193 Marijana Marelj UDK 811.163'36'37 Utrecht University* DOI: 10.4312/linguistica.56.1.193-209 IN THE SOUTH SLAVONIC GARDEN: LANDSCAPING THE LANDSCAPE OF ARGUMENTS AND NON-ARGUMENTS** 1. INTRODUCTION Over the last couple of decades, the phenomenon of Cognate Object Construction (COC) has been extensively discussed in theoretical linguistics. Though widely dis- cussed, even the status of the relevant nominals remains a matter of controversy. Re- searchers disagree not only on the syntactic and semantic status of what is referred to as Cognate Objects (COs), but also on the judgments regarding them: (1) a) *An uneventful life was lived by Harry. (Jones 1988: 91) b) A good life was lived by Susan. (Rice 1988: 210) In this paper, I address syntactic, semantic, and pragmatic puzzles that COs raise and also demonstrate that judgments as in (1) are not accidental or idiolectal. Rather, when viewed through the prism of languages under consideration here, they become congruent and can straightforwardly be explained. The core of my analysis is the following. Based on the data from Serbo-Croatian (SC) and Slovene (SLO), I identify two types of COs: i) arguments (ACOs) & ii) predi- cates (non-ACOs). The latter are first order predicates, while the former are arguments in the neo-Davidsonian sense. The notion of “cognatehood” is irrelevant in syntactic terms for either group, though it becomes relevant for the latter in a pragmatic sense. Syntactically, the non-ACOs are adjuncts; the fact that they cross-linguistically occur with unaccusatives is unproblematic. As for ACOs, I argue that verbs with which they appear are regular transitive and regular unergative verbs (see Marelj 2015). The paper is organised as follows: In section 2, I present core issues that theories on COs must address. In section 3, based on SC and SLO data, I discuss CO-tests. The results of this section lead me to identify two types of COs, while in section 4 I tackle non-ACOs and in section 5 I tackle ACOs. Finally, in section 6, I present the conclu- sion of this paper. * m.marelj@uu.nl ** I am grateful to the three anonymous reviewers of Linguistica for their valuable input, Marko Hladnik for the Slovene and Ora Matushansky for the Russia data. This research was partially supported by the Aspasia grant, awarded to me by the Netherlands Organisation for Scientific Research (NWO), which I gratefully acknowledge here. Any errors that remain are my sole re- sponsibility. Linguistica_2016_FINAL.indd 193 28.12.2016 8:57:53 194 2. MYSTERIOUS MYSTERIOUSNESS Crosslinguistically, it is examples like (2) – featuring intransitive verbs accompanied by etymologically related nouns – that are typically found in the literature illustrating COC. The morphological relatedness is not the only characteristic of the classical cases of COs. Though there is a well-known generalisation that COs appear only with uner- gatives (see Hale and Keyser 2002 a.o.), it is the unaccusative die as in (3) that often crops up at the most typical example of COC. (2) a) He dreams a dream. b) Živi život. [SC] lives.3sg life ‘He lives a life.’ (3) a) He died a natural death. b) Umro je prirodnom smrću. [SC] die.prt.m.sg AUX.3sg natural death ‘He died a natural death.’ 2.1 Scarsity of Syntactic Space The well-formed nature of examples such as (3) raises the question of the scarcity of syntactic space. Since the sole syntactic argument of unaccusatives (4b) begins its life as a complement of a verb, it is unclear how, to start with, “an additional complement” would enter the structure. (4) a) He [vP/VP he dreamt] [Unergative] b) He [VP [V’ died he] [Unaccusative] The fact that unergatives and unaccusatives are quite different is uncontroversial. One of the tests that seems to be cross-linguistically valid pertains to the fact that the past participial can be used attributively only with the internal arguments of relevant verbs. This prediction is borne out in (5). (5) a) the melted snow/the departed guest b) *the shouted child Slavonic languages behave on a par with English. For instance, a past participle in SC (active and passive alike) modifies internal arguments only (6):1 1 Under an analysis such as (4a), the question of syntactic space does not seem to arise for unega- tives. However, considering the fact that the prevelant view of unergatives is rooted in Hale and Keyser’s work (i), this problem arises for unergatives as well. I return to this issue in 5.2. (i) V V  V N V N laugh laugh Linguistica_2016_FINAL.indd 194 28.12.2016 8:57:53 195 (6) a) *plesani čovek danced man b) razbijeni/pali andjeo broken/fallen.prt.m.sg angel ‘a broken/fallen angel’ 2.2 What’s in a Modifier? COs do not differ only in their syntax. Smrt (death) cannot possibly occur unmodified: (7) #/*Umrl je smrti [SLO] die.prt.m.sg AUX.3sg death This behaviour is neither restricted to umreti (die) nor to SLO. However, the obliga- tory nature of modification does not hold in all instances of COs. This curious behav- iour has been left unexplained. 3. LANDSCAPING THE LANDSCAPE… As the data from SC and SLO are intended to be used as a sort of “litmus” test for the deliberation on the status of the data like (1), it is only prudent to start by submitting the relevant data to the standard tests on COs. 3.1 Classifying Cognate Objects Examples in (8) – (11) represent the most canonical tests on COs. As “die” and “dream” are typically found in the literature on English, below I use their equivalents in SC and SLO as well. (8) Passivisation: a) San je sanjan. [SC] Dream AUX.3sg dreamt.prt.m.sg ‘The dream was dreamt.’ b) *Užasna smrt je umrna horrible death AUX.3sg died.prt.f.sg c) Sanje so sanjane. [SLO] dream.pl AUX.3pl dreamt.prt.f.pl ‘The dreams were dreamt.’ d) *Strašna smrt je umrta horrible death AUX.3sg died.prt.f.sg Linguistica_2016_FINAL.indd 195 28.12.2016 8:57:53 196 (9) Pronominalisation: a) Svake noći je sanjao zastrašujući san. [SC] every night AUX.3sg dreamt.prt.m.sg terrifying dream ‘Every night he dreamt a terrifying dream.’ a’) Svake noći ga je sanjao. every night CL.acc.m AUX.3sg dreamt ‘Every night he dreamt it.’ b) Svake noći je umirao užasnom smrću. Every night AUX died horrible death ‘Every night he died a horrible death.’ b’) *Svake noći je njom umirao. Every night AUX.3sg CL.instr.f. died.prt.m.sg. c) Vsako noč je sanjal strašne sanje. [SLO] every night AUX.3sg dreamt.prt.m.sg terrifying dream ‘Every night he dreamt a terrifying dream.’ c’) Vsako noč jih je sanjal. every night CL.acc.f.pl AUX.3sg dreamt ‘Every night he dreamt them.’ d) Vsako noč je umrl strašne smrti. Every night AUX died.prt.m.sg horrible death ‘Every night he died a horrible death.’ d’) *Vsako noč jo je umrl. Every night AUX.3sg CL.gen.f died.prt.m.sg. (10) Definiteness restriction: a) Sanjao je taj/ovaj zastrašujući san. [SC] dreamt AUX.3sg this/that terrifying.acc dream.acc ‘He dreamt that/this terrifying dream.’ b) *Umro je tom/ovom užasnom smrću. die.prt.m.sg AUX.3sg this/that horrible.instr death.instr c) Sanjal je te/tiste strašne sanje. [SLO] dreamt.prt.m.sg AUX.3sg this/that terrifying.acc dream.acc He dreamt this/that terrifying dream. Linguistica_2016_FINAL.indd 196 28.12.2016 8:57:53 197 d) *Umrl je te/tste strašne smrti. die.prt.m.sg AUX.3sg this/that horrible.gen death.gen (11) Can be questioned:2 a) Šta je sanjao? [SC] what AUX.3sg dream.prt.m.sg ‘What did he dream?’ b) *Šta je umro? What.ACC AUX.3SG died.prt.m.sg c) Kaj je sanjal? [SLO] what AUX.3sg dream.prt.m.sg ‘What did he dream?’ d) *Kaj je umrl? what.acc AUX.3sg died.prt.m.sg Based on (8) – (11), the behaviour of sanjati (dream) i umreti (die) in SC and SLO seems consistent with the behaviour of their English equivalents. The data seem to require a sort of a “hybrid” approach, since some of the COs behave like run-of-the- mill arguments, while others behave like non-arguments/adjuncts. Before I address the two types of COs in SC and SLO, however, I must first address the relevance of the morphological case. 3.2 What’s in a Case? Though (8)–(11) seem to indicate that English, on the one hand, and SC and SLO, on the other hand, pattern fully alike, this is not entirely true. The data in these morpholog- ically robust languages seem to give rise to another generalisation; the “real” arguments appear in ACC, while the non-arguments appear in INSTR (SC) and GEN (SLO). At first blush, then, it seems that we have an additional morphosyntactic test to differenti- ate the argumental from the non-argumental instances of COs, thus having another piece of evidence to support that the morphological structure reflects the underlying structure. For instance, it is typically argued that non-verbal predicates bear INSTR as a predicate case (see Pereltsvaig 1999 for the discussion and references on Russian) (12). 2 I kindly ask (native) speakers of SC and SLO to grant me a bit of patience. Though they might exclaim at this point that what happens in (11b) and (11d) is an instance of a case mismatch that goes beyond the issues at stake here, I would like to remind them that leaving such examples in the text here underscores the relevance of exploring Slavonic languages like SC and SLO, since the traditional generalisation about COC in the literature (see Jones 1988, for instance) is that languages that express the morphological case overtly, such as Arabic, German, or Latin, select ACC, “a semantically empty Case to satisfy morphological requirements” (see Jones 1988: 109, but also Moltman 1989, for instance). Linguistica_2016_FINAL.indd 197 28.12.2016 8:57:54 198 (12) On stal učitelem [Russian] He become.prt.m.sg teacher.instr ‘He became a teacher.’ I argue here that a correlation between some “designated” predicate case and non- verbal predicates in South Slavonic simply does not hold. Contrary to standard views in the literature on Slavonic, I argue that there is no deeper relevance or meaning to the case marking here. The overt morphology here is simply idiosyncratic. Trivially, to begin with (quite like in other Slavonic languages), instrumental is not restricted to non-verbal predicates. For instance, verbs such as vladati (rule), upravljati (manage), dominirati (dominate) and trajati (last) in SC have INSTR internal arguments. Not only is the correlation not two-directional, but, more importantly, it is not one- directional either. Though there are a couple of exceptions as in (13), generally, non- verbal predicates in SC occur as NOM, not INSTR. (13) Postao je učitelj/učiteljem become.prt.m.sg. AUX. teacher.nom/teacher.instr ‘He became a teacher.’ Slovene underscores this point about the lack of the correlation in this respect even more strongly. Firstly, the canonical cases of non-arguments under consideration here in Slovene do not even consistently appear in the same “alleged” predicate case. Rath- er, some are genitive (14a), while others are prepositional instrumental (14b). (14) a) Umrl je naravne smrti. [SLO] die.prt.m.sg AUX.3sg natural death.gen ‘He died a natural death.’ b) Rasel je z neustavljivo rastjo.3 grew.prt.m.sg AUX.3sg with unstoppable growth.instr ‘He grew the unstoppable growth.’ Morevoer, the cases of INSTR non-verbal predicates like Russian (12) or even in SC (13) are always nominative.4 3 Cross-linguistically, it seems that the data on unaccusatives are met with varying reactions from native speakers. It is not obvious to me why this is so. Note that this is also the case with (14b), where R1 and my native spearker informant differ. 4 Needless to say, quite like SC, SLO also has “real” complements in genitive (ia) and preposi- tional instrumental case (ib): (i) a) Boji se Petra. [SLO] fear.3sg CL Peter.gen ‘He fears Peter.’ b) Upravljal je s tovarno. Managed prt.m.sg AUX.3sg with factory.instr ‘He managed a factory.’ Linguistica_2016_FINAL.indd 198 28.12.2016 8:57:54 199 (15) Postal je učitelj. [SLO] became AUX.3sg teacher.nom ‘He became a teacher.’ 4 ZOOMING IN ON NON-ACOS 4.1 Where Adverbs and Cognates Meet As any native speaker of SC and SLO can confirm (see also fn. 3), the ungrammatical- ity of (11b) and (11d) seems far more trivial than that of the English (16). INSTR- and GEN-nominals cannot be questioned using ACC wh-word šta/kaj (what.ACC) and wh- word čime (what.INSTR)/česa are required. (16) *What did he die? (Massam 1990:164) Note importantly, however, that the use of the morphologically “appropriate” wh- word will not improve the grammaticality of SLO and SC examples (17). (17) a) *Čime je umro? Užasnom smrću. [SC] what.instr AUX.3sg die.prt.m.sg terrible.instr death.instr b) *Česa je umrl? Strašne smrti. [SLO] what.gen AUX.3SG die.prt.m.sg horrible.gen death.gen The way to question the relevant SC and SLO examples is by using the wh-word “how”: (18) Kako je umro? Užasnom smrću. [SC] How AUX.3sg died.prt.m.sg horrible death.instr ‘How did he die? A horrible death.’ What does the use of this wh-element tell us about the syntax and semantics of non- ACOs? Wh-word kako (how) appears with adverbial modification – specifically, with manner adverbs. Consequently, interpretatively, non-ACOs seem to be comparable to manner adverbs (19). (19) Kako se ponaša? Pristojno/loše... [SC] how CL behave.3sg decently/badly ‘How does he behave? Decently/badly….’ The fact that examples such as (20a) can be paraphrased using sentences like (20b) further underscores the parallelism: Linguistica_2016_FINAL.indd 199 28.12.2016 8:57:54 200 (20) a) Smejao se grohotnim smehom. [SC] laugh.prt.m.sg CL loud.instr laugher.instr ‘He laughed a loud laughter.’ b) Grohotno se smejao. loudly CL laugh.prt.m.sg ‘He laughed loudly.’ The question that now arises is the following: How should this parallelism be un- derstood and formalised? 4.2 On Adverbs and Analogies Since we have noted the parallelism between manner adverbs and non-ACOs, we can explore whether or not an approach to the latter can be modelled based on the estab- lished approach to the former. The most elegant approach to manner adverbs originates in Davidson (1967). In Davidson’s original proposal, the event argument is accommodated by analysing the predicate as having one more argument place than is assumed in traditional analyses. The event argument (e) is existentially quantified, with the result that a sentence like (21a) takes on a logical structure, such as in (21b). Manner adverbs are added conjunc- tively and predicate the event argument directly, as in (21c). (21) a) Tristram ate a snowflake (carefully). b) ∃e eat (Tristram, a snowflake, e) c) ∃e [EAT (e, T, s) & carefully (e)] The neo-Davidsonian tradition (Higginbotham 1987, Parsons 1990) deviates from the original arguments into separate conjuncts as well: (22) ∃e [ working (e) & Agent (e)=T & Theme (e)=s & carefully (e)] The status of event argument modifiers seems to be exactly the right “description” of what non-ACOs in SC and SLO do. Indeed, my hypothesis is that the non-ACOs in SC and SLO are event predicates.5 Let us explore this hypothesis further. 4.3 Referentiality, or One More Look at the Pronominalisation Test Predicates are non-referential. For nominal predicates, this typically means that they are indefinite (see Higginbotham 1987 a.o.). Cannonical cases that illustrate this in- clude the bare NPs in languages like Dutch, which appear in predicative, but are barred from the argument positions. Though using bare NPs will not be a test in SC 5 See Moltmann (1989) for the same conclusion regarding English and German COs of verbs like “die”. Linguistica_2016_FINAL.indd 200 28.12.2016 8:57:54 201 and SLO, using pronominalisation as a test is a good way to establish the semantic status of a nominal.6 The rationale is straightforward: pronouns canonically refer to individuals and not to properties. Arguments always have referents and, consequently, can be pronominal- ised. Predicates do not have referents and hence cannot be pronominalised (23). (23) a) Postao je matematičar/*on [primary predicate] became.prt.m.sg AUX mathematician/he ‘He became a mathematician/*he’ b) Smatraju Mariju lepoticom/*njome [secondary predicate] consider.3.pl Marija.acc beauty.instr/her.instr ‘They consider Maria a beauty/*her’ Note now that the non-ACOs behave on a par with the predicates (24) and quite unlike the regular arguments (25) with respect to pronominalisation (recall also (9) please) and (impersonal) passivisation (recall also (8) please). (24) a) Nasmejao se grohotnim smehom/*njim. [INSTR-non-ACO] laughed.prt.m.sg CL loud laugher.instr/him.instr ‘He laughed a loud laughter/*it.’ b) *Grohotnim smehom je/se nasmejalo. loud laughter.instr AUX/CL laughed.prt.n.sg (25) a) On upravlja fabrikom/njom. [INSTR-argument] he manage.3SG factory/it-INSTR ‘He manages a factory/it.’ b) Fabrikom je upravljano (od strane radnika). factory.instr AUX managed prt.n.sg (by the workers) ‘The factory was managed by the workers.’ In what follows, I demonstrate how two important properties of non-ACOs (see sec- tion 2) follow from the simple hypothesis that non-ACOs are event predicates. 4. 4 Mystery of Modification The modification of non-ACOs need not always be appositive (26), but must always be present. The data below are from SC, but this is cross-linguistically true. 6 There is a known caveat here; it has been long established that there are pronouns the denotation of which is a property, rather than an individual (see Jespersen 1927; Williams 1983, a.o.), but [+human] pronouns are restricted to individuals. Linguistica_2016_FINAL.indd 201 28.12.2016 8:57:54 202 (26) Umro je smrću mučenika. die.prt.m.3sg AUX.3sg death.instr martyr.gen ‘He died a martyr’s death’ Though sentences like (7) are typically marked as ungrammatical in the literature, my assumption is that there is nothing wrong with their structure. I argue that they are pragmatically odd and that this oddness is the result of an interplay between their se- mantic status as predicates and their cognatehood. The relevant questions are the following: a) Is it relevant for grammar as such that non-ACOs are etymologically related to relevant verbs or is this relatedness an idiosyncratic feature with no relevance whatsoever? b) Why is the modification of ACOs obligatory? My hypothesis is that these two questions are directly related. Namely, I only par- tially agree with Perelsvaig’s (Perelsvaig 1999) conclusion that cognatehood is com- pletely and fully irrelevant and inconsequential. Specifically, I concur with her con- clusion that relatedness of relevant nominals is irrelevant in the cases of ACOs, the structural position and the denotation of which are no different than those of run-of- the-mill complements unrelated to the relevant verb. I argue, however, that the same rationale does not hold for non-ACOs. Their cognatehood becomes relevant since, in the absence of a modifier, an utterance like (27a) becomes pragmatically infelicitous, as it violates the Gricean Maxim of Quantity. (27) a) #Tristram died a death. b) Tristram died a horrible death. Just like manner modifiers, non-ACOs predicate the event argument directly. In the absence of the modifier, however, this result is uninformative, since the primary predi- cate (verb) and the secondary predicate (nominal) are identical (28a).7 The relevant part of the sentence that tells us how the event of dying is happening – be it “naturally” or “horribly” – rests on the modifier, not the nominal predicate (28b). (28) a) #Tristram je umro smrću. [SC] a’) ∃e [ dying(e, T) & death (e)] b) Tristram je umro užasnom smrću. b’) ∃e [ dying(e,T) & horrible death(e)] 7 My use of the notion «cognatehood» should be taken as a very crude shorthand. Lack of space prevents me from further discussing this, but rather than morphological or etymological related- ness, «identical» here should be understood roughly as «something that is presupposed» by the primary predicate. This is the reason why #Opomenuo je glasom/He warned her in a voice is infelicitous, whereas Opomenuo je tihim glasom/He warned her in a soft voice is fine. Linguistica_2016_FINAL.indd 202 28.12.2016 8:57:54 203 4.5 Scarcity of Syntactic Space Revisited Being modifiers, non-ACOs are expected to be adjuncts. This is a desireable outcome, as they cross-linguistically occur with unaccusatives. (29) vP/VP/VPUNACCUSATIVE v’/V’/VP ADJUNCT VUNACCUSATIVE COMPLEMENT It is completely irrelevant for the purposes of our discussion here if we think of an adjunct as a sister to V’, under the direct dominance of VP and the linearisation of which is different from that of arguments or if we treat them as creatures that involve a different plane altogether. Importantly, syntactically, the position that non-ACOs oc- cupy is never that of internal arguments.8 5. ZOOMING IN ON ACOS 5.1 True Complements Quite like regular complements, ACOs occur with strong determiners, like (30) for instance, and pass other tests other run-of-the-mill arguments do as well (recall (8) – (11)). (30) Sanjal je vse sanje (ki so kdaj obstajale). [SLO] dream.prt.m.3sg AUX.3sg every dream.acc (that every existed) ‘He dreamt every dream (that every existed)’ 5.2 The Syntax of ACOs So far, we have established that ACOs behave like regular complements. The question that arises now is: What is their syntactic structure? Since the dominant analysis of unergatives is the one developed by Hale and Keyser (Hale and Keyser 1993 et seq.), according to which the relevant unergatives involve a process of incorporation (31), this questions is far from trivial. (31) VP V N DO dance 8 (29), even as a simplification, is not the only possible structure, but for the purposes of our discus- sion, it is sufficient. Linguistica_2016_FINAL.indd 203 28.12.2016 8:57:54 204 Namely, if their analysis is correct, then there is no syntactic space for the ACO as the bare noun already occupies the position of the object. Hale and Keyser (1997) of- fer to solve this problem by using “index deletion”, which allows the trace left by the incorporation of dance into the semantically empty verb DO to be obliterated (invisible to syntax and interpretation). The process of the index deletion then allows the cognate object to be generated in the base position of the bare noun. For general theoretical reasons, this is not an optimal solution (see also Hale and Keyser 2002). Hale and Keyser (2002) modify the account of Hale and Keyser (1997) by splitting the group of cognate objects into those that are “zero-related” and those that do not share a root with the relevant verb, then arguing that only the former, but crucially not the latter, are the results of their incorporation mechanism. Under this account, then, one could expect “dance a dance” and “dance a Mazurka” to have different syntactic structures and “derivational history”. Only in the latter case could the verb “dance” be directly generated as a “fully-fledged” verb. Something like “dance a dance” should behave like an incorporation, whereas “dance a Mazurka” should behave like a regu- lar transitive verb. However, if put through the standard tests (as those illustrated in (8) – (11)), the two behave completely on a par. Furthermore, this theory provides no explanation as to why the verb “dance” would not be able to be directly generated in the position of the verb in the “zero-related” cases to begin with. Morevoer, one might say that the behaviour of the proverbial “die a gruesome death” and “smile a silly smile” actually argues against incorporation analysis for COs even more strongly than other pieces of data do. Discourse referentiality (32) is typically taken to be a diagnostic of N(oun) I(ncoproration) as a syntactic phenomenon: (32) ngii-moonahapnii mii dash ngii-giziibiiginigan [Ojibwe] n- gii- moonah -apnii -e mii dash 1- PST- dig -potato -VAI and then n- gii- giziibiiginig -an 1 PST- wash -3pl ‘I dug up potatoes, and then I washed them.’ (BJ, 2008-12-17, cited in Barrie & Matheu 2015:4) However, if discourse referentiality is characteristic of NI, then non-ACOs clearly do not seem to behave like NIs, as they are never referential (recall (9)) and conse- quently cannot be picked up in subsequent discourse (33): (33) a. *He smiled a happy smile and then her son smiled it too. An alternative to incorporation analyses is a transitive structure such as (34) (see also Pereltsvaig 1999).9 9 In this respect, the difference between Pereltsvaig 1999 and my account is that I argue that there is a subset of verbs that take ACOs but which are primitively unergative (see Marelj 2015 for discussion). Linguistica_2016_FINAL.indd 204 28.12.2016 8:57:54 205 (34) VP/vP/VP external VP/v’/ V’ argument V complement 5.3 The Relevance of the Semantic Status of ACOs? The reason we see the restriction in the examples in (36), but not in the examples in (35), is that the two types of nominal phrases are completely different semantically. Whereas non-ACOs are discussed in the previous section, 4.3 focuses on ACOs. (35) Pleše ples. [ACO] dance.3sg dance ‘S(he) dances a dance.’ (36) #Smeje se smehom. [non-ACO] laugh.3sg CL laugh ‘#He laughed a laugh.’ Under a neo-Davidsonian view, thematic roles describe the way a participant takes part in an event. Unlike predicates, arguments then relate to an event indirectly – via thematic roles (37):10 (37) a) Luka priča pravilno. Luka talk.3sg correctly c) ∃e [talking (e) & Agent (e, Luka) & correctly(e)] That the relation is mediated through a thematic role renders utterances in which ACOs appear always informative enough, regardless of whether ACOs are cognate or not.11 Hence, no modification of these arguments is required. Indeed, verbs that allow both ACOs and non-ACOs require modification only in the cases of non-ACOs: (38) Luka živi život /životom fudbalera /#živi životom. Luka live.3sg life-acc /life.instr footballer.gen /# live. life.instr 10 For ease of exposition, temporal information is absent from semantic formulae throughout. 11 “Informative enough” must here be understood against the background of the Neo-Davidsonian (specifically Parsons 1990) understanding of what the verb meaning is. Verbs denote one-place predicates of events and thematic roles are functions from events divided into individuals/par- ticipants of events. The ‘labels’ like Agent, Theme, Sentient inform us about the kind of involve- ment/the nature of the participation of an individual in the event under consideration. Linguistica_2016_FINAL.indd 205 28.12.2016 8:57:54 206 Last but not least, in accordance with the hypothesis that ACOs and non-ACOs are different, note that they can also co-occur as in (39) from SC:12 (39) (Posle dijagnoze raka)…. odlučila je da [živi život] [životom filmske zvezde] decided.prt.f.3sg AUX.3sg that live.3sg life.acc life.instr film star.gen ‘(After the cancer diagnosis)... she decided to live a life of a movie star.’ 5.4 Making Sense of Diametrically Opposed Judgments As evident from (38) and (39), some verbs can take both ACOs and non-ACOs. Within this analysis, one expects that ACOs of such verbs behave like regular arguments, while their non-ACOs are expected to behave like predicates. These predictions are borne out. As illustrated in (40a), for instance, while an ACO can appear with a strong determiner, a non-ACO (40b) cannot: (40) a) Sanjao je svaki san (koji je ikada postojao). [Argument] dream.prt.m.sg AUX every dream.acc (that ever existed) ‘He dreamt every dream that ever existed.’ b) *Sanjao je svakim snom (koji je ikada postojao).[Predicate] dream.prt.m.sg AUX every dream.instr (that ever existed) The existence of data like (38) – (40) leads me to propose that quite like SC, English also has verbs that allow either ACOs or non-ACOs and that different judgements re- flect these different options. While English lacks the initial “morphological clues” that one might be dealing with in two different types of nominals, under the relevant tests like e.g. passivisation, the data in English start behaving quite like the morphologically “robust” languages.13 12 As emphasised by R1, SLO behaves differently than SC here, as the counterpart of SC (39) is ungrammatical. What is interesting to note here is that it is not only the case that SLO disallows the co-occurrence of relevant ACOs and non-ACOs with verbs like dream or live, but also that it does not allow non-ACOs to appear with these verbs at all. It is not obvious why this is the case, but note that SLO has only a prepositional instrumental and its use is somewhat more restricted that in related languages. This seems to be relevant since the counterpart in Russian, for instance, is perfectly fine. (i) Ona prožila žisn’ žisn’ju spuerzvezdy She live.prf.dur life.acc life.instr superstar.gen 13 The poverty of the morphological case system in English is arguably also responsible for the in- ability of the two types of COs to co-occur together, giving rise to, for instance, a counterpart of (39) in English. Linguistica_2016_FINAL.indd 206 28.12.2016 8:57:54 207 6. CONCLUSION In this paper, I dealt with the phenomenon of cognate objects, viewed particularly through the prism of Serbo-Croatian and Slovene. The upshot of this analysis is that there are two types of COs: cognate arguments (ACOs) and cognate predicates (non- ACOs). The notion of “cognatehood” becomes relevant in the case of non-ACOs; their modification is required if utterances in which they appear are to be pragmatically fe- licitous. As far as ACOs are concerned, because the relation between an ACO nominal and the event variable is always mediated through a thematic role, utterances in which they occur are always informative enough. No modification of ACOs is required and their “cognatehood” is incosequential. The status of non-ACOs as adjuncts makes the syntax of unaccusatives with which they are licensed uproblematic. As for ACOs, I argue that some of the verbs with which they occur are primitively transitive, while others are primitively unergative. Crucially, there does not seem to be good reason to treat either ACOs or non-ACOs as incorporations. An examination of morphologically robust languages such as SC and SLO facilitates an understanding of some of the puz- zling properties of COCs cross-linguistically and also offers a means of explaining the disagreement regarding the judgments found in languages like English. References BARRIE, Michael/Eric MATHIEU (2015) “Noun Incorporation and Phrasal Move- ment.” Natural Language and Linguistic Theory 34/1, 1–51. DAVIDSON, Donald 1967. “Truth and Meaning.” In: D. Davidson (ed.), Inquiries into Truth and Interpretation. Oxford: Claredon Press, 17–36. HALE, Ken/Samuel Jay KEYSER (1993) “On Argument Structure and Lexical Ex- pression of Syntactic Relations.” In: K. Hale/S. J. Keyser (eds), The View from Building 20: Essays in Linguistics in Honor of Sylvain Brombeger. Cambridge, Mass: MIT Press, 53–110. HALE, Ken/Samuel Jay KEYSER (1997) “On the Complex Nature of Simple Predica- tors.” In: A. Alsina/J. Bresnan/P. Sells (eds), Complex Predicates. Standford: CSLI Publications, 29–66. HALE, Ken/Samuel Jay KEYSER (2002) Prolegomenon to a Theory of Argument Structure. Cambridge, Mass: MIT Press. HIGGINBOTHAM, Jim (1987) “Indefiniteness and Predication.” In: E. Reuland/A. ter Meulen (eds) The Representation of (In)definiteness. Cambridge, Mass: MIT Press, 43–70. JONES, Michael Allen (1988) “Cognate Objects and the Case-Filter.” Journal of Lin- guistics 24, 89–110. JESPERSEN, Otto (1927) A Modern English Grammar on Historical Principles, Vol- ume III (Syntax). København: Munksgaard. [Reprinted by Allen & Unwin 1954.] MARELJ, Marijana (2015) On Arguments and Predicates: In the Realm of COs. Paper presented at FDSL 11, University of Potsdam. Linguistica_2016_FINAL.indd 207 28.12.2016 8:57:54 208 MOLTMANN, Friederike (1989) “Nominal and Clausal Event Predicates.” In: Papers from the 25th Annual Regional Meetings of the Chicago Linguistic Society, 300–314. University of Chicago. PARSONS, Terence (1990) Events in the Semantics of English: A study in Sub-Atomic Semantics. Current Studies in Linguistics Series 21. Cambridge, Mass: MIT Press. PERELTSVAIG, Asya (1999) “Cognate Objects in Russian: Is the Notion “Cognate” Relevant for Syntax?” Canadian Journal of Linguistics 44/3, 267–291. RICE, Sally (1988) “Unlikely Lexical Entries.” In: Proceedings of the Fourteenth An- nual Meeting of the Berkeley Linguistics Society, 202–212. WILLIAMS, Edwin S. (1983) “Semantic vs. Syntactic Categories.” Linguistics and Philosophy 6, 423–446. Summary IN THE SOUTH SLAVONIC GARDEN: LANDSCAPING THE LANDSCAPE OF ARGUMENTS AND NON-ARGUMENTS This paper deals with morphological, syntactic, semantic, and pragmatic aspects of the so-called Cognate Object Construction with particular reference to Serbo-Croatian and Slovene. The relevance of an examination of such morphologically robust lan- guages is manifold. It facilitates an understanding of some of the puzzling properties of the construction cross-linguistically, offers a way of explaining the noted disagreement regarding judgments found in the literature on Germanic languages such as English and also presents a clear case where (contrary to the dominant view in the literature) mor- phology seems to deceive, rather than inform us, about syntax. Based on a barrage of tests, I argue that there are two types of cognate objects: arguments and non-arguments. Extending the treatment of modifiers within the Davidsonian tradition to the latter, I analyse them as first-order predicates. This allows me to capture their core properties, among which is the obligatory modification, something unaccounted for in the litera- ture. The semantic parallelism between the adverbial modifiers and non-ACOs extends to the syntax as well. Treating non-ACOs as adjuncts solves the problem of the scarcity of syntactic space that arises with unaccusative verbs that license them. ACOs, on the other hand, behave syntactically and semantically like run-of-the-mill arguments and a run-of-the-mill transitive syntax can be maintained (for a majority of them) instead. Keywords: argument, cognate, predicate, unaccusative, unergative Povzetek V JUŽNOSLOVANSKEM VRTU: RISANJE POKRAJINE GLAGOLSKIH ARGUMENTOV IN NE-ARGUMENTOV Članek obravnava morfološke, skladenjske, pomenske in pragmatične vidike t.i. zgradb s tavtološkimi predmeti s posebnim poudarkom na srbohrvaškem in slovenskem Linguistica_2016_FINAL.indd 208 28.12.2016 8:57:54 209 jeziku. Obravnava morfološko bogatih jezikov je pomembna iz več razlogov. Omogoča namreč razumevanje nekaterih zapletenih lastnosti teh zgradb v več jezikih, ponudi na- čin, kako razložiti znana razhajanja pri sodbah, ki jih najdemo v literaturi o germanskih jezikih (kot je npr. angleščina), ter služi kot jasen zgled, kako je morfologija lahko bolj zavajajoča kot informativna glede skladnje (proti prevladujočemu mnenju v literaturi). V članku na osnovi številnih testov zagovarjam trditev, da obstajata dva tipa tavtolo- ških predmetov, argumentni in neargumentni. Z razširitvijo obravnave modifikatorjev znotraj davidsonske tradicije na neargumente, so le-ti obravnavani kot predikati prvega reda. Ta pristop omogoča razlago njihovih temeljnih značilnosti, kot je npr. obvezna prisotnost prilastka. Pomenske vzporednice med prislovnimi določili in neargumentni- mi tavtološkimi predmeti je mogoče razširiti na skladnjo. Obravnava neargumentnih tavtoloških predmetov kot prislovnih določil namreč reši problem pomanjkanja skla- denjskega prostora, ki se pojavi pri netožilniških glagolih. V nasprotju z neargumen- tnimi tavtološkimi premeti pa se argumentni tavtološki predmeti obnašajo kot običajni argumenti z običajno skladnjo glede glagolske prehodnosti. Ključne besede: glagoski argument, tavtološki predmet, predikat, netožilniški glagol, neergativni glagoli Linguistica_2016_FINAL.indd 209 28.12.2016 8:57:54