Abstract

On the basis of binding facts, I argue that Serbo-Croatian (SC) does not project DP and that DP is not a universal property of language. I show that a number of binding contrasts between English and SC follow straightforwardly from independently motivated differences in their nominal structure, most notably from the assumption that DP is present only in English. I also discuss in detail the potential significance of this puzzling set of facts for the binding theory in general. Specifically, I propose that SC employs Condition C as defined in Lasnik 1989 and, in addition to the core binding conditions, a competitive mechanism adopted from Safir 2004, which regulates the distribution of reflexives, pronouns, and R-expressions. I also argue that the binding domains for pronouns and reflexives in SC need to be formulated differently.

1 Introduction

Among the interesting issues raised by the study of Serbo-Croatian (SC), and Slavic languages more generally, is the extent to which they pose a challenge to certain claims made about Universal Grammar that are based on the study of non-Slavic languages. One such claim, much discussed in the literature, is that DP is a universal projection and that all languages, including articleless languages like SC and most other Slavic languages, have overtly or covertly realized DP. Thus, the proponents of the so-called Universal DP Hypothesis (UDPH) argue that the structure of NP is universal, regardless of the presence/absence of overt articles in a language. According to this view, most notably represented by Bašić (2004) and Progovac (1998) for SC, the difference between languages with overt articles such as English and languages that lack articles such as SC is simply PF-based. That is, a D head exists even in languages like SC but it is not pronounced. For example, Bašić (2004:26) takes (1) to be the structure of the NP in SC.

(1)

  • graphic

  • ‘this talkative neighbor of his’

This position has not gone unchallenged, however. Authors like Baker (2003), Bošković (2005, 2008), Chierchia (1998), and Fukui (1988) have argued on independent grounds that DP is not a universal projection and that languages may differ with respect to whether they instantiate it. One of the most articulated proposals in this respect is made by Bošković (2005, 2008), who observes that languages without articles differ from languages with articles in surprising but quite systematic ways. The ways in which the two language groups consistently differ, according to Bošković (2008), are summarized in (2).1

(2)

  • a.

    ‘‘Left-branch extraction’’ is possible only in languages without articles.

  • b.

    ‘‘Adjunct extraction’’ is possible only in languages without articles.

  • c.

    (Japanese-style) scrambling is possible only in languages without articles.

  • d.

    Languages without articles disallow negative raising (i.e., strict negative polarity item licensing under negative raising), and languages with articles allow it.

  • e.

    Multiple wh-fronting languages without articles do not show superiority effects.

  • f.

    Clitic doubling is possible only in languages with articles.

  • g.

    Languages without articles do not allow transitive nominals with two genitives.

  • h.

    The majority superlative reading is possible only in languages with articles.

  • i.

    Head-internal relative clauses are island-sensitive in languages without articles, but not in those with articles.

  • j.

    Polysynthetic languages do not have articles.

Along the lines of Corver 1992, Bošković (2008) (see also Bošković 2005) proposes a DP/NP parameter whereby all of the noted differences are analyzed as a consequence of the lack of DP in languages without articles. According to this view, in languages without overt articles, the structure of the NPs is as in (3). Here, prenominal elements modifying the noun and agreeing with it in case, number, and gender are adjoined to NP.2

(3)

In this article, I present another argument, based on SC binding facts, in favor of the view of the second group of authors. I contend that the exactly opposite behavior of English and SC with respect to a number of binding phenomena can be straightforwardly accounted for under the assumption that DP is projected in English, but not in SC. I also show that this assumption goes a long way in explaining the complex binding situation in SC. My hope is that the new SC facts presented here will enable us to better comprehend the nature of the principles behind the binding theory in general. My goal, therefore, is to present the new SC data and a number of binding contrasts between English and SC, point out the relevance of these facts for the structure of NP, and then explore their consequences for the binding theory.

The article is structured as follows. In section 2, I present and discuss the noted binding facts, which are mainly related to the distribution of pronouns and R-expressions in SC, and use them to test predictions the above-mentioned two approaches make with respect to binding. I claim that only a view that assumes the lack of DP in SC, and allows prenominal modifiers to c-command out of their NPs, can handle the SC binding facts in a noncircular manner. I also show how my proposals neatly converge with other, independent facts from SC. In light of this discussion, in section 3, I examine implications of this analysis for the binding theory in general. Although the proposal that SC lacks DP is essential for explaining the main binding contrasts between English and SC, certain additional assumptions about general properties of the SC binding system are necessary in order to gain a clear picture of the full range of the facts in question. In particular, I argue that in SC Condition C should be defined as in Lasnik 1989 and that the binding domains for pronouns and anaphors in SC need to be formulated in different ways. In addition, I show that the competitive mechanism proposed by Safir (2004) is active in SC and that it can affect the language’s binding/coreference possibilities to a significant degree. I also discuss anti–subject orientation of pronouns and argue that SC facts support approaches on which the existence of anti–subject-oriented pronouns naturally follows from the distribution of subjectoriented anaphors (e.g., Burzio 1989, 1991, 1996, Hellan 1988, Safir 2004) and not from some independent principle of grammar.

2 The Universal DP Hypothesis and Binding

2.1 Universal DP Hypothesis: Background and Preliminaries

In this section, I will present a few puzzling binding paradigms from SC, which constitute the core of the article and which have not been given systematic attention in the literature so far. First, however, I will briefly go over some key theoretical underpinnings of the UDPH that are relevant to the discussion, in order to make my endpoint as clear as possible.

There are two arguments that proponents of the UDPH most commonly use in favor of the structure in (1) over the traditional NP analysis. First, only the structure in (1) directly derives the adjective-ordering restrictions from phrase structure, and doesn’t need to stipulate them by some external mechanism. Second, only (1) finds straightforward support in Kayne’s (1994) Antisymmetry view of syntax, since unlike the traditional NP-adjunction analysis of APs, (1) is compatible with Kayne’s approach, which allows only one specifier per projection and predicts that that specifier must be on the left.3

The first argument essentially comes from Cinque’s assumptions about phrase structure. Bašić (2004), for instance, follows Cinque (1994) in this respect and assumes that all attributive adjectives are generated in specifier positions of αPs, functional projections in the functional spine of DP. This is based on Cinque’s (1994) observation that the distribution of adjectives in NPs closely resembles the distribution of adverbs in VPs. The claim is that the strict ordering of adjectives in NPs reflects the fact that they are generated in specifiers of different, hierarchically ordered universal functional projections between D0 and NP, as shown in (4), a slightly expanded version of (1) (for a similar view, see Scott 2002 and references therein).

(4)

  • a.

    ovaj njegov veliki brbljivi sused

    this his big talkative neighbor

  • b.

    graphic

There are, however, some well-known general conceptual problems with this argument. For instance, as Bobaljik (1999) points out, taking the restrictions on adverbial/adjectival ordering to be a result of a fixed universal functional projection hierarchy in phrase structure leads to some nontrivial word order paradoxes, which necessarily leads to postulating multiple hierarchies and hence effectively diminishes the strength of the parsimony aspect of Cinque’s argument. Also, Bošković (2009) observes that the ordering restrictions on adjectives with respect to demonstratives and possessives can receive a principled account in terms of filtering effects of semantics. Bošković´ shows that possessives in SC stand in a freer ordering relation with respect to adjectives, in that they can both precede and follow them, whereas demonstratives necessarily precede both possessives and adjectives. Under the standard assumption that demonstratives are of type 《e,t〉,e〉, most adjectives are of type 〈e,t〉 ‭, and possessives are modificational, it is natural to assume that semantic composition requires demonstratives to be composed at the end, that is, after adjectives and possessives. Under this view, semantic composition essentially does not regulate the order of possessives and adjectives relative to each other in any way, which is consistent with the facts. However, while semantic composition allows possessives to be composed either after or before modifying adjectives, demonstratives must be composed after both possessives and adjectives, which overall matches the actual SC facts. The claim is then that since these ordering restrictions follow from semantic requirements, syntax can generate all the orders, but semantics will filter out the unacceptable ones. Bošković thus argues that the adjective-ordering restrictions follow directly from semantic composition and need not be imposed by syntax. Without going into any more details of the arguments for and against Cinque’s proposal, I will continue to assume that there is not enough conclusive evidence that assigning adjective-ordering restrictions to the phrase structure would be any less stipulative than analyzing them as a property of some syntax-external (semantic) mechanism (see also Ernst 2002 and Shaer 1998, among others, for arguments against Cinque’s view of adverbs, some of which can be extended to his treatment of adjectives).

The second argument is more directly relevant to the main research question of this article. For the theoretical argument about the position and number of specifiers per projection to carry weight, an account would need to adopt the Antisymmetry view of syntax entirely, with all the possible repercussions. In what follows, I show that adopting both the universal DP structure and the system proposed in Kayne 1994 is not tenable for SC. Furthermore, I show that a range of SC data raise serious problems for the UDPH, regardless of the Antisymmetry argument; that is, while these data should in principle not even exist under the UDPH, they are easily explained under the assumption that DP is not universal. Since under the UDPH the structure in (1) is the structure for NPs in both English and SC, the two languages are predicted not to exhibit any fundamental difference in their syntactic behavior. In the following section, I show that this prediction is not borne out and that the binding properties of English and SC differ quite systematically. I argue that methodologically and empirically, the most adequate way to account for these differences is to assume that DP is projected only in English. Such an approach, I claim, does not introduce unnecessary stipulations and is directly compatible with the above-mentioned crosslinguistic observations made by Boškovič (2008).

2.2 The Universal DP Hypothesis, Kayne 1994, and Serbo-Croatian

Assuming a standard DP structure as in (6) for English, the grammaticality of (5a–b) is as expected: being located in specifiers of subject DPs, the possessives hisi and Johni do not c-command Johni and himi, respectively, and thus do not violate Conditions C and B.

(5)

  • a.

    Hisi father considers Johni highly intelligent.

  • b.

    Johni’s father considers himi highly intelligent.

(6)

graphic

However, under Kayne’s Antisymmetry approach, specifiers are adjuncts and, by virtue of the definition of c-command given in (7), they c-command out of the category they are adjoined to/ are specifiers of.

(7) X c-commands Y iff X and Y are categories, X excludes Y, and every category that dominates X dominates Y. (X excludes Y if no segment of X dominates Y.)

Given this, (5a–b) would be incorrectly predicted to be ungrammatical under the structure in (6), since hisi and Johni are dominated only by a segment of the subject DP and therefore do c-command Johni and himi, violating Conditions C and B, respectively. To resolve this problem, Kayne makes two important assumptions. First, following Szabolcsi’s (1981, 1983, 1992) analysis of Hungarian possessives, Kayne assumes that the possessor is preceded by an independent D, much as in the Italian example in (8).

(8)

  • il mio libro

  • the my book

Kayne proposes that in English, too, the prenominal possessor is the specifier of a PossP, which in turn is dominated by a DP with a null D head, as in (9).

(9)

graphic

(5a–b) are then accounted for: the additional null DP projected above the possessor prevents hisi and Johni from c-commanding the coindexed elements outside the DP. Second, also following Szabolcsi, Kayne argues that the specifier of the null DP is exclusively an operator position, which—although essential to operator-variable binding of a pronoun—is irrelevant to Conditions A, B, and C of the binding theory. Kayne proposes that quantificational possessor phrases move up to this position at LF. Motivation for this movement comes from examples such as (10)–(11), where the QP every girl undergoes covert movement to the specifier of DP. Since from this position the QPs c-command the rest of the sentence, a bound variable interpretation of the pronoun she in (10) is legitimate. (11), on the other hand, is still excluded, since it is assumed that the operator cannot license a reflexive from this position (see Kayne 1994 and references therein for further details of the analysis).

(10) Every girl’s father thinks she is a genius.

(11) *Every girl’s father admires herself.

Returning to the question of how this relates to the structure of SC NPs, we see that (9) resembles (1) in one significant way: they both have a DP headed by a null D above the possessor. This projection plays a very important role in Kayne’s approach, since (i) it is necessary to explain the facts in (5a–b) in a way consistent with the assumption that ‘‘specifiers’’ c-command out of their projections and (ii) by making certain assumptions about the character of this projection’s specifier position, Kayne seems to be able to account for an interesting operator-variable paradigm in English within his framework.4 The question is then whether the DP headed by a null D in (1) plays a significant role in SC. If it does, and if the argument from Antisymmetry holds, we expect the binding facts in SC not to differ from those in English in any fundamental way; that is, the DP above the possessor should prevent illicit c-command relationships between the possessor and coindexed elements in the sentence. In this respect, consider the following SC constructions:

(12)

  • *Kusturicini najnoviji film gai je zaista razočarao.

  • Kusturica’s latest film him is really disappointed

  • ‘Kusturicai’s latest film really disappointed himi.’

(13)

  • *Njegovi najnoviji film je zaista razočarao Kusturicui.

  • his latest film is really disappointed Kusturica

  • ‘Hisi latest film really disappointed Kusturicai.’

(14)

  • *Jovanovi papagaj gai je juče ugrizao.

  • John’s parrot him is yesterday bitten

  • ‘Johni’s parrot bit himi yesterday.’

(15)

  • *Njegovi papagaj je juče ugrizao Jovanai.

  • his parrot is yesterday bitten John

  • ‘Hisi parrot bit Johni yesterday.’

(16)

  • *Markovai lopta gai je juče udarila u glavu.

  • Marko’s ball him is yesterday hit in head

  • ‘Markoi’s ball hit himi in the head yesterday.’

(17)

  • *Njegovai lopta je juče udarila Markai u glavu.

  • his ball is yesterday hit Marko in head

  • ‘Hisi ball hit Markoi in the head yesterday.’

These English and SC sentences clearly differ in acceptability. While all of the English examples are straightforward on the relevant readings (to the extent that backward anaphora of the sort illustrated in (13)/(15)/(17) is allowed in the language), none of the SC constructions are grammatical.5 This suggests that possessors in SC do c-command out of the subject NPs they are possessors of, and thus induce Condition B and C violations.6 If there were no essential difference between English and SC in the phrase structure of the nominal domain, and if the structure of SC NPs were as in (1), as the UDPH suggests, we would expect the two languages to behave similarly with respect to binding, contrary to fact. The data in (12)–(17), however, strongly suggest that there is no projection dominating the subject phrase that would block this illicit relation. To explain the contrast between SC and English, a UDPH approach to SC would have to make additional stipulations, and it would face serious difficulties in dealing with Boškovič’s generalizations (2a–j) in a principled manner. On the approach developed here, which is completely compatible with Bošković’s observations, the contrast in question comes for free and is a direct result of a deep structural difference between SC and English. I argue that unlike English, SC does not project a DP and that all prenominal modifiers (demonstratives, possessives, and adjectives) in this language are adjoined to the NP they modify.7 Since prenominal modifiers are dominated by segments (e.g., May 1985), they c-command out of their NPs (see (7)) and violate Conditions B and C in structures like (12)–(17). It is therefore important to note, in this respect, that both demonstratives and possessives are morphologically adjectival in SC, in that they agree with the noun they modify in case, number, and gender in the same way adjectives do. This is illustrated in (18) with respect to a partial case paradigm (see Bošković 2005 and Zlatić 1997 for details).

(18)

  • a.

    graphic

  • b.

    graphic

Moreover, SC possessives and demonstratives behave syntactically like adjectives in every respect, which is completely consistent with the proposed analysis (see Bošković 2005, 2010 and Zlatić 1997 for a number of arguments to this effect, which are based on the appearance of SC possessives and demonstratives in adjectival positions, stacking up, impossibility of modification, specificity effects, and so on; I return to this issue below).8

A particularly compelling argument against the UDPH analysis of SC comes in fact from constructions that involve both demonstratives and possessives. In order to account for the ungrammaticality of (12)–(17), one may argue for a ‘‘weaker’’ version of the UDPH. That is, it might be hypothesized that in languages like SC, DP is actually not always present, and that it is projected only when the specifier of DP (i.e., the demonstrative in (1)) is overtly realized. The prediction is then that (12)–(17) should improve significantly if the demonstrative is added to the subject NPs in these sentences. This, however, is not correct. Consider (19a–b), which are as unacceptable as (14) and (15).

(19)

  • a.

    graphic

    ‘This parrot of Johni’s bit himi yesterday.’

  • b.

    graphic

    ‘This parrot of hisi yesterday bit Johni.’

To be more precise, on this hypothetical, ‘‘weaker’’ variant of the UDPH approach to SC, the structure of the subject NP in (14) would be as in (20).

(20)

graphic

This modification of the UDPH would ultimately account for the unacceptability of (14). In particular, given Kayne’s proposal that specifiers c-command out of their phrases, (14) would violate Condition B since, by assumption, there would be no DP headed by a null D above the PossP in (20) that would prevent the object pronoun in (14) from being c-commanded by the possessor Jovanov ‘John’s’. By the same logic, the status of the rest of the paradigm in (12)–(17) would also be accounted for. The unacceptability of (19), however, directly challenges this alternative version of the UDPH. Since the demonstrative is overtly present in (19), which according to (1) should signal the presence of an underlying DP headed by a null D, (19) should be acceptable; that is, this DP should block the possessive from c-commanding into the structure and thus no binding violation should arise. However, (19) is as ungrammatical as (12)–(17), which clearly argues even against the alternative, ‘‘weaker’’ version of the UDPH analysis. The adjunct-based approach advanced here, on the other hand, predicts exactly this state of affairs. More precisely, adding a demonstrative to the subject in (12)–(17) should not affect the overall unacceptability of these constructions at all, since both the possessor and the demonstrative are adjuncts and they both c-command out of the subject NPs. Their relative order is here taken to be a result of semantic composition (not of intrinsically ordered functional projections), as discussed in the previous section (see also Bošković 2009). The same type of argument can be made with respect to adjectives. That is, as already pointed out, the order between adjectives and possessors in SC is relatively free, which would on at least some versions of the UDPH (see (1) and Cinque 1994, for instance) predict the absence of binding condition violations in cases where the possessor is linearly preceded by an adjective. More precisely, the functional projection αP that hosts the adjective in (21a) should be located above PossP by assumption (i.e., (21c)) and therefore prevent Jovanov ‘John’s’ from c-commanding the pronoun in the object position. Again, the expected contrast does not arise; such examples are as unacceptable as those in which the adjective follows the possessor, as shown in (21a–b), providing additional support for the view laid out here.

(21)

  • a.

    graphic

  • b.

    graphic

    ‘Johni’s favorite parrot bit himi yesterday.’

  • c.

    graphic

Now, it is important to show that (12)–(17) violate binding conditions and not something else. When the possessive clearly does not c-command the element coindexed with it, as in (22), no binding condition violations arise.

(22)

  • a.

    Papagaj kog je Jovani oduvek najviše voleo gai je juče ugrizao.

    parrot who is John always most loved him is yesterday bitten

    ‘The parrot which Johni has always loved the most bit himi yesterday.’

  • b.

    Onaj ko voli njegovogi papagaja voli i Jovanai.

    that who loves his parrot loves and John

    ‘The one who loves hisi parrot loves Johni as well.’

In the preceding pages, I have closely examined the often-cited Antisymmetry argument in favor of the UDPH, according to which a structure like (1) is conceptually superior to the traditional NP-adjunction view of the SC NP, and I have illustrated how (12)–(17) challenge it. The inevitable conclusion seems to be that (1), despite its elegance and appeal, requires some radical modification in order to deal convincingly with the facts. Although this discussion has focused on the Antisymmetry-based version of the UDPH, which deserves special attention since it makes a number of interesting predictions, it is important to point out that (12)–(17) are equally problematic for approaches that propose a more ‘‘standard’’ DP structure for SC. More specifically, even if we assume (6) (repeated here as (23)) as the structure of the SC DP, in which the possessive is located in the specifier of DP and does not c-command the element coindexed with it, (12)–(17) would still incorrectly be predicted to be grammatical.

(23)

graphic

Thus, irrespective of the Antisymmetry argument, the binding contrast between English and SC is quite puzzling from the UDPH perspective in general and in principle should not exist. On the other hand, this contrast is directly compatible with approaches that argue that DP is not universal since it is easily derivable from the structural ‘‘deficiency’’ of the SC NP; that is, the two main factors underlying this contrast are that (i) SC possessors c-command out of the phrase they modify because they are adjuncts, and (ii) there is no DP layer above NP in SC that would block possessors from c-commanding out of their NPs. Both assumptions are supported by strong independent evidence. For instance, assumption (ii) is supported by the fact that SC behaves like a typical DP-less language with respect to Bošković’s (2008) crosslinguistic generalizations: among other things, in contrast to English it is a scrambling language that allows left-branch extraction and adjunct extraction from NPs. The key factor underlying many of these properties, which I do not go into here for reasons of space, is the absence of DP (see Bošković 2008, 2010 for details). Assumption (i) is supported by a number of independent, language-specific properties of SC, as explicitly argued by other authors as well (see, e.g., Bošković 2005, Zlatić 1997). As an illustration, consider the following interesting characteristic of SC possessives:

(24)

  • a.

    *lepi čovekov pas

    beautiful man’s dog

    graphic

  • b.

    *taj dečakov pas

    that boy’s dog

    graphic

  • c.

    *Jovanov bratov pas

    John’s brother’s dog

    ‘John’s brother’s dog’

Unlike in English (and many other languages), in SC possessives cannot be modified by other possessives, demonstratives, or adjectives, as shown in (24). Thus, in (24a) the adjective lepi ‘beautiful’ can modify only the head noun pas ‘dog’, not the possessor čovekov ‘man’s’. Similarly, it is impossible for the demonstrative taj ‘that’ to modify the possessor in (24b); it can only pick out the noun pas ‘dog’. Finally, (24c) shows that a possessor cannot be further modified by another possessor, a construction that is, of course, perfectly fine in English. Recall, at the same time, that all of these elements are adjectival in the sense that they agree with the noun they modify in case, number, and gender, and that they are argued here to be adjuncts. While (24) is surprising for any UDPH approach to SC, the present analysis accounts for it straightforwardly on the rather natural assumption that adjunction to adjuncts is impossible, as proposed and discussed in many places (e.g., Chomsky 1986a, Saito 1994, Takahashi 1994). This, on the other hand, converges quite neatly with the binding facts presented in this section, which I take to be further support for the general approach advocated here (for other arguments of this kind, see Despić 2011). It is also worth noting that I have argued in this section only against the uncompromising version of the UDPH: namely, that all languages have the same structure in the nominal domain and that the apparent overt differences reflect only PF phenomena. That is, I do not necessarily argue against the possibility that some functional structure may be projected above SC NPs; I argue only that positing null projections must be empirically justifiable.9 More generally, my aim in this section has been to provide an account that will unify a range of seemingly disparate phenomena by pointing out that whether or not a language has DP may often affect its other general properties to a significant degree. At the same time, I hope to have shown that by fleshing out details of the nominal structure in a language like SC, this type of approach may also shed light on certain aspects of the English DP, which would otherwise go unnoticed.

Now, although the lack of DP is one key factor in understanding the nature of binding in SC and plays an important role in explaining the core SC binding facts, it cannot by itself explain the full range of data. To gain a complete understanding of binding in SC, we must properly spell out certain additional assumptions about the SC binding system in general. In the next section, therefore, I discuss binding in SC in more detail and explore the implications of the novel SC facts presented there for binding in SC and the binding theory in general.

3 Binding in Serbo-Croatian

3.1 More Contrasts: Condition C, Coreference, and Coindexation

In light of the above discussion, a particularly interesting question lurks behind (12)–(17): how, in fact, do native speakers of SC express the meanings of these unacceptable constructions, which are otherwise fairly easily expressible in English? Given the status of (12)–(17), and in particular the claim that in SC possessors c-command out of the NPs they modify, it is expected that a construction like the one in (25) should, like (15), violate Condition C.

(25)

  • Jovanovi papagaj je juče ugrizao Jovanai.

  • John’s parrot is yesterday bitten John

  • ‘Johni’s parrot bit Johni yesterday.’

Somewhat unexpectedly, however, (25) is grammatical. This suggests that (25) does not violate Condition C. The contrast between (25) and (15) becomes even more puzzling in light of (26a–b), which, under the current analysis, involve the same c-command relation between the two Rexpressions as (25), yet are ungrammatical.

(26)

  • a.

    *Jovani je juče ugrizao Jovanai.

    John is yesterday bitten John

    ‘Johni bit Johni yesterday.’

  • b.

    *Jovani obožava Jovanai.

    John adores John

    ‘Johni adores Johni.’

Furthermore, (27a–b) and (28a–b) are more degraded than (26a–b).10

(27)

  • a.

    **Oni je juče ugrizao Jovanai.

    he is yesterday bitten John

    ‘Hei bit Johni yesterday.’

  • b.

    **Oni obožava Jovanai.

    he adores John

    ‘Hei adores Johni.’

(28)

  • a.

    **Jovani je juče ugrizao njegai.

    John is yesterday bitten him

    ‘Johni bit himi yesterday.’

  • b.

    **Jovani obožava njegai.

    John adores him

    ‘Johni adores himi.’

The data in (12)–(17) and (25)–(28) raise a number of nontrivial questions, and the challenge lies in answering all of them within a restricted and internally consistent set of assumptions that are, at the same time, in line with the conclusions and predictions of section 2. My proposal consists of three key parts, which I outline here so that my endpoint will be clear as I flesh out supporting arguments in the coming pages.

First, I adopt Lasnik’s (1989) more restricted version of Condition C. Second, I argue that binding domains for pronouns and anaphors in SC should be distinguished. In other words, I assume that the standard binding conditions apply in SC; that is, Conditions B and C are syntactic conditions, which rule out derivations not conforming to them.

Third, I argue that SC also employs Safir’s (2004) Form to Interpretation Principle (FTIP), which regulates the distribution of reflexives, pronouns, and R-expressions. The FTIP essentially determines whether what Safir calls a dependent identity reading is possible with respect to some designated antecedent and different dependent forms available in a given syntactic context. In a nutshell, the effect of the FTIP is that a more dependent form always outcompetes a less dependent form to represent the dependent identity reading. Thus, in any context in which more or less dependent forms are in competition, this principle predicts complementary distribution between them. Since one of the main goals of competition approaches to binding is to derive Conditions B and C from various competitive algorithms, the analysis presented here obviously contradicts the main tenets of such approaches, given that it requires Conditions B and C independently. However, I will argue that an analysis balanced exactly this way is required to account for the full range of binding facts in SC, although it might not appear very parsimonious or conceptually appealing.

Consider first Lasnik’s (1989) Condition C. Lasnik observes that Condition C effects vary crosslinguistically and that the variation is parametric in an interesting way. In Thai, for instance, sentences like (26a–b) are fully acceptable. However, if the subject R-expression is replaced by a pronoun, (26a) becomes impossible, as much as (27a) is impossible in SC. On this basis, Lasnik concludes that Condition C, unlike Conditions A and B, involves reference to both the binder and the bindee. His version of Condition C is given in (29).

(29) An R-expression is pronoun-free.

Taking this definition to hold for SC as well, we can now account for the difference between (26) and (27): that is, only (27a–b) violate Condition C, and even though (26a–b) are unacceptable, this cannot be due to a Condition C violation; rather, it must be due to something else. Note that the ungrammaticality of (13)/(15)/(17) (repeated here as (30)/(31)/(32)) is still accounted for under this revised formulation of Condition C.

(30)

  • *Njegovi najnoviji film je zaista razočarao Kusturicui.

  • his latest film is really disappointed Kusturica

  • ‘Hisi latest film really disappointed Kusturicai.’

(31)

  • *Njegovi papagaj je juče ugrizao Jovanai.

  • his parrot is yesterday bitten John

  • ‘Hisi parrot bit Johni yesterday.’

(32)

  • *Njegovai lopta je juče udarila Markai u glavu.

  • his ball is yesterday hit Marko in head

  • ‘Hisi ball hit Markoi in the head yesterday.’

Following this logic, we can also assume that (28a–b) violate Condition B. Two questions remain, however: what do (26a–b) violate, and depending on the answer to that question, why is (25) grammatical?

One (perhaps obvious) way to go about this question, which needs to be considered here briefly, would be to assume Reinhart’s (1983) well-known Rule I (see also Grodzinsky and Reinhart 1993).

(33)

  • Rule I

  • NP A cannot corefer with NP B if replacing A with C, C a variable A-bound by B, yields an indistinguishable interpretation.

  • (Grodzinsky and Reinhart 1993:79)

The central thesis of Reinhart’s proposal is that only one type of relation between coreferring elements is syntactically represented and constrained by principles of grammar: namely, the relation of variable binding in the sense of formal logic. On this approach, coindexation has only the bound variable interpretation. Coreference, on the other hand, is a type of semantic relation, which is not represented on any syntactic level and can therefore not be directly licensed or ruled out by structural conditions. The coreference interpretation is obtained when the two elements bear different indices; when they are coindexed, the bound interpretation is obligatory. However, since coreference defined this way is too strong and general and would make many undoubtedly ungrammatical sentences acceptable, Reinhart introduces Rule I (33) to limit its distribution. Briefly, the logic behind this principle is that if a structure could allow bound variable anaphora, coreference is preferred only if it is motivated—that is, only if it is distinguishable from bound anaphora. In structures where both coreference and coindexation are in principle possible, (33) has the effect of allowing coreference only in contexts where it is distinguishable from the bound interpretation. The basic idea is that in standard cases, the easiest way to express coreference is by means of variable binding. When this option is avoided without relevant motivation that would give rise to a distinguishable interpretation, a lack of coreference intention is inferred (see also Heim 1998 for a reinterpretation of Reinhart’s approach).

Following this logic, one could argue that (26b), repeated here as (34), violates only Rule I, not any binding conditions. In other words, what (34) seems to be expressing without additional context can already be easily expressed using a bound variable construction, where the lower Rexpression is replaced with a reflexive (e.g., (35)).

(34)

  • *Jovani obožava Jovanai.

  • John adores John

  • ‘Johni adores Johni.’

(35)

  • Jovani obožava sebei.

  • John adores self

  • ‘Johni adores himselfi.’

Without a suitable context that would license an interpretation distinguishable from the one in (35), (34) is ungrammatical. In a proper context like (36), however, (34) improves considerably.

(36)

  • Znam šta Ana, Milan, i Jovan imaju zajedničko. Ana obožava Jovana, Milan

  • I.know what Ana Milan and John have common Ana adores John Milan

  • obožava Jovana, a i Jovan obožava Jovana.

  • adores John but and John adores John

  • ‘I know what Ana, Milan, and John have in common. Ana adores John, Milan adores

  • John, and John adores John.

  • (adapted from Evans 1980:356)

The context in (36) establishes a property that is shared by Ana, Milan, and John. When applied to (34) in isolation, the property of adoring John is indistinguishable from the bound variable interpretation of adoring oneself (i.e., (35) – John (λx (x adores x))). When applied to (34) in the context of (36), however, the property shared by Ana, Milan, and John is only the property of adoring John and not the property of adoring oneself. Therefore, in the context of (36), which gives rise to a distinguishable interpretation, Reinhart’s Rule I does not apply and (34) becomes acceptable. The problem for this type of explanation, however, is the acceptability contrast between (34) and (27b) (repeated here as (37) and (38)).

(37)

  • *Jovani obožava Jovanai.

  • John adores John

  • ‘Johni adores Johni.’

(38)

  • **Oni obožava Jovanai.

  • he adores John

  • ‘Hei adores Johni.’

This acceptability contrast is also reflected in the fact that when (38) is used in the context of (36) (e.g., (39)), the coreference reading is more difficult to obtain.11

(39)

  • ?*Znam šta Ana, Milan, i Jovan imaju zajedničko. Ana obožava Jovana, Milan

  • I.know what Ana Milan and John have common Ana adores John Milan

  • obožava Jovana, a i on obožava Jovana.

  • adores John but and he adores John

  • ‘I know what Ana, Milan, and John have in common. Ana adores John, Milan adores

  • John, and he adores John.’

This does not follow from Reinhart’s assumptions, since Rule I is intended to completely replace Condition C as redundant, given that names in Reinhart’s theory are excluded wherever reflexives and pronouns are possible. Thus, in principle whether an R-expression is anteceded by a pronoun or another R-expression should be irrelevant; that is, such constructions should be equally unacceptable.

At the same time, unlike (34)/(37), which becomes available in contexts that force the coreferential reading, (40) does not require any extra context. This fact also does not follow from Reinhart’s theory.

(40)

  • Jovanovi papagaj je juče ugrizao Jovanai.

  • John’s parrot is yesterday bitten John

  • ‘Johni’s parrot bit Johni yesterday.’

It therefore appears that Rule I and the simple distinction between coindexation and coreference do not make the right cut here, even though the logic behind them is appealing and seems to be on the right track. In order to deal with the matter at hand adequately, a more sophisticated and explicit type of competitive approach is needed. I will now argue that Safir’s (2004) FTIP provides such an approach.

3.2 Safir 2004 and the Form to Interpretation Principle

One of the main aims of Safir’s (2004) system is to derive complementarity between different dependent forms via the Form to Interpretation Principle in (41) and the hierarchy of dependent forms in (42).

(41)

  • Form to Interpretation Principle (FTIP)

  • If x c-commands y, and z is not the most dependent form available in position y with respect to x, then y cannot be directly dependent on x.

(42) SIG-SELF > pronoun-SELF > SIG > pronoun > R-expression

The FTIP compares competing derivations based on alternative numerations containing forms that are not equally dependent. Thus, a numeration containing the forms he, loves, him will result in the simplified LF in (43b). Since English has a form that is more dependent than pronoun in the hierarchy in (42) (i.e., the pronoun-SELF), a competing derivation will be the one in (44), which is based on a numeration containing he, loves, himself.

(43)

  • a.

    Numeration: he, loves, him

  • b.

    LF: [he [loves him]]

(44)

  • a.

    Numeration: he, loves, himself

  • b.

    LF: [he [loves himself ]]

Since the comparison determines that him is not the most dependent form available in object position, the FTIP determines that the pronoun cannot be dependent on (i.e., coindexed with, in Reinhart’s terms) the subject in (43b).

I argue in this section that in terms of the division of labor between different mechanisms governing the binding system in SC, the empirically most accurate approach is to assume that in addition to the standard Conditions B and C, SC employs the FTIP. Consider first how this proposal accounts for the acceptability of (45), which is given here for comparison with its full alternative paradigm.

(45)

  • Jovanovi papagaj je juče ugrizao Jovanai.

  • John’s parrot is yesterday bitten John

  • ‘Johni’s parrot bit Johni yesterday.’

(46)

  • *Jovanovi papagaj je juče ugrizao sebei.

  • John’s parrot is yesterday bitten self

  • ‘Johni’s parrot bit himselfi yesterday.’

(47)

  • *Jovanovi papagaj gai je juče ugrizao.

  • John’s parrot him is yesterday bitten

  • ‘Johni’s parrot bit himi yesterday.’

(48)

  • *Svoji papagaj je juče ugrizao Jovanai.

  • self’s parrot is yesterday bitten John

  • ‘Himselfi’s parrot bit Johni yesterday.’

(49)

  • *Njegovi papagaj je juče ugrizao Jovanai.

  • his parrot is yesterday bitten John

  • ‘Hisi parrot bit Johni yesterday.’

To make the argument explicit, I need to say a few words concerning the nature of SC reflexives. SC uses two kinds of reflexive pronouns: sebe and svoj. Both sebe and svoj are generalized to all persons. The possessive form svoj takes on various forms since it always agrees with the noun it modifies in gender, number, and case. Most importantly, for the purposes of this discussion, these two reflexive pronouns are similar to Norwegian seg selv and Japanese zibun-zisin in that they are strictly subject-oriented and local. As illustrated in (50), sebe and svoj can be anteceded only by a local subject.12

(50)

  • a.

    Jovani je pričao Markuj o sebii/*j/svomi/*j bratu.

    John is talked Marko about self self’s brother

    ‘John told Marko about himself/his brother.’

  • b.

    Jovani je rekao da je Markoj video sebe*i/j/svog*i/j brata.

    John is told that is Marko seen self self’s brother

    ‘John said that Marko saw himself/his brother.’

Given these properties of the SC reflexives, we can assume that (46) is ungrammatical because sebe is strictly subject-oriented and cannot be anteceded by the possessor of the subject (which on this account is an adjunct). (47), on the other hand, is a Condition B violation, as discussed in section 2. (48) is a Condition A violation, and (49) a Condition C violation (assuming Lasnik’s (1989) definition of Condition C). So, all the potential alternatives to (45) that would involve a pronoun or a reflexive are excluded on independent grounds. This is, however, not true for (26a) (repeated here as (51)).

(51)

  • *Jovani je juče ugrizao Jovanai.

  • John is yesterday bitten John

  • ‘Johni bit Johni yesterday.’

(52)

  • Jovani je juče ugrizao sebei.

  • John is yesterday bitten self

  • ‘Johni bit himselfi yesterday.’

Like (45), (51) does not violate Conditions B and C, but unlike (45), it does have a successful potential alternative that involves the reflexive sebe (i.e., (52)). This suggests that (45) is grammatical because all of its alternatives with reflexives or pronouns (i.e., (46)–(49)) are ungrammatical, while (51) is unacceptable because there exists a grammatical alternative to it.

Furthermore, with this type of approach we can explain the acceptability contrast between (51) and (53)–(54).

(53)

  • **Oni je ugrizao Jovanai.

  • he is bitten John

  • ‘Hei bit Johni.’

(54)

  • **Jovani je ugrizao njegai.

  • John is bitten him

  • ‘Johni bit himi.’

In contrast to (51), which violates only the FTIP, (53) and (54) violate the FTIP and a binding condition each; that is, (53) violates Condition C and the FTIP (the alternative with the reflexive sebe in place of Jovana ‘John’ is available), while (54) violates Condition B and the FTIP (njega ‘him’ can also be successfully replaced with sebe). Thus, since it violates only the competition principle, (51) is less degraded than (53)–(54).13 The hierarchy of dependent elements in SC would therefore include only three elements:

(55) sebe > pronoun > R-expression

The most highly dependent element in (55) is the reflexive sebe, which is a local and strictly subject-oriented anaphor, and which in this sense corresponds to SIG-SELF in (42) (there are no pronoun-SELF and SIG-type anaphors in SC).

Thus, instead of mandating separate domains for each dependent form in such a way that complementary distribution between them is accidental, Safir (2004) develops a system in which the complementarity is derived by principles that select the ‘‘best available’’ form-to-interpretation match. For instance, pronouns are (for the most part) excluded in exactly those environments where anaphors are available, and this complementary distribution was achieved in traditional approaches to binding by positing separate conditions (i.e., Conditions A and B). Safir’s approach, however, allows us to eliminate Condition B and its descendants as an independent principle regulating pronouns in the theory of anaphora. This aspect of Safir’s approach (and competitionbased binding theories in general) clearly contradicts my assumption that Condition B (as well as Condition C) is necessary to explain SC facts. That is, both Conditions B and C are necessary to exclude (47) and (49), respectively. While I believe that the principles behind Safir’s theory are universal, I will argue that (a particular version of ) Condition B is needed to account for the full set of facts in SC, and that the effects of the competition between pronouns and reflexives in this language are often obscured by binding conditions. In particular, I will show that exactly in cases in which neither pronouns nor reflexives violate binding conditions, the morphological form of the dependent element in question becomes crucial, as predicted by the FTIP. In the next section, I present these cases and justify my position with respect to Condition B in SC.

3.3 Condition B in Serbo-Croatian

It is certainly not controversial to assume that binding domains for anaphors and pronouns are not identical. It is well-known that theories that assume Conditions A and B to hold in the same domain and thus predict anaphors and pronouns to stand in complementary distribution (e.g., Chomsky 1981) are empirically challenged by the fact that across languages the predicted complementarity does not hold in a variety of configurations. A number of attempts have been made to distinguish the domains of ConditionA and B and at least partly resolve the problem of overlapping distribution. Some approaches defined ‘‘governing categories’’ differently for anaphors and pronouns (e.g., Chomsky 1986b, Huang 1983), while others argued that the domain of Condition B effects (i.e., ‘‘disjoint reference’’) should be formulated in terms of a predicate’s arguments (e.g., Hellan 1988, Sells 1986). Also, it has been proposed by many that the disjoint reference principle is sensitive to semantic interpretation of arguments (e.g., Pollard and Sag 1992, 1994, Reinhart and Reuland 1993, Williams 1994). For instance, Reinhart and Reuland (1993) specifically argue that Condition B should be formulated on semantic predicates and Condition A on syntactic predicates (see Kiparsky 2002 for further discussion of these matters).

Here, I offer evidence that domains for anaphors and pronouns in SC should also be formulated differently. Recall first that I argued in section 2 that a structure like (56) violates Condition B.

(56)

  • *Jovanovi papagaj gai je juče ugrizao.

  • John’s parrot him is yesterday bitten

  • ‘Johni’s parrot bit himi yesterday.’

In contrast to (56), (57) is perfectly acceptable; in this example, the object pronoun is embedded in an NP.

(57)

  • ✓Jovanovi papagaj je juče ugrizao njegovogi brata.

  • John’s parrot is yesterday bitten his brother

  • ‘Johni’s parrot bit hisi brother yesterday.’

On the other hand, example (59), in which the object R-expression is a possessive, is as unacceptable as (58).

(58)

  • *Njegovi papagaj je juče ugrizao Jovanai.

  • his parrot is yesterday bitten John

  • ‘Hisi parrot bit Johni yesterday.’

(59)

  • *Njegovi papagaj je juče ugrizao Jovanovogi brata.

  • his parrot is yesterday bitten John’s brother

  • ‘Hisi parrot bit Johni’s brother yesterday.’

Now, note again that the absence of DP in SC is what essentially creates this state of affairs and thus indirectly brings about a series of interesting questions about binding; that is, the contrasts observed in the constructions above would not exist if SC were like English since structures like (56)/(58) would be good. To account for these facts, I propose that SC employs the following version of Condition B, which essentially implies distinct binding domains for pronouns and reflexives:

(60)

  • Condition B

  • A pronoun is free in its own predicate domain (i.e., phrase). (An element is free if it is not c-commanded by a coindexed NP.)

According to (60), the pronoun in (56) is c-commanded by an element (i.e., the possessive) within its own predicate domain (i.e., the whole sentence). When the pronoun is embedded in an NP, as in (57), there is no Condition B violation since there is no element coindexed with the pronoun that c-commands it within that NP.14 This does not apply to (59) because Condition C, as defined here, is not sensitive to locality domains. At the same time, the pronominal possessive in (57) cannot be replaced by the reflexive possessive svoj, because svoj is strictly subject-oriented.

(61)

  • *Jovanovi papagaj je juče ugrizao svogi brata.

  • John’s parrot is yesterday bitten self’s brother

The question is then whether or not the acceptability of (57) should be related to the fact that (61) is impossible. Consider in this respect the following examples:

(62)

  • a.

    ??Jovani je udario njegovogi prijatelja.

    John is hit his friend

  • b.

    Jovani je udario svogi prijatelja.

    John is hit self’s friend

    ‘Johni hit hisi friend.’

There are two pieces of information that are important here. First, it is fairly well-known that native speakers of SC often produce constructions like (62a), even though traditional grammars argue that they are unacceptable (e.g., Stevanović 1962:97). However, native speakers never produce (63) with the indicated coindexation.

(63)

  • **Jovani je udario njegai.

  • John is hit him

  • ‘Johni hit himi.’

Second, constructions like (62a) become fully acceptable when the possessive pronoun is anteceded by a coordinated NP.

(64)

  • Fuji Heavy Industries Ltdi i Sumitomo Corpj su predstavili njihovi+j zajednički

  • Fuji Heavy Industries Ltd and Sumitomo Corp are introduced their joint

  • samostalni robotski sistem za čišćenje podova u Sumitomo zgradi u Osaki.

  • independent robotic system for cleaning floors in Sumitomo building in Osaka

  • ‘Fuji Heavy Industries Ltdi and Sumitomo Corpj introduced theiri+j joint independent

  • floor-cleaning robotic system in the Sumitomo building in Osaka.’

  • (http://www.otpornik.info/zanimljivosti/.../101-robot-usisivac.html)

The present approach accounts for (62a) quite naturally. (62a) does not violate Condition B, given the definition in (60), and its relative unacceptability is the result of a competition between reflexives and pronouns: namely, a more dependent form svoj ‘self’s’ is available in this construction and it does not outcompete the less dependent form njegov ‘his’. And exactly in cases like this, njegov becomes fully acceptable when the coreference reading is forced. Consider the following examples from Marelj 2011:207:

(65)

  • a.

    graphic

    Laurence hates his neighbor but and Tristram too

  • b.

    graphic

    Laurence hates self’s neighbor but and Tristram too

    ‘Laurence hates his neighbor and Tristram does too.’

The strict reading, indicating coreference (or ‘‘covaluation’’ in Marelj’s terms), arises with the use of njegov in (65a), and the sloppy reading, indicating coindexation, is restricted to the use of svoj in (65b). (66) exhibits similar effects (Marelj 2011:208).

(66)

  • a.

    graphic

    only Lucie respects her husband

  • b.

    graphic

    only Lucie respects self’s husband

(66a) entails that other women do not respect Lucie’s husband, while (66b) entails that unlike Lucie, other women do not respect their own husbands. Thus, when the pronoun does not violate Condition B, it becomes perfectly available in contexts with coreferential interpretation—contexts in which reflexives are generally not available.

However, any approach that attempts to seriously investigate issues of pronoun/reflexive complementarity needs to accommodate cases of coreference in one way or another (see the discussion around (36) and (39)). It is well-established that overlaps in the distribution of pronominal and reflexive forms often involve the representation of distinct interpretations, and (65) and (66) are just another example of that.15 Structures like (64), on the other hand, are particularly interesting because they are not limited to coreference. That is, these structures allow pronouns in places in which reflexives are possible and at the same time they have bound variable interpretation.

(67) Context

  • Samo nekoliko autora je juče predstavilo svoje najnovije knjige. Recimo,

  • only few authors are yesterday presented self’s latest books for.instance

  • ‘Only a few authors presented their latest books yesterday. For instance,’

  • Čomskii i Lasnikj su predstavili njihovui+j najnoviju zajedničku knjigu dok

  • Chomsky and Lasnik are presented their latest joint book while

  • Polardk i Sagm nisu.

  • Pollard and Sag are.not

  • ‘Chomskyi and Lasnikj presented theiri+j latest ( joint) book while Pollardk and Sagm did not.’

In the first conjunct, the pronoun is assigned the same referent as ‘Chomsky and Lasnik’, whether it is bound by ‘Chomsky and Lasnik’ or coreferential with it. The interpretation of the pronoun in the elided VP is crucial, though. The elided njihovu ‘their’ can be assigned the same referent as ‘Pollard and Sag’; that is, the sentence can have the sloppy interpretation. In order to license ellipsis, I assume a ‘‘parallelism’’ requirement that the elided element be identical (in certain relevant respects) to the ‘‘antecedent’’ VP. Thus, (64) and (67) have bound variable interpretation and are not cases of obligatory coreference.

The crucial difference between (64)/(67) and (62a) is that the subjects in (64)/(67) are coordinated NPs and therefore interpreted as plural. SC reflexives sebe and svoj are underspecified for φ-features; for example, they do not have distinct singular and plural forms. SC pronouns, on the other hand, do have separate singular and plural forms (e.g., njegov ‘his’ and njihov ‘their’). I will assume that this morphological contrast makes SC pronouns much more accessible for the collective interpretation of the antecedent. At the same time, SC reflexives tend to support distributive readings.16 The adjective zajednički ‘joint’ in (64)/(67) unambiguously presupposes the collective reading of the subject antecedent and the pronominal form becomes clearly available. Safir’s (2004) approach is directly relevant for these examples, since one of its general goals is to explain why pronouns may express reflexive relationships if the morphology of a language has no dedicated reflexive form available. On this approach, if a language happens not to have a dedicated reflexive form, then by the FTIP the pronoun will display the familiar absence of Condition B effects. For instance, in Danish simple reflexives cannot have plural antecedents, and consequently a plural pronoun replaces the reflexive for the local bound reading, as predicted by the competitive theory (Safir 2004:72, originally from Vikner 1985).

(68)

  • a. John læste sin/*hans artikel.

  • John read SIN/his article

  • b. John og Mary l+ste *sine/deres artikler.

  • John and Mary read SIN/their paper

In Danish, the SIG form for possessives, sin, only obviates pronouns when its antecedent is singular, as in (68a). In (68b), sin is not acceptable and hence does not obviate the plural nonanaphoric pronoun.17

I believe that this analysis can successfully account for the SC facts in question as well. Since SC reflexives are underspecified for number, and since they strongly tend to support a distributive interpretation, the pronoun becomes available exactly when a collective interpretation is forced. In other words, because of their morphological simplicity (namely, the fact that they do not have plural forms), SC reflexives become irrelevant for the purposes of competition with pronouns when the antecedent has the collective reading. Collective interpretation does not, however, entail the lack of a bound variable interpretation, and it is therefore not surprising that the structure in (67) licenses the sloppy reading.

It is clear that SC facts support competition approaches to pronouns and reflexives, and the question is then whether the competition in question is sufficiently significant to derive Condition B as well, that being one of the ultimate goals of such approaches. I believe, given the facts discussed so far, that Condition B is a principle of its own in SC and that it cannot be dispensed with. At the same time, the data above strongly suggest that pronouns and reflexives do compete in this language and that Condition B often camouflages effects of the competition, which become visible exactly in situations in which Condition B is not violated. For instance, in contrast to (67), (69) is ungrammatical because it violates Condition B on this approach, which makes it impossible to conclude anything about the relation between pronouns and reflexives.

(69)

  • *Čomskii i Lasnikj su predstavili njihi+j (zajedno).

  • Chomsky and Lasnik are presented them (together)

  • ‘Chomsky and Lasnik presented themselves.’

One could possibly come up with a context that would support a non–bound variable reading and make this sentence (relatively) acceptable, but it would then be a case of coreference and would not tell us much about the principles that underlie the competition between anaphors and pronouns. Admittedly, the present analysis, which insists on the existence of both binding conditions and a competitive principle such as the FTIP in SC, does not seem very parsimonious and conceptually appealing, but it accounts for many fairly complex SC facts in a rather straightforward way.

As shown throughout this section and section 3.2, however, one nontrivial advantage of the present approach is that it provides an explicit means of dealing with different levels of (un)acceptability of a wide range of examples. If a language employs both the FTIP and the binding conditions, it is natural to expect that constructions that violate both of these principles should be judged worse than those that violate just one of them. For example, as already mentioned, (70b) violates both Condition B and the FTIP and is therefore worse than (70a), which violates only the latter.

(70)

  • a.

    ??Jovani je udario njegovogi prijatelja.

    John is hit his friend

    ‘Johni hit hisi friend.’

  • b.

    **Jovani je udario njegai.

    John is hit him

    ‘Johni hit himi.’

At the same time, structures violating just one condition should improve more easily in the right context than structures violating more than one. As already shown, this holds for cases in which only the FTIP is violated (e.g., (36)). However, structures that violate only, say, Condition B behave similarly. Given that binding conditions are irrelevant for coreference as long as there is enough pragmatic force to support interpretations distinguishable from the bound variable reading, it is expected that constructions like (71), which by assumption violate only Condition B (not the FTIP), should improve relatively easily in the right context. (72) illustrates this point.

(71)

  • *Jovanovi papagaj gai je juče ugrizao.

  • John’s parrot him is yesterday bitten

  • ‘Johni’s parrot bit himi yesterday.’

(72)

  • Znam šta Milanov magarac i Jovanov papagaj imaju zajedničko. Milanov

  • I.know what Milan’s donkey and John’s parrot have common Milan’s

  • magarac je juče ugrizao Jovana, a i Jovanov papagaj ga je ugrizao.

  • donkey is yesterday bitten John and but John’s parrot him is bitten

  • ‘I know what Milan’s donkey and John’s parrot have in common. Milan’s donkey bit

  • John yesterday and John’s parrot bit him too.’

On the bound reading, Jovanovpapagajgaje ugrizao would be interpreted as (John (λx (x’s parrot bit x))), which is clearly not the intended meaning of (72), in which John is bitten by both his own parrot and Milan’s donkey. Finally, although coreference is possible in special contexts, pragmatic accommodation of this sort is irrelevant for bound readings; that is, violations of the binding conditions and/or the competitive principle are always characterized by the lack of bound interpretation.

At this point, an interesting example brought to my attention by an anonymous reviewer should be addressed.

(73)

  • Jovan je razočaran. Njegov omiljeni papagaj ga je juče ugrizao.

  • John is disappointed his favorite parrot him is yesterday bitten

  • ‘John is disappointed. His favorite parrot bit him yesterday.’

It is possible in (73) for njegov ‘his’ and ga ‘him’ in the second sentence to refer to the same individual, John in this case, who is introduced in the preceding sentence. Yet it is not clear that this sentence should not violate Condition B (e.g., see (71)). Another interesting property of (73) is that even though njegov ‘his’ and ga ‘him’ can refer to the same individual, the structure in question does not allow the bound interpretation. Consider first (74).

(74)

  • Jovanovi papagaj je juče ugrizao njegovui majku, dok Markov papagaj nije.

  • John’s parrot is yesterday bitten his mother while Marko’s parrot is.not

  • ‘Johni’s parrot bit hisi mother yesterday, while Marko’s parrot did not.’

Here, the sloppy reading under which Marko’s parrot did not bite Marko’s mother is allowed. However, (75) does not license the sloppy reading.

(75)

  • Jovan je razočaran. Njegov papagaj ga je juče ugrizao, dok Markov

  • John is disappointed his parrot him is yesterday bitten while Marko’s

  • papagaj nije.

  • parrot is.not

  • ‘John is disappointed. His parrot bit him yesterday, while Marko’s parrot did not.’

(75) minimally differs from (73), but the sloppy reading is not possible. That is, the only reading available here is that Marko’s parrot did not bite John. The sentence cannot mean that Marko’s parrot did not bite Marko. Although (73) is not predicted by the present system, the fact that this structure is good but at the same time limited to coreference (even without any special context) shows that it is quite exceptional and therefore may be a result of some independent principles of SC. In light of this, I leave a more detailed exploration of (73)/(75) for future work.18

Finally, the analysis developed here also accounts for the contrast between (76) and (77). Following Reinhart and Reuland (1993) (see also Marantz 1984), we can assume that the PPs in these two constructions are not of the same type: in (76) the pronoun and the antecedent are thematic arguments, whereas in (77) the PP is not selected by the verb; it is a separate predicate and forms a binding domain for the pronoun on its own.

(76)

  • **Jovani se raspravlja sa njimi.

  • John argues with him

  • ‘Johni argues with himi.’

(77)

  • ??Jovani je osetio zmiju nedge blizu njegai.

  • John is felt snake somewhere near him

  • ‘Johni felt a snake somewhere near himi.’

(76) violates both Condition B and the competitive principle, while (77) violates only the latter, since sebe ‘self’ is available. When the pronoun is embedded in an NP, as in (78), (76) significantly improves, since as predicted it no longer violates Condition B. (77), on the hand, does not violate Condition B to begin with and embedding the pronoun in an NP, as in (79), does not change its status significantly.

(78)

  • ??Jovani se raspravlja sa njegovimi ocem.

  • John argues with his father

  • ‘Johni argues with hisi father.’

(79)

  • ??Jovani je osetio zmiju nedge blizu njegovei kuče.

  • John is felt snake somewhere near his house

  • ‘Johni felt a snake somewhere near hisi house.’

3.4 Anti–Subject Orientation of Pronouns

The preceding discussion raises some issues regarding a frequent proposal that pronouns in SC and Slavic in general are ‘‘anti–subject-oriented’’ and that an independent principle of grammar is responsible for this. The anti–subject orientation of pronouns has been discussed by many authors (see, e.g., Burzio 1989, 1991, Hellan 1988, Hestvik 1992, Safir 2004, Vikner 1985), and the central empirical motivation for this proposal is that in many languages pronouns are required to be free from closest subjects whereas in English they are not. At the same time, in these languages pronouns may be anteceded by a subject if another subject or a tensed clause boundary intervenes, which seems to be true of SC and many other Slavic languages as well. Therefore, the term anti–subject orientation comes from the fact that there is no requirement that the pronoun be free from a higher object, even if this object is closer than the subject. On this approach, the fact that the pronoun in (80) cannot be anteceded by the subject is due to an independent principle that prevents the pronoun from being anteceded by the subject.

(80)

  • *Jovani je predstavio Marka njemui.

  • John is introduced Marko him

  • ‘Johni introduced Marko to himi.’

The anti–subject orientation proposal has two essential properties, each of which is falsified here: (i) pronouns cannot be anteceded by subjects, and (ii) the fact that they cannot be anteceded by subjects is completely independent from the distribution of reflexives. Structures like (64) immediately challenge (i) since pronouns can clearly be anteceded by subjects. Also, a pronoun can be anteceded by the subject exactly when the reflexive is unavailable, which argues against (ii). The distinction between (81) and (82) is useful: the pronoun competes with the reflexive only in (81), and exactly when the reflexive is excluded from the competition because of the subject orientation requirement, the sentence becomes acceptable, (82).

(81)

  • ??Jovani je juče ugrizao njegovogi brata.

  • John is yesterday bitten his brother

  • ‘Johni bit hisi brother yesterday.’

(82)

  • ✓Jovanovi papagaj je juče ugrizao njegovogi brata.

  • John’s parrot is yesterday bitten his brother

  • ‘Johni’s parrot bit hisi brother yesterday.’

This strongly suggests that the anti–subject orientation of pronouns is contingent on the availability of subject-oriented reflexives. Another fact that appears accidental under the anti–subject orientation view is that (83) and (84) differ in acceptability.

(83)

  • **Markoi voli njegai.

  • Marko loves him ‘

  • Markoi loves himi.’

(84)

  • ??Markoi voli njegovogi psa.

  • Marko loves his dog ‘

  • Markoi loves hisi dog.’

On the analysis advocated here, on the other hand, this contrast is not surprising. The SC facts discussed in this article thus provide strong support for approaches on which the anti–subject orientation of pronouns is essentially governed by the syntactic distribution of strictly subject-oriented anaphors (e.g., Burzio 1989, 1991, Hellan 1988, Safir 2004) and not by some independent principle.

4 Summary and Outlook

In this article, I have argued for the following points:

(85)

  • a.

    SC lacks DP.

  • b.

    SC possessors are adjuncts c-commanding out of the NP they modify.

  • c.

    SC employs Condition B, Condition C, and a competitive principle, which are defined as follows:

    • i.

      graphic

    • An R-expression is pronoun-free.

    • ii.

      Condition B

    • A pronoun is free in its own predicate domain (i.e., phrase). (An element is free if it is not c-commanded by a coindexed NP.)

    • iii.

      • A.

        graphic

      • If x c-commands y, and z is not the most dependent form available in position y with respect to x, then y cannot be directly dependent on x.

      • B.

        Hierarchy of dependent elements

      • SIG-SELF > pronoun-SELF > SIG > pronoun > R-expression

      • C.

        Hierarchy of dependent elements in SC

      • sebe > pronoun > R-expression

I have focused in particular on the interaction between (85a) and (85b), on the one hand, and (85c), on the other, by emphasizing the relevance of a number of binding contrasts between SC and English for the structure of their respective nominal domains. I have argued that the absence of DP in SC and the ability of its possessives to c-command out of the NP they modify are the key factors underlying these binding contrasts. It is important to clarify again that by arguing against the UDPH, I have not argued against the DP hypothesis in general; that is, my central argument is that the DP hypothesis does not apply to all languages and that this point of variation can, if properly investigated, elucidate the nature of a number of other, seemingly unrelated types of phenomena (in this particular case, binding). Also, the assumption that SC (as an articleless language) lacks DP, together with my general agenda, should not be mistaken for an attempt to claim that languages without articles completely lack any kind of functional projections in the nominal domain or that functional projections in general cannot be null (i.e., that they must have some morphological exponent). As in other, similar works (e.g., Baker 2003, Bošković 2005, 2008, 2010, Chierchia 1998, Despić 2011), my more general point has been to show that Universal Grammar offers a wider range of possibilities than suggested by the UDPH, where SC and English stand at opposite ends of the spectrum. One should, however, not take this proposal to imply that all DP-less languages should behave like SC with respect to binding, since the SC phenomena discussed here are also governed by the peculiar nature of prenominal (adjectival) possessives (which even within the Slavic family display significant variation (e.g., Corbett 1987)); that is, it is certainly possible that possessives in certain DP-less languages are not adjoined to NP, but occupy its specifier (or even complement) position.

In the second part of the article, I have focused on fleshing out the principles that underlie the binding properties of SC. I have argued that in addition to Conditions B and C, which rule out derivations not conforming to them, SC employs a competitive principle, Safir’s (2004) FTIP, which regulates the distribution of reflexives, pronouns, and R-expressions. I have proposed a version of Condition B for SC that implies distinct binding domains for pronouns and reflexives, and I have presented arguments in favor of Lasnik’s (1989) definition of Condition C.

Notes

For helpful discussion of the material presented here and related ideas, I am particularly grateful to Jonathan Bobaljik, Željko Bošković, Jon Gajewski, Jairo Nunes, Mamoru Saito, Yael Sharvit, Daiko Takahashi, Masahiko Takahashi, and Susi Wurmbrand, as well as the audience at FASL 17 (Yale University). The article has also benefited greatly from the comments of anonymous LI reviewers. For assistance with data collection, I thank Ðorđe Despić, Helena Despić, Igor Markićević, and especially Alen Bešić. This material is based in part upon work supported by the National Science Foundation under grant NSF SBE #0920888 (PI Bošković, Co-PI Gajewski). Any errors are my own responsibility.

1 See Bošković 2008, 2010 for detailed discussion (which I cannot go into here for reasons of space), including illustrations of the generalizations in (2) and precise definitions of the phenomena these generalizations refer to (e.g., what is meant by scrambling in (2c) is long-distance scrambling of the kind found in Japanese; see also Corver 1992 and Uriagereka 1988 regarding (2a)).

2 Alternatively, they can be treated as multiple specifiers of NP (see Bošković 2005 for a detailed discussion of this alternative).

3 This is on the assumption that adjectives do not take NPs as their own complements, as proposed by Abney (1987).

4 A number of plausible proposals about structures like (10) and (11) have been made, and my goal here is not to assess different theories; rather, it is to briefly summarize some of the arguments that, in Kayne’s analysis, motivate the structure in (9).

5 A written questionnaire was administered to 25 informants through electronic mail. Informants were asked to evaluate the sample sentences on a 5-point scale, ranging from totally unacceptable through three intermediate levels to fully acceptable. Grammaticality judgments collected in this survey directly support the claim made here. Twenty speakers found cases like (12)/(14)/(16) completely unacceptable, while 5 found them unnatural but possible in certain contexts. (In section 3, I discuss contexts in which these examples become acceptable, since it is an issue that is directly relevant to my proposal.) On the other hand, 24 speakers found cases like (13)/(15)/(17) completely unacceptable. One speaker found them marginally possible only in a context where njegov receives emphatic stress. The overall picture is that such constructions may become relatively acceptable with emphatic stress, suggesting that notions like contrastive focus/topic may affect grammaticality judgments to a certain degree. However, in out-of-the-blue contexts these constructions are clearly unacceptable in SC, unlike in English, and this is the contrast that this study focuses on.

6 It could be argued that the ungrammaticality of (12)/(14)/(16) might be due to the fact that the pronoun in these examples is a clitic, and that clitics usually refer to an already established discourse referent. The question is then whether the Condition B–like effects in constructions of this type really violate Condition B or some other, pragmatic principle. After all, Fiengo and Higginbotham (1981) observe that even in English (i) is ungrammatical when the pronoun him is unstressed and reduced to ’im.

(i) *Johni read [DP books about ’imi].

Also, pronominal clitics in general are known to sometimes alter binding possibilities (e.g., Kayne 2002), and it is generally accepted that there is no delay-of-Condition-B effect in language acquisition in languages with clitic pronouns (e.g., Escobar and Gavarró 2001, Hamann, Kowalski, and Philip 1997, McKee 1992, Padilla 1990). This phenomenon, sometimes called the clitic exemption effect, also seems to show that clitics (and weak pronouns) display exceptional behavior with respect to Condition B effects. However, the speakers I consulted (see footnote 5) found examples like (ii) equally ungrammatical (or even more): 21 of them found cases like (ii) completely unacceptable, while 4 found them marginally possible with emphatic stress on the pronoun njega. This almost exactly mirrors the (un)acceptability of (12)/(14)/(16) (see footnote 5). The issue of emphatically stressed pronouns does not arise in these examples, since clitics cannot bear (emphatic) stress.

(ii)

  • *Kusturicini najnoviji film je zaista razočarao njegai.

  • Kusturica’s latest film is really disappointed him

  • ‘Kusturicai’s latest film really disappointed himi.’

Note finally that the ungrammaticality of (12)–(17) cannot be due to the type of verb used; that is, these constructions are equally unacceptable even though razočarati ‘to disappoint’ is a psych verb, in contrast to ugristi ‘to bite’ and udariti ‘to hit’.

7 The proposal that SC NP modifiers are adjuncts is by no means new; see, for example, Bošković 2005 and Zlatić 1997.

8 See also Fukui 1988 for relevant discussion of Japanese.

9 For a detailed discussion of these matters, including alternative ways of avoiding the binding condition violations in (12)–(17), see Despić 2011:chap. 2 and 4; for a discussion of similar issues in Japanese, see Takahashi 2011.

10 When asked to compare cases like (26a–b) with cases like (27a–b) and (28a–b), my informants (see footnotes 5 and 6) reported the following judgments: 16 speakers found constructions like (26a–b) to be less degraded than (27a–b), while 17 speakers found them less degraded than (28a–b); the other informants did not find much difference between them. In general, the informants judged constructions like (26a–b) as unnatural and unacceptable, but no informant found them worse than the type of examples in (27)–(28). To indicate that they are in this respect worse than (26a–b), I mark (27a–b) and (28a–b) in a somewhat unconventional way with **.

11 The coreference reading could ultimately be available for (39), but it requires much more pragmatic force than (36).

12 A number of different types of proposals have been made to derive the strict subject orientation of certain reflexives (via movement, φ-feature underspecification, etc.), and most of them are directly compatible with the main points of this discussion; that is, I don’t see that my main points would be affected by making any particular choice among these approaches. See Despić 2011, Zlatić 1997, and references therein for Condition A and reflexives in SC; see also Despić 2011:chap. 3 for extensive discussion of a crosslinguistic correlation between definiteness marking and reflexive pronouns and its implications for the binding theory.

13 Note also that it is expected on this approach that examples like (51) and (37), which violate only the FTIP, would become more easily accessible in the right context (see (36)) than examples like (53) and (38), which violate both the FTIP and Condition C (e.g., (39)). I come back to this issue in section 3.3.

14 Probably the most compelling argument against a coargument approach to the binding theory concerns exceptional- Case-marking (ECM) constructions.

(i) *Johni believes himi to like Kathy.

An approach to Condition B violations based on a constraint on coreference between coarguments encounters difficulty with (i) because him is an argument of like, and John is an argument of believe: John and him are not coarguments, yet (i) is ungrammatical. Reinhart and Reuland (1993) argue that such cases are not Condition B violations per se, but violations of a separate syntactic condition on the formation of A-chains. It is impossible to evaluate the strength of this argument in SC, simply because SC lacks ECM (and more generally raising) infinitives.

15 See Safir 2004:sec. 3.3.3 for an overview of strategies for apparent noncomplementarity of distribution, which among other things includes cases in which interpretations are distinct.

16 See Avrutin 1994 for a discussion of similar examples in Russian.

17 Note that on the present approach there would be no Condition B violation in (68).

18 Given my assumptions about the FTIP, the right prediction here is that if no anaphor is possible as the object in the antecedent clause of the second sentence in (75), then (75) should permit a bound reading and hence a sloppy reading under ellipsis, contrary to fact. Thus, in this sense the exceptional behavior of (75) is a problem for my account, but no more than it is for any alternative approach (to the extent that any other approach could explain it together with all the other facts discussed in the article, within an internally consistent set of assumptions). As for why this construction has only the coreferential reading, I can at this point only speculate that this might have something to do with how pronouns are updated in discourse in languages like SC. It might be the case that when the R-expression Jovan is introduced in the first sentence of (75), both the possessive pronoun and the object pronoun of the next sentence establish a coreference relation with it independently, giving rise to the coreference interpretation. Adequately engaging this proposal, however, is a project I need to leave to future work; it remains open whether, or in what way, the exceptional behavior of (73)/ (75) reflects anything more fundamental.

References

Abney
,
Steven
.
1987
.
The English noun phrase in its sentential aspect
.
Doctoral dissertation, MIT, Cambridge, MA
.
Avrutin
,
Sergey
.
1994
.
Psycholinguistic investigations in the theory of reference
.
Doctoral dissertation, MIT, Cambridge, MA
.
Baker
,
Mark
.
2003
.
Lexical categories: Verbs, nouns, and adjectives
.
Cambridge
:
Cambridge University Press
.
Bašić
,
Monika
.
2004
.
Nominal subextractions and the structure of NPs in Serbian and English
.
Master’s thesis, University of Tromsø
.
Bobaljik
,
Jonathan David
.
1999
.
Adverbs: The hierarchy paradox
.
Glot International
4
:
27
28
.
Bošković
,
Željko
.
2005
.
On the locality of left branch extraction and the structure of NP
.
Studia Linguistica
59
:
1
45
.
Bošković
,
Željko
.
2008
.
What will you have, DP or NP?
In
NELS 37
, ed. by
Emily
Elfner
and
Martin
Walkow
,
101
114
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Bošković
,
Željko
.
2009
.
More on the no-DP analysis of article-less languages
.
Studia Linguistica
63
:
187
203
.
Bošković
,
Željko
.
2010
.
On NPs and clauses
.
Ms., University of Connecticut, Storrs
.
Burzio
,
Luigi
.
1989
.
On the non-existence of disjoint reference principles
.
Rivista di Grammatica Generativa
14
:
3
27
.
Burzio
,
Luigi
.
1991
.
The morphological basis of anaphora
.
Journal of Linguistics
27
:
81
105
.
Burzio
,
Luigi
.
1996
.
The role of the antecedent in anaphoric relations
. In
Current issues in comparative grammar
, ed. by
Robert
Freidin
,
1
45
.
Dordrecht
:
Kluwer
.
Chierchia
,
Guglielmo
.
1998
.
Reference to kinds across languages
.
Natural Language Semantics
6
:
339
405
.
Chomsky
,
Noam
.
1981
.
Lectures on government and binding
.
Dordrecht
:
Foris
.
Chomsky
,
Noam
.
1986a
.
Barriers
.
Cambridge, MA
:
MIT Press
.
Chomsky
,
Noam
.
1986b
.
Knowledge of language
.
New York
:
Praeger
.
Cinque
,
Guglielmo
.
1994
.
On the evidence for partial N-movement in the Romance DP
. In
Paths towards Universal Grammar
, ed. by
Guglielmo
Cinque
,
Jan
Koster
,
Jean-Yves
Pollock
,
Luigi
Rizzi
, and
Raffaella
Zanuttini
,
85
110
.
Washington, DC
:
Georgetown University Press
.
Corbett
,
Greville
.
1987
.
The morphology/syntax interface: Evidence from possessive adjectives in Slavonic
.
Language
63
:
299
345
.
Corver
,
Norbert
.
1992
.
Left branch extraction
. In
NELS 22
, ed. by
Kimberly
Broderick
,
67
84
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Despić
,
Miloje
.
2011
.
Syntax in the absence of Determiner Phrase
.
Doctoral dissertation, University of Connecticut, Storrs
.
Ernst
,
Thomas
.
2002
.
The syntax of adjuncts
.
Cambridge
:
Cambridge University Press
.
Escobar
,
Linda
, and
Anna
Gavarró
.
2001
.
The acquisition of clitics and strong pronouns in Catalan
. In
Clitics in phonology, morphology and syntax
, ed. by
Birgit
Gerlach
and
Janet
Grijzenhout
,
161
180
.
Amsterdam
:
John Benjamins
.
Evans
,
Gareth
.
1980
.
Pronouns
.
Linguistic Inquiry
11
:
337
362
.
Fiengo
,
Robert
, and
James
Higginbotham
.
1981
.
Opacity in NP
.
Linguistic Analysis
7
:
395
421
.
Fukui
,
Naoki
.
1988
.
Deriving the differences between English and Japanese
.
English Linguistics
5
:
249
270
.
Grodzinsky
,
Yosef
, and
Tanya
Reinhart
.
1993
.
The innateness of binding and coreference
.
Linguistic Inquiry
24
:
69
101
.
Hamann
,
Cornelia
,
Odette
Kowalski
, and
William
Philip
.
1997
.
The French ‘delay of Principle B’ effect
. In
Proceedings of the 21st annual Boston University Conference on Language Development
, ed. by
Elizabeth
Hughes
,
Mary
Hughes
, and
Annabel
Greenhill
,
205
219
.
Somerville, MA
:
Cascadilla Press
.
Heim
,
Irene
.
1998
.
Anaphora and semantic interpretation: A reinterpretation of Reinhart’s approach
. In
The interpretive tract
, ed. by
Uli
Sauerland
and
Orin
Percus
,
205
246
.
MIT Working Papers in Linguistics 25
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Hellan
,
Lars
.
1988
.
Anaphora in Norwegian and the theory of grammar
.
Dordrecht
:
Foris
.
Hestvik
,
Arild
.
1992
.
LF movement of pronouns and antisubject orientation
.
Linguistic Inquiry
23
:
557
594
.
Huang
,
C.-T. James
.
1983
.
A note on the binding theory
.
Linguistic Inquiry
14
:
554
561
.
Kayne
,
Richard
.
1994
.
The antisymmetry of syntax
.
Cambridge, MA
:
MIT Press
.
Kayne
,
Richard
.
2002
.
Pronouns and their antecedents
. In
Derivation and explanation in the Minimalist Program
, ed. by
Samuel David
Epstein
and
T. Daniel
Seely
,
133
166
.
Oxford
:
Blackwell
.
Kiparsky
,
Paul
.
2002
.
Disjoint reference and the typology of pronouns
. In
More than words: A festschrift for Dieter Wunderlich
, ed. by
Ingrid
Kaufmann
and
Barbara
Stiebels
,
179
226
.
Berlin
:
Akademie Verlag
.
Lasnik
,
Howard
.
1989
.
Essays on anaphora
.
Dordrecht
:
Kluwer
.
Marantz
,
Alec
.
1984
.
On the nature of grammatical relations
.
Cambridge, MA
:
MIT Press
.
Marelj
,
Marijana
.
2011
.
Bound-variable anaphora and Left Branch Condition
.
Syntax
14
:
205
229
.
May
,
Robert
.
1985
.
Logical Form: Its structure and derivation
.
Cambridge, MA
:
MIT Press
.
McKee
,
Cecile
.
1992
.
A comparison of pronouns and anaphors in Italian and English language acquisition
.
Language Acquisition
2
:
21
54
.
Padilla
,
Jose
.
1990
.
On the definition of binding domains in Spanish: Evidence from child language
.
Dordrecht
:
Kluwer
.
Pollard
,
Carl
, and
Ivan
Sag
.
1992
.
Anaphors in English and the scope of binding theory
.
Linguistic Inquiry
23
:
261
303
.
Pollard
,
Carl
, and
Ivan
Sag
.
1994
.
Head-Driven Phrase Structure Grammar
.
Stanford, CA
:
CSLI Publications
.
Progovac
,
Ljiljana
.
1998
.
Determiner phrase in a language without determiners
.
Journal of Linguistics
34
:
165
179
.
Reinhart
,
Tanya
.
1983
.
Anaphora and semantic interpretation
.
London
:
Croom Helm
.
Reinhart
,
Tanya
, and
Eric
Reuland
.
1993
.
Reflexivity
.
Linguistic Inquiry
24
:
657
720
.
Safir
,
Ken
.
2004
.
The syntax of anaphora
.
Oxford
:
Oxford University Press
.
Saito
,
Mamoru
.
1994
.
Additional-wh effects and the adjunction site theory
.
Journal of East Asian Linguistics
3
:
195
240
.
Scott
,
Gary-John
.
2002
.
Stacked adjectival modification and the structure of nominal phrases
. In
Functional structure in DP and IP
, ed. by
Guglielmo
Cinque
,
91
120
.
Oxford
:
Oxford University Press
.
Sells
,
Peter
.
1986
.
Coreference and bound anaphora: A restatement of the facts
. In
NELS 16
, ed. by
Stephen
Berman
,
Jae-Woong
Choe
, and
Joyce
McDonough
,
434
446
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Shaer
,
Benjamin
.
1998
.
Adverbials, functional structure, and restrictiveness
. In
NELS 28
, ed. by
Pius N.
Tamanji
and
Kiyomi
Kusumoto
,
391
408
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Stevanović
,
Mihailo
.
1962
.
Gramatika srpskohrvatskog jexika za više razrede gimnazije
.
Četvrto izdanje. Obod
:
Cetinje
.
Szabolcsi
,
Anna
.
1981
.
The possessive construction in Hungarian: A configurational category in a nonconfigurational language
.
Acta Linguistica Academiae Scientiarum Hungaricae
31
:
261
289
.
Szabolcsi
,
Anna
.
1983
.
The possessor that ran away from home
.
The Linguistic Review
3
:
89
102
.
Szabolcsi
,
Anna
.
1992
.
Subordination: Articles and complementizers
. In
The structure of Hungarian
, ed. by
István
Kenesei
and
Csaba
Pléh
,
123
137
.
Approaches to Hungarian 4
.
Szeged
:
JATE
.
Takahashi
,
Daiko
.
1994
.
Minimality of movement
.
Doctoral dissertation, University of Connecticut, Storrs
.
Takahashi
,
Masahiko
.
2011
.
Some theoretical consequences of case-marking in Japanese
.
Doctoral dissertation, University of Connecticut, Storrs
.
Uriagereka
,
Juan
.
1988
.
On government
.
Doctoral dissertation, University of Connecticut, Storrs
.
Vikner
,
Sten
.
1985
.
Parameters of binder and binding category in Danish
.
Working Papers in Scandinavian Syntax
23
:
1
61
.
Williams
,
Edwin
.
1994
.
Thematic structure in syntax
.
Cambridge, MA
:
MIT Press
.
Zlatić
,
Larisa
.
1997
.
The structure of the Serbian noun phrase
.
Doctoral dissertation, University of Texas, Austin
.