboljši iskalec si [PP od vseh iskalcev] [RC kar smo best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. of all-gen.pl. seeker-gen.pl. that BE-1st.pl.pres. jih imeli.] 3rd.pl.acc.pron have-L-participle ‘You are the best seeker of all the seekers that we have had.’ In (14), the plural resumptive pronoun has a plural antecedent which can either be overt or covert. With a ki-type relative, the prepositional phrase containing the plural antecedent must not be deleted, as illustrated by the difference between the grammati- cality judgments of examples (10) and (12). Unlike ki, the relativizer kar licenses the PF-deletion of the prepositional phrase ‘of all the seekers’. In LF, however, the PP is present, giving jih, the plural clitic pronoun, a semantic antecedent. If this is true, then one must assume that the PP ‘of all the seekers’ is an autonomous constituent, that is to say that it occurs in a specifier position to which the relative clause is adjoined. If it were not a constituent, then it could not undergo deletion. In order to determine whether or not ‘of all the seekers’ is indeed a constituent, one can apply a movement test, topicalizing the PP. As shown below, the kar-phrases allow the plural head to be topicalized, while the ki-type relatives do not allow this movement. (15) [Od vseh iskalcev]j si najboljši iskalec tj [RC kar smo of all-gen.pl. seeker-gen.pl BE-2nd.sg.pres. best-Nom.sg.masc. seeker-Nom.sg.masc. that BE- 1st.pl.pres. jih imeli.] 3rd.pl.acc.pron have-L-participle ‘Of all the seekers, you are the best seeker that we have had (them).’ (16) *[Od vseh iskalcev]j si najboljši iskalec tj [RC ki smo of all-gen.pl. seeker-gen.pl BE-2nd.sg.pres. best-Nom.sg.masc. seeker-Nom.sg.masc. that BE-1st.pl.pres. jih imeli.] 3rd.pl.acc.pron have-L-participle ‘Of all the seekers, you are the best seeker that we have had (them).’ The movement in (16) is not permitted because, as shown in (fig.A), the phrase ‘of all the seekers’ is not a constituent. These data suggest that, in kar-phrases, the PP ‘of all the seekers’ forms a constituent, which leads inevitably to the question of the position that the PP occupies. One possibility that has been raised for the analysis of relative clauses is that they are not complements of the head, but that they are adjoined to a higher projection, essentially as an adjunct modifier (Alexiadou et al. 2000). For re- strictive relative clauses, it has been assumed (if you are to take the adjunction model) that the clause adjoins to the NP head, but in the case of the kar-type relatives, the 29 Slovenski jezik – Slovene Linguistic Studies 6 (2007) head is embedded in a PP. In the kar-type relatives, then, the derivation would be as shown in (fig. B) below, if we adhere to the adjunction model. Figure B The analysis illustrated above follows the base-generated external head hypothesis (Chomsky 1977, Browning 1991), which asserts that the head of the relative clause is base generated outside of the CP and that it is linked to the relative clause by an interpretive relation (Alexiadou et al. 2000). This relation could play out in one of three ways: i. there could be a silent operator that is co-indexed with the head and moves into SpecCP, ii. there could be a [-wh] pronoun in SpecCP, or iii. there could be a [+wh] pronoun that moves into SpecCP. Given the argument in section 3 that kar is a complementizer rather than a wh- pronoun, the only satisfactory possibility is that there is an operator in SpecCP that was moved from the object position. It seems awkward that a silent operator should leave behind an overt trace after raising to SpecCP, but there appears to be no other solution. Furthermore, under this analysis, the original problem arises again: why is the resumptive pronoun plural? One possibility is that kar itself is a complex head, meaning ‘of which’. This is not entirely implausible, as the word kar does impart a partitive meaning. The word as a whole is co-indexed with the PP ‘of all the seekers’, but one of its constituent parts, namely ‘which’, is co-indexed with the N ‘seekers’. Thus, kar ends up with two indices, one bearing a semantic partitive feature and an- other bearing a plural syntactic feature. The index that it shares with the noun ‘seek- ers’ is shared by the operator, whose trace, therefore, also shares the same index. This explains both the partitive nature of the construction and the plurality on the resump- tive pronoun. There are a number of problems with this analysis. First, it assumes that a relative head can be embedded within a prepositional phrase external to the relative clause. The relationship between the head and the relative clause is, then, obscured. There would be no c-command relationship between the head of the relative clause and the relative clause itself, so the mechanism for agreement is not entirely clear. Further- Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 295 more, there is no clear way to determine the mechanism by which either a ki-type relative or a kar-type relative is selected. This leads to the most glaring oversight of the account: it says nothing about the major generalization that kar-type relatives oc- cur most often with superlatives, a point that will be addressed fully in the following section. Nevertheless, this analysis, in spite of its flaws, does reveal an important in- tuition - namely that there seems to be an underlying partitive reading to the kar-type relative clauses. 5. Pseudo-Relativity Theory The notion that there is something inherently partitive about the kar-phrase is one that needs to be specified in the syntax. In the previous model, this was done by adding a partitive PP and a corresponding relative pronoun. That analysis, however, did not give any insight into the distribution of kar-phrases in the language. One salient property of the kar-phrase is that it appears more often than not with superla- tives. Even other elements that quantify over a variable do not necessarily license the kar-phrase. (17) *Kupila sem veliko knjig, kar sem jih našla v knjigarni. bought-L-participle AUX many books-gen.pl., that AUX them-3rd.pl. found-L-particple in book- store-loc.sg. ‘I bought many books that I found at the bookstore.’ Native speakers judged this sentence to be deviant, noting that it would be grammati- cal simply if the word kar were replace with ki. Several other sentences using non- superlative quantifiers along with the kar-phrase were also judged ungrammatical. Thus, in order to properly investigate the kar-phrase construction, it is necessary to make some reference to the analysis of degree words (i.e. superlatives and compara- tives). Andrews (1975) and Alexiadou et al. (2000) point out that clauses selected by comparative adjectives function similarly to relative clauses. (18) a. more books [than John can read] b. as many books [as John can read] c. too many books [for John to read] (Alexiadou et al. 2000) They note the similarity that holds between the two constructions, citing specific commonalities. Like a headed restrictive relative clause, all three are DPs. They ap- pear in the same position as a relative clause would, namely on the right periphery. Both in comparative constructions as well as relative clause constructions, there ap- pears to be a transformation akin to wh-movement. Alexiadou (2002) claims that it would seem reasonable to analyze this as adjunction (that is to say that the than- clause, as-clause, and for-clause are adjuncts), “except that the selection of the clause by the degree word fails to be captured” (Alexiadou et al. 2000). Alexiadou et al. give the following analysis of comparatives, in which the clause selected by the degree determiner is extraposed: [NP [Det Deg + S] N]→[NP [Det Deg] N S] 296 Slovenski jezik – Slovene Linguistic Studies 6 (2007) It is certainly conceivable that a similar relationship holds between the superla- tive and the kar-phrase, namely that the superlative selects a phrase headed by kar, and that, like with comparatives, the selected phrase is extraposed into the right pe- riphery. The syntax of sentence (3), then would be as follows: [NP [Det Najboljši + kar smo jih imeli] iskalec]]→[NP [Det Najboljši] iskalec kar smo jih imeli] [NP [Det Best + that we have them had] seeker]]→[NP [Det Best] seeker that we have them had] ‘Best seeker that we have had (them).’ The question that inevitably arises from this analysis pertains to the nature of the selection of the kar-phrase. What sort of phrase does a superlative select? This question was partially answered in Section 4 of this paper, in that the PP presented in the underlying (LF) structure of the kar-phrase is a partitive prepositional phrase. Just as the comparative selects a CP containing the C0 ‘than’, a superlative selects as its complement a partitive CP headed by kar. Unlike the analysis given in section 4, the partitive reading is not realized by means of a constituent external to the CP, but rather it is an inherent component of the selected CP. Given the analysis as stated so far, one potential pitfall is that it does not show what motivates the requirement of a pronoun in the CP. In the previous analysis laid out in Section 4, the motivation was clear: relative clauses in Slovene that contain invariable complementizers must also contain resumptive pronouns across the board, so the presence of a resumptive pronoun in the kar-phrase is unremarkable. In the pseudo-relative analysis, in fact, a slightly stronger justification emerges. The pseudo-relative is so called because it bears a strong resemblance to a rela- tive clause. Like a relative clause, it contains a gap that represents a variable. In every relative clause, there will be a variable that appears as a result of movement of the relative head. In comparatives, while there doesn’t seem to be a movement out of the clausal complement to the comparative, there is a gap in the semantic structure that, as with a relative clause, is determined by the head noun. I have [more [books [than John can read (books)]]] (Alexidou et al. 2000) The gap, which I have filled in here with the word books offset in parentheses, is also a variable. In fact, the meaning of the gap would be (x number of books), but to sim- plify, I will say that the word ‘books’ is the assignment of the variable. The analysis of the Slovene superlative does not differ greatly from this analysis of the English comparative. The superlative takes the kar-phrase as a complement, and the phrase has a variable. The kar-phrase functions much like the partitive PP ‘of all the seekers’ would, but as it takes the form of an entire clause, it must, then, have a variable in it that the superlative quantifies over. He is [the best [seeker [PPpartitive of all the seekers] (no gap) He is [the best [seeker [CPpartitive that we have had (seekers)]]] (gap) The variable in these clauses is always plural because the entire clause functions as a partitive CP, and the embedded NP in a partitive must be both plural and definite (de Hoop 2003), and thus the variable in the kar-phrase must and does satisfy both Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 297 of these conditions. Under this analysis, the plural pronoun in the kar-phrase is not a resumptive pronoun, as it is in the ki-type relative clauses, but rather it represents a variable. It is important to note here that the ki-type relative, being a “true” relative, has the structure laid out roughly in fig.A. When it occurs with a superlative, as in (12), the ki-type relative is an adjunct to the relative head modified by the superlative, and its features are identical to those of the relative head. The conclusion that can then be drawn is that the syntactic structure of ki-type relatives never matches the syntactic structure of kar-phrases. 6. Some Lingering Issues There are some problems with this analysis. If the kar-phrase is not a relative clause, but a clausal complement of the superlative, then one would expect to find it uniquely in combination with superlatives. While it is overwhelmingly the case that they appear with superlatives, it is not an absolute requirement, as shown in example (19) and (20). (19) To so vsi iskalci, [RC kar smo jih imeli.] 5 that-dem.pron. BE-3rd.pl.pres. all-Nom.pl.masc. seekers-NOM.pl.masc. that BE-1st.pl.pres. 3rd. pl.acc.pron have-L-participle ‘Those are all the seekers that we have had (them).’ (20) Edini iskalec si, [RC kar smo jih imeli.] only-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the only seeker that we have had (them).’ Examples (19) and (20) both contain kar- phrases without superlatives, which given this analysis, should not be a possibility. Moreover, it should be possible with any superlative, which turns out not to be true. The following example has been judged as ungrammatical. (21) *Prvi iskalec si, [RC kar smo jih imeli.] first-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the first seeker that we have had (them).’ There seems to be a possible semantic explanation for this, namely that if there is one seeker who is the first and only seeker, then ‘first’ cannot select a partitive. In fact, when the matrix verb of this sentence, namely the copula, is made past tense, most native speakers have judged it to be grammatical. So perhaps this is a pragmatic prob- lem rather than a syntactic one. On the other hand, the reason why sentences (19) and (20) are considered by many native speakers to be well-formed remains a mystery. 5 The grammaticality judgements of (19), (20), and (21) vary. While some native speakers find them entirely acceptable, others have judged them ungrammatical. Nevertheless, I feel compelled to include these examples as problems to my proposed theory, as some speakers do consider them grammatical. My thanks to Marko Snoj, Marina Zorman and to the anonymous reviewer who pointed this out. 298 Slovenski jezik – Slovene Linguistic Studies 6 (2007) Additionally, the following sentence is also grammatical: (22) Najslabše pivo je, kar sem ga kdaj pila.6 worst-Nom.sg.neut. beer-Nom.sg.neut. BE-3rd.sg.pres. that BE-1st.pl.pres. 3rd.sg.acc.pron ever drink-L-participle ‘It is the worst beer that I have ever drunk.’ In this example, the clitic pronoun within the kar-phrase is singular, matching the number feature of the relative head. The construction shown in sentence (22) only works for mass nouns. In fact, when the singular clitic in the kar-phrase in (22) is replaced by the plural clitic pronoun jih, the sentence takes on the meaning shown in (23). (23) Najslabše pivo je, kar sem jih kdaj pila. worst-Nom.sg.neut. beer-Nom.sg.neut. BE-3rd.sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron ever drink-L-participle ‘It is the worst beer that I have ever drunk (of a certain variety of beers).’ Here, the understanding is obligatorily that the beer being drunk is the worst can of beer of a particular brand. Essentially, it would be equivalent to saying ‘This is the worst Union that I have ever drunk’ (Union being a brand of Slovene beer). This is often the reading when one pluralizes a mass noun, i.e. ‘wines’ refers to varieties of wine. This gives some support to the notion that in (22), the fact that ‘pivo’ is a mass noun somehow accounts for the singular clitic pronoun in the kar-phrase, although the exact mechanism for this is unclear. I leave this issue for further research. There are also some curious facts about the ordering of these clauses in sentences with multiple clauses. If the approach in Section 4 were correct, then one would ex- pect that the two types of relative clauses could appear in the same sentence in any order, since theoretically, a sentence can have an infinite number of relative clauses (i.e. The dog that ate the cat that ate the mouse that ate the cheese...) and they can appear in any configuration with respect to each other. That is to say that there are no rules determining the ordering of different types of relative clauses. In Slovene, both the embedding and the ordering of relative clauses seem to have critical effects. (24) ??Najboljši iskalec si, [RC kar smo jih imeli.] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle [RC ki sem ga trenirala.] that BE-1st.sg.pres. 3rd.sg.acc.pron train-L-participle ‘You are the best seeker that we have had (them) that I have trained him.’ (25) *Najboljši iskalec si, [RC ki sem ga trenirala.] best-Nom.sg.masc. seeker-Nom.sg.masc. BE-2nd.sg.pres. that BE-1st.sg.pres. 3rd.sg.acc.pron train-L-participle [RC kar smo jih imeli.] that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the best seeker that I have trained (him) that we have had (them).’ 6 This sentence was uttered spontaneously by a native speaker, Alja Ferme. Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 299 Both sentences are degraded, which could be explained by some sort of complemen- tizer harmony in Slovene, but it is an inelegant solution and likely the wrong one. Under the relative clause hypothesis, then, there doesn’t seem to be a good reason for (24) and (25) to be ruled out. What is somewhat mysterious in terms of the solution presented in this section is the difference in judgements between the two sentences. Several native speakers claimed that while both of these sentences are degraded, that (24) is slightly better than (25). This might be explained by the extraposition of the kar-phrase. The kar- phrase is extracted from the degree phrase and moves to the right periphery over the head noun. It is conceivable that the kar-phrase cannot extrapose over another clause, as it would be a violation of a minimality condition. So the structure shown below would be impossible. [NP [Det Najboljši + kar smo jih imeli] [iskalec [ki sem ga trenirala]]]→ [NP [Det Najboljši t] [[[iskalec [ki sem ga trenirala]] kar smo jih imeli]] [NP [Det Best + that we have had them] [seeker [that I have trained him]]]→ [NP [Det Best t] [[[seeker [that I have trained him]][that we have had them]] ‘Best seeker that I have trained (him) that we have had (them).’ Nevertheless, other native speakers claimed that (25) was a more grammatical sen- tence than (24), saying that if it were only slightly rephrased, as in (26), it would be better. (26) Ti si najboljši iskalec [RC ki sem ga trenirala.] you BE-2nd.sg.pres. best-Nom.sg.masc. seeker-Nom.sg.masc. that BE-1st.sg.pres. 3rd.sg.acc.pron train-L-participle [RC kar smo jih imeli.] that BE-1st.pl.pres. 3rd.pl.acc.pron have-L-participle ‘You are the best seeker that I have trained (him) that we have had (them).’ (26) shows that the kar-phrase can, for some speakers, extrapose over the NP ‘iskalec, ki sem ga trenirala’. Other speakers treat this as a minimality violation. 7. Conclusion The two approaches given in sections 3 and 4 illustrate two ways to solve a prob- lem of feature mismatching in Slovene. The problem, although superficially simple, involves an analysis of the entire ‘kar’-phrase construction, which as it turns out, is not so straightforward. While textbooks and grammars have traditionally referred to these as ‘relative clauses’ by virtue of their function and their similarity to ‘ki’-type relative clauses, an analysis of them as relative clauses will be fraught with problems, unless there emerges an entirely new analysis of them. Chomsky (1995) says “we still have no good phrase structure theory for such simple matters as attributive adjectives, relative clauses, and adjuncts of many different types.” Perhaps under a more com- prehensive analysis of the structure of relative clauses, the behavior of ‘kar’-phrases will fall into place. The second (‘pseudo-relative’) approach taken appears to be the more explana- tory model. It does not require a great number of mechanical stipulations and it ac- 300 Slovenski jezik – Slovene Linguistic Studies 6 (2007) counts for the data fairly well. It, too, has some drawbacks, but they are few and far less serious in comparison with the problems of the relative clause analysis. While there are still a number of issues to be addressed, it seems that the pseudo-relative approach remains a promising means to a solution. Bibliography Albretti, Andrea. 1995. Colloquial Slovene: A Complete Language Course. London: Routledge. Alexiadou, Artemis, Paul Law, André Meinunger, and Chris Wilder. 2000. Introduc- tion. The Syntax of Relative Clauses: 1–51. Amsterdam, Philadelphia: John Ben- jamins. Andrews, Avery. 1985. Studies in the Syntax of Relative and Comparative Clauses. New York: Garland. Bianchi, Valentina. 1999. Consequences of Antisymmetry: Headed Relative Clauses. Berlin: Mouton de Gruyter. Bianchi, Valentina. 2004. Resumptive Relatives and LF Chain: The Structure of CP and IP. The Cartography of Syntactic Structures, Vol. 2., edited by L. Rizzi: 76– 114. Oxford: Oxford University Press. Chierchia, Gennaro and Sally McConnell-Genet. 1990. Meaning and Grammar. Cambridge: MIT Press. Chomsky, Noam.1995. The Minimalist Program. Cambridge: MIT Press. De Hoop, Helen. 2003. Partitivity. The Second Glot International State-of-the-Ar- ticle Book. The Latest in Linguistics, edited by Lisa Cheng and Rint Sybesma: 179–212. Berlin: Mouton de Gruyter. Derbyshire, William W. 1993. Basic Reference Grammar of Slovene. Columbus, Ohio: Slavica. Heim, Irene and Angelika Kratzer. 2005. Semantics in Generative Grammar. Malden, Massachusetts: Blackwell. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Lencek, Rado L. 1982. The Structure and History of the Slovene Language. Colum- bus, Ohio: Slavica. Rowling, J.K. 2003. Harry Potter in jetnik iz Azkabana. Trans. Jakob J. Kenda. Lju- bljana, Slovenia: EPTA. [Original title: Harry Potter and the Prisoner of Azka- ban, 1999.] Toporišič, Jože. 2000. Slovenska slovnica. Maribor: Obzorja. Prispelo oktobra 2006, sprejeto januarja 2007 Received October 2006, accepted January 2007 Pravi in nepravi oziralni stavki v slovenščini Slovenščina ima več vrst oziralnih stavkov, ki univerzalni slovnici lahko pomagajo osvetliti zgradbo oziralnih stavkov nasploh. Tako ima različne slovnične konstrukcije glede na to, ali se oziralni stavek nanaša na argument ali adjunkt, za Vrinda Subhalaxmi Chidambaram, Relative and Pseudo-Relative Clauses in Slovene 301 proste oziralne stavke in za korelative. Poleg tega ima tudi posebeno konstrukcijo, ki se uporablja samo skupaj s presežnikom, to je kar-stavek. To konstrukcijo so vse dosedanje slovenske slovnice štele med oziralne stavke, jezikoslovci je pa niso po- drobneje obravnavali ne v okviru tvorbene slovnice ne kako drugače. Za analizo kar- stavka imamo dve možnosti: štejemo ga lahko za oziralni stavek ali pa za neki drug tip stavka. Če ga želimo analizirati kot oziralni stavek, moramo v opis uvesti obsežen dodaten sintaktični tehnični aparat. Ta rešitev ni elegantna in na koncu nekateri pro- blemi še vedno ostanejo odprti. Če se odločimo za drugo pot, predlagam, da kar-sta- vek namesto za oziralni stavek štejemo za partitivni vsebinski odvisnik k presežniku. Na ta način z lahkoto odkrijemo dejstva o kar. Zaimek vloženega stavka razumemo kot spremenljivko, ki jo presežnik kvantificira. Ker je vloženi vsebinski odvisnik par- titiven, mora biti samostalniška besedna zveza v njem določna in v množini (de Hoop 2003). S tem smo razložili, zakaj je zaimek v množinski obliki. Še vedno pa, kot je prikazano v točki 7, nekaj vprašanj v zvezi s to analizo in samo zgradbo ostaja odprtih za nadaljnje raziskovanje. Relative and Pseudo-Relative Clauses in Slovene The Slovene language has a rich variety of relative clauses types, which could shed light on the general structure of relative clauses in terms of a UG analysis. It has specific constructions for relativizing arguments, for relativizing adjuncts, for free relatives and correlatives. Moreover, it has a construction that is used exclusively with the superlative, namely the kar-phrase. This construction has been classified as a relative clause by every grammar of the Slovene language to date, and there has not been any linguistic analysis of it, in a generative framework or otherwise. There are two possibilities to explore with the kar-phrase: either it is a relative clause or it is some other type of clause. In order to analyze this construction as a relative clause, it is necessary to posit a large amount of extra technical syntactic machinery so that the facts about the kar-phrase fall out. This solution was inelegant and ultimately still problematic. In the second solution, I propose that the kar-phrase is not a relative clause but a partitive CP selected by the superlative degree word. From this, the facts about kar fell out naturally. The pronoun in the embedded clause is analyzed as the variable that the superlative quantifies over. And as the embedded CP is partitive, the NP within it must be plural and definite (de Hoop 2003), thus explaining why the pro- noun appears in the plural. Yet, as shown in part 7, some issues regarding this analysis and these constructions in general still remain for further research.