Steven Franks The Slovenian Orphan Accusative, Empty ... orphan accusative… · where the expected accusative is overridden by the genitive case: the genitive can be assigned instead

Steven FranksThe Slovenian Orphan Accusative, Empty Pronouns and Noun Phrase Structure¹

1 IntroductionThis contribution examines the Orphan Accusative (OA) construction found in Slovenian, with the aim of understanding its unique properties within an articu-lated model of noun phrase structure.

Perlmutter and Orešnik’s seminal 1973 paper, henceforth P&O, laid out the core OA facts, identified the issues raised, and isolated the factors necessary for an insightful analysis. Indeed, that paper was such a comprehensive study that it has yet to be superceded by any more complete account of the phenom-enon, and it remains the model for what little work on the construction that has been done since. Recently, however, Rappaport (2009) and Peti-Stantić (2009) reopened the discussion with interesting conference presentations about the OA. The former updated the account in P&O and proposed generalizations of their ideas; the latter offered some new empirical observations and raised additional questions about the construction. It is these three studies, although primarily P&O, that serve as my point of departure.

My renewed interest in the OA derives from two related concerns. In work such as Franks (1994, 2001), Franks and Pereltsvaig (2004) and Franks and Rudin (2005), I have explored the structure of the extended projection of the Slavic nominal domain. Any adequate analysis of the OA needs to be couched within that discussion. There is also an important question that Bošković’s recent work on typological differences between what he calls DP and NP languages brings to the fore: Where does Slovenian fit in terms of that dichotomy? Since I have suggested in Franks (2007) that Slovenian may be developing into a DP language, it may well be that the OA can be associated with the additional struc-ture afforded by the DP type. It is issues such as these that provide the conceptual context in which this paper has been written.

1 While I have always been fascinated by Perlmutter and Orešnik’s (1973) discussion of the Orphan Accusative, my interest was rekindled by two recent conference presentations in 2009, those of Gilbert Rappaport and Anita Peti-Stantić. I thank both these scholars for sharing their ideas and materials with me, as well as Lanko Marušič, Rok Žaucer, Tatjana Marvin, Don Reindl, and especially Peter Jurgec for help with the Slovenian data. I also thank Anastasia Giannakidou and Željko Bošković for valuable feedback on an earlier version of this paper. All misinterpreta-tions of the facts remain of course my own.

130 Steven Franks

2 The core phenomenon and its Slavic contextThis section presents the core OA facts, reviews the problems they pose for lin-guistic analysis and identifies the most likely direction for a solution.

2.1 Some basic Slovenian data

As in other languages, verbs canonically take accusative objects:

(1) a. Kateriacc kruhacc hočete? ‘Which bread do you want?’ b. Hočem beliacc kruhacc. ‘I want the white bread.’

(2) a. Kateroacc hišoacc hočete? ‘Which house do you want?’ b. Hočem novoacc hišoacc. ‘I want the new house.’

However, when the object noun is missing, the adjective appears in the genitive if the noun is masculine, as in (3a). With a missing feminine noun, as in (3b), only the accusative form is possible:

(3) a. Hočem belegagen/*beliacc. ‘I want the white one’. b. Hočem *novegen/novoacc. ‘I want the new one’.

It is this peculiar use of the genitive form in canonical accusative contexts such as (3a) which Perlmutter and Orešnik (1973) dubbed the “Orphan Accusative”.²

The OA is not restricted to post-verbal position, but can appy to any accusa-tive phrase; nor is it restricted to particular classes of adjectives. In (4a, b), for example, the OA is used after prepositions which govern the accusative, while (4c–e) show that any element with adjectival declension can give rise to the OA:

(4) a. Za kateregagen/*kateriacc si se odločila: za belegagen/*beliacc ali za modregagen/*modriacc? ‘Which one did you pick: the white one or a blue one?’

b. princip, v kateregagen/*kateriacc verjamem … ‘the principle in which I believe …’

c. Hočem tegagen. ‘I want this one’.

2 The Slovenian terminology is “navezovalni tožilnik” or “navezovalna končnica”.

The Slovenian Orphan Accusative, Empty Pronouns and Noun Phrase Structure 131

d. Hočem kateregagen koli. ‘I want any one’.

e. Hočem drugegagen. ‘I want another one’.

The OA never applies when the missing noun is feminine, dual or plural. Inter-estingly, when the missing noun is neuter the OA can apply. There is however some vacillation and speaker uncertainty. P&O state that both are possible in the example reproduced in (5a), while Rappaport (2009) cites (5b, c):³

(5) a. Kateroacc prosoacc hočete? Navadnegagen/navadnoacc. ‘Which millet do you want? Ordinary.’

b. določilo, ki je spolni odnos med dvema moškima obravnaval kot kaznivo dejanje, za kateregagen je bila predvidena zaporna kazen od enega do petih let, …. ‘… the amendment that treated sexual relations between two men as a criminal act for which a prison sentence from one to five years is provided, ….’ (ISZ ZRC SAZU corpus)

c. Se zaveda življenja, kakršnegagen živi. ‘(He) is conscious of the (kind of) life which (he is) leading.’ (Toporišič) These are the basic facts.

2.2 Some fundamental questions

The questions posed by P&O (1973:422) about the OA and repeated below are exactly what one should ask:(6) Perlmutter and Orešnik’s questions about the Orphan Accusative 1. Why is it that the special form found in the OA is not just an arbitrary

ending? 2. Why is the OA form the same as that of the genitive case? 3. Why do all constituents with adjectival endings, rather than just quanti-

fiers or demonstratives or just adjectives that refer to transient properties, have a special form for the OA?

4. Why is it that masculines have a special OA form and feminines do not? 5. Why is it that the accusative has a special form, rather than some other

case?

3 Unless otherwise indicated, I have retained translations from the indicated source.

132 Steven Franks

6. Why is there a special form in the absence of the head noun, and not some other environment?

7. Why Slovenian – what is particular to Slovenian in the OA phenomenon, and what is more general?

In their paper, P&O set out to address these questions and, to no mean degree, succeed in doing so. However, even if on revisiting P&O’s paper the inevitable conclusion is that they largely said it all in the first place,⁴ it is nonetheless a worthwhile exercise to add to this list of questions new ones that situate the issue within a more contemporary framework, in the hope of leading to new ways of understanding the Slovenian OA. In particular, we now have available a more explicit theory of phrase structure as well as a more explicit conception of types of “missing” material. This should enable us to approach P&O’s questions from a fresh perspective.

Like P&O, I am concerned with how the OA “works”. However, I believe that the leading question must be their second one and that any answer to this must be driven by solutions to the following more general questions:

(7) a. What is the internal structure of the extended nominal projection in Slove-nian?

b. What is the nature of the “missing” nominal part?

(7a) asks what projections one should posit above NP for Slovenian. While the vast amount of research into extended nominal projections over the past few decades has established the need for functional projections above NP, there is also considerable variation across grammars. So, for example, case features and specificity features can be instantiated not just on NP, but also on a distinct K(ase)P and/or a distinct D(eterminer)P;⁵ the names of the functional categories are not as important as their independent existence and their specific proper-ties. In Section 7 below I return to this variation, showing how the need for dif-ferent extended projections is compellingly demonstrated by the broad range of nominal structures in South Slavic. (7b) on the other hand is more of a question about how the OA is generated and, as such, is I believe far easier to answer. A variety of mechanisms exist to render material interpretable to Logical Form (LF) silent on the Phonetic Form (PF) side. In particular, we can wonder at what point in the derivation the adjectival material is “orphaned”. Is the missing nominal

4 And here I am basically in concurrence with Rappaport’s conclusion.5 This is of course reminiscent of variation with regard to the instantiation of verbal functional features such as agreement and tense within the clause.


material elided in the course of the derivation (and, if so, before, after, or as part of Spell–Out) or is it silent pro from the start (and, if so, what are the categorial and other properties of this pro)? I will eventually argue that, although the OA instantiates agreement with a proN, Slovenian also countenances nominal ellipsis – a fact which obfuscates the data since simple ellipsis does not introduces any agreement complexities.

Finally, if the reason why the form of the orphaned adjective(s) is genitive(-like) must be the leading question, then P&O’s last question of how Slovenian is different must be the closing one. Whatever our answers to (7) turn out to be, the reason Slovenian is special is surely going to have to be a consequence of those answers. So I will need in this paper also to examine how other languages – specifically, other South Slavic languages, since they are minimally distinct from Slovenian – differ such that the answers to (7) for Slovenian do not accidentally implicate them as well.

2.3 One obvious first step in formulating a coherent answer

In the Slavic languages there are well-established alternations between accusa-tive and genitive. These alternations are of two types. First, there are situations where the expected accusative is overridden by the genitive case: the genitive can be assigned instead of accusative under negation (the so-called genitive of nega-tion), with verbs which are able to govern either case, and in certain quantified structures (the so-called genitive of quantification). In these, the genitive extends across the paradigm and is associated with the LF of the structure (marking, e.g., scope, negation, quantification, intensionality or individuation/definiteness/specificity). Relevant to the discussion at hand is the second type of alternation. These are situations in which the expected accusative is replaced by a genitive morphological form, as in the “animacy” rule. This affects only certain nominal classes and seems to be morphologically rather than semantically driven. To see the difference between these two types of alternations compare Russian (8) and (9):⁶

(8) a. Ja ždu vašacc otvetacc. ‘I am waiting for an answer.’ [more likely specific] b. Ja ždu vašegogen otvetaacc. ‘I am waiting for an answer.’ [non-specific]

6 There is much discussion of the contrast in (8); see for example Kagan (2007).

134 Steven Franks

(9) a. Ja vižu vašuacc mamuacc. ‘I see your mom.’

b. Ja vižu vašegoacc/gen papuacc. ‘I see your dad.’

c. Ja vižu vašuacc mat’acc/nom. ‘I see your mother.’

d. Ja vižu vašegoacc/gen otcaacc/gen. ‘I see your father.’

e. Ja vižu vašacc/nom domacc/nom. ‘I see your house.’

Significantly, in (8b) the entire NP complement appears in the genitive case, whereas in (9b) and (9d) a genitive form appears even though the accusative is called for syntactically.⁷ Hence in (9b) we have what appears to be a mixed phrase, with a genitive suffix on the modifier even though the noun is unambiguously accusative. The flip side of this can be seen in (9c), where it is the adjective that is unambiguous. This paradigm is part of a more general morphological rule, tradi-tionally known as the Accusative Prediction Rule. P&O’s version is given below:

(10) Accusative Prediction Rule a. For animates, the accusative is like the genitive.

b. For inanimates, the accusative is like the nominative.

Clearly, whatever is going on in Slovenian must be somehow parasitic on some more general Slavic case phenomenon. I will concur with P&O that the OA exploits (10), that is, it uses the “genitive” morphological form in “accusative” syntactic contexts, rather than some semantically sensitive replacement of syn-tactic accusative by genitive.⁸ In short, the phenomenon reflects a paradigmatic rather than syntagmatic solution, whereby animate (and certain other) mascu-line nouns employ the morphological genitive in accusative contexts.

3 Brief review of some previous discussionsThis section recapitulates essential aspects of three discussions of the OA phe-nomenon.

7 In the examples in (9), the first case indicated is what is syntactically governed (always accusa-tive), while the second is the form that is employed (either nominative or genitive, depending on the rule in (10)).8 As discussed in Section 3.3, Peti-Stantić (2009) does however attempt to build an argument that semantics should be implicated in the OA.


3.1 Perlmutter and Orešnik (1973)

As stated, P&O’s original account offered what I believe to have been basically the right insight. Their account exploits the rule in (10). What is most important about the Accusative Prediction Rule is that it is only with nouns and adjectives that lack a distinct morphological accusative that the syntactic accusative is mapped onto morphological nominative or genitive. Crucially, then, this is not lexical syncretism per se but rather a morphological rule.⁹ The Accusative Pre-diction Rule in (10) is invoked if and only if the lexical entry of the item being sent to Spell–Out lacks an accusative form. The paradigm has a gap in it and (10) provides an instruction to the morphology about how to remedy the situation. The puzzle posed by the Slovenian OA is thus why (10a) overapplies, mapping adjectives which modify missing inanimate nouns into the genitive rather than the accusative form – as it would if the accusative N were actually present.

In this context, P&O conclude that the OA phenomenon must have to do with the form and interpretation of Slovenian pronouns. In particular, they observe the following two facts:

(11) a. Slovenian pronouns behave morphologically as if they were animate.

b. Slovenian pronouns allow an Identity of Sense (I/S) interpretation, not just the standard Identity of Reference (I/R) interpretation.

Fact (11a) refers to the ability of the Slovenian masculine accusative clitic pronoun ga to refer to an inanimate masculine (or neuter) noun, despite the fact that it is technically genitive-like in form. Although P&O do not emphasize this, coopting the genitive form appropriate for animates to inanimate accusative pronominals is a more general Slavic phenomenon,¹⁰ one which operates wherever the para-digm is missing a distinct accusative form. Consider, for example, Russian (12a), Slovak (12b), or Bulgarian (12c):¹¹

(12) a. Stane videl sinij avto/sinjuju mašinu i Tone tože ego/ee videl. [Russian] ‘Stane saw a blue car and Tone also saw it/*one.’

9 Hence, in case feature systems such as I have proposed in Franks (1995, 2002), the identity between nominative and accusative or genitive and accusative is not evidence for feature neu-tralization syncretism.10 As noted in the next subsection, Rappaport (2009) also makes this point in rejecting its spe-cial significance for Slovenian.11 Thanks to M. Shardakova and M. Shrager, M. Mullek and P. Kosta, and R. Slabakova and R. Pančeva, respectively, for confirming these Russian, Slovak, and Bulgarian judgments.

136 Steven Franks

b. Pavol videl modré auto/modrý automobil i Peter ho tiež videl. [Slovak] ‘Pavol saw a blue car and Peter also saw it/*one.’

c. Ivan vidja sin automobil i Petâr go vidja sâšto. [Bulgarian] ‘Ivan saw a blue car and Peter also saw it/*one.’

For Russian (12a) I provide masculine (indeclinable) avto as well as feminine mašinu, with masculine/neuter pronoun ego or feminine ee, respectively; for Slovak (12b) I provide neuter auto as well as masculine autombil, with pronomi-nal clitic ho able to refer back to either; Bulgarian (12c) just shows masculine autombil, with pronominal clitic go.

Fact (11b) is much more unusual. Ordinarily pronouns are interpreted as core-ferential with referring expressions in the sentence or discourse, i.e., they express specific tokens. This “strict” reading is what P&O call I/R. In Slovenian, however, clitic pronouns are also able to refer back to the unindividuated properties of the antecedent, i.e., they can also express the general type. This “sloppy” reading is what P&O call I/S.¹² As a consequence, pronouns in Slovenian are not necessarily referential, as shown by the data set in (13) from P&O:

(13) a. Stane je videl plav avto in tudi Tone ga je videl. ‘Stane saw a blue car and Tone also saw it/one.’

b. Stane ima rjav površnik in tudi Tone ga ima. ‘Stane has a brown overcoat and Tone also has one.’

c. Stane ima pametnega otroka in tudi Tone ga ima. ‘Stane has a smart child and Tone also has one.’

d. Stane ima pametno ženo in tudi Tone jo ima. ‘Stane has a smart wife and Tone also has one.’

As the translations indicate, the I/S interpretation corresponds to English one. The eventual analysis developed in Section 6 will capitalize on this point. For now, it suffices to observe that this special ability of Slovenian clitic pronouns to express the sense of the antecedent is typologically striking in that pronouns in other Slavic languages lack the I/S interpretation. Thus, in the parallel examples from Russian, Slovak, and Bulgarian given in (y), only the I/R interpretation is

12 I thank Ž. Bošković (p. c.) for pointing out that in the East Asian language literature pro is claimed to be NP ellipsis precisely because of the availability of the sloppy reading. See Bošković (2011) for some relevant arguments.


possible. As indicated, translations with one instead of it are not possible here – the I/S interpretation is absolutely unavailable for overt pronouns in these other languages.

P&O offer a specific exploitation of the Slovenian facts to derive the OA, as follows:

(14) a. The noun is replaced by an appropriate pronoun.

b. Concord takes place between the modifiers and the pronoun.

c. The pronoun is deleted, orphaning the modifiers.

This is a highly derivational system, with each rule feeding the next one. Since pronouns are animate, the result of (14a) is that concord in (14b) will lead to use of the genitive form of the modifiers, given the combination of (11a) and (10a). However, this only happens when the pronoun is ga. The reason is that, unlike in say Russian, ga is the only form in the Slovenian pronominal clitic paradigm where there is no accusative form distinct from the genitive. Elsewhere, the Accu-sative Prediction Rule is simply not invoked. Because ga is the masculine singular form, this is the place where (10) can apply, giving rise to the Orphan Accusative.

Of course, as in other Slavic languages, ga also serves as the neuter singular accusative/genitive pronoun. Hence one might expect the OA to arise with missing neuter nouns as well. And indeed, this is what P&O (pp. 436–8) report. They comment that speakers are uncertain about the accusative of neuter animates (e.g., both navadno dekle ‘ordinary girl’ and navadnega dekleta are accepted), since there is a “conflict between the generalization that neuter endings are like masculine endings in the singular … and the generalization that the accusative of neuters is like the nominative throughout the paradigm”. Neuter proso ‘millet’ induces the same confusion in (5a), suggesting that neuter and masculine ga must in some way be distinguished. The two are different, however, in that dekle ‘girl’ refers to a female entity, hence also conflicts with a second robust gener-alization, namely that words referring to females are grammatically feminine.¹³

13 Also, according to L. Marušič (p. c.), it is marginal to use ga to refer back to a female individ-ual, especially, as D. Reindl (p. c.) points out, when the antecedent is distant from the pronoun (although there is much variation):(i) Še nobeno dekle mi ni bilo tako všeč kot tole, zapomnu si jo/??ga bom za celo življenje. ‘I haven’t liked any girl so much as this one, I’ll remember her for my entire life.’

138 Steven Franks

3.2 Rappaport (2009)

In answer to their 7th question, P&O (p. 457) state that is the fact that pronouns are animate which renders Slovenian special: “… under the correct analysis the only thing that is particular to Slovenian is the marking [+animate] on pronouns”. Rappaport takes issue with this claim, pointing out that this is hardly unique to Slovenian. Rappaport is of course absolutely correct that this is generally true in Slavic,¹⁴ so that the animacy of pronouns per se can hardly be the key factor giving rise to the OA. He argues instead that the special property of Slovenian pronouns is their capacity to display I/S, as we saw in (13). This is I will argue an important but not the crucial part of the story.

Rappaport proposes a “lexical rule which changes the category of an adjec-tive to N and fills its external argument slot with the feature [pronominal]”. He further states that the “external argument” of an A is an NP not a DP, providing the structure in (15):

(15) DP D NP Adj NP ← External argument of adjective … N …

Note that this is the traditional structure, taking his “Adj” to be an AP. Rappaport concludes, and I concur, that the I/S possibility in addition to the standard I/R possibility shows that Slovenian pronouns can be either “NP level” or “DP level”. What then is the difference between NP and DP? Roughly put, DP serves to house referential information whereas NP serves to house lexical information. So the

14 There is another animacy effect on Slavic pronouns worth mentioning: whereas clitic pro-nouns can have either animate or inanimate antedecents, full pronouns are generally limited to referring back to animate antecedents; cf. e.g. Despić (this volume). Speakers of Polish, Czech, Croatian, Serbian, Slovenian, Macedonian and Bulgarian who I have consulted also report this correlation, although certain factors can ameliorate the strong preference to interpret full forms as implying animate antecedents. While this phenomenon deserves fuller investigation, since the relevant factors seem to be pragmatic and since I ultimately see the OA as morphologically driven, I do not attempt in this paper to relate the two animacy effects. There may, nonetheless, be some interesting connection to be exploited. As argued, below pronouns require a morpho-logical value for [±animate], so there may be something about the full form that leads to semantic animacy as well.


standard I/R interpretation derives from a pronoun having D properties, while the I/S interpretation derives from a pronoun having N properties.¹⁵

While this distinction is on the right track, I disagree about the implementa-tion. First, following Bošković (2005, 2008, 2009), not all referential nominal expressions are DPs. In Slavic languages such as Bosnian/Croatian/Serbian and Russian, he argues, they can be maximally NPs. These languages lack articles but still express semantic features of the extended nominal projection such as defi-niteness/specificity. Second, I will argue that Slavic overt clitic pronouns such as Slovenian ga are K (for Kase), rather than N. Third, the relationship between A and N in (15) is one in which the AP is an adjunct, not one in which the NP is an argument. Before presenting the analysis, however, I turn to one more recent study of the OA.

3.3 Peti-Stantić (2009)

Peti-Stantić’s talk is exploratory, her primary argument being that one should probe the Slovenian OA facts more carefully before deciding what is at the root of the phenomenon. This is definitely a valuable recommendation, since there is still an enormous amount of empirical work to be done. Another argument she makes is that Slovenian needs to be more carefully compared to neighbor-ing Croatian. This is also a valuable suggestion.¹⁶ Peti-Stantić also claims that various semantic factors are relevant to the viability of the OA, as well as the type of adjective. Unfortunately, I have found considerable disagreement about her data. I nonetheless summarize them below.

In her presentation Peti-Stantić makes several observations. First, Slovenian does not apply the animacy rule as restrictively as some other languages. There are various types of nouns that do not denote animates, but which function as if [+animate] so that (10a) rather than (10b) systematically applies. Peti-Stantić cites the following from Herrity (2000:34):

15 Clearly, the traditional term “pronoun” fails to express such a distinction. Items such as ga, it, one and silent pro are all pro-forms, but what they substitute for is an empirical matter. In Sec-tion 6.1 I argue that all pro-forms are heads, and as such can instantiate various part-of-speech categories.16 Indeed, given the transitional nature of dialect communities across the entire South Slavic area, correlation of properties such as I/S and the OA among Slovenian and Croatian speakers would be a fascinating research project.

140 Steven Franks

(16) a. names of cars: ford, opel, folksvagen, golf, mercedes, rolsrojs, fiat, reno b. diseases named after animals: rak, volk c. certain instruments and devices named after animals: petelin, francoz,

skobec, robot d. creative works named after their author: Rembrandt ‘a Rembrand’, Picasso

‘a Picasso’ e. names of chess pieces and playing cards: kmet, kralj, as, pagat, fant f. names of wines: vipavec, jeruzalemčan, bizeljčan g. names of mushrooms: jurček, goban, turek, ciganček h. names of sport teams: Partizan, Železničar i. certain terms for money: tisočak, stotak j. terms for dead persons or animals: mrtvec, pokojnik, mrlič, mrtvak,

piščanec k. a few isolated words: zmaj ‘kite’, konjiček ‘rocking horse’, metuljček

‘bowtie, butterfly’

For P&O, this apparent extension of [+animate] in Slovenian provides grist for their mill, (11a) being a similar extension. Peti-Stantić takes issue with the idea that (16) supports the claim that what is special about Slovenian is its cavalier use of animacy or that this has anything to do with why Slovenian developed the OA. She argues, first, that many of the examples in (16) reflect simple anthro-pomorphization and, second, that other Slavic languages show similar exten-sion of animacy (as reflected in the use of the genitive-accusative for inanimate nouns), but nonetheless lack the OA.¹⁷ This is of course all true, but does not I

17 Polish is notorious for extension of the genitive-accusative. Interestingly, it does optionally exhibit something like the Slovenian OA. When certain masculine singular inanimate nouns are orphaned, speakers allow genitive adjectives in accusative contexts:(i) a. Jan kupił nowyacc samochódacc. ‘Jan bought a new car.’ b. A ja chcę kupić staregogen. ‘And I want to buy an old (one).’(ii) a. Ja widzę dużyacc samochódacc. ‘I see a big car.’ b. A małegogen też widzisz? ‘And do you also see a small (one).’ Like Slovenian, Polish only does this with masculine singular nouns, which presumably relates to the fact that, in Polish as in Slovenian, the masculine singular clitic pronoun go is the only one that is neutral between genitive and accusative (i.e, there is no distinct accusative form so the genitive substitutes, regardless of animacy). Feminine ją, neuter je, and non-virile plural je are unambiguously accusative, the genitive forms being jej, go, and ich, respectively (virile ich is irrelevant since it invokes the Accusative Prediction Rule regardless). Like English one and unlike Slovenian, the Polish OA only applies when the missing nominal is non-specific, otherwise the accusative is used. I have no idea however why speakers accept it with some nouns and not others; clearly this is a problem worthy of further study. I thank I. Dębowska-Wosik for help with the data.


believe invalidate P&O’s larger point that not everything grammatically animate is semantically animate.

Peti-Stantić also somewhat misinterprets P&O’s proposals. She states that “it remains unclear why ga should be marked [+animate], and ju, for example, should not”.¹⁸ But of course, for P&O all pronouns are marked [+animate], it is just that the Accusative Prediction Rule in (10) only applies in the absence of a distinct accusative form. Hence the Slovenian feminine singular accusative pronoun jo is also [+animate] in accordance with (11a), although that fact is irrelevant to (10).¹⁹ Moreover, if (10b) Identity of Sense were the key factor, rather than (10b) animacy, it would be no problem for Bosnian/Croatian/Serbian feminine accusative ju/je also to be [+animate].

Since Peti-Stantić concludes that the “animacy story is not convincing”, she suggests that the OA may have to do with “extension of … some other semantic category in Slovene”. Possible candidates one might consider include animacy, definiteness, partitivity, individuation (suggested by R. Alexander (p. c.)), or the genitive of negation. However, unlike the other Slavic genitive–accusative alter-nations mentioned in Section 2.3 above, there is really no evidence that the Slo-venian OA is semantically driven.²⁰ In the end Peti-Stantić argues that the most relevant factor in determining the felicity of the OA is the type of adjective, claim-ing that “the so called Orphan Accusative occurs as a standard possibility with qualitative and possessive adjectives, but never with relational”. While there is some truth to this claim, it is I will argue below epiphenomenal, in that the problem actually derives from the semantically vacuous nature of the missing expression.²¹

Let us now consider that claim, turning to the new data which she provides. Peti-Stantić (2009) offers the example sets in (17)–(19). Following one tradi-

18 Here she is citing the Croatian feminine singular form ju, which alternates with je and which in Slovenian is jo.19 Recall that the rule only applies when there is no distinct accusative entry for a lexical item; if there is, then (10) is blocked from applying. This is nothing other than an instance of Panini’s Principle, probably the most pervasive and enduring linguistic “traffic rule” that there is.20 This caveat is refined later. There are no semantic restrictions in the sense Peti-Stantić has in mind, although below I show that the OA can only involve adjunct material – no overt material can be interpreted if it requires access to lexical properties of the missing noun.21 A. Giannakidou (p. c.) observes that the more evaluative the adjective, the higher it will be and the more acceptable the OA. Under my eventual account this observation can be turned on its head, in that the more classifying the adjective, the more intrinsic to the noun it will be (and also of course lower), hence the less available the OA.

142 Steven Franks

tional terminology for types of adjectives, she refers to those in (17) as qualitative, in (18) as possessive, and in (19) as relational (classifying).²²

(17) A: Kupila sem kavč. ‘I bought a couch.’ B: Za kateregagen/*kateriacc si se odločila: za belegagen/*beliacc ali za modregagen/*modriacc? ‘Which one did you pick: the white one or a blue one?’

(18) A: Vzela sem dežnik. ‘I took an umbrella.’ B: Kateregagen/*Kateriacc, mojegagen/*mojacc ali svojegagen/*svojacc? ‘Which one, mine or yours?’

(19) a. A: Kupili smo nov stroj. ‘We bought a new machine.’ B: *Kateregagen/Kateriacc, *pralnegagen/pralniacc ali *pomivalnegagen/

pomivalniacc? ‘Which one, the (clothes) washing one or the (dish) washing one? b. A: Naredila je poskus. ‘She conducted an experiment.’ B: *Kateregagen/Kakšenacc? *Kemičnegagen/Kemičenacc ali *biološkegagen/biološkiacc? ‘Which?/What kind? Chemical or biological?’

c. A: Prinesla je stol. ‘She brought a chair.’ B: *Kateregagen/Kateriacc? *Lesenegagen/Leseniacc ali*plastičnegagen/

plastičniacc? ‘Which one? The wooden one or a plastic one?’

22 This may not be the best adjectival typology to employ here. The traditional definitions are in-deed somewhat difficult to pin down. Qualitative adjectives provide information about the quali-ties of the nouns they modify; they can be graded. Relational adjectives are more commonly re-ferred to as “classifying” adjectives, which classify the noun by placing it into a class or category; they cannot be graded. Thus the former will not be understand as intrinsic to the noun but the latter may be, as in (19a). I will argue below that this is the relevant criterion. Gradability might be relevant to the judgments Peti-Stantić reports, except that possessive adjectives pattern with qualitative adjectives with respect to the OA, as in (18), even though not gradable. The reader is referred to Marušič and Žaucer (this volume: Section 5.1) for a description of the traditional morphological, syntactic and semantic distinctions among these types of adjectives.


The examples in (17) and (18) are fairly standard illustrations of the OA and require no comment. Peti-Stantić’s claim is that the accusative is the only pos-sible alternative in (19) because the adjectives are relational.²³ I will argue that, to the extent that there is an effect in (19), the correct characterization of why the genitive-like OA form does not arise does not have to do with this subtle, if traditional, terminology, but rather derives from a completely different property of certain adjective–noun collocations. At this point simply note that none of the Slovenian speakers I consulted agree with all the judgments reported in (19), all stating that (19a) is indeed correct but in (19b) the genitive is perfectly normal and in (19c) the accusative seems dubious.

For the sake of completeness, there is one additional claim made by Peti-Stantić that I have not been able to corroborate. Colloquial Slovenian exhibits a curious particle-like definite marker ta, described in considerable detail by Marušič and Žaucer (this volume) as marking what they call “type definite-ness”). Peti-Stantić reports that ta is incompatible with the OA, so that when ta is present, as in (20b), only the accusative is possible:

(20) a. Prosim belegagen /*beliacc. ‘Give me the white one.’

b. Prosim *ta belegagen/ta beliacc . ‘Give me the white one.’ [type-definite ‘white’]

This would indeed be an interesting fact, if verifiable, one which could shed considerable light on the relevance for the OA of the internal structure of the extended nominal projection in Slovenian. Unfortunately, as noted, speakers I consulted do not share these judgments, finding (20a) and (20b) equally felici-tous, both requiring the genitive form belega.

An interesting possibility nonetheless presents itself. Let us accept the judg-ments reported by Peti-Stantić but let us also imagine that speakers fall into two groups, consistently using either the accusative or the genitive with classifying adjectives as in (19) and with adjectives preceded by ta as in (20). In this context, note that under the analysis of Marušič and Žaucer (this volume), ta can change a qualitative adjective into a classifying adjective. Thus, it is to be expected that speakers who report the judgments in (19) would also disallow the genitive in (20b). Although in Section 6.2.2 below I return to this issue briefly, the discrep-

23 This is of course not an explanation, just an attempt at a description. My account below, on the other hand, offers a well-grounded explanation for why the OA fails with some adjectives traditionally characterized as “relational”. Interesting, Marušič and Žaucer conclude that “the distinction between the classifying subtype and the definite qualitative subtype is blurred/hard to define and may actually not even exist”.

144 Steven Franks

ancy in reported judgments surely requires further investigation. Not only, as I show below, are there two distinct ways of obtaining nominal gaps in Slovenian (pro and ellipsis), but it is also probable that different speakers employ these in different ways.

4 Some additional examplesTo recapitulate and complete the relevant OA data, here are some additional rep-resentative examples from the literature to illustrate the extent of the phenom-enon. First, in a sequence of adjectives modifying a missing noun, all of them go into the OA:

(21) a. Moji sosedi imajo dva avta, enegagen rdečegagen in enegagen zelenegagen. (Herrity)

‘My neighbors have two cars, a red one and a green one.’

b. Ima samo enegagen staregagen rjavegagen. ‘He has only one old brown one.’ (P&O) Next, note that appositive adjectives also go into the OA:

(22) a. Mlada žena zahtevala, naj ji prodajo parfum, prav tistegagen, ki je v izložbi. (Rappaport)

‘The young woman wanted them to sell her perfume, precisely that one in the display window.’

b. Se en prizor si je zamislil, najprisrčnejšegagen, najneizraznejšegagen. (Rappaport) ‘It’s quite a spectacle you’ve made up, a most hearty, a most distinctive

one.’

c. Videl sem velik zemljevid, obsegajoč hrvatski okraj, in majhnegagen, obsegajočegagen slovenski okraj. (P&O)

‘I saw a large map, comprising a Croatian district, and a small one, com-prising a Slovene district.’

P&O’s (22c) combines a canonical missing nominal example, in majhnega [zemljevid ‘map’ → ga → ∅] ‘and a small one’, with an appositive participle, obsegajočega ‘comprising’, which itself takes a nominal complement.

Finally, depictive secondary predicates behave like modifying adjectives, not appositives, in that they do not exhibit the OA:


(23) Včeraj smo našli stol pomazanacc s krvjo. (P&O) ‘Yesterday we found the chair stained with blood.’

5 A proposalP&O’s account, which is opaque in that it employs pronominalization and then deletes the pronoun after concord, is easily updated by postulating an appropri-ate empty category. What is required is a null pronoun with which modifiers can agree. Crucially, the category of this Slovenian null pronoun cannot be D or K, as standard pronouns are, but rather must be lower. I therefore posit a null proN in Slovenian. Just like other Slavic pronouns, proN is grammatically [+animate]. However, because it is a noun, proN is expected to behave just like English ‘one’. It is I argue the existence of this silent lexical item in Slovenian that gives rise to the OA.

There are (at least) two factors which confound the judgments. The first is straightforward, namely, that like other Slavic languages, Slovenian also has simple nominal ellipsis. I take this to be deletion (non-pronunciation) of syntacti-cally present material in the mapping to PF. So, instead of proN, it is possible to have an NP which has been elided in the course of Spell–Out, hence does not cul-minate in acoustic instantiation. Adjectives will necessarily agree with the elided NP,²⁴ giving rise to the expected “nominative” accusative instead of the OA. I will take such judgments as a sign that simple ellipsis has taken place.²⁵ The second factor is more complex and requires an excursus into the syntax and semantics of the English lexical item one.

5.1 Digression on the status of English one

This section examines the properties of English one in order to provide a back-drop for the analysis of Slovenian proN. The traditional GB account and a mini-malist recasting of it are presented.

24 Assuming “Bare Phrase Structure” (Chomsky 1995), this result should obtain whether adjec-tival modifiers head an AP projection above NP or are adjoined to NP; cf. also Bošković (2005, 2008, 2009).25 Indeed, R. Žaucer (p. c.) comments about Peti-Stantić’s (19b, c) that these “feel a bit … ellipti-cal, i.e., like there’s an unpronounced eksperiment [or stol] there”.

146 Steven Franks

5.1.1 One under government and binding

One–substitution, as in (24), provided a classic GB argument for the existence of intermediate level X– bar categories.

(24) a. I like this student from France better than that one [=student] from Italy.

b. *I like this student of chemistry better than that one of physics.

c. I like this student of chemistry better than that one [=student of chemistry].

d. I like this student of chemistry from France better than that one [=student of chemistry] from Italy.

The ungrammaticality of (24b) was taken to imply that one–substitution targets a constituent higher than N0, i.e., N’, since N’ includes the noun plus its comple-ments. Thus, not only is (24a) acceptable because from France is an adjunct, but (22c, d) are also acceptable with one substituting for student of chemistry. Some-what more tenuous complement versus adjunct minimal pairs are provided in (25) and (26):

(25) a. ?*David’s argument that the world is flat is not as strange as Elisabeth’s one that there are dinosaurs on a remote Caribbean island.

b. David’s argument that you denied is not as strange as Elisabeth’s one that you believed.

(26) a. ?*The decision that we must leave was less surprising than the one that we must stay.

b. The decision that David made was less surprising than the one that Elisa-beth made.

Judgments about the awkwardness of examples such as (25a) or (26a) may vary because nouns are reluctant to take true complements. But the point here is that one –substitution to some extent distinguishes complement clauses, which are typically degraded, from relative clauses, which are invariably fine.

Any N’ was regarded as a viable target for one –substitution. Consider (27a):

(27) a. Jane kissed Fred’s big black dog and Jean kissed Sam’s small one [=black dog or=dog].

b. [NP Fred’s [N’ big [N’ black [N’ [N dog ]]]]]


Assuming a structure as in (27b), it was argued that one could replace either the N’ black dog or the N’ dog.²⁶ On the other hand, one cannot refer to the entire NP. That is, one in (28) cannot mean ‘it’, i.e., ‘Fred’s big black dog’, only another dog:

(28) Jane kissed Fred’s big black dog and Jean kissed one too.

The important point here is that this set of facts shows that English one works just like the Slovenian null pronoun proN. In short, both exhibit canonical I/S rather than the I/R typical of traditional pronouns. This, it turns out, is hardly surpris-ing, since, when updated to reflect more recent assumptions about phrase struc-ture, it becomes clear that both proN and one necessarily lack referential features. With the addition of functional projections above NP and the adoption of X–bar syntax for all categories, a structure as in (27b) becomes otiose. Instead, Fred’s is necessarily excluded as that part of the DP projection external to NP, and all the other material is included as under NP. This provides us with a point of departure for the minimalist analysis.

5.1.2 One under minimalism

Minimalism eschews X’ as a syntactically accessible node. It is invisible because X–bar status is not a primitive, but rather relationally defined, as neither a head/X0 (=a category which does not project at all) nor a phrase/XP (=a category which does not project any further). This is the essence of Chomsky’s (1995) Bare Phrase Structure. In this system, lexical items drive the derivation. Trees are built from the bottom up by selecting an item from the inventory of lexical resources known as the “Numeration”,²⁷ and then merging this with some other element to result in a binary branching structure. If Merge of YP takes place to satisfy the lexical needs of the head X, then the YP is an argument of X and X projects (with one fewer lexical requirement in need of saturation); otherwise YP is an adjunct to X (the projection of which remaining identical).

Imagine that a noun has been selected and merged with phrasal (non-pro-jecting or maximal) material (which as noted in fn. 86 would already have been

26 It could also replace big black dog, although the meaning would be the incongruous ‘small big black dog’; note that the lowest N’ is gratuitous, the correct generalization, as discussed below, being of course that one is itself an N.27 More precisely, the Numeration is relevant for Initial Merge (aka External Merge), the tree is for (Re)merge (aka Internal Merge, i.e., movement), and another workspace is for sidewards movement as well as for merging complex heads or specifiers.

148 Steven Franks

constructed in a separate work space but according to the same principles) and the N projects. Eventually, when the projection of N combines with another head X, then that head X projects instead. Consider the very nice student of physics from France:

(29) D(max)

N(max)

P(max)A(max) P(max)

N(max) N(max)

N

N

NA P PD Advthe very nice student of physics from France

Note that only lexical categories are represented in (29); the superscript “(max)” is merely intended as a graphic device to help identify maximal projections. This is a representational notion. It is on the other hand not possible to determine in the course of the derivation whether a particular node X is maximal or not until something else merges with it. This point will be essential in understand the workings of English one and its Slovenian counterpart proN.

In order to build (29), successive applications of Merge must combine two nodes into one, projecting the head. In (29), every binary pair of nodes that have merged is set off to facilitate identification. Note that the relative scope of the AP very nice and the PP from France could be different, with very nice merging after from France rather than before, but because these are adjuncts the order in which they merge (thus their relative scope) is technically free.²⁸ Crucially, the same is not true of the complement of physics, which is an argument of student. When an adjunct merges with X, its meaning combines with that of X in a purely compositional manner. When a complement merges with X, the meaning of [X+complement] is a function of the argument structure of X.

28 This is, the linear order in (29) could mean ‘the student of physics who is very nice from France’ (answering Which very nice student of physics?), as depicted, or ‘the student of physics from France who is very nice’ (answering Which student of physics from France?).


Within the minimalist Bare Phrase Structure context, a straightforward account of one -substitution presents itself. Following Franks (1999), the pronoun one is just that, a noun. As a noun, it receives its sense from its antecedent – cru-cially, I/S not I/R. This antecedent can be any available node of the same category, namely, N. Consider (30), in which one can substitute for any projection of N:

(30) a. Jane kissed this big black dog and Jean kissed that one.

b. [D this [N big [N black [N dog ]]]]

The nominal projections dog, black dog, and big black dog are formally indistin-guishable, all being N, so that one can mean just ‘dog’, ‘black dog’ or ‘big black dog’. Moreover, as an N, one has no referential features. This is what differenti-ates it from words like it, he, or she, which lead to I/R. Traditional pro-nouns are of course really pro-Ds, i.e., they substitute full, referential DPs. Compare one and it in (31):

(31) a. Jane kissed [D Fred’s [N big [N black [N dog]]]] and Jean kissed one too.

b. Jane kissed [D Fred’s big black dog] and Jean kissed it too.

Furthermore, any semantically appropriate N in the discourse is accessible. That this is not a matter of syntactic anaphora is demonstrated by the examples in (32). The accessibility of the Ns textbook and linguistics books show that formal rela-tions such as c-command are irrelevant:

(32) a. A: Which [N textbook] did you end up buying? B: The one by Radford.

b. Elisabeth decided to read [D the [N dust jackets of [D the library’s [N dull [N linguistics books]]]]], after realizing that all the interesting ones had already been checked out.

Finally, and most importantly, as a pro-form one can have no theta-roles of its own to assign. This is why, as argued in Franks (1999), one never takes a comple-ment: if it did, then that complement would have no way to receive a theta-role, hence it could not be interpreted.²⁹ This point will be crucial to our understand-ing of how Slovenian proN works. Like English one, it is a lexical N and, as such,

29 Panagiotidis (2003a) offers the same account of English one as in Franks (1999); see also Panagiotidis (2003b).

150 Steven Franks

can merge with any adjunct to create a higher N projection, but it can never merge with a complement. This is an essential property of pro-forms:

(33) John will bake a pie and Mary will (bake/*do (so)) a cake.³⁰

A pro-form, be it overt or covert, lacks a thematic grid, and is exclusively identifi-able by virtue of its dicourse antecedent. Thus, the ability to take a complement distinguishes pro from an ellipsis site. No new arguments can ever be introduced by any pro-form.

5.2 Back to Slovenian

We now return to the Slovenian OA in light of this account of English one.

5.2.1 Obtaining the Orphan Accusative

The Slovenian lexicon countenances a phonetically silent noun, proN, the sense of which is determined through matching with an overt N established in the dis-course. In accordance with P&O’s (11a), proN bears the morphological feature [+animate].³¹ Since, as emphasized above, (11a) pertains not just to Slovenian, it must be the existence of proN itself that makes Slovenian special. And because this lexical item is available, it can be selected and merged (optionally) with adjuncts until another head is selected to merge with it. Whenever those adjuncts are adjectival, they undergo concord with proN, giving rise to the OA. On the other hand, if proN were ever to merge with a phrase that needed to be theta-marked as an argument, then that phrase would never receive a theta-role and the result would be semantically ill-formed. There is nothing syntactic (in the sense of deri-vation) blocking the OA when it is unacceptable, rather the problems are interpre-tative (in the sense of LF legibility). So the restrictions on the viability of the OA are indeed semantic, as Peti-Stantić argues, but along very different lines: to the extent that the modifier is understood as intrinsic to the meaning of the missing

30 Although ellipsis of bake is indicated here by outline font, this may not be the correct account of gapping, since (33) could involve movement of a cake out of the VP, with subsequent deletion of the entire VP. The point nonetheless remains that the pro-form do cannot assign the theme theta-role of bake.31 Why Slovenian proN should be “rich” in this sense is somewhat puzzling. As discussed in Sec-tion 7.5 below, pronouns seem to require a value for [±animate].


noun, the gap must be treated as PF ellipsis of the N rather than a syntactically present proN. Hence it comes as no surprise that the accusative rather than the genitive is used in her (19a), repeated below, where pralni and pomivalni go part-and-parcel with stroj:

(34) A: Kupili smo nov stroj. ‘We bought a new machine.’ B: *Kateregagen/Kateriacc, *pralnegagen/pralniacc ali *pomivalnegagen/pomivalniacc? ‘Which one, the (clothes) washing one or the (dish) washing one?

Variation in judgments is also expected, since there is a competition between a proN analysis and an ellipsis analysis in the speaker’s mind. ProN is in general greatly preferred but, when it is not viable because the adjective can only be inter-preted by virtue of the missing noun, then ellipsis is the only alternative.³²

To illustrate, consider (35) from this perspective.

(35) Stane ima star rjav površnik, Tone pa ima novega. (based on P&O) ‘Stane has an old brown overcoat, and Tone has a new (brown) one.’

In (35), proN can either have the sense površnik ‘overcoat’ or rjav površnik ‘brown overcoat’, since both are realistic nominal antecedents in the discourse context. Relevant structures for the antecedent and the OA component are given in (36):

(36)N(max) N(max)

N A(max)

A(max)A(max)

star

nowega pron

rjav provršnik

N

N

a. b.

(36a) provides the syntactic range of antecedents for proN, although only površnik and rjav površnik are sensible, ‘new’ and ‘old’ being situationally incompatible under most scenarios.

32 Note that this view is incompatible with the standard minimalist conception of a Numeration, to my mind not an undesirable consequence: pro, overt pronouns, and fully specified lexical items surely compete on a pragmatic level.

152 Steven Franks

5.2.2 Thoughts on ellipsis

I have proposed that there are two ways in which a missing N can be analyzed, either as PF ellipsis of a syntactic NP or with a syntactically present but phono-logically null proN. Only the latter gives rise to the OA. Variation in judgments is the consequence of pragmatic competition between the two options, and it is likely that diverse factors come into play in deciding between them. The ellipsis option presumably also provides an avenue for explaining the use of accusative with neuter and semantically female antecedents, such as proso ‘millet’ and dekle ‘girl’, since as noted the genitive arising through agreement with a proN comes up against morphological generalizations.

R. Žaucer (p. c.) suggests that discourse analysis supports my claim that the accusative implicates ellipsis. He reports the dialog in (37a) on a television program, but states that the exchange in (37b) would also have been possible:

(37) a. Waiter: Kaj boš? Kaj prnesem? ‘What will you have? What shall I bring?’ Customer: En pir. ‘A beer.’ Waiter: Kašngagen pa? ‘Which one?/What kind?’ Customer: Ta velzgagen./*Ta velkacc. ‘A large one./A pint.’

b. Waiter: Kaj boš? Kaj prnesem? ‘What will you have? What shall I bring?’ Customer: En pir. ‘A beer.’ Waiter: Kašnacc? ‘Which?/What kind?’ Customer: Ta velkacc./*Ta velzgagen. ‘Large./A pint.’

In (37a) the silent noun is proN, whereas in (37b) it is pir ‘beer’. On the other hand, a language such as Russian, which lacks proN, only has the ellipsis option:

(38) Waiter: Kakoeacc/Kakojacc/Kakujuacc pivoacc/sendvičacc/vodkuacc xotite? ‘What kind of beer/sandwich/vodka do you want? Customer: Bol’šoeacc/Bol’šojacc/Bol’šujuacc. ‘A large one.’

Žaucer points out that since pir in (37) occurs immediately before the question word, both alternatives are viable. However, when pir does not occur immediately before the question word, only the long form is acceptable. That is, in (39), if a customer simply asks for a large beer without the antecedent pir, then ellipsis is impossible, only proN can be used, hence the OA is the only possibility:

(39) Živjo, dej mi engagen/*enacc ta velzgagen/*ta velkacc. ‘Hi, give me a large one [=a pint].’


Žaucer adds that for him the contrast between (37) and (39) is “crystal clear”. The discourse context in which (39) is uttered can only mean the customer is requesting a pint of beer. This shows quite clearly that the OA involves deep, not surface anaphora: since it does not need a linguistic antecedent, the structure must involve proN rather than ellipsis. Hence only the genitive is acceptable in (39).

Similarly, Slovenian has certain “lexicalized” ta+adjective collocations with nominal meanings, which I would analyze as appearing inside an NP headed by proN. Examples provided by Žaucer include: ta kratek ‘a shot’, ta beu ‘quisling/WWII Home Guard’, ta rdeč ‘commie’, ta mau ‘little one’. Interestingly, the OA is obligatory with these:

(40) Živjo, dej mi engagen/*enacc ta kratkegagen/*ta kratekacc. ‘Hi, give me a short one [=a shot].’

The accusative form reflecting ellipsis is inapplicable because there is never a noun in the first place that could be ellided. Instead, the only analysis available to the speaker is the proN one.

The fact that such lexicalized adjectival expressions require the OA suggests that nominal use of adjectives may involve an adjective modifying syntactic proN rather than actual (“zero” derivational morpheme) nominalization of the adjec-tive. It is a common phenomenon in all Slavic languages for an adjective to be lex-icalized so that it functions as a noun; this is especially true of adjectives based on verbal participles, such as Russian moroženoe ‘ice cream’ (literally, that which has been frozen) or trudjaščiesja ‘workers’ (literally, those who are laboring), and surnames, such as Tolstoj or Dostoevkij. These are surely substantivized adjec-tives in the sense that they are categorially nouns, they just retain their adjecti-val declensions. On the other hand, as A. Giannakidou (p. c.) points out, many languages, including English, have a substantivization process which seems to transform adjectives into generic (and typically human) nouns, such as the fol-lowing:

(41) a. The blind deserve our admiration.

b. The (stinking) rich should remember the (homeless) poor.

In this light, she notes as a diagnostic the fact that these adjectives resist the com-parative (and superlative) degree. Consider the following, from Giannakidou & Stavrou (1999):

154 Steven Franks

(42) a. I plusii sinithos ksexnun apo pu ksekinisan. the rich usually forget3.pl from where started3.pl ‘The rich usually forget where they started from.’

b. *I plusioteri / *i pjo plusii ksexnun apo pu ksekinisan. the richer / the more rich forget3.pl from where started3.pl ‘??The richer forget where they started from.’

In Greek, as in English, any attempt to create a comparative form of such generic substantivized adjectives is degraded. With respect to the Slovenian OA, Gianna-kidou thus wonders whether “we are dealing with this nominalized kind of empty noun”.

I think not. OA adjectives are not generic, nor are they restricted to animates. And I have shown that the gender restriction is no more than an artifact of Slove-nian morphology. In sum, the syntactic status of this proN is the same as English one. I have argued that the syntactically present proN, which gives rise to the OA, must be distinguished from from a PF-elided NP, which does not. Unfortunately, the kind of nominalization in (41) and (42) seems to me to be yet another phe-nomenon. While Giannakidou & Stavrou (1999) show that this is not ellipsis, it is also clearly different from the Slovenian OA construction. On the one hand, comparatives and superlatives pose absolutely no problem for the OA. In example (22b) above the adjectives were both superlative forms. We see OA comparatives in (43a) and (43b), with both synthetic zanimivejšega ‘interesting-er’ and analytic bolj plavega ‘more blue’:

(43) a. Kakšenacc/kateriacc članek hočeš prebrati? Zanimivejšegagen. ‘What kind of/which article do you want to read? ‘A/The more interesting

one.’

b. Za kateriacc površnik si se odločila? Za bolj plavegagen/*bolj plavacc? ‘Which overcoat did you pick? The bluer one.’

On the other hand, Slovenian also uses adjectives in the same way as (41) and (42), in which case comparatives and superlatives are comparably degraded. Slo-venian (44) has the same status as Greek (42) above:

(44) Bogati/??Bogatejši/?*Bolj bogati pozabljajo od kod prihajajo. ‘The rich/??richer/??more rich forget where they started from.’

I thus do not believe that the OA phenomenon can be assimilated to any more familiar substantivization phenomenon.


Let us return finally to Peti-Stantić’s proposal that, at least for the speakers who provide such judgments, classifying/relational adjectives are incompatible with the OA. A simple solution presents itself: for these speakers, proN does not occur with such adjectives. Of course, even if this were to turn out to be empiri-cally correct, it begs the question of “Why?”. Moreover, we would still expect such speakers to produce the judgments in (39) and (40).

5.2.3 Complement–adjunct asymmetries and the OA

If this general account of the OA as employing a proN comparable to English one is correct, we should find similar complement–adjunct asymmetries associated with the OA. And, as the examples below show, this is indeed the case. In (45), the antecedent of proN, tečaj ‘course’, has a complement fizike ‘(of) physics’, whereas in (46) it does not; additionally, the adjunct na univerzi ‘at (the) university’ in (46b) is irrelevant to interpreting proN:

(45) a. *Peter je naredil tečaj fizike, ampak Janez ‘*Peter passed a physics course, but Janez je naredil enegagen proN matematike. passed a mathematics one.’

b. *Peter je naredil težek tečaj fizike, ampak ‘*Peter passed a difficult physics course, but anez je naredil lahkegagen proN matematike. Janez passed an easy mathematics one.’

(46) a. Peter je naredil težek tečaj, ampak Janez ‘Peter passed a difficult course, but Janez je naredil lahkegagen proN. passed an easy one.’

b. Peter je naredil težek tečaj na univerzi, ‘Peter passed a difficult physics course at university, ampak Janez je naredil lahkegagen proN na srednji šoli. but Janez passed an easy one at school.’

(47) illustrates the (admittedly slippery) complement clause versus relative clause contrast:

156 Steven Franks

(47) a. ?*Razumem dober predlog da bo Janez prišel in tudi sijajnegagen proN da bo prinesel pivo.

‘?*I understand the good proposal that Janez come and also the great one that he bring beer.’

b. Razumem dober predlog, ki ga je Janez imel in tudi sijajnegagen proN ki ga je Janez ponovil.

‘I understand the good proposal that Janez made and also the great one that Janez repeated.’

(48) illustrates various viable positions for proN in more detail:

(48) a. V sredo bomo zapili Petrov zagovor, in v petek pa Janezovegagen proN. ‘We will celebrate Peter’s defense on Wednesday and Janez’s one

[=defense] on Friday.’

b. V sredo bomo zapili Petrov zagovor doktorata, v petek pa Janezovegagen proN.

‘We will celebrate Peter’s dissertation defense on Wednesday and Janez’s one [=dissertation defense] on Friday.’

c. V sredo bomo zapili Petrov zagovor doktorata u veliki dvorani, in v petek pa Janezovegagen proN u mali dvorani.

‘We will celebrate Peter’s dissertation defense on Wednesday in the great hall and

Janez’s one [=dissertation defense] on Friday in the small hall.’

d. Prebrali smo Petrov doktorat, napisan v francoščini, in tudi Janezovegagen proN, napisanegagen v ruščini.

‘We read Peter’s dissertation, written in French, and also Janez’s one [=dis-sertation], written in Russian.

Speakers report that the accusative is also acceptable here, although dispreferred:

(49) ?Prebrali smo Petrov doktorat, napisan v francoščini, in tudi Janezovacc dok-torat, napisanacc v ruščini.

‘We read Peter’s dissertation, written in French, and also Janez’s, written in Russian.’

In keeping with the present analysis, I treat the accusative possibility as reflecting ellipsis. Further confirmation for an account in which Slovenian has two compet-ing ways of deriving such gaps is the fact that the choice of genitive or accusative


on the possessive adjective Janezovega/Janezov and on the appositive adjective napisanega/napisan necessarily covaries.

To complete the paradigm, (50) shows that Slovenian proN cannot take a complement:

(50) *V sredo bomo zapili Petrov zagovor kriminalcev, v petek pa Janezovega proN nedolznezev.

‘*On Wednesday we will celebrate Peter’s defense of criminals and on Friday Janez’s one of innocents.’

Like English one, Slovenian proN has no argument structure and so cannot assign a theta-role to the complement nedolznezev ‘innocents’. In general, then, the acceptability of the OA parallels the acceptability of English one.³³

6 Extensions and speculationsIn this final section, I offer some speculations about how the Slovenian nominal domain might fit into more general patterns of nominal structure and whether similarities can be found between the OA and morphological phenomena in other languages. I also briefly consider the nature of the Accusative Prediction Rule in (10) and the special status of pronouns.

6.1 DP versus NP languages (Bošković 2005, 2008, 2009) and beyond

Nominals have features for case and specificity, inter alia, but how these features are expressed can differ from language to language. The variation is a matter of whether features are borne by a separate functional head or not. This is com-parable to clausal features such as tense and agreement, which can require an independent TP or AgrP in order to be expressed or not. There are thus canonical “DP languages” such as French or English and canonical “NP languages” such as

33 This seems a more reliable diagnostic than some intuitive notion of what should be a complement to a noun. For example, in (i) both the OA and its English translation with one sound fine:(i) Ali hočeš kupit tale zanimiv film o živalih ali tistega dolgočasnega o letalih? ‘Do you want to buy this interesting movie about animals or that boring one about air-

planes?’

158 Steven Franks

Latin or Russian. Bošković, in a series of recent publications and presentations, explores this traditional theme, elaborates upon it, and takes it in a host of novel directions. Without getting into the details, one could posit rough structures for, say, French and Russian as follows:

(51) D(max)

Dle

N(max)

a.

A(max) N

N(max)

A(max) N

b.[French] [Russian]

It seems morphologically appropriate to regard a pronominal clitic such as French le also as a D head, i.e., as an element which, when exhaustively also Dmax, also serves as a direct object clitic in that language. Pronominal clitics in the Slavic languages, on the other hand, resemble instead the case morphemes of these languages. Consequently, Franks & King (2000), Franks & Rudin (2005) and Franks (2009), among others, argue that pronominal clitics such as Bosnian/Croatian/Serbian ga are K0 heads, with a nominal structure as in (52a). Unlike French or English, definiteness is not independently realized. Bulgarian, on the other hand, is a DP language as in (52b) although, contra Bošković, what is crucial is not where in the extended projection definiteness features are realized vis-à-vis case features, but just that both are realized independently.³⁴

(52)

K(max)

Kga

N(max)

a.

A(max) N

go

D N(max)

N(max)

K D(max)

b.[Bosnian/Croatian/Serbian] [Bulgarian]

There are many associated properties that distinguish Bosnian/Croatian/Serbian from Bulgarian. The most obvious one, of course, is the fact that it is only in the latter language that definiteness is morphologically marked. In Bulgarian defi-

34 See Franks & Rudin (2005) or Franks (2009) for justification of the DP–over–KP structure, which supercedes the KP-less analysis of Bulgarian in Franks (2001). Also, in Franks (2009) I show that in Macedonian the clitic is evolving into an agreement marker, not moving from within KP but being generated above VP in the first place.


niteness is expressed as inflection on the highest head in D’s complement,³⁵ as in (53), from Franks (2001):

(53) a. [kniga-ta] ‘the book’

b. [interesna-ta] kniga ‘the interesting book’

c. [goljama-ta] interesna kniga ‘the big, interesting book’

d. [dosta glupava-ta] zabeležka ‘the quite stupid remark’

e. [polučena-ta sâs mâka] stipendija ‘the scholarship received with pain’

f. [verni-jat na [demokratični-te idei]] prezident ‘the president faithful to democratic ideas’

g. [kupeni-te včera] knigi ‘the books bought yesterday’

The most salient syntactic property, discussed at length in Bošković (2005), is the possibility of left-branch extraction in Bosnian/Croatian/Serbian, as in (54a), versus its absence in attempted Bulgarian (54b):

(49) a. Kakva si kupio [e kola]? ‘What kind of car did you buy?’

b. *Kakva prodade Petko [e kola]? ‘What kind of car did Peter sell?’

Another generalization which may be relevant for Slovenian is that clitic doubling is only available in DP languages. Bošković (2008) adds this property, which follows from the account in Franks & Rudin (2005) and developed in Franks (2009). Essentially, doubling in my view is only available when the extended nominal projection provides two sites for referential features, so that the nominal can be doubly instantiated. Hence in Bulgarian (55) the DP extracts out of KP, leaving the clitic behind, which can subsequently move to V as a non-branching maximal projection:

(55) a. Tjah nikoj ne gi pazi. ‘Nobody is guarding them.’

b. Ivan go târsjat. ‘They are looking for Ivan.’

35 I refer to “D’s complement” rather than just “NP” because my previous work on the Bulgarian KP/DP adopted an Abney-style AP/QP–over–NP phrase structure for DP languages. This is not crucial to the analysis at hand.

160 Steven Franks

Franks & Rudin (2005) suggest that Spanish to some extent embodies the flip side of Bulgarian. Consider the following examples, all drawn from Franco (2000):

(56) a. Pedro la ha visto a Sandra. ‘Pedro (her) has seen Sandra’

b. ¿A quién le pegaste? ‘(To) whom did you (him) hit?’

In Romance, the clitic is a D0. In French this is simply as in (51a), the DP version of Bosnian/Croatian/Serbian (52a). In Spanish, however, since the clitic doubles an associate (which appears with the preposition a), there must be another maximal projection above NP. Presumably this is an instance of DP–over–KP, as in Spanish (57), which is thus a DP version of Bulgarian (52b):

(57) D(max) [Spanish] D K(max)

le K N(max)

Of course, in standard Spanish KP splits into P+DP in the morphology, so that the preposition a appears, either to case mark the NP or as a realization of case features themselves.³⁶

I believe that something similar may be taking place with Slovenian, i.e., that it is developing from an NP language, with a general structure as in Bosnian/Cro-atian/Serbian (52a), into a DP language, with the general structure of Bulgarian (52b).³⁷ However, while the Slovenian nominal domain has additional functional structure, the evidence is at best mixed that it should be equated with the DP of

36 Interestingly, Franco (2000) notes that varieties of Southern Cone Spanish countenance clitic doubling even when the associate does not have a preposition:(i) La comí la torta. ‘I ate (her) the cake.’However, this is most likely fully fledged object agreement, not clitic doubling, as argued in Franks (2009) for Macedonian.37 Bošković (2009) similarly concludes his paper on Slovenian noun phrases with an ambiva-lent footnote, stating that he “wouldn’t rule out the possibility that we are starting to witness a change here, i.e., the beginning of the emergence of a DP system, with the change starting with indefinite articles …”. In personal communication he concurs with my claims in this paper, sug-gesting that the possibility of left-branch extraction may be taken as a litmus test for the presence of additional functional structure, as in Slovenian.


Bošković’s typology, so for now I will simply refer to this category as “FP” rather than DP. This is shown in (58):

(58) K(max) [Slovenian] K F(max)

F N(max)

A(max) N

Pronominal clitics are ordinarily the realization of case and definiteness fea-tures in K (or D, e.g. for Romance). This is why clitics are canonically definite, giving rise to the standard I/R interpretation. On the other hand, as remarked by Rappaport (2009), the I/S interpretation “shows that the personal pronoun in Slovenian is lexicalized at both the DP level (like other Slavic languages and English) and the NP level (unlike other Slavic languages, like English one)”. I concur, but with three technical caveats that pertain to what is meant by “the DP/NP level”: (i) given their case morphology, the appropriate higher category is K, rather than D; (ii) the relevant lower level is FP rather than NP; and (iii) clitics are lexical items not phrases, even though, given Bare Phrase Structure as outlined in Section 6.1.2, they are typically simultaneously also phrases.

For one thing, Slovenian indeed generally eschews left-branch extraction, patterning like DP languages.³⁸ Compare Slovenian (59b), which behaves like Bulgarian (59c) and not like Bosnian/Croatian/Serbian (59a):

(59) a. Vukina odlazi ćerka. ‘Vuk’s daughter is leaving.’

b. *Milojkina odhaja hči. ‘Milojka’s daughter is leaving.’ [cf. Milojkina hči odhaja.]

c. *Novata prodade kola. ‘S/he sold the new car.’ [cf. Prodade novata kola.]

However, as Bošković shows, the impossibility of left-branch extraction does not depend on DP per se, but rather on higher blocking functional structure which, presumably, the KP–over–FP complex also can instantiate.

38 See Bošković (2009: fn. 20) for further critical discussion of left-branch extraction in Slove-nian.

162 Steven Franks

On the other hand, although Slovenian does not have obligatory determin-ers, it does have some DP-like elements. For one thing, as mentioned earlier, colloquial Slovenian has a special definite marker ta although, as discussed by Marušič & Žaucer (this volume), this ta element clearly occurs within AP and marks what they analyze as “type-definiteness”.³⁹ More striking, however, is the that fact Slovenian, like English, regularly employs en ‘a’ as an indefinite article, which – again like English – it distinguishes from independent eden ‘one’. This suggests that what I have called “FP” should more perspicuously be labelled “IndefP”, since its one obvious occupant is the indefinite article. The realization of indefiniteness, as Bošković (2009) also suggested (cf. fn. 96), is likely an early stage in the development of a fully-fledged DP. Let us therefore make this descrip-tive move.

Finally, also as discussed by Marušič & Žaucer (2009), dialects of Slove-nian – like Bulgarian and Macedonian – exhibit clitic doubling. Consider their (60), from Gorica Slovenian:

(60) a. Ma to megen menegen ne briga. ‘But I don’t care about this.’

b. Bi mudat mogu njemudat pustit. ‘I should have left that for him.’

c. Jaz se gagen njegagen spomnim še iz srednje šole. ‘I remember him already from high school.’

d. Lahko jih acc/gen pa njih acc/gen vpraša. ‘He can ask them.’

Marušič & Žaucer (2009) note that this doubling phenomenon (although pos-sibly not in all dialects that have it) is restricted to pronominal associates. As in Bulgarian, where doubling I would argue correlates with the rise of DP, the devel-opment of doubling in Slovenian suggests a more finely articulated extended nominal projection for Slovenian as well, to allow for both the clitic and its asso-ciate. The issue of why it is limited to occurring with pronominal associates may provide some hints as to the details.

39 Interestingly, Marušič & Žaucer (this volume) place ta APs under FP, above NP and above APs that cannot take ta, in a way that is reminiscent of the FP structure for Slovenian in (58), which I have adopted here.


To address this issue one must first raise the question of the category of the clitic itself. It could either originate as an N, as Rappaport claimed, subsequently moving to Indef0 and then to K0. This is represented in (61):

(61) K(max)

Kga

Indef(max)

Indef N(max)

ga

gaA(max) N

Alternatively, it could merge in Indef0 and move to K0, as in (62a), or just merge directly as a K0, as in (62b). Finally, at least in the structures in (61) or (62a), it could be ultimately realized in a lower head position instead of moving all the way to K0.

(62) K(max)

Kga

ga

Indef(max)

a.

Indef N(max)

A(max) N

ga

Indef N(max)

K

K Indef(max)

b.[Slovenian, final]

As argued in Franks & King (2000) inter alia, all clitics are universally pure instantiations of functional features: they can only have grammatical content, never lexical content. If so, the option in (62a) should be chosen over (61) for Slo-venian.⁴⁰ Working with this structure further, we could imagine that clitic dou-bling in this language arises as follows. First, suppose that the nj(e)– element

40 It is however conceivable that ga instantiates just the formal features of N, e.g., [+N, –V, –fem, –pl, …], although this would be incompatible with the analysis in (63) which treats nj(e)– as an N0.

164 Steven Franks

that initiates the third person full forms is in fact N0.⁴¹ It could then raise to the clitic to form a tonic pronominal, as in (63a) or (63b):⁴²

(63) K(max)

K

nje+ga

nje

Indef(max)

a.

Indef N(max)

A(max) N

nje+ga

A(max) N

K(max)

K N(max)

b.[intermediate stage,of doubling dialect]

nje

In the absence of IndefP, nj(e)– will simply adjoin to ga in K, as represented in (63b). This will be the account of tonic pronouns in Slavic languages of the NP type, in which there is no doubling. It is also I would argue appropriate for the Slovenian non-doubling norm, since material in the functional category I am calling Indef cannot be interpreted as definite. That is, definiteness is still an ancillary feature of KP in Slovenian, occurring along with case features, as it does in most other Slavic languages. So even if ga merges in Indef it will move to K.

Clitic doubling in Gorica Slovenian (61) can then be analyzed as having the following structure, with two head movements:

(64)

A(max) N

Indef(max)

Indef N(max)ga

K(max)

K

nje

nje + ga

41 This proposal is reminiscent of derivations in Progovac (2000). However, her Bosnian/Croa-tian/Serbian structure is in my view overly exploded and is not amenable to the kind of nominal typology cogently demonstrated in Bošković’s work; see Despić (this volume) for discussion.42 Arguably this nj(e)– is the same morpheme as the –nj which appears after certain prepo-sitions in many of the Slavic languages. This sort of N–to–P movement, as in Polish nań ‘on him/?her?’, is largely archaic. It is moreover interesting to note that, as reported in Franks & King (2000: 152), this form is no longer gender bound.


Nj(e)– moves to ga and, as elsewhere, ga moves to K, the locus of case and defi-niteness features.⁴³ Note that in (64) both copies of ga are realized, the one that nj(e)– incorporates into and the one in K0. Following Nunes (2004:43 ff.), both are realized precisely because of the incorporation: he argues persuasively that “clitic duplication is a by-product of morphological restructuring” (Nunes 2004:45). That is, the two instances of ga in (64) are distinct: because of the restructuring, the chain is disrupted and they cannot be treated as copies of the same element. Each is thus linearized independently.

6.2 More on the I/S versus I/R contrast

How might this account relate to the I/S versus I/R contrast exhibited by Slovenian pronominal clitics? In the spirit of Rappaport (2009), this distinction should in some way depend on where in the structure the clitic is interpreted. Given (62a), the I/S reading should derive from interpreting the clitic in Indef0 and the stan-dard I/R reading should derive from interpreting it in K0. Since all Slavic pronomi-nal clitics are pure instantiations of case and phi-features, one might argue that they must always be K0, whether merging there, as in Bosnian/Croatian/Serbian, or moving up to K0, as in Slovenian. If we take it to be a general truth that all Slavic pronominal clitics are K heads because they express case features, then the I/S reading arises from interpreting the clitic in its lower position, Indef0, rather than its ultimate position, K0. Since we know independently that Spell–Out to LF does not necessarily reflect PF, this proposal is theoretically sound. Consider, for example, the lower binding of the reflexive herself in English (65):

(65) a. [CP Which pictures of herself does [TP Mary–herself [VP think [CP which pictures of herself (that) [TP Bill [VP likes which pictures of herself]]]]]]?

b. [CP Which pictures of herself does [TP Bill [VP think [CP which pictures of herself (that) [TP Mary–herself [VP likes which pictures of herself]]]]]]?

My account thus differs from Rappaport’s in that an I/S clitic, although it origi-nates as an Indef0 (his N), undergoes head movement to K0, where it is expressed phonetically.

43 One possible technical problem is that, contrary to fact, (63a) and (64) imply that Slovenian njega should be able to stay in Indef and admit the I/S reading if the ga which nj(e)– adjoins to were the [–definite] one. A straightforward solution is for nj(e)– itself to be [+definite]. This makes sense both historically and in light of the previous footnote.

166 Steven Franks

As in (63b), is also possible that clitics with the I/R interpretation undergo initial merge under K0; the nj(e)– morpheme would still move up to support it. The correct analysis, I believe, depends on the overall viability of the I/S inter-pretation and the extent to which its acceptability correlates with other proper-ties of an individual’s grammar. More careful study of clitic doubling systems in Slovenian dialects may provide a clue. One correlation is that clitic doubling is incompatible with I/S, meaning that the clitic in K0 must be interpreted there when the movement chain is broken by the restructuring of nj(e)+clitic in K0. If so, there may also be a correlation between clitic doubling dialects and the avail-ability of the I/S interpretation in those dialects, since under the analysis in (64) these require the clitic to be merged in Indef0 but to move to K; cf. also fn. 102.

In languages which lack IndefP, clitics undergo initial merge under K, which leads to the standard referential interpretation of clitic pronouns. It should be noted that not only is this of course the general position where pronominal clitics are introduced in Slavic (modulo Macedonian, if Franks 2009 is correct), but also that across Slavic the I/S interpretation is extremely rare. Recall from (12) above that an informal survey of speakers of other Slavic languages indicates that the I/S interpretation is virtually always rejected. Here is a representative example from Czech:⁴⁴

(66) Jan vidí hnědý kabát a Petr ho vidí taky. ‘Jan sees a brown coat and Petr sees it/*one too.’

However, some individuals consulted did not completely reject the I/S reading in all examples. Judgments are subtle, and my guess is that reluctant acceptance of what appears to be I/S parallels use of I/R it in English sentences such as (67):

(67) John ordered the lamb and Peter ordered it too.

This is deceptive, because here the lamb does not refer to a specific piece of lamb in the kitchen, but rather the item on the menu; indeed it could be generic, without the article:

(68) John ordered lamb and Peter ordered it too.

44 The Czech example was provided by V. Vaníček. In general, I have verified the impossibility of the I/S reading with speakers of Bosnian/Croatian/Serbian, Polish, Bulgarian, Macedonian, Russian, Czech, and Slovak.


Nonetheless, some I/S may be possible beyond Slovenian. While speakers of Serbian and most speakers of Croatian who I have consulted reject the I/S inter-pretation, consider the following perplexing set of Croatian judgments, due to A. Peti-Stantić (p. c.):

(69) a. Stane ima (smeđi) kaput, a i Tone ga ima. ‘Stane has a (brown) coat and Tone has one too.’

b. Stane ima (*pametno) dijete, a i Tone ga ima. ‘Stane has a (smart) child, and Tone has one too.’

c. Stane ima (*dragog) prijatelja, a i Tone ga ima. ‘Stane has a (dear) friend, and Tone has one too.’

She suggests that inanimates can admit I/S regardless of modification but ani-mates only can if unmodified. Be that as it may, such judgments indicate that the possibility of I/S in a grammar does not necessarily imply the OA, which is unique to Slovenian.

On the other hand, not all Slovenian speakers agree about the I/S interpre-tation, despite all reporting robust use of the OA. In particular, R. Žaucer does not accept it even in P&O’s famous examples in (12), repeated below (with more colloquial morphology and Žaucer’s judgments embedded into the translations):

(70) a. Stane je vidu plav avto in tud Tone ga je vidu.⁴⁵ ‘Stane saw a blue car and Tone also saw it/*one.’

b. Stane ma rjav plašč in Tone ga ma tud.⁴⁶ ‘Stane has a brown coat and Tone has it/?*one too.’

c. Stane ma pametnga otroka in Tone ga ma tud. ‘Stane has a smart child and Tone has him/??one too.’

d. Stane ma pametno ženo in Tone jo ma tud. ‘Stane has a smart wife and Tone has her/??one too.’

He remarks that, although judgments are affected by the availability of only one sensible way of understanding the sentence, the only appropriate way to say

45 Žaucer points out that definite ‘the blue car’ would be a much more likely interpretation in (70a), which would be compatible only with the I/R reading, but even if an indefinite interpre-tation is artificially imposed on plav avto (or it were replaced by en (plav) auto ‘a blue car’, his judgments remain the same.46 Plašč is a more standard term for ‘coat’ than P&O’s površnik in (12b). Note that, whatever is

168 Steven Franks

such things is with VP-ellipsis, which gives rise to absolutely perfect sentences. Compare (71a, b) with (70c, d):⁴⁷

(71) a. Stane ma pametnga otroka in Tone tud/lih tko. ‘Stane has a smart child and Tone also/likewise.’

b. Stane ma pametno ženo in Tone tud/lih tko. ‘Stane has a smart wife and Tone also/likewise.’

Some further potential I/S examples which Žaucer rejects are in (72):

(72) a. Peter je pojedu tri piškote in tud Tone jih je (pojedu). ‘Peter ate three biscuits and Tone (ate) them too.’ [not different biscuits!]

b. Peter je fentu dva pajka in tud Tone jih/ju je (fentu). ‘Peter killed two spiders and Tone killed them two spiders.’ [not different

spiders!]

c. Peter je grizu en jabuk in tud Tone ga je (grizu). ‘Peter chewed on an apple and Tone also chewed on it.’ [not different

apples!]

In general, Žaucer only allows a different referent for the pronoun in those con-texts where English does too, so this is not I/S. These contexts involve kind-denot-ing objects:⁴⁸

(73) a. Peter je jedu polento in tud Tone jo je. ‘Peter ate polenta and Tone ate it too.’

b. Js nosm najkice že ceu živlejne in Peter jih tud (nos že ceu živlejne). ‘I’ve been wearing Nikes already all life and Peter has been wearing them

too.’

going on in this example, Žaucer’s hesitance in completely ruling it out may derive from the same “generification” effect as credibly leads Peti-Stantić to accept the Croatian examples in (69). 47 Indeed, some speakers of other Slavic languages at first commented “I suppose you could say that” but subsequently corrected themselves, offering more felicitous alternatives. For example, B. Volková made such a comment about Czech (66), but later added that “grammatically it … doesn’t make sense, but since the only possible pragmatic interpretation is that it is a different coat, the interlocutor will make room for that”. Allowance of course must be made for “poetic license”, especially – as in this case – when the consultant is herself a poet!48 Similarly, pronouns occur in “donkey”-sentences in both English and Slovenian. See also examples (67) and (68).


One might suspect that the I/S reading arises only with non-referential ante-cedents, but Žaucer also rejects it in intensional contexts:

(74) a. Peter je hotu brat knjigo in Tone jo je tud hotu brat. Peter wanted to read a book and Tone also wanted to read it/*one.’

b. Peter si je želeu spoznat manekenko in Tone si jo je tud želeu spoznat. ‘Peter wanted to meet a model and Tone also wanted to meet her/*one.’

Taking the structures posited for Slovenian to be correct, one would be led to conclude that Žaucer’s Slovenian grammar simply lacks [ –definite] pronominal clitics; it is on the other hand possible that Peti-Stantić’s Croatian grammar (to some extent) has them. That is, I/S arises through an innovation under which pronominal clitics, although they retain case and phi-features, lose their definite-ness. This possibility in all likelihood became available because, as an emerging DP language, Slovenian has begun to divorce definiteness from DP per se and now countenances an IndefP projection. So there can now exist in Slovenian [–definite] pronouns alongside the familiar [+definite] ones. Finally, we can take Bulgarian, with a full DP as in (52b), to represent a stage in which definiteness has its own dedicated projection and pronominal clitics in Bulgarian instantiate pure case.

6.3 Connecting I/S and the OA

These are complex problems requiring further detailed cross-linguistic study. My point, nonetheless is clear: the OA can exist in a grammatical system that does not countenance the I/S reading.⁴⁹ The former, in my account, depends on proN and the latter on [–definite] pronouns. But nothing I have said explains why Slo-venian should countenance both the OA and the I/S. What these seem to share is the IndefP projection. With respect to I/S the role of IndefP is obvious: pronomi-nal clitics with the I/S reading must be introduced as Indef0. But what does the OA innovation, which depends on the availability of proN, have to do with IndefP? My answer to this question again relies on proposed typological connections with Bulgarian.

49 Given Peti-Stantić’s Croatian judgments the converse, although not nearly as clear, may also be true. This may however correlate with the fact that Croatian actually diplays some OA-like ef-fects, as discussed in Section 7.5.1 below. It is, moreover, also my impression that the closer Croa-tian speakers are to the Slovenian area, the more likely they are to accept examples that evoke both the I/S and the OA. This would of course require empirical substantiation.

170 Steven Franks

Recall that the Bulgarian extended nominal projection has the KP–over–DP structure, which arose as postpositive definite articles evolved from the demon-stratives, with the clitics retained as K0 (and subsequently reanalyzable as Agr0, a process which is much more advanced in Macedonian; cf. Franks 2009). Now, while the present paper is not about the syntax of clitics per se, one significant generalization is that pronominal clitics are overt when their complements are null. This may well have to do with the relationship, first noted in Lobeck (1990), between Spec–head agreement and complement ellipsis:

(75) a. Mary said John is leaving but I don’t know when [ip he is leaving].

b. *Mary said John is leaving but I don’t think that [ip he is leaving].

c. We are looking for John’s [np book].

d. *We are looking for the [np book].

Agreement licenses silence. Clitic doubling may well be a reflection of the same generalization. Returning to the Bulgarian structure in (52b), in Franks & Rudin (2005) we argued that in clitic doubling the associate has vacated the KP/DP complex. In that paper we discussed various motivations for movement, the most significant being topicalization. Indeed, the basic function of clitic doubling in Bulgarian is to mark topics:

(76) a. Vanja ne ja vâlnuvat tezi nešta. ‘Vanja these things don’t excite (her).’

b. Na tjax toj im dava točno tolkova, kolkoto i na drugite. ‘To them he gives (them) exactly as much as to the others.’

One appealing reason for unifying the phenomena in (76) and (75) is that both involve deletion in the mapping to PF of a phrase which is the complement of a head that has undergone Spec–head agreement. In English (75a, c), the agree-ment depends on the features of C0 and D0, respectively, but in Bulgarian (76) agreement is introduced via movement of the associate. Specifically, the analy-sis of clitic doubling in Bulgarian (52b) is that the complement to D0 moves out, through SpecKP, inducing agreement, which is morphologically realized as the clitic in K0. Given this, we can treat even those Bulgarian clitics that have no overt associate as involving the movement of a (silent) topical associate, in short, a pro topic. This is the Bulgarian connection.

Bulgarian and Slovenian both have more articulated nominal structures, with two projections above NP. For Bulgarian, it is this additional structure that


allows for clitic doubling, since the extra layer enables the NP complement to escape without violating antilocality. This is represented in (77):

(77)

go D(max)

D

N(max)

K(max)

N(max) K

NP

Top(max) [Bulgarian]

Antilocality prohibits movement from complement to specifier position, intui-tively because such movement would not create a closer/new relationship between the probe and its goal. The KP–over–DP structure schematized in Bul-garian (77), however, allows NP to circumvent antilocality. And, while passing through SpecKP, the topic NP also participates in Spec–head agreement with K0; cf. Bošković (2005) for some relevant discussion of the interaction between the structure of NP and antilocality.

This account of Bulgarian suggests that antilocality should also be exploited in the analysis of the Slovenian OA. We began this section with the question of how to take advantage of IndefP in order to relate I/S and proN. I/S followed directly from the possibility of situating pronominal clitics in Indef0, but for proN to depend on IndefP implies that it too needs an escape hatch that is not overly local. In short, Slovenian proN must also vacate KP, something like (78):

(78)

K Indef(max)

Indef

proN

K(max)

pron K

pron

Top(max) [Slovenian]

172 Steven Franks

The general idea is that an additional intervening functional projection, tentatively IndefP, is what allows both for clitics to have the I/S reading and for silent proN topics. The Slovenian OA is a consequence of the latter. This account gives us a handle on P&O’s big question: Why Slovenian? It arises through the interaction of a more articulated nominal structure, relatively rich case morphol-ogy, quirks of pronominal declensional paradigms, and the Accusative Prediction Rule. Of course, the account still leaves unresolved certain questions, foremost among which is how exactly proN leads to the OA.⁵⁰ I consider this question more carefully in the Section 7.5 below. First, however, let us turn to some typological issues.

6.4 Is there anything like the Slovenian OA in other languages?

Fascination with the Slovenian OA construction is largely due to its rarity. Indeed it is very hard to find comparable phenomena in the grammatical system of any other language. Interestingly, Bosnian/Croatian/Serbian, its closest neighbor, does display some relative pronoun properties that suggest an affinity to the OA.

50 Another concerns the history of the OA, since the analysis predicts this to have arisen as a consquence of changes in the internal structure of the Slovenian noun phrase. While the OA is mentioned in Slovenian grammars as early as 1899 (L. Marušič, p. c.), the older (16th century onwards) candidates for OA status I am aware of – i.e., bare direct object adjectives in the geni-tive – have all turned out to be actually genitive. They are quantificational in nature and/or the complement to an indefinite pronoun (typically kar or kai/kaj ‘what’):(i) a. inu če bo eden kaj umazanigagen pustil … ‘and if one (of us) leaves something unwashed …’ b. Inu kar ie kuli dobrigagen, suetigagen, velikigagen, možnigagen v nebesih oli na zemli, tu

vse se od samiga Boga pride. ‘Whatever is good, holy, great, (or) possible in heaven or on earth, this all came from God

himself.’The appositives parallel contemporary Slovenian (ii) and the predicate adjectives parallel (iv), in contrast to agreeing (iii):(ii) a. nekaj hudegagen ‘something (specific) bad’ b. mnogo hudegagen ‘plenty bad’ c. Kupi mi karkoli dobregagen. ‘Buy me something (free choice) good.’(iii) a. To je dobronom. ‘That is good.’ b. Peter je dobernom. ‘Peter is good.’(iv) a. Nekaj je dobregagen. ‘Something is good.’ b. Mnogo je dobregagen. ‘Much is good.’Thanks especially to D. Reindl and L. Marušič (p. c.) for discussion.


In this section I discuss these and briefly mention a few other examples further afield.

6.4.1 Agreement with overt pronouns in Slovenian and Bosnian/Croatian/Serbian

P&O’s insight that some kind of agreement with a pronoun is involved in the OA finds support in the fact that similar agreement occurs when the pronoun is overt. They provide the following examples, where the pronominal clitic and a quanti-fier agree:

(79) a. Želel sem jo pojesti vsoacc. ‘I wanted to eat it (feminine) all up.’

b. Želel sem ga pojesti vsegagen/*vesacc. ‘I wanted to eat it (masculine) all up.’

Crucially, the floated quantifier vs – ‘all’ in (74b) appears in the genitive rather than the accusative, despite the fact that the clitic ga refers to something inani-mate. Also, compare (80b) with (23) above, repeated as (80a):

(80) a. Včeraj smo našli stolacc pomazanacc s krvjo. ‘Yesterday we found the chair stained with blood.’

b. Včeraj smo ga našli pomaznegagen/*pomazanacc s krvjo. ‘Yesterday we found it [=chair] stained with blood.’

Even though ga in (80b) is identical in reference to stol ‘chair’ in (80a), and as an adjective pomazn– ‘stained’ should have a choice of forms, it opts for the animate possibility, in agreement with ga. Presumably, the only difference between (79b, 80b), on the one hand, and the OA, on the other, is that the OA implicates a null proN, something peculiar to Slovenian.

If so, and everything else being more or less equal, we might expect to find agreement with overt pronouns even in languages that lack proN. Strikingly, Bosnian/Croatian/Serbian displays a similar agreement pattern here as Slove-nian. Consider the following paradigm:⁵¹

51 Thanks are due to W. Browne (p. c.) for pointing this out during the discussion of Rappaport’s 2009 SLS presentation.

174 Steven Franks

(81) a. Ostavio sam stol prevrnutanacc. ‘I left the table overturned.’

b. Ostavio sam ga prevrnutoggen/*prevrnutanacc. ‘I left it [=table] overturned.’

Bosnian/Croatian/Serbian (81) behaves exactly like Slovenian (80), in that the genitive case of the adjective in (81b) presumably reflects ga being treated as [+animate].

A second type of possibly related Bosnian/Croatian/Serbian example is given in (82), where the relative pronoun exhibits two acceptable agreement pos-sibilities: the genitive, reflecting morphological animacy, or the (nominative-like) accusative, reflecting inanimacy:

(82) stol kojegagen/kojiacc sam prevrnuo … ‘the table which I overturned …’

Why should this be, especially when the secondary predicate in (81) must agree with ga when it is present?

To answer this question, let us consider the two relativization strategies avail-able in Bosnian/Croatian/Serbian (along with many Slavic and other languages). First, one can employ an interrogative-like relative pronoun, which receives appropriate case and does not cooccur with a (resumptive?) clitic:⁵²

(83) a. učitelj kogagen (*ga) Lucija voli … ‘the teacher who Lucija (him) loves …’

b. stol kojiacc sam (*ga) prevrnuo… ‘the table I overturned (it) …’

Second, one can employ the invariant complementizer što. Here, interestingly, the object clitic can cooccur, (often) optionally if the antecedent is inanimate but (almost) always obligatorily if it is animate:⁵³

(84) a. učitelj što *(ga) Lucija voli … ‘the teacher that Lucija (him) loves …’

b. stol što sam (ga) prevrnuo… ‘the table that I overturned (it) …’

In Franks (in press), I argue for movement of a null operator ∅ to the SpecCP of the relative clause. Alternatively, à la Kayne, što-relativization could involve movement of the actual NPs učitelj and stol from object position. Regardless of

52 Cf. Franks (1995:82–83; in press) for discussion as well as Gračanin-Yüksek (2009).53 For discussion of the caveats, see Franks (in press) or Gračanin-Yüksek (2009), as well as Section 7.5 below.


how this relativization strategy is formalized, the parallelism between (84b) and (82) should be obvious.

6.4.2 PF effects of proN in other languages?

Extensive efforts to identify morphological consequences of missing pronouns in other languages only serve to highlight the obscurity of the Slovenian OA. The only well-known example I am aware of can be found in Spanish, where the defi-nite article lo instead of el is used with proN:

(85) a. lo grande [NP proN ] ‘the big one’

b. el grande [NP libro] ‘the big book’

c. el grande [NP libro] ‘the big one [= book]’

The traditionally “neuter” lo is used only with (presumably genderless) proN. Oth-erwise, agreement is required, whether the NP is overt or, as in (85c), elided.

In Koyra Chiini, a Songhai language spoken in Timbuktu, Mali, when an adjec-tive is used without a noun, the adjective must take a prefix i–. (86) is adapted from Dryer (2004), citing Heath (1999), who calls i– an “absolute” prefix:

(86) i- jeeno [NP proN ] di ‘the old one’ absol-old def

Dryer also cites Herault (1978) that in Adioukrou, a Kwa language spoken in Côte d’Ivoire, an NP with an adjective but no noun requires a definite article, but the definite article is optional if there is a noun present.

6.5 Final reflections

This paper has explored the Orphan Accusative construction from the perspec-tive of Slovenian itself, variation in South Slavic nominal phrases, and principles of Universal Grammar. It was argued that what makes Slovenian special in this regard is the availability in its lexicon of a proN, which is for all intents and pur-poses a silent version of English one. ProN is a silent non-referential topic. As such, it moves to SpecTopP through SpecKP. This movement is made possible despite antilocality because Slovenian is developing into a DP language and has an intermediate IndefP projection. Indef0 is the locus of the indefinite article

176 Steven Franks

en and also of clitics when they have the I/S interpretation. This analysis was coupled with P&O’s traditional observation that pronouns behave morphologi-cally as if they were animate (in this language and elsewhere in Slavic). ProN thus interacts with the Accusative Prediction Rule in (10) and repeated below, to give rise to the OA effect by virtue of modifier agreement with it:

(87) Accusative Prediction Rule

a. For animates, the accusative is like the genitive.

b. For inanimates, the accusative is like the nominative.

But why should a rule such as (87) exist in the first place?

6.5.1 The Accusative Prediction Rule and the role of surface morphology

It seems to me that ultimately surface morphology must be implicated in the OA. It is of course no accident that (87) only applies when there is no autono-mous accusative form. Thus, this is a rule of PF implementation that instructs the morphology to look to a different cell in the paradigm just in case, in seeking an accusative form, the rule encounters a gap. In general, then, (87) will pertain to masculine and neuter singular substantives. The rule itself is modulated by the feature [±animate], and all items which exhibit agreement with the nominal head will be affected. Thus, for example, in Russian (9b), repeated below, although the noun papa, by virtue of its declensional class, has an independent accusa-tive, the adjective vaš– ‘your’ does not. However, since in (88) it is [+masculine, +animate], the genitive form vašego is accessed:

(88) Ja vižu vašegoacc/gen papuacc. ‘I see your dad.’

So far so good. But why do pronouns fail to reflect natural animacy, opting to behave as if [+animate] instead? This mysterious fact is crucial not only to the Slovenian OA, but to the workings of Slavic pronouns in general. Neither P&O nor Rappaport have anything particular to say about this, beyond the stipulation itself.

I suggest that the reason is because, when the referent of the pronoun is inanimate, (87b) actually fails to give any result at all. Slavic pronouns are, as dis-cussed above, essentially case morphemes. But I contend that there are no nomi-


native case morphemes in Slavic, these being invariably phonetically null.⁵⁴ As proposed in Franks (1995), the –a, –o, zero, –a and –i of Russian kniga ‘book’, zerkalo ‘mirror’, stol ‘table’, zerkala ‘mirrors’ and knigi ‘books’, respectively, are in fact gender/number class markers, not case endings. This is why caseless verbal participles and caseless (“short form”) predicate adjectives utilize them: these items agree only in gender/number. The Accusative Prediction Rule has no dif-ficulty here, however, because it is able to return a concrete form despite the lack of morphological case. With pronouns, however, there is a problem: since there is neither substantive stem nor functional material desinence, there can be no morphological exponent at all. All Slavic languages solve this problem with sup-pletion, coopting a demonstrative stem for the third person nominative.⁵⁵ This is typically neutral n– or distal t – (also proximal v –), e.g., Slovenian on ‘he’ or Bul-garian toj ‘he’. My claim, then, is that P&O’s (11a), which states that “Slovenian pronouns behave morphologically as if they were animate” is the by-product of a morphologically impoverished paradigm: it is resorted to just in case there is no nominative in the paradigm at all. Instead of ineffability, the situation is rescued by a morphological rule which changes the feature value [–animate] to [+animate], allowing access to the masculine/neuter singular genitive form in the pronominal paradigm. Interestingly, if this scenario is on the right track, it should technically be only those pronouns which lack a distinct accusative entry that are intrinsi-cally [+animate]. Hence, whenever a pronoun does have a distinct accusative – as do many Slavic pronouns, depending on the language – (87) is not invoked and Spell–Out has no occasion to avail itself of the rule assigning [+animate]. This means that, unlike ga, the Slovenian clitic pronoun jo ‘her, it’, for example, can be either [+animate] or [–animate]; unfortunately, since animacy plays no role in the grammar here that I can think of, there is no way to tell. Thus, calling the rule a “lexical redundancy rule”, as Rappaport (2009) does, is not exactly accurate: this is not a property of the Slavic lexicon and there is no redundancy rule which would guarantee “[+pronoun] ↔ [+animate]”. Rather, this should be thought of in the terms of Distributed Morphology: the feature [–animate] is liter-ally changed to [+animate], thereby enabling (87) to return an output form.

Such superficial morphological effects are commonplace cross-linguistically, and indeed, this is precisely the sort of mechanical result Distributed Morphology was designed to achieve. Let us briefly consider again the problem of clitics in što-relatives discussed at the end of Section 7.4.1. One might think that, since učitelj

54 Bošković (p. c.) speculates out that my contention about nominative morphology in Slavic might follow from the absence of TP, a claim he has made for more general typological reasons.55 First and second person, already being naturally [+animate], present no problem for (87) (al-though they too are suppletive in the nominative).

178 Steven Franks

in (84a) is naturally animate, morphological animacy is obligatory, forcing the genitive ga. If so, one would conclude that in (84b), with referentially inanimate stol, morphological animacy is optional, so that ga is optional. Closer examina-tion, however, reveals that this is not what is going on, and that very superfi-cial aspects of the antecedent’s case paradigm are in fact driving the system. The correct observation, due to Gračanin-Yüksek (2009) and discussed in Franks (in press), is that accusative pronominal clitics can be dropped in što-relative clauses in Bosnian/Croatian/Serbian just in case the form of the nominal head of the relative clause is appropriate to the gap in the relative clause. For example, učitelj ‘teacher’ in (84a) is distinct from accusative-genitive učitelja, whereas stol ‘table’ in (84b) is not. However, since želja ‘desire’ is distinct from accusative želju in (84a), the accusative clitic je cannot be dropped. On the other hand, although it is equally feminine and inanimate, ljubav ‘love’ in (89b) belongs to a different paradigm which has no distinct accusative form, so je can be dropped.

(89) a. [Želja što sam *(je) osjetio] bila je jaka. ‘The desire that I felt was strong.’

b. [Ljubav što sam (je) osjetio] bila je jaka. ‘The love that I felt was strong.’

The only way that this generalization can be implemented is by literally consult-ing the surface morphological form of the antecedent želja or ljubav.

6.5.2 Agreement and the case of proN

Little has been said about the agreement of modifiers with proN; like other schol-ars who have studied the Slovenian OA, I have simply assumed that whatever agreement mechanisms are at work in general also pertain to this construction. That may be, but since I have also argued that, as a topic, proN must move, an account of modifier agreement that derives from the movement (possibly involv-ing Spec–head agreement) could surely be developed. This would depend, however, on more detailed assumptions about the position of modifiers in the nominal projection than I care to make in this paper. I therefore put the possibil-ity aside for future research.

Let us return, finally, to the outstanding problem posed by the Orphan Accu-sative: Why is it that proN – which after all has no surface morphological form – ends up as [+animate]? Here I can only guess at what is going on, since at some level, it just does. In the GB tradition it has long been established that empty cat-egories have case and phi-features. For example, as demonstrated through predi-


cate adjective agreement, we can tell that arbitrary PRO in Italian is masculine and plural and that in Russian it is dative and masculine singular:

(90) a. Come smetter di [PRO essere nervosim.pl]? ‘How to stop being nervous?’

b. Nevozmožno [PRO perejti ètot most samomudat.m.sg]. ‘It is impossible to cross this bridge by oneself.’

One possibility is that proN is simply [+animate].⁵⁶ Alternatively, whenever proN is [–animate] by virtue of the natural gender of its antecedent, the morphology ends up changing that feature value to [+animate]. In the system I have outlined, one might imagine a solution in which (87a) applies first, seeking a nominative form if [–animate]. Only if that fails does the morphological rule change the value to [+animate] in attempt to exploit (87b). With overt pronouns this strategy is suc-cessful, providing the genitive entry as its output. Spell–Out of proN, of course, will find nothing morphological to match with, be it [–animate] or [+animate]. Perhaps, then, whenever proN is [–animate], since the first clause of the Accu-sative Prediction Rule returns nothing, the morphological rule that changes the value to [+animate] is invoked. This is of course a vain attempt, since (87b) returns nothing as well. ProN nonetheless remains animate, as evidenced by the agreement of adjectives and is, in the end, exactly the Orphan Accusative effect grappled with in this paper.⁵⁷

ReferencesBaltin, Mark & Jeroen van Craenenbroeck (2008). On becoming a pronoun: Towards a unified

theory of ellipsis. Talk given at 10th CUNY/SUNY/NYU miniconference. CUNY Graduate Center, NY.

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: Proceedings of the 37th annual meeting of the North East Linguistic Society, vol. 1, ed. by Emily Elfner & Martin Walkow, 101–114. Amherst: University of Massachusetts, Graduate Linguistic Student Association.

56 This is the appeal of simply forcing the issue with a lexical redundancy rule.57 Note that under this scenario every proN ends up [+animate], regardless of case or phi-features. Once again, I can think of no way to test this.

180 Steven Franks

Bošković, Željko (2009). The NP/DP analysis and Slovenian. In Proceeding of the University of Novi Sad Workshop on Generative Syntax 1, pp. 53–73.

Bošković, Željko (2011). Rescue by PF deletion, traces as (non) interveners, and the that-trace effect. Linguistic Inquiry 42: 1–44.

Chomsky, Noam. (1995). Bare phrase structure. In Evolution and Revolution in Linguistic Theory, ed. by Héctor Campos & Paula Marie Kempchinsky. Georgetown University Press, pp. 51–109.

Déchaine, Rose-Marie & Martina Wiltschko (2002). Decomposing pronouns. Linguistic Inquiry 33: 409–442.

Despić, Miloje (this volume). Intensifiers, focus and clitic pronouns: Do pronouns move to D in Serbo-Croatian?

Dryer, Matthew (2004). Noun phrases without nouns. Functions of Language 11.1: 43–76.Franco, Jon (2000). Agreement as a continuum: The case of Spanish pronominal clitics. In Clitic

Systems in European languages, ed. by Frits Beukema & Marcel den Dikken, pp. 147–89.Franks, Steven (1994). Parametric properties of numeral phrases in Slavic. Natural Language

and Linguistic Theory 12.4: 570–649.Franks, Steven (1995). Parameters of Slavic Morphosyntax. Oxford University Press.Franks, Steven (1999). X’-invisibility, derivational c–command and one –substitution. Linguistic

Analysis 29: 1–8. Franks, Steven (2001). The internal structure of Slavic NPs, with special reference to

Bulgarian. In Generative Linguistics in Poland: Syntax and Morphosyntax, ed. by Adam Przepiórkowski & Piotr Bański, pp. 5–69.

Franks, Steven (2002). A Jakobsonian feature based analysis of the Slavic numeric quantifier genitive. Journal of Slavic Linguistics 10: 141–181.

Franks, Steven (2007). Deriving discontinuity. In Studies in Formal Slavic Linguistics, ed. by Franc Marušič & Rok Žaucer. Peter Lang: Frankfurt, pp. 103–120.

Franks, Steven (2009). Macedonian pronominal clitics as object agreement markers. In A Linguist’s Linguist: Studies in South Slavic Linguistics in Honor of E. Wayles Browne. ed. by Steven Franks, Vrinda Chidambaram & Brian Joseph. Bloomington, IN: Slavica, pp. 189–221.

Franks, Steven (in press). Dynamic Spell–Out as Interface Optimization. To appear in Peter Kosta & Lilia Schürcks, eds. Selected Papers from FDSL 8. Berlin: Peter Lang.

Franks, Steven & Asya Pereltsvaig (2004). Functional categories in the nominal domain. In Formal Approaches to Slavic Linguistics: The Ottawa Meeting, ed. by Olga Arnaudova et al, pp. 109–128.

Franks, Steven & Tracy Holloway King (2000). A Handbook of Slavic Clitics. Oxford University Press.

Franks, Steven & Catherine Rudin (2005). Bulgarian clitics as Kº Heads. In Formal Approaches to Slavic Linguistics: The Columbia Meeting, ed. by Steven Franks, Frank Gladney & Mila Tasseva-Kurktchieva, pp. 104–116.

Giannakidou, Anastasia & Melita Stavrou (1999). Nominalization and ellipsis in the Greek DP. The Linguistic Review 16: 295–331.

Gračanin-Yüksek, Martina (2009). On a matching effect in headed relative clauses. Paper presented at Formal Approaches to Slavic Linguistics 18, Cornell University, Ithaca, NY, May 2009.

Heath, Jeffrey. 1999. A Grammar of Koyra Chiini: the Songhay of Timbuktu. Berlin: Mouton de Gruyter.


Herault, G. 1978. Elements de Grammaire Adioukrou. Abidjan: Institut de Linguistique Appliquée.

Herrity, Peter (2000). Slovene. A Comprehensive Grammar. Routledge Grammars.Kagan, Olga (2007). A modal analysis of genitive case in Russian. In Linguistic Investigations

into Formal Description of Slavic Languages, ed. by Peter Kosta & Lilia Schürcks. Berlin: Peter Lang, pp. 217–226.

Lobeck, Anne (1990). Functional heads as proper governors. In Proceedings of NELS 20, GLSA, UMass, Amherst, pp. 348–362.

Marušič, Franc & Rok Žaucer (2009). On clitic doubling in Gorica Slovenian. In A Linguist’s Linguist: Studies in South Slavic Linguistics in Honor of E. Wayles Browne. ed. by Steven Franks, Vrinda Chidambaram & Brian Joseph. Bloomington, IN: Slavica, pp. 281–296.

Marušič, Franc & Rok Žaucer (this volume). A definite article in the AP – evidence from colloquial Slovenian.

Nunes, Jairo (2004). Linearization of Chains and Sideward Movement. Cambridge, MA: The MIT Press.

Panagiotidis, Phoevos (2003a). One, empty nouns, and Θ-assignment. Linguistic Inquiry 34: 281–292.

Panagiotidis, Phoevos (2003b). Empty nouns. Natural Language and Linguistic Theory 12: 381–432.

Perlmutter, David & Janez Orešnik (1973). Language-particular rules and explanation in syntax. In A Festschrift for Morris Halle, ed. by Stephen Anderson & Paul Kiparsky. Holt, Rinehart and Winston: New York, pp. 419 –459.

Peti-Stantić, Anita (2009). Which case is the so-called “orphan accusative” in Slovene? Talk presented at American Association for the Advancement of Slavic Studies, Boston.

Progovac, Ljiljana (1998). Determiner phrase in a language without determiners. Journal of Linguistics 34: 165–179.

Rappaport, Gilbert (2009). The “orphan accusative” in Slovene: Grammatical features lexicalized. Talk presented at Slavic Linguistics Society 4, Zadar.

Toporišič, Jože (2004). Slovenska slovnica. Založba Obzorja: Maribor.Zwicky, Arnold M. (1987). The Slovenian orphan accusative, component interfaces, and

covert grammatical categories. OSU Working Papers in Linguistics 35: 29–38. Indiana University – Bloomington

Documents

Steven Franks The Slovenian Orphan Accusative, Empty ... orphan accusative… · where the expected accusative is overridden by the genitive case: the genitive can be assigned instead