The Dative Dispositional Construction in Russian Dganit Jenia Kim Tal Siloni

The article sheds new light on the so-called dative dispositional con- struction in Russian. We revise the characterization of the classes of verbs able to feed the construction, showing that its input is best de- fined in terms of the Theta System (Reinhart’s (2002, 2016) approach to thematic relations). To resolve controversies in the literature regard- ing the licensing conditions of the construction and its possible inter- pretations, we ran two surveys whose design and findings we report. We then discuss the properties of the construction comparatively, de- fine the operation deriving it, and provide evidence that the operation is lexical. This has consequences regarding the nature of the lexical component. Keywords: dative dispositional construction, argument structure, Ex- periencer, Theta System

1 Introduction Verbal alternations are pervasive across languages, and their study not only sheds light on ques- tions of argument structure, but also has consequences regarding the architecture of grammar and the division of labor between its components. This article concentrates on a verbal alternation common in Slavic languages. Specifically, it deals with the dative dispositional alternation in Russian, illustrated in (1b). (1a) involves a basic, unergative verb, while (1b) realizes the related verbal form in the dative dispositional construction (DDC).1 As is obvious and will be discussed in detail, the two verbal forms in (1) differ in both morphosyntactic structure and meaning.2 (1) a. Ja ploxo rabotaju. I.NOM badly work.PRES.1SG ‘I work badly.’ / ‘I’m working badly.’

For helpful comments, we would like to thank Irena Botwinik, Julia Horvath, Roni Katzir, Fred Landman, Aya Meltzer-Asscher, Eric Reuland, three anonymous LI reviewers, and the audience at the Interdisciplinary Colloquium of the Department of Linguistics at Tel Aviv University. We are grateful to Elena Shpigner, Slava Kim, and Senja Krol for their help with writing and distributing the questionnaires, and Lawrence Vriend, Julie Fadlon, and Simon Vriend for their help with the statistical analysis of the results. 1 The DDC is also known in the literature as the feel-like construction, the dative impersonal reflexive construction, the involuntary state construction, the dative existential disclosure, the desiderative inversion, and the dative habitual construction. 2 All examples provided in this article are in Russian unless otherwise specified. The abbreviations used throughout ס feminine, GEN ס distributive, FM ס dative, DIST ס auxiliary, DAT ס agreement, AUX ס accusative, AGR ס are ACC neuter, NOM ס negation, NEU ס masculine, NEG ס instrumental, MS ס infinitive, INST ס gerund, INF ס genitive, GER ס question, SG ס nonpast tense, QUEST ס plural, PRES ס perfective, PL ס past tense, PERF ס nominative, PAST ס singular. Linguistic Inquiry, Volume 51, Number 2, Spring 2020 237–279 ᭧ 2019 by the Massachusetts Institute of Technology https://doi.org/10.1162/ling_a_00341 237

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 238 DGANIT JENIA KIM AND TAL SILONI

b. Mne ploxo rabotaetsja. I.DAT badly work.PRES.3SG.SJA ‘I don’t feel like working.’ / ‘I can’t work (due to my psychological circumstances).’ / ‘I evaluate my working negatively.’ The Russian DDC has received much attention in the literature (e.g., Benedicto 1995, Franks 1995, Maru'i? and äaucer 2006, Zeldowicz 2011, Rivero and Arregui 2012). Still, there is no consensus regarding its basic properties, that is, the environments in which the DDC is licensed and the interpretations it allows. Further, it is standardly assumed that pairs such as (1a–b) are derivationally related, but there is no agreement about what the relation is. For example, Franks (1995) suggests that the DDC is derived by a lexical operation on the base verbal entry’s ␪-grid. In contrast, Benedicto (1995), Maru'i? and äaucer (2006), and Rivero and Arregui (2012) propose that the derivation is syntactic, the VP projected by the base entry being embedded under a null (modal or psych) head. Up to now, nonsystematic searches in corpora and intuitions of authors and/or of a small number of speakers have been used in studies of the DDC, but these have not resulted in a clear description of the construction. Therefore, we carried out two surveys with a large number of native Russian speakers to resolve the basic controversies. Further, we explored and revised the definition of the set of verbs feeding the DDC. The new findings have led to a new analysis of the construction and its derivation, and have consequences regarding the nature of the lexical component. The article is organized as follows. In section 2, we present the basic morphosyntactic and semantic characteristics of the DDC, introduce the controversies regarding its distribution and interpretation, and briefly discuss the two surveys we carried out and their results. Section 3 deals with the set of verbs feeding the formation of the dative dispositional alternant. We show that the common view taking the input to be the set of unergative verbs is inaccurate, and we offer a revised characterization of the input. Section 4 focuses on the operation forming the DDC. We present representative syntactic analyses of the construction and show that in contrast to their claims, the input verb does not merge as is, nor does its argument. We then argue that our findings point in the direction of a lexical derivation. We define the operation and draw the theoretical consequences our analysis bears. Section 5 concludes. The online appendices discuss the two surveys in detail and give the full list of items they included (https://www.mitpressjournals.org /doi/suppl/10.1162/ling_a_00341).

2 The Dative Dispositional Construction 2.1 Morphosyntactic and Semantic Characteristics Compare (1a), repeated in (2a), with its corresponding DDC (1b), repeated in (2b). (2) a. Ja ploxo rabotaju. I.NOM badly work.PRES.1SG ‘I work badly.’ / ‘I’m working badly.’

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 239

b. Mne ploxo rabotaetsja. I.DAT badly work.PRES.3SG.SJA ‘I don’t feel like working.’ / ‘I can’t work (due to my psychological circumstances).’ / ‘I evaluate my working negatively.’ The members of the alternation in (2) differ morphologically. First, unlike the verb in (2a), the verb in the DDC (2b) is suffixed with -sja, the so-called reflexive morpheme (parallel to the clitic se found in other Slavic languages and Romance). The suffix is realized as -sja following a consonant, and as -s’ following a vowel. This morphological marking is found on different types of verbs in Russian, most notably on reflexives, reciprocals, unaccusatives, middles, and passives (for a survey, see Gerritsen 1990). Second, while verbs in Russian agree with their nominative subject in number and person in the present (“nonpast”) tense (2a) and in number and gender in the past, the verb in the DDC invariably shows the “default” or “impersonal” 3rd person singular inflection in the present (2b), and singular neuter in the past. In fact, there is no nominative argument in (2b). While the sole argument in (2a) is a nomina- tive subject, in the DDC (2b) the corresponding argument bears dative case. Schoorlemmer (1993) and Benedicto (1995) argue that the dative is the subject, as it can control into gerunds (3c) and antecede anaphors (4c), like nominative subjects ((3a) and (4a), respectively), and unlike dative indirect objects ((3b), (4b)).

(3) a. Jak ?itaju PROk sidja u okna. I.NOM read.PRES.1SG sit.GER by window ‘I am reading while sitting by the window.’

b. Ma'ak mnej ?itaet PROk/*j sidja u okna. Masha.NOM I.DAT read.PRES.3SG sit.GER by window ‘Masha is reading to me while she is sitting by the window.’

c. Mnek ne ?itaetsja PROk sidja u okna. I.DAT NEG read.PRES.3SG.SJA sit.GER by window ‘I can’t read / I don’t feel like reading / I evaluate my reading negatively, when I’m sitting by the window.’

(4) a. Jak ?itaju v svoejk komnate. I.NOM read.PRES.1SG in own room ‘I read in my own room.’ / ‘I’m reading in my own room.’

b. Ma'ak mnej ?itaet v svoejk/*j komnate. Masha.NOM I.DAT read.PRES.3SG in own room ‘Masha is reading to me in her own room.’

c. Mnek ne ?itaetsja v svoejk komnate. I.DAT NEG read.PRES.3SG.SJA in own room ‘I can’t read / I don’t feel like reading / I evaluate my reading negatively, when I’m in my room.’ However, as observed by Greenberg and Franks (1991) and Komar (1999), nonsubjects can also control into gerunds and antecede anaphors in Russian. In (5a), the anaphor is bound by a genitive

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 240 DGANIT JENIA KIM AND TAL SILONI

noun phrase; in (5b), it is bound by an accusative one. In (6a), a genitive noun phrase controls into the gerund, and in (6b), an accusative one does so.

(5) a. Skol’ko u neei bylo s soboji/*neji deneg? how.much at her.GEN was.AGR with self/*her money ‘How much money did she have with her(self)?’ (Komar 1999:254, (16))

b. On zastal menjai v svoeji komnate. he.NOM found me.ACC in own room ‘He found me in my room.’ (Greenberg and Franks 1991:77, (22)) (6) a. Proxodja mimo krasivogo cvetnika, na ix ustalyx licax pojavljaetsja crossing.GER by beautiful garden on their.GEN tired faces appears dobraja blagodarstvennaja ulybka. kind thankful smile ‘While crossing the beautiful garden, on their tired faces appears a kind, thankful smile.’ (Greenberg and Franks 1991:78, (23)) b. Projdja pervyj obÅig, izdelie pokryvajut glazur’ju. going.through.GER first firing article.ACC cover.3PL glaze ‘When going through the first firing, they cover the article with glaze.’ (Greenberg and Franks 1991:78, (24)) It thus seems that control into gerunds and binding of anaphors are not reliable diagnostics of subjects. Moreover, there is some evidence showing that the dative noun phrase in the DDC does not behave as a subject, as explained below. The dative noun phrase of the DDC is not the only dative noun phrase argued to be a subject in Russian. Important for our purposes are the so-called dative subjects of infinitives. These datives display properties undoubtedly associated with subjects. Moore and Perlmutter (2000) show that they can be a controlled PRO (as in (7a), where the predicate odnomu ‘alone.DAT’ reflects the dative case of PRO) and occupy the subject position of a raising verb (7b). Moore and Perlmutter further show that unlike the dative subject of an infinitive, the dative in the DDC can neither be a controlled PRO (8b) nor occupy the subject position of a raising verb (8b).

(7) a. Borisi sdelal vse vozmoÅnoe, ?toby PROi rabotat’ odnomu. Boris.NOM did all possible in.order work.INF alone.DAT ‘Boris did everything possible to work alone.’ (Moore and Perlmutter 2000:397, (40b)) b. Im ne na?at’ rabotat’ odnim. they.DAT NEG begin.INF work.INF alone.DAT ‘It’s not (in the cards) for them to begin to work alone.’ (Moore and Perlmutter 2000:399, (44a))

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 241

(8) a. *Borisi xodit k psixiatru dlja togo ?toby PROi rabotat’sja lu?'e. Boris.NOM goes to psychiatrist for in.order work.INF.SJA better ‘Boris goes to a psychiatrist in order to be able to work better.’ (Moore and Perlmutter 2000:398, (42b)) b. *Im ne na?at’ rabotat’sja lu?'e. they.DAT NEG begin.INF work.INF.SJA better ‘It’s not for them to begin to work better.’ (Moore and Perlmutter 2000:400, (45)) These data suggest that the dative noun phrase in the Russian DDC is not a subject; rather, the construction is impersonal. Besides these structural distinctions, the DDC denotes a different type of eventuality than its basic counterpart. While the basic version in (2a) describes an activity of working, the DDC in (2b) describes a psychological state related to this activity—for example, not feeling like working (the precise interpretations of the DDC are discussed in section 2.3). Accordingly, the role of the participant in the eventuality is different: while the nominative subject in (2a) is an Agent, the dative in (2b) is understood as the Experiencer of the psychological state (e.g., Bene- dicto 1995, Franks 1995, Maru'i? and äaucer 2006, Rivero and Arregui 2012). This semantic shift singles out the DDC from other structurally similar constructions in Russian.3 The role of the adverb is also different: while in (2a) ploxo ‘badly’ describes the quality of the event of working, in the DDC (2b) it describes a negative disposition toward this eventuality or a subjective negative evaluation of it, as will be further discussed in section 4.3.2. Although the Russian DDC has been widely discussed in the literature, there is still disagree- ment regarding certain basic properties of the construction—specifically, the environments in which it is licensed and its interpretations. To put our analysis on a solid empirical footing, we conducted two surveys to shed light on these controversial issues. In sections 2.2 and 2.3, we introduce the unresolved questions and briefly discuss our surveys and their findings. A detailed discussion of the surveys appears in online appendices A and B.

2.2 Licensing Environments 2.2.1 The Controversy According to Pariser (1982) and Franks (1995), the Russian DDC re- quires modification by an adverbial element (9a), including negation (9b). These modifiers de- scribe the disposition of the dative Experiencer toward the eventuality or the Experiencer’s subjec- tive evaluation of it. According to these authors, the sentence in (9c), which does not involve any modification, is ungrammatical or at least infelicitous. (9) a. Mne xoro'o rabotaetsja. I.DAT well work.PRES.3SG.SJA ‘I feel like working.’ / ‘I can work.’ / ‘I evaluate my working positively.’

3 There are additional structurally similar constructions in Russian that do not involve this semantic change and are less productive. The present article is devoted to the DDC. For a more detailed survey of similar constructions, see Gerritsen 1990.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 242 DGANIT JENIA KIM AND TAL SILONI

b. Mne ne rabotaetsja. I.DAT NEG work.PRES.3SG.SJA ‘I don’t feel like working.’ / ‘I can’t work.’ / ‘I evaluate my working negatively.’ c. */??Mne rabotaetsja. I.DAT work.PRES.3SG.SJA However, while there is consensus regarding the grammaticality of the adverbial (9a) and negated (9b) DDC, the views about the “bare” type (9c) vary significantly. According to Maru'i? and äaucer (2006), bare DDCs are licensed in Russian in downward-entailing environments—not only under negation, but also in yes/no questions (10), in relative clauses in the restrictor of universal quantifiers (11), and in the antecedent of conditionals (12). (10) Tebe rabotaetsja v pjatnicu? you.DAT work.PRES.3SG.SJA in Friday ‘Do you feel like working on Fridays?’ / ‘Are you able to work on Fridays?’ (http://otvet.bigmir.net/question/751663/, last accessed July 2017) (11) V blogax mogut i dolÅny pisat’ vse komu pi'etsja. in blogs can and must write.INF all who.DAT write.PRES.3SG.SJA ‘Any person who feels like writing can and must write a blog.’ (http://yerevan.ru/2010/04/02/foto-progulka-po-erevanu-prodolzhenie/, last accessed February 2014) (12) I ja mogu 10 ty'’ vydat’ esli mne pi'etsja... and I.NOM can 10 thousand deliver.INF if I.DAT write.PRES.3SG.SJA ‘I also can deliver ten thousand [words a day], if I feel like writing.’ (http://vk.com/topic-13125990_22236637, last accessed July 2017) Fici (2011) suggests that bare DDCs are even less restricted and can be found without any special context, as in (13). A similar view is held by Gerritsen (1990): according to her, neither an adverb nor negation is obligatory; in their absence, positive modification is implied. (13) Stranno, kogda ja byla odinokoj . . . mne pisalos’. strange when I.NOM was lonely I.DAT write.PAST.SG.NEU.SJA ‘Strange, when I was lonely . . . I felt like writing / I could write / I experienced my writing positively.’ (Jaxontova, Iz vospominanij starogo teni'evca (1996–1998). From the Russian National Corpus: http://www.ruscorpora.ru) Instances of bare DDCs (in all kinds of environments) can indeed be found in Web corpora and in the Russian National Corpus, as shown in (10)–(13). However, these DDC instances are rare and usually appear in a context that indirectly provides the information associated with the adverb/ negation. For example, in (13) positive modification is deduced owing to the presence of stranno ‘strangely’, which makes it clear that the feeling described by the DDC contrasts with the feeling of loneliness. In the absence of modification, a positive one is assumed.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 243

The grammatical status of different DDC types cannot be determined on the basis of authors’ intuitions or sporadic attested examples, as these practices have led to contradictory descriptions. We therefore carried out a survey of judgments with a large number of native speakers to obtain more solid data. In sections 2.2.2 and 2.2.3, respectively, we present and discuss the findings. Further details about the participants and the survey are presented in online appendix A. 2.2.2 Empirical Findings: Questionnaire I The first survey compared the acceptability ratings of four DDC types: negated DDC, adverbial DDC, bare DDC, and yes/no question DDC (the last one represents a bare DDC in a downward-entailing environment). These four DDC types are referred to as experimental conditions A–D, as shown in table 1. The experimental items were presented in this survey without context. Given that context may indirectly provide the same type of information that adverbial modification does, presenting the DDC sentences in context would not allow us to determine whether or not a genuinely bare DDC and a yes/no question DDC are possible. Participants were asked to rate the acceptability of sentences representing these conditions on a scale of 1 to 5. The conditions A–D were checked with 12 different verbs. Participants in this survey were 89 native speakers of Russian. The results (i.e., median scores per DDC type) are given in table 2. As the table shows, experimental conditions A and B (the negated and adverbial DDCs) received high acceptability ratings (median scores of 4 or higher on a 1-to-5 scale), meaning that these DDC types are perceived by native speakers as grammatical. Conditions C and D received median scores of 2.5 and 3, respectively. This shows that the bare DDC, whether a declarative sentence or a yes/no question, is perceived as less acceptable than the negated and adverbial variants. Participants did not categorically disallow the bare DDC, which is consistent with the fact (mentioned in section 2.2.1) that sporadic examples of use can be found in corpora (though rarely, as will be discussed in the next paragraph). However, the grammatical status of the bare DDC is clearly marginal and different from that of the negated and adverbial DDCs. Importantly, many participants did not give consistent judgments regarding bare DDCs; of those who did, the majority (71% for condition C) consistently rated them as

Table 1 The experimental conditions Condition DDC type Example A Negated Mne segodnja ne tancuetsja. I.DAT today NEG dance.PRES.3SG.SJA B Adverbial Mne segodnja xorosˇo tancuetsja. I.DAT today well dance.PRES.3SG.SJA C Bare Mne segodnja tancuetsja. I.DAT today dance.PRES.3SG.SJA D Yes/No question Tebe segodnja tancuetsja? you.DAT today dance.PRES.3SG.SJA

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 244 DGANIT JENIA KIM AND TAL SILONI

Table 2 Median scores per DDC type Condition A Condition B Condition C Condition D Negated DDC Adverbial DDC Bare DDC Yes/No question DDC Median score 4 4.5 2.5 3

unacceptable. The fact that the scores for the bare DDC gravitate away from the endpoint of the scale may also be related to the type of fillers used in this study. The ungrammatical fillers were clearly bad (median score: 1, mean score: 1.16) and included sentences that would never be uttered in Russian. Compared with such sentences, bare DDCs are “not as bad” and therefore were rated higher.4 Our findings are supported by an analysis we performed on a sample of corpus data. In a tagged web corpus on Sketch Engine (name: ruTenTen11; size: over 18 billion tokens),5 about 7,000 occurrences of DDCs with rabotaetsja ‘work.SJA’ were found. According to our analysis, only 1.4% of this sample (100 instances) are bare DDCs and none is a yes/no question DDC.6 2.2.3 Discussion The survey shows that the negated DDC and the adverbial DDC are perceived by native speakers as grammatical. This is not surprising. As mentioned above, there is general consensus in the literature that these two DDC types are possible in Russian, and the results of the survey confirm it. The bare DDC, on the other hand, scored significantly lower in the question- naire, both as a declarative and as a yes/no question. The survey’s results make clear that it is not considered as acceptable as the canonical DDC types. In addition, the bare DDC is rare in corpus tokens (1.4% in our sample), usually appears in context, and is interpreted as involving positive evaluation/disposition. That is, the construction requires modification, and in the absence of modification (which is rare), a positive one is assumed.7

4 Likewise, the grammatical fillers were undoubtedly grammatical “everyday” sentences, for example, simple sen- tences with a -sja unaccusative or reflexive verb (median score: 5, mean score: 4.8). Compared with them, DDC construc- tions are much less frequent, which may have had an effect on participants’ evaluations, resulting in medians 4–4.5. preloadedסDetails about the corpus can be found at https://the.sketchengine.co.uk/corpus/corp_info?corpname 5 .1סsubcorpora&1סrutenten11_8&struct _attr_stats/ 6 The verb rabotaetsja ‘work.SJA’ is a good candidate for the analysis since it is almost exclusively reserved for the DDC. The initial search for rabotaetsja returned 7,968 results. Finding the bare DDCs in such a large corpus is not straightforward because the word order in Russian sentences is flexible, meaning that adverbs are not necessarily adjacent to the verb. To find the bare DDC instances, we searched for all sentences where the two slots before rabotaetsja were not occupied by an adverb, negation, or ‘how’ and the two slots after rabotaetsja were not occupied by an adverb (as negation or a wh-phrase cannot follow the verb). This query returned 1,930 items, which we then manually analyzed to exclude (a) instances where an adverb/negation was present in the sentence, but farther away from the verb; (b) instances with adverbs that were not tagged as such and therefore not filtered out by the query; (c) instances that were not DDCs. Of the 1,930 manually analyzed results, 11% were not DDCs. Taking this sample to be representative, we assumed that the same percentage of the initial 7,968 results were not DDCs as well. This led to the following calculation: out of 7,091 DDCs with rabotaetsja found in the corpus, 100 instances are bare DDCs (1.4%). 7 Sentence (i) is one of the rare examples we found with a genuinely bare DDC. It is interpreted as involving positive modification.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 245

An adequate analysis of the Russian DDC must account for the fact that modification is required. Other Slavic languages, such as Slovenian and Bulgarian, freely allow bare DDCs, as illustrated in (14) and (15) (the clitic se marks the alternation). (14) Gabru se ple'e. Gaber.DAT SE dance.PRES.3SG ‘Gaber feels like dancing.’ (Slovenian; Maru'i? and äaucer 2006:1095, (2)) (15) Raboti mi se. work.PRES.3SG I.DAT SE ‘I feel like working.’ (Bulgarian; Rivero and Milojevic´-Sheppard 2003:148n8) The role of modification in the Russian DDC is discussed in section 4.4.2.

2.3 Interpretations 2.3.1 The Controversy At least four types of meanings are associated with the DDC construction in the literature; see (16a–d) and (17a–d). The controversy among authors revolves around two questions: which of these meanings are actually available for the Russian construction, and which types of DDC are associated with which interpretations? (16) Mne ne rabotaetsja. I.DAT NEG work.PRES.3SG.SJA a. Disposition: ‘I don’t feel like working.’ / ‘I’m not in a working mood.’ b. Capability: ‘I can’t work.’ c. Evaluation of the activity: ‘I feel that my work is going badly.’ d. Evaluation of the participant’s mental state: ‘I am working and not enjoying it.’ / ‘I am working and (cannot help) feeling bad about it.’ (17) Mne xoro'o rabotaetsja. I.DAT well work.PRES.3SG.SJA a. Disposition: ‘I feel like working.’ / ‘I’m in a working mood.’ b. Capability: ‘I can work.’ / ‘I can work well.’ c. Evaluation of the activity: ‘I feel that my work is going well.’ d. Evaluation of the participant’s mental state: ‘I am working and enjoying it.’ / ‘I am working and (cannot help) feeling good about it.’

(i) Medvedev dlja menja – politik sover'enno drugoj formacii, ?em te, s kotorymi mne Medvedev for me politician completely different formation than those with whom I.DAT rabotalos’ i rabotaetsja po sej den’. work.PAST.SG.NEU.SJA and work.PRES.3SG.SJA to this day ‘For me, Medvedev is a politician of a completely different order than those with whom I successfully worked and work to this day.’ (http://www.politonline.ru/comments/9669.html; last accessed December 2018)

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 246 DGANIT JENIA KIM AND TAL SILONI

Specifically, the disagreement revolves around three main points. First, does the DDC neces- sarily involve a factual eventuality? While meanings (c) and (d) involve an actual event of working in the real world, meanings (a) and (b) do not entail such an event. Rivero and Arregui (2012) argue that there are two variants of the DDC in Slavic languages, which differ from each other with regard to the factuality feature. In Russian and West Slavic languages (e.g., Polish, Czech, Slovak), the construction describes an involuntary mental state regarding an actual eventuality; specifically, meaning (d) (e.g., ‘I am working and cannot help feeling good about it’) is associated with the DDC. In South Slavic (Slovenian, Bulgarian, and Serbian/Croatian), on the other hand, the construction has a ‘feel like’ interpretation (meaning (a)) and does not imply an actual eventual- ity. Rivero and Arregui (2012) further claim that the two types of meanings fail to coexist in a single language, meaning that the Russian and West Slavic DDC cannot have the ‘feel like’ interpretation, except for the case of negation, as mentioned in the next paragraph. Other authors do not consider the Russian DDC as necessarily factual. Benedicto (1995) and Franks (1995), for example, attribute the nonfactual meanings (a) and (b) to both the adverbial and the negated types. Second, does the negated DDC have specific interpretations unavailable for the adverbial type? According to Fici (2011), the negated DDC is ambiguous: it can be understood as synony- mous with the DDC with a negative adverb (e.g., ploxo ‘badly’), but it can also have the disposi- tional ‘feel like’ interpretation (a), which is unavailable for the adverbial type. Similarly, Maru'i? and äaucer (2006) claim that the dispositional meaning is unavailable for the adverbial DDC; that meaning is restricted to downward-entailing environments, including negation. Rivero and Arregui (2012:308) also mention negation as having “intriguing effects” on the Russian DDC. Specifically, it can cancel out the factuality of the construction and introduce a dispositional interpretation. Slobodchikoff (2008) promotes a completely opposite view. According to her, the negated DDC (the only type she discusses) does not have a dispositional interpretation; rather, it has the capability interpretation (meaning (b)). The capability meaning is also discussed by Pariser (1982), Benedicto (1995), and Maru'i? and äaucer (2006). Under their analyses, this meaning is attributed to both the negated and the adverbial variants. However, they disagree regarding the exact interpretation of capability, as elaborated directly. Third, is the capability meaning of the DDC specialized for psychological circumstances? According to Benedicto (1995), the DDC refers to psychological circumstances. This means that (16), for instance, expressing one’s inability to work, can only be uttered in a situation in which the reasons for the inability have to do with one’s mental state; it cannot be used when the reasons are of a different kind (physical, external, etc.). In contrast, Pariser (1982) claims that the (in)capability can be understood as arising from any circumstances outside of one’s own control, including external ones. We carried out a survey with a large number of speakers to shed light on the above questions. In sections 2.3.2 and 2.3.3, respectively, we present and discuss the results of the survey. Further details about the survey are presented in online appendix B. 2.3.2 Empirical Findings: Questionnaire II Our second survey was designed to compare the appropriateness ratings for the five meanings presented in table 3. These are the meanings (a)–(d)

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 247

Table 3 Putative DDC meanings Condition Putative DDC meaning A Disposition (‘feel like’) B Capability (due to external/physical circumstances) C Capability (due to psychological circumstances) D Evaluation of the activity E Evaluation of the participant’s mental state

discussed in section 2.3.1, but with the capability interpretation split into two kinds: psychological circumstances (condition C) and physical or external ones (condition B). The survey checked the availability of these meanings in the two canonical DDC types: the negated DDC and the adverbial DDC. The conditions were checked with three verbs: rabotat’ ‘work’, ?itat’ ‘read’, and spat’ ‘sleep’. These verbs were chosen on the basis of the results of questionnaire I, in which the negated and adverbial constructions with these verbs were judged as highly acceptable. To elicit judgments regarding the availability of a meaning in a DDC sentence, we used the following task. Participants were directed toward the intended meaning of each test sentence by a short context and a paraphrase. These were followed by the DDC sentence and the task. Partici- pants were asked to rate, on a scale of 1 to 5, the appropriateness of using the given DDC sentence to express the intended meaning in the given situation. Table 4 offers an example. Participants in this survey were 284 native speakers of Russian. Table 5 presents the results, that is, the median ratings regarding how appropriate it is to use the negated/adverbial DDC to express each of the five types of meanings. Since three different verbs were used in this survey, we also summarized the results by item to check whether the use of a specific verb affected

Table 4 A question checking the availability of meaning B in a negated DDC Context Grandma has poor eyesight. She wanted to read the newspaper, but her eyes immediately got tired and she stopped reading after the first paragraph. Paraphrase Grandma wants to say that she cannot read today. She says: Utterance Mne segodnja ne ?itaetsja. I.DAT today NEG read.PRES.3SG.SJA Task Your task is to determine whether the sentence Mne segodnja ne ?itaetsja is appropriate for this situation and expresses the meaning that Grandma intends. Mark (1) if you think that the sentence is absolutely inappropriate for the situation and/or does not express what the person wants to say. Mark (5) if you think that the sentence is absolutely natural in this situation and expresses exactly what the person wants to say. If your judgment about the sentence is somewhere between these extremes, mark one of the middle responses (2), (3), or (4).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 248 DGANIT JENIA KIM AND TAL SILONI

Table 5 The appropriateness (median ratings) of using a negated/adverbial DDC to express each meaning type Condition/Meaning type Negated DDC Adverbial DDC A: Disposition (‘feel like’) 4* 2** B: Capability (due to external/physical circumstances) 2 2 C: Capability (due to psychological circumstances) 5 4 D: Evaluation of the activity 5 5 E: Evaluation of the participant’s mental state 3 4 * When excluding the verb spat’ ‘sleep’ (for reasons explained in the text), the median rating is 5. ** When excluding the verb spat’ ‘sleep’ (for reasons explained in the text), the median rating is 4.

informants’ judgments. This by-item summary revealed an idiosyncratic behavior of the verb spat’ ‘sleep’; both the negated and the adverbial DDC with spat’ received exceptionally low ratings in condition A. These findings suggest that a DDC with the verb spat’ cannot express the dispositional ‘feel like’ meaning, while the DDCs with the other two verbs can. Excluding the verb spat’ ‘sleep’, the median rating for condition A_negation is 5 and for condition A_adverb is 4. The results thus suggest the following conclusions: 1. Both DDC types are appropriate to express meanings A, C, and D (disposition, capability (psychological), and evaluation of the activity, respectively), given that the median ratings for these conditions are 4–5 on a scale of 1 to 5.8 2. Neither DDC type can express meaning B (capability due to external/physical circum- stances). 3. The adverbial DDC can express meaning E (evaluation of the participant’s mental state while performing an activity).9 2.3.3 Discussion The results presented in section 2.3.2 provide some answers to the unresolved questions raised in section 2.3.1. First, it turns out that the negated DDC and the adverbial DDC with ploxo ‘badly’ are interpreted similarly.10 This finding has consequences for the analyses of Maru'i? and äaucer (2006), Fici (2011), and Rivero and Arregui (2012), who claim that the

8 The dispositional meaning is not available for a DDC with the verb spat’ ‘sleep’. Whether this is an idiosyncrasy of this specific verb or a restriction on verbs of its type (see section 3.3.1) is a question for future experimental research. 9 Condition E with the negated DDC received the median score 3; respondents’ ratings for this condition were scattered more or less evenly among the values of the scale: 1, 25%; 2, 21%; 3, 17%; 4, 16%; 5, 21%. This suggests that respondents rated these examples randomly; that is, they did not have clear and consistent judgments regarding the appropriateness of a negated DDC to express meaning E. 10 The only case where results differed for the adverbial and the negated DDCs involved meaning E. However, since there were no clear judgments regarding condition E with negation, we cannot conclude that there are differences in interpretation between the negated DDC and the adverbial DDC.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 249

dispositional meaning is restricted to the negated DDC. Excluding spat’, our results show that the dispositional meaning can be expressed by both the negated and the adverbial DDC. Further, the availability of the dispositional meaning and the capability meaning for both DDC types is at odds with Rivero and Arregui’s (2012) claim that the Russian DDC is necessarily factual. The DDC allows both factual interpretations (e.g., ‘I feel that my work is going badly’) and nonfactual ones (e.g., ‘I don’t feel like working’, ‘I can’t work’). The results also provide a clear answer about which type of circumstance is involved in the (in)capability interpretation. The DDC is appropriate to express (in)capability due to psychological circumstances, but cannot express (in)capability due to physical or external reasons. These findings strongly support Benedicto’s (1995) claim that the DDC is specialized for psychological circum- stances. This means that the English translation that uses the modal ‘can’ (e.g., ‘I can’t work’) is not adequate to express the type of capability expressed by the Russian DDC. Rather, the capability expressed by the DDC is closer to the dispositional meaning; a certain mental state (mood) is necessary to be able to perform the activity. To sum up, the meanings that speakers associate with the Russian DDC are as follows: (18) Mne ne rabotaetsja. I.DAT NEG work.PRES.3SG.SJA a. Disposition: ‘I don’t feel like working.’ / ‘I’m not in a working mood.’ b. Capability: ‘I can’t work due to my psychological circumstances.’ c. Evaluation of the activity: ‘I feel that my work is going badly.’ (19) Mne ploxo rabotaetsja. I.DAT badly work.PRES.3SG.SJA a. Disposition: ‘I don’t feel like working.’ / ‘I’m not in a working mood.’ b. Capability: ‘I can’t work due to my psychological circumstances.’ c. Evaluation of the activity: ‘I feel that my work is going badly.’ d. Evaluation of the participant’s mental state: ‘I am working and not enjoying it.’ The interpretation of the construction is further discussed in section 4.4. In the next section, we characterize the type of predicates participating in the DDC.

3 The Input to the DDC In the previous section, we described the structural and semantic characteristics of the DDC. Next, we examine what type of verbs can feed it. The commonly accepted view, presented in section 3.1, is that the Russian DDC is fed by unergative verbs. In section 3.2, we argue that this characterization is not accurate; we revise it in section 3.3.

3.1 The Existing Characterization of the Input The Russian DDC is more restricted than its counterparts in other Slavic languages regarding the verbs it allows. It has been observed that the Russian DDC disallows realization of internal arguments, thus excluding transitive verbs (e.g., Schoorlemmer 1993, Franks 1995). This contrasts

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 250 DGANIT JENIA KIM AND TAL SILONI

with languages such as Slovenian and Polish, which, in addition to the dative noun phrase, can realize an accusative internal argument. This is illustrated for Slovenian in (20).11 The parallel sentence in Russian is ungrammatical, as shown in (21a). The verb jest’ ‘eat’ can appear in the DDC only if its internal argument is not realized, as in (21b). Transitive verbs whose internal argument is not optional (22a) like stroit’ ‘build’ cannot participate in the Russian DDC at all (22b).12 (20) Danes dopoldne se mi je jedlo jagode. today morning SE I.DAT be.3SG eat.NEU strawberries.ACC ‘This morning I felt like eating strawberries.’ (Slovenian; Rivero and Milojevic´-Sheppard 2003:92, (1b)) (21) a. *Mne v?era ne jelos’ klubniku. I.DAT yesterday NEG eat.PAST.SG.NEU.SJA strawberries.ACC b. Mne v?era ne jelos’. I.DAT yesterday NEG eat.PAST.SG.NEU.SJA ‘I didn’t feel like eating yesterday.’ / ‘I couldn’t eat yesterday.’ / ‘I evaluate my eating yesterday negatively.’

11 In addition, Slavic languages (e.g., Slovenian) also exhibit a passive dispositional construction, where the internal argument shows nominative case. Certain Slavic languages (e.g., Serbian/Croatian and Bulgarian) exhibit only the passive variant, which is illustrated in (i) (Rivero and Milojevic´-Sheppard 2003, Maru'i? and äaucer 2006). (i) Cˇ etjaxa mi se knigi. read.PAST.3PL I.DAT SE books.NOM ‘I felt like reading books.’ (Bulgarian; Rivero and Milojevic´-Sheppard 2003:146, (84b)) 12 There are rare occurrences of a passive dispositional construction of the type in (i): the internal argument appears in a morphological form ambiguous between the accusative and nominative; but given that the verb shows agreement with the internal argument (and not default agreement), the argument must be in the nominative, and the construction must be classified as passive. (i) . . . ?toby emu lu?'e pisalis’ stixi o Trojanskoj vojne. . . . so.that he.DAT better write.PAST.PL.SJA poems about Trojan war ‘ . . . in order to be able to write poems about the Trojan War.’ (Sketch Engine, ruTenTen11 corpus) We have also found sparse instances with the verb pit’ ‘drink’ occurring with an internal argument in the singular accusative/nominative form (ii). (ii) V Åaru mne xoro'o p’jotsja zeljonyj ?aj. in heat I.DAT well drink.PRES.3SG.SJA green tea ‘When it’s hot, I feel like drinking green tea.’ / ‘When it’s hot, I enjoy drinking green tea.’ (http://kuking.net) As the internal argument is in the singular, the agreement on the verb is ambiguous between agreement with the internal argument and default agreement. But given the existence of cases such as (i), there is no reason to analyze examples such as (ii) as involving an accusative internal argument. We have not found a single instance of the DDC involving an unambiguously accusative internal argument. Furthermore, on the basis of our online searches in the Sketch Engine ruTenTen11 corpus and the Russian National Corpus, and as also mentioned by an anonymous reviewer, even the passive instances are very rare and restricted to these two verbs (‘write’ and ‘drink’) in the same “verb – internal argument” combinations (‘write’, for instance, appears only with stixi ‘poems’ and ni?ego ‘nothing’). Hence, we believe that the generalization that the DDC cannot be fed by a transitive verb is solid.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 251

(22) a. Ja stroju *(dom). I.NOM build.PRES.1SG house ‘I’m building a house.’ b. *Mne ne stroitsja (dom). I.DAT NEG build.PRES.3SG.SJA house Franks (1995) observes that not only direct objects but also prepositional arguments are excluded from the Russian DDC (23a), as opposed to nonargumental PPs (23b). He concludes that “a Russian verb may not enter into the . . . construction if it has any internal arguments” (p. 365). (23) a. *Mne ne rabotaetsja nad etoj zada?ej. I.DAT NEG work.PRES.3SG.SJA on this problem (Franks 1995:365, (64a)) b. Mne ne rabotaetsja pri takix uslovijax. I.DAT NEG work.PRES.3SG.SJA under such conditions ‘I don’t feel like working under such conditions.’ / ‘I can’t work under such condi- tions.’ / ‘I feel that my work is going badly under such conditions.’ (Franks 1995:366, (68a)) The ban on internal arguments is not limited to direct and indirect objects. Schoorlemmer (1993) shows that intransitive unaccusative verbs, whose subject is believed to be an internal argument, are also not admissible in the construction, as shown in (24).13 She concludes that DDC formation in Russian is “productive for all unergative intransitive verbs” (p. 155). (24) a. *Vase ne rastetsja. Vasja.DAT NEG grow.PRES.3SG.SJA (Schoorlemmer 1993:158, (48b)) b. *Vase ne padaetsja. Vasja.DAT NEG fall.PRES.3SG.SJA Though the observations that the Russian DDC allows only one argument and that this argument cannot be internal are correct, in the next section we will show that not all intransitives with an external argument take part in the construction; rather, only a subset of them do.

3.2 Problems with the Existing Characterization Evidence that not all unergatives can serve as input to the DDC comes from emission verbs. These verbs are described in the literature as nonagentive unergatives, which express nonvoluntary emission of sound, light, smell, or substance (e.g., Levin and Rappaport Hovav 1995, Reinhart 2002, Potashnik 2012). Some Russian examples include zvonit’ ‘ring’, gremet’ ‘thunder, rattle’, vizÅat’ ‘squeal’, svistet’ ‘whistle’, potet’ ‘sweat’, vonjat’ ‘stink’, droÅat’ ‘tremble’, sverkat’ ‘spar-

13 Evidence that rasti ‘grow’ and padat’ ‘fall’ are unaccusative is presented in (27)–(28).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 252 DGANIT JENIA KIM AND TAL SILONI

kle’, blestet’ ‘shine’, cvesti ‘bloom’. Crosslinguistic evidence suggests that emission verbs are unergative; they fail unaccusativity diagnostics, like agentive unergative verbs.14 In Russian, this can be shown with the distributive po-phrase diagnostic. Pesetsky (1982), Borik (1995), and Schoorlemmer (2004) show that the distributive po-phrase can appear with direct objects (25b), subjects of passives (26b), and subjects of unaccusatives ((27b), (28b)), but not with subjects of transitives (29b) or subjects of agentive unergatives (30b). Crucial for present purposes is the observation that the subject of an unaccusative always licenses the distributive po-phrase. As shown in (31b) and (32b), subjects of emission verbs disallow it, pairing up with subjects of unergative verbs.15 (25) Direct object a. My polu?ili knigu. we.NOM receive.PAST.PL book.ACC ‘We received a/the book.’ b. My polu?ili po knige. we.NOM receive.PAST.PL DIST book.DAT ‘We received a book each.’ (26) Subject of passive a. Na na'ej verfi vypuskajutsja lodki. in our shipyard produce.PRES.3SG.SJA boats.NOM ‘At our shipyard boats are produced.’ b. Na na'ej verfi vypuskajetsja po dve lodki v god. in our shipyard produce.PRES.3SG.SJA DIST two.NOM boat.GEN in year ‘At our shipyard two boats are produced every year.’ (Kuznetsova 2005:170, (2)) (27) Subject of unaccusative a. V gor'ke ros cveto?ek. in pot grow.PAST.SG.MS flower.NOM ‘There was a flower growing in the pot.’ b. V kaÅdom gor'ke roslo po cveto?ku. in each pot grow.PAST.SG.NEU DIST flower.DAT ‘There was a flower growing in each pot.’

14 See Potashnik 2012 for an overview of the evidence that emission verbs are unergative in English, Italian, Dutch, and Hebrew. 15 Providing a few examples where subjects of transitives and unergatives license the occurrence of the distributive po-phrase, Kuznetsova (2005) suggests that three semantic conditions underlie the licensing of the phrase. Abstracting away from details, we note that our examples take Kuznetsova’s conditions into account. Nonetheless, emission verbs disallow the distributive po-phrase ((31b), (32b)). Moreover, Kuznetsova’s study does not cast doubt on the generalization that all unaccusative verbs license the distributive po-phrase (rather, it casts doubt on the claim that unergatives (and transitives) never do). Hence, the fact that emission verbs fail the diagnostic strongly suggests that they are unergative, a conclusion reinforced by crosslinguistic evidence. The same holds for reflexive, reciprocal, and subject-Experiencer verbs, which are illustrated in (34)–(36).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 253

(28) Subject of unaccusative a. S dereva upalo jabloko. from tree fall.PAST.SG.NEU apple.NOM ‘An apple fell from the tree.’ b. Po jabloku upalo s kaÅdogo dereva. DIST apple.DAT fall.PAST.SG.NEU from each tree ‘An apple fell from each tree.’ (Babby 1980:45) (29) Subject of transitive a. KaÅdyj rebenok polu?il knigu. every child receive.PAST.SG.MS book.ACC ‘Every child got a book.’ b. *V kaÅdom klasse po rebenku polu?il(-o) knigu. in every class DIST child.DAT receive.PAST.SG.NEU book.ACC (30) Subject of unergative a. Na dereve pela ptica. on tree sing.PAST.SG.FM bird.NOM ‘There was a bird singing in the tree.’ b. *Na kaÅdom dereve pelo po ptice. on each tree sing.PAST.SG.NEU DIST bird.DAT (Kuznetsova 2005:171, (3c)) (31) Subject of emission verb a. V kamere potel uznik. in cell sweat.PAST.SG.MS prisoner.NOM ‘There was a prisoner sweating in the cell.’ b. *V kaÅdoj kamere potelo po uzniku. in each cell sweat.PAST.SG.NEU DIST prisoner.DAT (32) Subject of emission verb a. V komnate vonjajut botinki. in room stink.PRES.3PL shoes.NOM ‘There are shoes stinking in the room.’ b. *V kaÅdoj komnate vonjaet po botinku. in each room stink.PRES.3SG DIST shoe.DAT The diagnostic confirms that Russian emission verbs are unergative. Under the existing characteri- zation of the input to the DDC, they are predicted to allow it. This prediction is not borne out. (33) a. *Emu ne poteetsja. he.DAT NEG sweat.PRES.3SG.SJA b. *Emu ne vonjaetsja. he.DAT NEG stink.PRES.3SG.SJA

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 254 DGANIT JENIA KIM AND TAL SILONI

The difference between emission verbs and unergatives that do feed the DDC (e.g., rabotat’ ‘work’) seems to lie in their argument structure. As mentioned above, emission verbs are “non- agentive”; we will argue in section 3.3.3 that this makes them unsuitable as input for the DDC. Besides emission verbs, there are other types of unergatives that do not give rise to the DDC alternation: reflexives, reciprocals, and subject-Experiencer verbs. In (34)–(36), the (a) examples present relevant types of verbs, and the (b) sentences show that the subjects of such verbs do not license the distributive po-phrase. This means that they are unergative.16 The (c) sentences show that these verbs are nonetheless disallowed in the DDC. (34) Reflexives a. V prudu kupalsja mal’?ik. in pond wash.PAST.SG.MS.SJA boy.NOM ‘There was a boy bathing in the pond.’ b. *V kaÅdom prudu kupalos’ po mal’?iku. in each pond wash.PAST.SG.NEU.SJA DIST boy.DAT c. *Mne v?era ne kupalos’. I.DAT yesterday NEG wash.PAST.SG.NEU.SJA (35) Reciprocals a. Na skamejke celovalas’ para. on bench kiss.PAST.SG.FM.SJA couple.NOM ‘There was a couple kissing on the bench.’ b. *Na kaÅdoj skamejke celovalos’ po pare. on each bench kiss.PAST.SG.NEU.SJA DIST couple.DAT c. *Nam ne celovalos’. we.DAT NEG kiss.PAST.SG.NEU.SJA (36) Subject-Experiencer verbs a. Ministr volnovalsja iz-za stat’ji. minister.NOM.MS worry.PAST.SG.MS.SJA because article ‘The minister was worrying because of the article.’ b. *Iz-za stat’ji, v kaÅdom ofise volnovalos’ po ministru. because article in each office worry.PAST.SG.NEU.SJA DIST minister.DAT c. *Mne ne volnovalos’. I.DAT NEG worry.PAST.SG.NEU.SJA Why are these types of verbs excluded from the DDC? Two explanations immediately come to mind. The first is that only base (i.e., underived) entries constitute a possible input to the DDC, while reflexive, reciprocal, and subject-Experiencer verbs all are derived entries. The second is

16 For arguments in favor of the unergative derivation of these classes of verbs in other languages, see Reinhart and Siloni 2005 (for reflexives), Siloni 2012 (for reciprocals), and Pesetsky 1995 and Reinhart 2002 (for subject-Experiencer verbs).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 255

that a morphological “identity avoidance” rule is at work (e.g., Yip 1998). The latter explanation has to do with -sja affixation. In the DDC, the verb is obligatorily marked with -sja, which is also the suffix used by reflexive, reciprocal, and subject-Experiencer verbs (see the (a) examples above). According to “identity avoidance,” homophonous morphemes cannot appear adjacent in the same word; hence, double -sja affixation is banned and the form is blocked. These two explanations seem to be able to account for the fact that the verbs in (34)–(36) are excluded from the DDC. However, they cannot rule out emission verbs as a possible input (33) because these verbs are neither derived entries nor suffixed with -sja. The morphological account also fails to explain why verbs of these classes that are not suffixed with -sja, such as the reciprocals besedovat’ ‘converse’ (37a) and sporit’ ‘argue’ (38a), are excluded from the DDC, as shown in (37b) and (38b). In section 3.3, we suggest that it is the argument structure of (both the -sja-marked and -unmarked) reciprocals, reflexives, and subject-Experiencer verbs and of emission verbs that makes them unsuitable inputs to the DDC (see footnote 18 for further evidence that “identity avoidance” does not suffice to exclude the former from the DDC). (37) a. My v?era ne besedovali. we.NOM yesterday NEG converse.PAST.PL ‘We didn’t converse (with each other) yesterday.’ b. *Nam v?era ne besedovalos’. we.DAT yesterday NEG converse.PAST.SG.NEU.SJA (38) a. My v?era ne sporili. we.NOM yesterday NEG argue.PAST.PL ‘We didn’t argue (with each other) yesterday.’ b. *Nam v?era ne sporilos’. we.DAT yesterday NEG argue.PAST.SG.NEU.SJA In sum, the fact that emission verbs, reflexives, reciprocals, and subject-Experiencer verbs do not participate in the DDC leads us to conclude that it is not the set of unergatives that defines the input to the DDC, but a subset thereof. In section 3.3, we will propose an alternative characterization of the set.

3.3 Redefining the Input 3.3.1 Verbs Admissible in the DDC Three sets of verbs appear in the Russian DDC: (a) intransi- tives with an Agent, (b) intransitives whose argument is either an Agent or an Experiencer, (c) “object drop” transitives with an Agent. The first type includes verbs such as rabotat’ ‘work’, begat’ ‘run’, guljat’ ‘stroll’, xodit’ ‘walk’, plavat’ ‘swim’, prygat’ ‘jump’, 'utit’ ‘joke’, pet’ ‘sing’, tancevat’ ‘dance’, pljasat’ ‘dance’. A verb of this type, guljat’ ‘stroll’, is illustrated in (39a), occurring in the DDC in (39b). (39) a. Ja guljaju (v parke). I.NOM stroll.PRES.1SG in park ‘I am walking around (in the park).’

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 256 DGANIT JENIA KIM AND TAL SILONI

b. Na ulice doÅdi i xolod, tak ?to mne ne guljaetsja. on street rains and cold so I.DAT NEG stroll.PRES.SG.SJA ‘It’s rainy and cold on the street, so I don’t feel like walking around.’ (http://golodanie.su/forum/archive/index.php/t-12087.html, last accessed August 2017) These verbs denote activities, and their argument is an Agent, a human (more generally, animate) entity perceived as volitionally causing the eventuality described by the verb. It is widely accepted that this argument merges externally (e.g., the distributive po-phrase test in (30)). The second type of verb includes intransitives such as spat’ ‘sleep’, sidet’ ‘sit’, leÅat’ ‘lie down’, otdyxat’ ‘rest, be on vacation’, Åit’ ‘live’. This type is illustrated in (40) by spat’ ‘sleep’.17 (40) a. V?era ja ne spal. yesterday I.NOM NEG sleep.PAST.SG.MS ‘I didn’t sleep yesterday.’ b. Nesmotrja na ustalost’, mne ne spalos’. despite tiredness I.DAT NEG sleep.PAST.SG.NEU.SJA ‘Despite my tiredness, I could not sleep / I felt that my sleeping was going badly.’ (Volkov, Smjatenie Anastasii (1988). From the Russian National Corpus: http:// www.ruscorpora.ru) These verbs denote physical states, and their argument is also necessarily animate. However, unlike the subject of an agentive intransitive, this argument is not necessarily understood as causing the eventuality. In many contexts, it is most naturally understood as the Experiencer of the state described by the verb—for example, the state of being asleep in John was tired and now he is sleeping. In other contexts, however, this argument is understood as volitionally causing the eventuality—for example, in John sleeps a lot in order to have energy for his football training. The derivation of such verbs is also unergative, as shown by the distributive po-phrase test.18

17 The results of the surveys reported in section 2 offer reason to suspect that the participation of this group of verbs in the DDC is somewhat more restricted than that of the other two groups. In questionnaire I, we noticed that for two of the three verbs from this group the adverbial DDC is much more acceptable than the negated type. In questionnaire II, we noticed that the dispositional meaning was not available for the verb spat’ ‘sleep’, as discussed in section 2.3.3. Despite these peculiarities, it is clear both from the surveys and from corpora that this set of verbs serves as input to the construction. Systematic research into the restrictions is left for future study. 18 The verbs ulybat’sja ‘smile’ and smejat’sja ‘laugh’, which belong to the second type, tend not to appear in the DDC. These verbs have a -sja suffix as part of their form (they do not have an alternant without -sja). Their exclusion from the DDC may be related to “identity avoidance,” which bans adjacent affixation of homophonous morphemes (as mentioned in section 3.2). Indeed, one does find sporadic online examples with haplology, that is, instances of ‘smile’ and ‘laugh’ in the DDC that omit one -sja morpheme, the remaining -sja carrying the semantics of both, as illustrated in (i). Importantly, we have not found any such DDC examples with other -sja verbs, specifically, reflexives, reciprocals, or subject-Experiencer verbs. This suggests that the latter are impossible inputs for (additional) independent reasons, as we indeed argue in section 3.3.3. (i) Zdes’ mne legko smejotsja i dy'itsja. here I.DAT easily laugh.PRES.3SG and breathe.PRES.3SG.SJA ‘I feel that here I can laugh easily and breathe easily.’ (http://www.trud.ru/article/26-08-2003/61062_stanislav_govoruxin_nam_nuzhen_glotok_nadezhdy.html, last accessed April 2019)

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 257

(41) a. V komnate spal mal’?ik. in room sleep.PAST.SG.MS.SJA boy.NOM ‘There was a boy sleeping in the room.’ b. *V kaÅdoj komnate spalo po mal’?iku. in each room sleep.PAST.SG.NEU DIST boy.DAT The third verb type acceptable in the DDC is “object drop” transitive verbs, such as pisat’ ‘write’, ?itat’ ‘read’, ku'at’ ‘eat’, jest’ ‘eat’, pit’ ‘drink’, as illustrated in (42). (42) a. Ja ?itaju (gazetu). I.NOM read.PRES.1SG newspaper.ACC ‘I am reading (a newspaper).’ b. V samoljote mne xoro'o ?itaetsja. in airplane I.DAT well read.PRES.3SG.SJA ‘In airplanes, I feel like reading / I can read / I feel that my reading is going well.’ (http://motorka-lara.livejournal.com/53859.html, last accessed February 2016) These are two-place predicates taking an Agent and an optional internal argument. As mentioned in section 3.1, these verbs can participate in the DDC only if the internal argument is dropped. Transitives whose internal argument is not optional are excluded from the DDC (see, e.g., (22b)). It is important to emphasize that formation of the DDC with verbs from the above three sets is productive; that is, it can apply to neologisms, as illustrated in (43). The verbs used in (43) are relatively new in Russian (they are derived from nouns or verbs borrowed from foreign languages). They appear in the DDC because they belong to one of the classes mentioned above: direktorstvovat’ ‘function/serve as a manager’ is an Agent unergative, and improvizirovat’ ‘impro- vise’ and filosofstvovat’ ‘philosophize’ are “object drop” transitive verbs. (43) a. Xoro'o li emu direktorstvuetsja? well QUEST he.DAT manage.PRES.3SG.SJA ‘Does he experience being a manager positively?’ / ‘Is being a manager going well for him?’ b. Segodnja kompozitoru ?to-to ne improviziruetsja. today composer.DAT for.some.reason NEG improvise.PRES.3SG.SJA ‘For some reason, the composer cannot improvise today.’ / ‘For some reason, the composer’s improvising is not going well today.’ c. Segodnja na seminare nam xoro'o pofilosofstvovalos’. today on seminar we.DAT well PERF-philosophize.PRES.3SG.SJA ‘We experienced our philosophizing at the seminar today positively.’ / ‘Our philoso- phizing at the seminar today went well.’ (Gerritsen 1990:174–175, (262), (263), (265)) Tables 6–7 summarize and exemplify the classes of verbs that can feed the DDC and the ones that cannot. Regarding transitive verbs, the generalization is quite straightforward: only “object drop” transitives are admissible. The situation with intransitive verbs is more puzzling.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 258 DGANIT JENIA KIM AND TAL SILONI

Table 6 Intransitive verbs Participate in the DDC Do not participate in the DDC Unergatives Agent/ Subject- Agent Experiencer Emission Experiencer unergatives unergatives Unaccusatives verbs verbs Reflexives Reciprocals rabotat’ spat’ rasti potet’ volnovat’sja myt’sja obnimat’sja ‘work’ ‘sleep’ ‘grow’ ‘sweat’ ‘worry’ ‘wash’ ‘hug’ begat’ otdyxat’ padat’ vonjat’ udivljat’sja brit’sja vstre?at’sja ‘run’ ‘rest’ ‘fall’ ‘stink’ ‘be surprised’ ‘shave’ ‘meet’ pet’ ‘sing’

As is evident from table 6, both unaccusatives and certain types of unergatives are excluded, and it is unclear what the difference is between the unergatives that can feed the DDC and those that cannot. It may be that only unergatives with a specific type of ␪-role are eligible, but an attempt to reach a generalization regarding the ␪-role in question encounters problems. First, the common denominator underlying the Agent/Experiencer duality in the interpretation of the argument of verbs like spat’ ‘sleep’ cannot be straightforwardly defined in terms of ␪-roles. Second, as the argument of reflexive and reciprocal verbs is perceived as volitionally causing the eventuality, these verbs are agentive; however, they cannot feed the DDC, in contrast with other agentive unergatives. It is thus not clear why they cannot appear in the DDC, while other agentive intransi- tives can. In sum, it seems impossible to define what distinguishes the types of unergatives that partici- pate in the alternation from those that do not. There seems to be no generalization that would single out the ␪-role of unergatives that can feed the construction. In the next sections, we show

Table 7 Transitive verbs Participate in the DDC Do not participate in the DDC Agent- Cause- Object drop Theme Cause-Theme Experiencer Experiencer- Agent-Theme transitives transitives transitives transitives Theme transitives pisat’ ‘write’ stroit’ ‘build’ otkryvat’ ‘open’ volnovat’ ‘worry’ ljubit’ ‘love’ ?itat’ ‘read’ varit’ ‘cook’ napolnjat’ ‘fill’ udivljat’ ‘surprise’ nenavidit’ ‘hate’ jest’ ‘eat’ ?istit’ ‘clean’ razbivat’ ‘break’ znat’ ‘know’

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 259

that adopting the Theta System, a theoretical framework in which ␪-roles can be decomposed into feature clusters, resolves the problem and allows a straightforward definition of the input. 3.3.2 The Theta System The Theta System (Reinhart 2002, 2016, Everaert, Marelj, and Siloni 2012) is a model of representation of thematic information. The system views ␪-roles not as grammatical primitives but as conventionalized labels for feature clusters. The feature composition of roles is based on their semantics. Two atomic features, [c] and [m], underlie the set of ␪-roles. The feature [c] determines whether or not the argument in question is necessarily responsible for causing the denoted event (or change). The feature [m] determines whether or not the mental state of the argument in question is relevant for the denoted event, that is, whether the event or[ם] involves volition and intention of the argument. Each of these features can be positively valued or be unvalued, so certain ␪-clusters are fully specified (i.e., the value of [מ] negatively both their features is defined) and others are underspecified or unary (i.e., the value of one of the features is not defined).19 The fully specified and the unary clusters are shown in table 8. The clusters do not directly correspond to traditional ␪-roles, and many of them have varying contextual interpretations, as explained below. The labels in table 8 are presented for convenience; they m] areמ] c] andמ] represent the role that each cluster is most typically associated with. The roles not relevant for the current study and therefore we do not discuss them. The unary clusters are undefined with respect to the value of one of their features and are thus compatible with either value. This allows them to have a greater range of interpretations [mםcם] depending on the utterance they are realized in. For example, the difference between a c] (Cause) cluster is that the former necessarily involves volition andם] Agent) cluster and a)

Table 8 ␪-clusters Label Causes the event Mental state relevant Fully specified clusters m] Agent Yes Yesםcם] m] Theme No Noמcמ] m] Cause, Instrument Yes Noמcם] m] Experiencer No Yesםcמ] Unary clusters c] Cause Yes Undefinedם] m] Sentient Undefined Yesם] c] Goal, Benefactor No Undefinedמ] m] Subject Matter, Target of Emotion Undefined Noמ]

19 The system also includes an “empty cluster,” which is undefined regarding both features. The empty cluster [ ] is argued to be relevant for lexical middle formation (Marelj 2004), light verbs (Ackema and Marelj 2012), and lexical reciprocal verbs (Siloni 2012). This cluster is not relevant for the current discussion.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 260 DGANIT JENIA KIM AND TAL SILONI

intention, which is expressed via the positive value of the [m] feature. So verbs such as feed, / m] cluster, can only realize it as an animate entity: The father /*The spoonםcם] which have a c] cluster, on the other hand, is unspecified for mental stateם] The hunger fed the baby. The* and thus can be interpreted as an inanimate Cause (e.g., natural force), an Instrument, or an Agent, depending on the specific sentence: for example, The storm / The stone / Max broke the window.20 m] cluster labeled Sentient in the Theta System has not been identified as a role differentם] The from the Experiencer role in other frameworks. Reinhart (2002) associates this cluster with the subject of verbs such as laugh, cry, sleep, see, hear, love, know, believe. The argument realizing m] cluster is usually interpreted as experiencing a mental state, but the difference betweenם]a m] and a Sentient is that the latter, unlike the former, can be interpreted inםcמ] an Experiencer m] cluster. In section 3.3.1, we notedםcם] certain contexts as causing the eventuality, that is, as a that Agent/Experiencer unergatives, such as spat’ ‘sleep’ and sidet’ ‘sit’, participate in the DDC, while subject-Experiencer unergatives, such as volnovat’sja ‘worry’ and udivljat’sja ‘wonder’, .[mםcמ] m] cluster, while the latter haveם] do not. In terms of the Theta System, the former have a Reinhart’s (2002) arguments in favor of the distinction between these two sets are mainly theory- internal (they involve case and mapping considerations). However, Horvath and Siloni (2011) m] external roleם] provide independent support for the distinction: they show that verbs with a m] role cannot. Theםcמ] can serve as input to causativization in Hungarian, while verbs with a Russian DDC turns out to contribute additional independent evidence for the distinction: while m] verbs cannot. We discuss this issue in sectionםcמ] ,m] verbs can serve as input to the DDCם] 3.3.3. One of the achievements of the Theta System is that the feature composition of its clusters gives rise to natural classes of verbs regarding merging. Clusters with positively valued features -c]—always merge exterם] m], and Causeם] m], Sentientםcם] clusters)—that is, Agent [ם]) only m], Subjectמcמ] clusters)—that is, Theme [מ]) nally. Clusters with negatively valued features only c]—always merge internally. The mixed clusters (involving a positivelyמ] m], and Goalמ] Matter m] and Experiencerמcם] valued feature and a negatively valued one)—that is, Instrument cluster is [ם] cluster, and internally when a [ם] m]—merge externally in the absence of aםcמ] present. For example, consider the transitive-intransitive alternation in examples (44a–b) and -cluster that merges exter [ם] 45a–b). In both cases, the transitive in the (a) example involves a) cluster is not present in the intransitive derivations in the (b) examples, which realize [ם] nally. This only the argument that was internal in the (a) examples. The distributive po-phrase diagnostic in the (c) examples shows that the mapping of the intransitive alternant in the two (b) examples is distinct: while the argument of the intransitive razbit’sja ‘break.SJA’ merges internally, the argu- ment of the intransitive volnovat’sja ‘worry.SJA’ is external. (44) a. Ma'a razbila okno. Masha.NOM break.PRES.SG.FM window.ACC ‘Masha broke the window.’ m] cluster; the difference between them is thatמcם] Both an inanimate Cause and an Instrument correspond to a 20 an Instrument never causes the event by itself, but requires an explicit or implicit Agent (Reinhart 2002, Siloni 2002).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 261

b. Okno razbilos’. window.NOM.NEU break.PRES.SG.NEU.SJA ‘The window broke.’ c. V kaÅdoj komnate razbilos’ po oknu. in each room break.PAST.SG.NEU.SJA DIST window.DAT ‘A window broke in each room.’ (45) a. Ego povedenie volnuet Ma'u. his behavior.NOM.NEU worry.PRES.3SG Masha.ACC ‘His behavior worries Masha.’ b. Ma'a volnuetsja. Masha.NOM worry.PRES.3SG.SJA ‘Masha worries.’ c. *V kaÅdom ofise volnovalos’ po ministru. in each office worry.PAST.SG.NEU.SJA DIST minister.DAT This difference between the two (c) examples is naturally explained by the cluster composition [mמcמ] of their argument. The argument of the unaccusative razbit’sja ‘break.SJA’ realizes a cluster (Theme): it does not bring about the event and its mental state is irrelevant. Since it is a cluster, it always merges internally (44c). The argument of the subject-Experiencer verb [מ] m] (Experiencer); in the absence ofםcמ] volnovat’sja ‘worry.SJA’, in contrast, is a mixed cluster .(cluster, it merges externally (45c [ם]a (m] (typically, an Instrument), is argued by Potashnik (2012מcם] ,The second mixed cluster to be the role realized by the subject of emission verbs (see discussion in section 3.2).21 This cluster also demonstrates the mapping pattern of mixed clusters: it merges externally in the absence [ם] cluster, as shown by the po-phrase test in (46b), and internally (as a PP) when a [ם]ofa cluster is present, as illustrated in (46c).22 (46) a. V kvartire zvonil telefon. in apartment ring.PAST.SG.MS telephone.NOM ‘There was a telephone ringing in the apartment.’ b. *V kaÅdoj kvartire zvonilo po telefonu. in each apartment ring.PAST.SG.NEU DIST telephone.DAT c. Ma'a zvonila v kolokol / v dver’. Masha ring.PAST.SG.FM in bell / in door ‘Masha rang the bell / the doorbell.’

21 Potashnik’s (2012) proposal differs from Reinhart’s (2002) analysis of this class; under the latter, emission verbs m] cluster. To account for their unergative derivation, Reinhartמcמ] are viewed as “theme unergatives,” that is, as having a m]. His analysisמcם] introduces a special constraint. Potashnik offers empirical evidence showing that their role is eliminates the need to assume a special constraint for the mapping of emission verbs. 22 The two-place alternant in (46c) is possible only for a subset of emission verbs. For more discussion, see Potashnik 2012.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 262 DGANIT JENIA KIM AND TAL SILONI

The two additional classes of intransitives mentioned in the previous sections are reflexive and reciprocal verbs. Reinhart and Siloni (2005) and Siloni (2012) argue that these verbs are derived from their transitive alternants by an operation that takes the two ␪-roles of the verb (e.g., an Agent and a Theme) and forms one complex ␪-role that retains the semantic properties of both original roles. For example, the transitive entry wash (as in John washed the dishes) with an (m] turns into the intransitive entry wash (as in John washedמcמ] m] and a Themeםcם] Agent m]]. The complex cluster is assigned to the sole argumentמcמ][mםcם]] with one complex ␪-cluster of the intransitive, resulting in the interpretation of this argument as being in the relation Agent and Theme to the event. With the basic foundations of the framework in mind, we resume discussion of the set of verbs that can serve as input to the DDC. As will become clear in the next section, the feature clusters suggested by the Theta System allow a natural definition of the DDC input. 3.3.3 The Revised Characterization of the Input In sections 3.1, 3.2, and 3.3.1, we examined different types of verbs, showing which classes permit the DDC. We showed that the commonly accepted generalization that the input to the Russian DDC is the set of unergatives is inaccurate. But we provided no clear generalization instead. We now show that in the Theta System, the sets of verbs feeding the DDC form a natural class. Tables 6–7 are repeated in tables 9–10, with the addition that they specify the ␪-clusters cluster can [ם] that the verbs are equipped with. As is clear from table 9, intransitives with a or a mixed cluster cannot. Transitive verbs with a [מ] feed the DDC, while intransitives with a cluster can participate in the DDC only if they allow “object drop.” Transitive verbs whose [ם] ,m] argument disallow “object drop” and hence do not appear in the DDCם] c] orם] subject is a

Table 9 Intransitive verbs Participate in the DDC Do not participate in the DDC Unergatives Agent/ Subject- Agent Experiencer Emission Experiencer unergatives unergatives Unaccusatives verbs verbs Reflexives Reciprocals [mםcם]] [mםcם]] [mםcמ] [mמcם] [mמcמ] [mם] [mםcם] [[mמcמ] [[mמcמ] rabotat’ spat’ rasti potet’ volnovat’sja myt’sja obnimat’sja ‘work’ ‘sleep’ ‘grow’ ‘sweat’ ‘worry’ ‘wash’ ‘hug’ begat’ otdyxat’ padat’ vonjat’ udivljat’sja brit’sja vstre?at’sja ‘run’ ‘rest’ ‘fall’ ‘stink’ ‘be surprised’ ‘shave’ ‘meet’ pet’ ‘sing’

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 263

Table 10 Transitive verbs Participate in the DDC Do not participate in the DDC Cause- Object drop Agent- Agent-Theme Cause-Theme Experiencer Sentient-Theme Theme transitives transitives transitives transitives transitives [mמcמ] ,[mם] [mםcמ] ,[cם] [mמcמ] ,[cם] [mמcמ] ,[mםcם] ([mמcמ]) ,[mםcם] pisat’ ‘write’ stroit’ ‘build’ otkryvat’ ‘open’ volnovat’ ‘worry’ ljubit’ ‘love’ ?itat’ ‘read’ varit’ ‘cook’ napolnjat’ ‘fill’ udivljat’ ‘surprise’ nenavidit’ ‘hate’ jest’ ‘eat’ ?istit’ ‘clean’ razbivat’ ‘break’ znat’ ‘know’

as can be seen in table 10. Thus, the input to DDC formation in Russian can be defined as in (47). (47) Input to DDC formation in Russian A verb can give rise to the DDC if and only if it is an intransitive verb, including .cluster [ם] object drop” transitive verbs, with a“ The definition in (47), unlike the definition that limits the set to unergatives, distinguishes between clusters that must be external and those that can be external but do not have to be. The former are clusters that are external inherently due to their feature composition, while the latter are clusters that are not inherently external, but rather merge externally in certain environments. Indeed, Horvath and Siloni (2011) observe that under the Theta System the term externally mapped clusters) and [ם]) ␪-cluster lumps together two distinct kinds of roles: inherently external roles roles that de facto merge externally. As discussed in section 3.3.2, certain ␪-roles—specifically [ם] m])—merge internally in derivations that include aםcמ] m] andמcם]) the mixed clusters cluster. Therefore, although they are not inherently [ם] cluster and externally in the absence of a external roles (i.e., they are not external by their cluster composition), in intransitive derivations they merge externally. The definition of the input to DDC formation as unergative verbs includes cluster and verbs with a mixed cluster. In contrast, the revised definition [ם] both verbs with a proposed here defines the input more precisely as the set of intransitive verbs (including “object cluster. We resume discussion [ם] drop” transitives) that have an inherently external role, that is, a of this point in section 4.2. In section 4.3.1, we further discuss the exclusion of transitive verbs from the DDC. In section 4, overall, we take up the formation of the DDC.

4 Formation In light of the discussion thus far, any analysis of the DDC must explain three main characteristics of the alternation, summarized in (48).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 264 DGANIT JENIA KIM AND TAL SILONI

(48) a. While the argument of the base entry is an Agent or a Sentient, the DDC involves an Experiencer (section 2.1). b. The DDC denotes a type of eventuality different from that denoted by the base entry. While the latter is either an activity (e.g., working) or a physical state (e.g., sitting), the DDC alternant denotes a psychological state regarding this eventuality. As is clear from the results of questionnaire II (section 2.3.3), the psychological state described by the DDC can be either (a) disposition or psychological capability to participate in this eventuality or (b) a subjective evaluation of the eventuality (or of the participant’s mental state during the eventuality). c. While adverbial modification (including negation) is usually optional, it is obligatory in the DDC (sections 2.2.2–2.2.3). Further, the interpretation is “shifted.” For exam- ple, while the adverb ploxo ‘badly’ usually modifies the eventuality, in the DDC it describes a negative disposition toward the eventuality or a subjective negative evaluation of it (section 2.1). We first discuss representative syntactic analyses of the DDC, with particular attention to the above points, and show that these analyses are untenable as far as the Russian DDC is concerned (section 4.1). We then turn to our proposal (sections 4.2–4.3).

4.1 Representative Syntactic Approaches Many of the analyses proposed in the literature derive the DDC by means of inserting additional syntactic structure above the VP projected by the lexical verb (i.e., the base verb/entry). Benedicto (1995) and Rivero and Arregui (2012) attribute the characteristic psychological interpretation of the DDC (48b) to the presence of a null modal head; the latter embeds a clausal complement (AgrP/TP) that includes the base verb and its argument, which is implicit.23 The dative noun phrase in the DDC is analyzed as the subject of the modal, hence its Experiencer flavor (48a). The adverbial is also viewed as an argument of the modal (Rivero and Arregui 2012), hence the shift in its meaning (48c). Maru'i? and äaucer (2006) offer a similar analysis, but with a null psych verb (rather than a null modal) above the embedded projection. They suggest that the DDC is a biclausal structure with a covert GIVE predicate, which is interpreted as a psych verb taking a clausal complement. This structure, as schematized for Russian in (49), is the null parallel of the Slovenian ‘feel like’ construction, which includes an overt nonactive verb ‘give’, as shown in (50). Under this analysis, the base verb’s ␪-role is realized as PRO, and the dative noun phrase is the Experiencer argument of the null psych verb, that is, the subject of the matrix clause.

23 More precisely, according to Benedicto (1995), the ␪-role of the base verb is realized syntactically as pro. Rivero and Arregui (2012) suggest that the subject position of the embedded clause is saturated by a variable (the “reflexive pronoun”) that is interpreted as the Agent of V.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 265

(49) Mne ne GIVE [PRO rabotaetsja]. I.DAT NEG work.PRES.3SG.SJA ‘I don’t feel like working.’ (50) Danes se mi ne da delat. today SE I.DAT NEG give.3SG work ‘I don’t feel like working today.’ (Slovenian; Maru'i? and äaucer 2006:1148, (84a)) The two analyses share the view that a VP headed by the base verb is present in the structure of the DDC and that the semantic effects that characterize the construction are due to the presence of the higher null head. Maru'i? and äaucer (2006) show that the DDC in Slovenian allows two temporal adverbials, one associated with the dispositional eventuality denoted by the null predicate (‘yesterday’ in (51)) and the other modifying the embedded verb (‘tomorrow’ in (51)). (51) V?eraj se mi ni 'lo jutri domov. yesterday SE I.DAT AUX.NEG.PAST go tomorrow home ‘Yesterday, I didn’t feel like going home tomorrow.’ (Slovenian; Maru'i? and äaucer 2006:1098, (13)) This, however, is impossible in the Russian DDC. While the biclausal construction involving an overt desiderative predicate allows for two incompatible temporal adverbials in Russian (52a), just as in the Slovenian DDC, the Russian DDC disallows this cooccurrence (52b). (52) a. V?era mne ne xotelos’ segodnja tancevat’. yesterday I.DAT NEG want.PAST.SG.NEU.SJA today dance.INF ‘Yesterday I didn’t feel like dancing today.’ b. *V?era mne segodnja ne tancevalos’. yesterday I.DAT today NEG dance.PAST.SG.NEU.SJA The fact that incompatible temporal adverbs are impossible in the Russian DDC (52b) could perhaps follow from a different variant of the syntactic analysis. Suppose that in the Russian DDC, the embedded verb projects a smaller bit of structure lacking a tense operator. It could then follow that unlike in the biclausal structure, in the DDC the matrix head and the embedded one are bound by the same temporal operator, which would explain why contradicting temporal modifiers are impossible. However, incompatible nontemporal adverbs cannot cooccur in the Russian DDC either, as discussed directly. Rivero and Arregui (2012) show that the DDC in Polish and Czech allows two semantically incompatible manner adverbials, such as ‘well’ and ‘badly/terribly’, as illustrated in (53a–b). These sentences are possible because one adverb (‘badly’ in (53)) modifies the embedded verb, and the other one (‘well’) modifies the modal head. If the Russian DDC involved a structure similar to the one in Polish and Czech, as these authors suggest, the same pattern of modification would be expected to be possible in Russian as well. However, Russian disallows two incompatible manner adverbs, as illustrated in (53c).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 266 DGANIT JENIA KIM AND TAL SILONI

(53) a. Dobrze Jankowi tanc´zylo sie fatalnie. well Jan.DAT danced.NEU SE terribly ‘Jan could not help enjoying his awful dancing.’ (Polish; Rivero and Arregui 2012:316, (45a)) b. Jankovi se dobrˇe tan?ilo bylbje. Jan.DAT SE well danced.NEU badly ‘Jan could not help enjoying his awful dancing.’ (Czech; Rivero and Arregui 2012:316, (45b)) c. *Emu xoro'o tancevalos’ Åutko. he.DAT well dance.PAST.SG.NEU.SJA terribly Further, the verb in the Russian DDC does not even license the same set of modifiers as the corresponding base verb. Some manner adverbs, like krasivo ‘beautifully’, which can modify the base verb both in (54a) and in the biclausal desiderative construction (54b), are impossible in the DDC (54c). (54) a. On krasivo tanceval. he.NOM beautifully dance.PAST.SG.MS ‘He danced beautifully.’ b. Emu xotelos’ krasivo tancevat’. he.DAT want.PAST.SG.NEU.SJA beautifully dance.INF ‘He wanted to dance beautifully.’ c. *Emu krasivo tancevalos’. he.DAT beautifully dance.PAST.SG.NEU.SJA In sum, the base verb of the Russian DDC is not detectable by syntactic diagnostics, which means that it does not project in the syntax as is; the eventuality it denotes is merely implied. This view is reinforced by the interpretation of the DDC in (23b), repeated as here as (55). This example involves the nonargumental PP modifier pri takix uslovijax ‘under such conditions’. The PP necessarily modifies the whole predicate, namely, the psychological state (disposition or evaluation) as well as the implied working event (possible or actual), showing that the DDC involves a single syntactic predicate, the DDC alternant. (55) Mne ne rabotaetsja pri takix uslovijax. I.DAT NEG work.PRES.3SG.SJA under such conditions ‘I don’t feel like working under such conditions.’ / ‘My psychological state makes me unable to work under such conditions.’ / ‘I feel that my work is going badly under such conditions.’ (Franks 1995:366, (68a)) Similarly, in (56) the PP v Moskve ‘in Moscow’, independently of its position in the sentence, necessarily modifies the syntactic head rabotaetsja ‘work.SJA’, involving both the psychological state (disposition or evaluation) and the implied working event (possible or actual). This is ex- pected if the construction involves a single verbal head in the syntax.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 267

(56) (V Moskve) mne xoro'o rabotaetsja (v Moskve). in Moscow I.DAT well work.PRES.3SG.SJA in Moscow ‘In Moscow I feel like working.’ / ‘I am psychologically capable of working in Mos- cow.’ / ‘When I work in Moscow, I experience it positively.’ In sum, there is evidence that the base verb does not project in the Russian DDC. That is, Russian, unlike West Slavic (e.g., Polish and Czech) and South Slavic (e.g., Slovenian) languages, does not derive its DDC by insertion of syntactic structure above the base VP. Two additional facts point in this direction. First, given that the base verb does not project in the syntax, we do not expect its argument to be syntactically present either. This means that the sole argument in the Russian DDC is the dative Experiencer. Consider then rationale clauses. The two main views regarding their control properties are that (a) their PRO subject is controlled by the matrix (explicit or implicit) Agent (e.g., Roeper 1987) and (b) their PRO subject is controlled by the eventuality they refer to (e.g., Williams 1985). The agentive verb rabotat’ ‘work’ (57b) [m] cluster undefined regarding its [cם] and the verb spat’ ‘sleep’ (58b), which has a Sentient value and thus compatible with agentive interpretation, allow a rationale clause. They also allow such a clause when they are embedded in the biclausal desiderative construction, as shown by the (c) examples. In contrast, the corresponding DDCs disallow rationale clauses, as shown by the (a) examples. Regardless of which approach to rationale clauses is on the right track, the fact that the Russian DDC disallows them is predicted by our conclusion that the construction syntactically realizes neither the base verb nor its argument; instead, it realizes the DDC alternant and its (dative) Experiencer.24 (57) a. *Mne xoro'o rabotaetsja ?toby ugodit’ na?al’niku. I.DAT well work.PRES.3SG.SJA in.order.to please.INF boss.DAT b. Ja xoro'o rabotaju ?toby ugodit’ na?al’niku. I.NOM well work.PRES.1SG in.order.to please.INF boss.DAT ‘I work well in order to please the boss.’ c. Mne xo?etsja xoro'o rabotat’ ?toby ugodit’ na?al’niku. I.DAT want.PRES.3SG.SJA well work.INF in.order.to please.INF boss.DAT ‘I feel like working well in order to please the boss.’ (58) a. *Mne xoro'o spitsja ?toby produktivno rabotat’. I.DAT well sleep.PRES.3SG.SJA in.order.to productively work.INF b. Ja xoro'o splju ?toby produktivno rabotat’. I.NOM well sleep.PRES.1SG in.order.to productively work.INF ‘I sleep well in order to work productively.’ c. Mne xo?etsja xoro'o spat’ ?toby produktivno rabotat’. I.DAT want.PRES.3SG.SJA well sleep.INF in.order.to productively work.INF ‘I feel like sleeping well in order to work productively.’

24 The PRO subject of the rationale clause can be controlled neither by the mental state denoted by the DDC nor by its Experiencer.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 268 DGANIT JENIA KIM AND TAL SILONI

Moreover, as is well-known, Agent-oriented adverbs require an (explicit or implicit) Agent. Unlike sentences involving the corresponding base verb ((59b), (60b)) and unlike biclausal desiderative constructions ((59c), (60c)), the Russian DDC ((59a), (60a)) disallows Agent-oriented adverbs. This shows that the Agent of the base verb is not present in syntax (nor is it present in the semantic representation exclusively, a point we will return to in section 4.3.1). (59) a. *Mne naro?no ploxo rabotaetsja. I.DAT deliberately badly work.PRES.3SG.SJA b. Ja naro?no ploxo rabotaju. I.NOM deliberately badly work.PRES.1SG ‘I deliberately work badly.’ c. Mne xo?etsja naro?no ploxo rabotat’. I.DAT want.PRES.3SG.SJA deliberately badly work.INF ‘I feel like working badly deliberately.’ (60) a. *Mne naro?no ne spalos’. I.DAT deliberately NEG sleep.PAST.SG.NEU.SJA b. Ja naro?no ne spala. I.NOM deliberately NEG sleep.PAST.SG.FM ‘I deliberately didn’t sleep.’ c. Mne xo?etsja naro?no ne spat’. I.DAT want.PRES.3SG.SJA deliberately NEG sleep.INF ‘I feel like deliberately not sleeping.’ In conclusion, the above data show that the base verb does not project as is in the Russian DDC, nor does its argument.25 In light of this fact, two options arise: (a) It is not the base verb itself that merges; rather, it is the bare root, which as such can neither be modified (hence, it is undetectable by diagnostics) nor give rise to the projection of its argument. Subsequently, the root is associated with (raises to) a psychological stative head that introduces the Experiencer argument, thereby forming the DDC alternant syntactically. (b) The operation deriving the DDC alternant from the base entry applies prior to the merger of syntactic structure (i.e., in the lexicon), and the output of the operation is inserted into syntactic structure. The building blocks of lexical operations are not visible to the syntax. Hence, the base entry is not syntactically detectable. In the next section, we present evidence in favor of the latter option: the Russian DDC is derived prior to the merger of syntactic structure. We show that the input to the formation of the DDC alternant cannot be defined in syntactic terms; hence, the “root option” mentioned in (a) should be discarded.

25 An additional Agenthood diagnostic proposed in the literature is licensing of an Instrument (see Reinhart 2002, Siloni 2002). See footnote 30 for some discussion.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 269

4.2 Lexical Formation: The Input Set In contrast to the syntactic proposals discussed in the previous section, Franks (1995) suggests that the DDC alternant in Russian is formed in the lexicon, while in other Slavic languages, such as Polish (West Slavic) and Serbian/Croatian (South Slavic), it is formed in the syntax. His argument is based on the observation that the Russian variant is more restricted than its counter- parts in other Slavic languages; in Russian, unlike in the latter, verbs with an internal argument cannot feed the DDC (section 3.1). According to Franks, this restriction could follow from general lexical word formation rules, or it could be explicitly stipulated in lexical terms referring to the input’s argument structure. Under this view, the restriction on DDC alternant formation in Russian is a result of its being a lexical operation. In other Slavic languages, the DDC is formed in the syntax, which is, so the argument goes, the productive engine. Hence, the operation is productive. As detailed in section 3.1, the input to the Russian DDC was defined in previous studies as the set of unergative verbs, that is, intransitives whose argument is external (including “object drop” transitives). This definition refers to a syntactic characteristic: external vs. internal. We have shown that this is not accurate. If it were, the set could be defined in syntactic terms: the set of one-place verbs (roots) associated with an external argument and no internal one. The restriction on the input then could not serve as evidence that the operation is lexical: the difference between Russian on the one hand and West Slavic and South Slavic on the other would be that in the former, but not in the latter, the input is limited to verbs with an external argument and no internal one. Nonetheless, we argue that Franks’s conclusion that the formation operation is lexical, not syntactic, is correct, but on somewhat different grounds. The more precise characterization of the input set offered in section 3.3.3 turns out to provide direct evidence for a lexical formation operation in Russian. Recall that in section 3.2, we rejected the claim that all unergative verbs can feed the DDC, using evidence from emission verbs and from subject-Experiencer verbs, reflexives, and reciprocals. We showed that these verbs do not constitute suitable inputs, even though they have an external argument (and no other arguments). Instead, we defined the input cluster. Such a characterization of the input is compatible [ם] as the set of intransitives that have a only with a lexical formation operation, since the feature composition of ␪-clusters is accessible in the lexicon, but not in the syntax. The syntactic component does not “see” the feature makeup cluster [ם] of ␪-clusters and therefore cannot distinguish between entries whose external role is a (these can feed the DDC) and entries whose external role is mixed (as discussed in sections 3.3.2–3.3.3) (these cannot feed the DDC). For the syntax, the relevant distinction is between internal and external merger. In other words, the fact that it is not the actual merger in syntax that matters, but the internal composition of the ␪-cluster, provides direct evidence that the formation is lexical. On any syntactic derivation of the Russian DDC alternant, it would be unclear what the common denominator is among verbs such as ‘work’ and ‘sleep’ that makes them able to give rise to the DDC, while verbs such as ‘sweat’ and ‘stink’ cannot do so. There would be no straightforward explanation for the fact that the former can accommodate a dispositional/evalua- tive meaning alongside the related Experiencer, while the latter cannot.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 270 DGANIT JENIA KIM AND TAL SILONI

A parallel argument is advanced by Horvath and Siloni (2011), who argue that causativization in Hungarian is a lexical operation. Among other things, they show that the operation is sensitive to the feature composition of the external role of the input. Notice in passing that the fact that the feature makeup of the external role is accessible to the operation forming the Russian DDC alternant as well as to the one forming the Hungarian causative verb shows also that information about the external cluster (role) is associated with verbal entries in the lexicon. Finally, the claim that the operation is lexical in Russian fits into the bigger picture of intralanguage consistency described in the literature regarding the locus of derivation of certain verbal alternations. Reinhart and Siloni (2005) argue that the formation of reflexive and reciprocal verbs, as well as middle formation, apply in certain languages in the lexicon and in others in the syntax (as described by their lexicon-syntax parameter). Siloni (2012) entertains the idea that the properties of the morphological outfit that a language uses for certain argument structure operations determine where they apply. Systematic variations across languages and predicate classes are explained in terms of the distinct locus of formation (see, e.g., Franks 1995, Marelj 2004, Reinhart and Siloni 2005, Horvath and Siloni 2008, 2011, Meltzer-Asscher 2011, Hron 2012, Siloni 2012). Further, diverse studies provide psycholinguistic and neurolinguistic evidence that certain opera- tions on argument structure apply in the lexicon (e.g., Shetreet, Friedmann, and Hadar 2009, 2010, Shetreet and Friedmann 2012, Meltzer-Asscher et al. 2013, Siloni et al. 2018). At any rate, Russian is argued to form reflexives, reciprocals, and middles (which all use the -sja morpheme) in the lexicon.26 The lexical formation we advance for the Russian DDC alternant (which is marked by -sja, too) converges with the proposed classification. In sum, in section 4.1 we showed that the base verb does not project in the Russian DDC, nor does its argument: neither is detectable in the syntax. In this section, we explained why the cluster provides [ם] fact that the formation of the Russian DDC alternant targets only verbs with a direct evidence that the DDC alternant is formed by a lexical operation: since the internal composi- tion of ␪-clusters is not the type of information that the syntactic component has access to, the set of verbs feeding the DDC does not constitute a natural class in syntactic terms. That is, under a syntactic derivation of the Russian DDC, it would be unclear why certain classes of unergatives can feed it, but not others. Finally, our proposal is coherent with previous proposals locating parallel operations in Russian in the lexicon. The fact that empirical evidence shows that certain operations on argument structure (in the present case, the formation of the Russian DDC alternant) must apply before the merger of syntactic structure—that is, in the lexicon—undermines the rather standard assumption nowadays that the syntax is the sole component where various predicates are formed. We now turn to the formation of the Russian DDC.

26 Hron (2012) offers a thorough survey of Slavic languages with regard to reflexive, reciprocal, and middle formation. He also argues that their formation is lexical in Russian (as well as in other East Slavic languages, such as Ukrainian and Belarusian), while in West and South Slavic languages it is syntactic.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 271

4.3 The Formation of the Russian DDC The operation forming the DDC alternant in Russian is lexical, in concert with the findings presented in sections 4.1–4.2 and given the evidence that the lexicon allows for operations on argument structure (such as reflexivization and reciprocalization; see the literature cited in section 4.2). As summarized in (48), any proposal regarding the formation of the Russian DDC should explain (a) the fact that the Agent of the input appears as the Experiencer of the output, (b) the interpretation of the output, and (c) the obligatoriness of adverbial modification. In sections 4.3.1–4.3.2, we discuss these issues. cluster (role) of the input verb [ם] Feature Adjustment In section 4.1, we showed that the 4.3.1 is not syntactically present in the Russian DDC. Franks (1995) and Fehrmann, Junghanns, and Lenertova´ (2010) suggest that the Agent role of the input is modified so that it is understood as the Experiencer of the DDC.27 But it is unclear how an Agent can become an Experiencer under the traditional view of ␪-roles as primitives.28 Under the Theta System, however, a lexical operation involving role adjustment (role modification) is possible, as argued by Reinhart (2002) and Hor- vath and Siloni (2011) and as explained directly, since roles constitute feature clusters. We suggest that DDC alternant formation involves adjustment of the value of the [c] feature .cluster [ם] of the input’s cluster. The operation negatively revalues the [c] feature of the input’s m] cluster (Sentient), theם] m] cluster (Agent) or aםcם] This cluster can originally be either a role of verbs such as spat’ ‘sleep’. In the former case, the [c] feature is positively valued, and in the latter, is unspecified. In either case, the operation revalues the [c] feature, assigning it a m] cluster (Experiencer), as schematized inםcמ] negative value, thus forming an entry with a (61). The suffix -sja is the morphological marking licensing the operation.29 V stands for the input of the operation, and V-SJA for its output. (61) Feature adjustment in the formation of the Russian DDC alternant [mםcמ]m] N V-SJAםcם]a. V ([mםcמ] m] N rabotat’sjaםcם] ’e.g., ‘work’: rabotat) [mםcמ]m] N V-SJAם]b. V ([mםcמ] m] N spat’sjaם] ’e.g., ‘sleep’: spat) cluster of the [ם] The feature system enables us to offer a straightforward mechanism to turn the -m]. This mechanism has already been proposed in the literaםcמ] input verb into an Experiencer

27 More precisely, according to Franks’s proposal, the Experiencer is realized by the dative noun phrase. Fehrmann, Junghanns, and Lenertova´ (2010) suggest that the Experiencer is implicit, the dative being an adjunct coindexed with it and licensed by the modal operator. 28 Suggesting that the Agent role is removed and instead an Experiencer is inserted seems to us ad hoc; we therefore do not pursue this direction. 29 Recall that the -sja suffix is found on various types of verbs in Russian. The notation V-SJA used in this section refers specifically to the DDC alternant and not to any other verbs suffixed with -sja.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 272 DGANIT JENIA KIM AND TAL SILONI

ture: Horvath and Siloni (2011) suggest that the lexical operation of causativization includes revaluation of the [c] feature, assigning it a negative value. -m] (Experiencer) is by and large associated with verbs denoting psychologםcמ] The ␪-cluster ical (mental) states, such as ‘worry’ (see section 3.3.2). Similarly, the resulting DDC alternant, V-SJA, does not denote the activity or physical state denoted by the input verb. Rather, V-SJA is understood as a psychological state experienced by its argument. It now must be explained what it means to experience working or dancing as a psychological state. Under one of the readings available for the DDC, the psychological state is a subjective evaluation of the eventuality denoted by V (the input): for example, ‘I feel that my work is going well’ (or ‘I feel well while working’). In this case, an actual eventuality denoted by V is entailed by the mental state V-SJA. It is also entailed that the Experiencer of the mental state is the Agent (or Sentient) of the entailed eventuality (e.g., the one that is working). This “double” interpretation of the argument has already been described by, for example, RuÅi?kova (1971) and Rivero and Arregui (2012) (regarding the Slovak construction). In the other readings of the DDC (e.g., ‘I feel like working’ and ‘I can work due to my psychological circumstances’), the psychological state is a disposition toward the possibility of performing the activity denoted by V (the input) or undergoing the state V denotes. Here, V-SJA also entails the eventuality denoted by V whose Agent (or Sentient) is the Experiencer of V-SJA, the difference being that it is not an actual eventuality in the real world, but a possible one. We define the above information by means of a meaning postulate governing the entailment relation- ship between the output V-SJA and its input (V). Meaning postulates place constraints on the relation between lexical items, without assuming syntactic decomposition (see Chierchia and McConnell-Ginet 1990). The meaning postulate governing the interpretation of the verbal entry m] is summarized in (62). The entailment relationships between the output and inputםcמ] V-SJA are given in (a) and (b). (62) The DDC alternant: Meaning postulate V-SJA denotes a psychological state regarding the eventuality denoted by the input V

Subjective evaluation of V Disposition toward V a. V-SJA entails an actual eventuality V a. V-SJA entails a possible eventuality V b. [–c+m] of V-SJA is [(+c)+m] of V b. [–c+m] of V-SJA is [(+c)+m] of V

As shown in section 4.1, the input verb is not detectable by adverbial modification in the Russian DDC. Under the analysis proposed here, this is so because the input verb does not merge as is. Rather, the (lexical) output, V-SJA, does, and its lexical ingredients are invisible to the syntax. The eventuality the input verb denotes is only entailed by V-SJA ((a) in (62)). As the input does not merge, its internal argument(s) cannot merge either because there is no corresponding VP. Nor can the internal argument(s) appear within the projection of V-SJA, as they are not its arguments. Hence, among transitive verbs only those allowing “object drop” are possible inputs. Similarly, the Agent of V is undetectable by Agent-oriented adverbs, as it is not present in the

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 273 m] cluster of V-SJA ((b) in (62)). Modifiersםcמ] syntax. Rather, it is a meaning ingredient of the m] argument. They can modify neither the input verb norםcמ] can only modify V-SJA and its its argument.30 m] role as itsם(cם)] The lexical formation of the DDC, then, takes a verb equipped with a m] role and a dative case feature. The operationםcמ] input (V), forming V-SJA equipped with a can be described as follows: (63) Formation of the DDC alternant in the lexicon -m], where V-SJA denotes a psychological state regardםcמ] m] N V-SJA DATם(cם)]V ing the actual or possible eventuality denoted by V. The DDC in (64a), for example, is derived as follows. In the lexicon, the operation produces the DDC alternant rabotat’sja (equipped with dative case) from the input rabotat’, negatively revalu- ,(m] cluster (Experiencerםcמ] m] cluster, thereby forming aםcם] ing the [c] feature of the input’s as shown in (64b). (64c) shows the lexical array to be merged in the syntax (abstracting away m] cluster and dative case, a dativeםcמ] from irrelevant details), namely, the DDC output with a noun phrase, and an adverb (at what stage of the derivation the phonological material is inserted is irrelevant for the present proposal). (64d) schematizes the syntactic structure of (64a). (64) a. Mne ploxo rabotaetsja. I.DAT badly work.PRES.3SG.SJA ‘I don’t feel like working.’ / ‘I can’t work (due to my psychological circumstances).’ / ‘I evaluate my working negatively.’ b. Lexicon: The DDC alternant formation 31 [mםcמ] m] N rabotat’sja ‘work.SJA’ DATםcם] ’rabotat’ ‘work

30 The DDC seems to allow an Instrument with a few verbs, most notably pisat’ ‘write’ (i), although not all speakers accept such sentences (see Gerritsen 1990). (i) Emu ne pi'etsja takim tupym karanda'om. he.DAT NEG write.PRES.3SG.SJA such.INST dull.INST pencil.INST ‘He doesn’t feel like writing with such a dull pencil.’ / ‘He feels that his writing is going badly with such a dull pencil.’ (Pariser 1982:135, (195)) The admissibility of an Instrument is argued to be a diagnostic detecting an (explicit or implicit) Agent (Reinhart 2002, Siloni 2002). If the DDC licensed an Instrument, that finding would contradict the results presented in section 4.1, namely, that the Agent is not present in the DDC. However, Gerritsen (1990) notes that in (i) the instrumental phrase can be understood as the reason for the psychological state described in the sentence, that is, ‘He doesn’t feel like writing / feels that his writing is going badly because of the dullness of the pencil’. Gerritsen interprets the instrumental phrase here as describing an external factor that evokes the disposition/evaluation. If this proposal is on the right track, the instrumental phrase is possible here because it relates to the psychological state and not to the entailed event of writing. This would be compatible with the conclusion that the Agent of the input is not present in the DDC (section 4.1). 31 As for the nature of the relation between the input and the output of the operation, two options come to mind: (a) the DDC alternants are lexically derived anew upon each retrieval (use) or (b) the DDC alternants are listed as lexical entries, the relation between them and their input being lexically specified. Dealing with this issue is beyond the scope of this article. See Siloni et al. 2018 for some discussion of the issue. Siloni et al. argue that unaccusatives and passives that are derived in the lexicon must be listed as lexical entries, in light of evidence from the domain of phrasal idioms.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 274 DGANIT JENIA KIM AND TAL SILONI m], mne DAT, ploxo͖םcמ] c. Lexical array: ͕rabotaetsja DAT 32 d. Syntax: [TP pro ...[VP mne ploxo [V rabotaetsja]]] 4.3.2 The Role of the Adverb in the DDC The new lexical entry formed by the operation, V- m] undergoes: a subjective evaluationםcמ] SJA, denotes a mental state that the Experiencer regarding the event denoted by V or a disposition toward it. The DDC describes whether the evaluation or disposition is negative or positive, but the information regarding the positive or negative nature of the mental state is not encoded in V-SJA itself. In contrast, genuine Experiencer verbs, such as worry, do encode a specific mental state inherently. Similarly, biclausal construc- tions, which are formed in the syntax by means of an overt (contentful) modal, also include information about the mental state. Since in the Russian DDC the information regarding the nature of the mental state has to be specified, the construction involves adverbial modification that provides it. In other words, the role of the adverb in the DDC is to modify the mental state that V-SJA denotes by specifying whether the experienced psychological state is positive or negative. Negation is interpreted as negative qualification, similar to ploxo ‘badly’, as shown by the results of questionnaire II (section 2.3.2).33 This explains why the Russian DDC requires modification. Recall that (a) bare DDCs are rare in corpora; (b) when they do occur, they are interpreted as involving positive modification, mostly supported by the context; and (c) when they are presented with no context (see section 2.2), informants judge them as marginal/unacceptable. This is so because without modification, it is unclear what kind of mental state the Experiencer experiences, and it is meaningless to talk about a mental state such as evaluation and disposition without specifying what they are. The V- SJA predicate, we propose, describes a feeling on the positive-to-negative scale regarding a possible or an actual eventuality. Adverbial modification specifies where the feeling is on this scale. The choice of adverbs found in the DDC is accordingly limited to adverbs that are able to specify the relevant information. The list in (65) presents the 20 most common adverbs appearing in DDCs with rabotaetsja ‘work.SJA’, as found in a tagged Web corpus on Sketch Engine (see footnote 5 for details about the corpus).

(65) Most common adverbs modifying DDCs with rabotaetsja ‘work.SJA’ legko, legcˇe ‘(more) easily’ komfortno ‘comfortably’

32 If indeed, as suggested in section 2.1 on the basis of examples (7)–(8), the dative Experiencer is not a sub- ject—namely, it does not raise to the subject position—this position may be occupied by an expletive pro. This raises the question, why does this dative Experiencer not move to the subject position? A possible explanation, along lines proposed by Fehrmann, Junghanns, and Lenertova´ (2010), could be that the Experiencer role in fact is assigned to a variable in the semantic representation, the dative being an adjunct coindexed with it. In this regard, it is relevant to mention that the dative Experiencer seems optional (see online appendix A, footnote 1). We leave further questions regarding the dative Experiencer for future research. 33 Pariser (1982) and Gerritsen (1990) also mention that negation in the DDC imparts a negative flavor to the experience, on a par with ploxo ‘badly’.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 275

tjazˇko, tjazˇelo ‘with difficulty’ veselo, veselee ‘(more) cheerfully’ xorosˇo, lucˇsˇe ‘well’, ‘better’ neprosto, nelegko ‘not easily’, ‘with challenges’ produktivno ‘productively’ normal’no ‘fine, all right’ slavno ‘well’ ploxo, xuzˇe ‘badly’, ‘worse’ plodotvorno ‘fruitfully’ prijatno, prijatnee ‘pleasantly’ zamecˇatel’no ‘remarkably’ otlicˇno ‘excellently’ zdorovo ‘excellently’ udobno ‘comfortably’ neploxo ‘fine, all right’ spokojno ‘calmly, easily’ prekrasno ‘wonderfully’ trudno ‘with difficulty’ These adverbs each represent a point on a positive-to-negative scale. Many of them (e.g., ‘well’, ‘badly’, ‘excellently’) are suitable to express any of the DDC meanings, that is, positive/negative disposition/capability, or positive/negative evaluation (either of the entailed activity (actual or possible) or of the mental state of the participant). Others are more specific to one of the mean- ings—for example, ‘productively’, which describes a positive subjective evaluation of the activity, or ‘cheerfully’, which describes a positive evaluation of the mental state. Agent-oriented adverbs such as vnimatel’no ‘attentively’ and ostoroÅno ‘carefully’ are disal- lowed in the DDC (66b), although they can modify the corresponding (non-DDC) verb (66a). This is so because they are inappropriate to anchor the state denoted by V-SJA on the positive- to-negative scale; nor can they modify the corresponding input verb referring to its Agent, as both do not project (section 4.1). (66) a. Ja ostoroÅno tancevala. I.NOM carefully dance.PAST.SG.FM ‘I danced carefully.’ b. *Mne ostoroÅno tancevalos’. I.DAT carefully dance.PAST.SG.NEU.SJA Adverbs such as legko ‘easily’, trudno/tjaÅelo ‘with difficulty’ are possible in the DDC (67b), although they cannot modify the corresponding (non-DDC) verb (67a). Such adverbs are possible in the DDC as they modify the psychological predicate V-SJA (67b). (67) a. *Ja trudno pisala. I.NOM with difficulty write.PAST.SG.FM

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 276 DGANIT JENIA KIM AND TAL SILONI

b. Mne trudno pisalos’. I.DAT with.difficulty write.PAST.SG.NEU.SJA ‘I experienced difficulty while writing.’ Finally, the adverbs that are found both in the DDC and with the corresponding (non-DDC) verb (e.g., xoro'o ‘well’, ploxo ‘badly’) are interpreted differently in the two cases. While (68a) necessarily states something regarding the objective quality of the work, (68b) can easily be uttered even when the quality of work is pretty bad (see Gerritsen 1990, Benedicto 1995). This shift in the meaning reflects the difference between modifying the verb that denotes the event of, say, working itself as in (68a) and modifying the corresponding V-SJA, which denotes a psychological state referring to this event (68b). (68) a. Ja xoro'o rabotala. I.NOM well work.PAST.SG.FM ‘I worked well.’ (i.e., the quality of the work was good) b. Mne xoro'o rabotalos’. I.DAT well work.PAST.SG.NEU.SJA ‘I felt like working.’ / ‘I felt psychologically able to work.’ / ‘I evaluated my working positively (either because I felt it was going well, because I enjoyed it, or both).’ The above facts are expected under the analysis proposed here. Only adverbs (including negation) that can specify whether the evaluation of the eventuality denoted by V or disposition toward this eventuality is positive or negative are admissible in the Russian DDC.

5 Conclusion In this article, we put forth a new analysis of the dative dispositional construction (DDC) in Russian. We revised the characterization of the input to the DDC, showing that the construction is fed by a subset of the set of unergative verbs. The common denominator for the verbs in this set cannot be formulated under the traditional view of ␪-roles as atomic primitives. Under the Theta System, a framework that views ␪-roles as clusters comprising two features positively or negatively valued or unvalued (Reinhart 2002, 2016, Everaert, Marelj, and Siloni 2012), the generalization is obvious. The operation applies to a natural class of ␪-clusters, that is, clusters consisting of only positively valued features. This lends additional support to the representation of ␪-roles as clusters of features. The revised characterization of the input has consequences regarding the component of grammar where the DDC alternant is formed. An operation sensitive to the internal composition of ␪-clusters is unlikely to apply in the syntax, as the internal makeup of ␪-clusters is illegible to the syntactic component. This conclusion converges with results of standard diagnostics failing to detect the presence of the base verb (and its argument) in the syntactic structure of the DDC. The conclusion that DDC alternant formation is lexical in Russian conforms to previous proposals that the Slavic family splits regarding the locus of formation of certain verbal alternations, with Russian exerting the lexical option. If this is correct, it provides additional support for the view of the lexicon as an active component of grammar where certain verbal instantiations are formed.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 277

We also reported the findings of two surveys we conducted to resolve controversies in the literature regarding the licensing conditions and interpretations of the Russian DDC. The first survey showed that the Russian DDC necessarily involves an adverb or negation; the bare DDC, which is available in other Slavic languages (e.g., Slovenian and Bulgarian), was judged as unacceptable/marginal. In corpora, the bare DDC is very rare, interpreted as involving positive modification, usually supported by context. The second survey revealed that the Russian DDC expresses two types of psychological state: either a subjective evaluation of an actual eventuality or a disposition toward the possibility of participating in a potential eventuality. Moreover, it showed that the two meanings are available for both the negated and the adverbial DDC. We argue that the Russian DDC alternant is formed by a lexical operation, whose input is cluster (role), and its output (a) denotes a psychological state regarding [ם] an unergative with a the (actual or possible) eventuality denoted by the input and (b) is equipped with an Experiencer role derived by feature adjustment of the input’s role. Adverbial modification (or negation) is required in the Russian DDC in order to anchor the psychological state denoted by the output on a positive-to-negative scale.

References Ackema, Peter, and Marijana Marelj. 2012. To have an empty theta role. In The Theta System: Argument structure at the interface, ed. by Martin Everaert, Marijana Marelj, and Tal Siloni, 227–250. Oxford: Oxford University Press. Babby, Leonard. 1980. Existential sentences and negation in Russian. Ann Arbor, MI: Karoma. Benedicto, Elena E. 1995. (Relativized) minimality, datives, and reflexive suffixes. In Proceedings of the 19th Annual Penn Linguistics Colloquium, article 3. Penn Working Papers in Linguistics 2(2). https:// repository.upenn.edu/pwpl/vol2/iss2/3. Borik, Olga. 1995. Sintakti?eskij priznak neakkuzativnosti glagola (na materiale russkogo jazyka) (The syntactic marker of verbal unaccusativity (on Russian data)). Master’s thesis, Moscow State Univer- sity. Chierchia, Gennaro, and Sally McConnell-Ginet. 1990. Meaning and grammar. Cambridge, MA: MIT Press. Everaert, Martin, Marijana Marelj, and Tal Siloni. 2012. The Theta System: An introduction. In The Theta System: Argument structure at the interface, ed. by Martin Everaert, Marijana Marelj, and Tal Siloni, 1–19. Oxford: Oxford University Press. Fehrmann, Dorothee, Uwe Junghanns, and Denisa Lenertova´. 2010. Two reflexive markers in Slavic. Russian Linguistics 34:203–238. Fici, Francesca. 2011. Ob odnoj modal’noj funkcii refleksivnyx konstrukcij (On one modal function of reflexive constructions). In Slovo i jazyk: Sbornik statej k vos’midesjatiletiju akademika Ju. D.

Apresjana (Word and language: Collection of essays for the 80th anniversary of J. D. Apresjan), 645–652. Moscow: C˝ ®. Franks, Steven. 1995. Parameters of Slavic morphosyntax. Oxford: Oxford University Press. Gerritsen, Nelleke. 1990. Russian reflexive verbs: In search of unity in diversity. Leiden: Brill. Greenberg, Gerald R., and Steven Franks. 1991. A parametric approach to dative subjects and the second dative in Slavic. Slavic and East European Journal 35:71–97. Horvath, Julia, and Tal Siloni. 2008. Active lexicon: Adjectival and verbal passives. In Current issues in generative Hebrew linguistics, ed. by Sharon Armon-Lotem, Gabi Danon, and Susan Rothstein, 105–134. Amsterdam: John Benjamins. Horvath, Julia, and Tal Siloni. 2011. Causatives across components. Natural Language and Linguistic Theory 29:657–704.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 278 DGANIT JENIA KIM AND TAL SILONI

Hron, David. 2012. Arity operations: Domain of application; a comparative study of Slavic languages. Doctoral dissertation, Tel Aviv University. Komar, Eric S. 1999. Dative subjects in Russian revisited: Are all datives created equal? In Formal Ap- proaches to Slavic Linguistics: The Seattle Meeting, ed. by Katarzyna Dziwirek, Herbert Coats, and Cynthia M. Vakareliyska, 245–264. Ann Arbor: Michigan Slavic Publications. Kuznetsova, Julia. 2005. Against the Russian distributive po-construction as a diagnostic for unaccusativity. In Formal Approaches to Slavic Linguistics: The South Carolina Meeting, ed. by Steven Franks, Frank Y. Gladney, and Mila Tasseva-Kurktchieva, 170–180. Ann Arbor: Michigan Slavic Publications. Levin, Beth, and Malka Rappaport Hovav. 1995. Unaccusativity: At the syntax–lexical semantics interface. Cambridge, MA: MIT Press. Marelj, Marijana. 2004. Middles and argument structure across languages. Doctoral dissertation, Utrecht University. Maru'i?, Franc L., and Rok äaucer. 2006. On the intensional FEEL-LIKE construction in Slovenian: A case of a phonologically null verb. Natural Language and Linguistic Theory 24:1093–1159. Meltzer-Asscher, Aya. 2011. Adjectival passives in Hebrew: Evidence for parallelism between the adjectival and verbal systems. Natural Language and Linguistic Theory 29:815–855. Meltzer-Asscher, Aya, Julia Schuchard, Dirk-Bart den Ouden, and Cynthia K. Thompson. 2013. The neural substrates of complex argument structure representations: Processing ‘alternating transitivity’ verbs. Language and Cognitive Processes 28:1154–1168. Moore, John, and David M. Perlmutter. 2000. What does it take to be a dative subject? Natural Language and Linguistic Theory 18:373–416. Pariser, Jon R. 1982. Dative-reflexive constructions in contemporary Russian. Doctoral dissertation, UCLA, Los Angeles, CA. Pesetsky, David. 1982. Paths and categories. Doctoral dissertation, MIT, Cambridge, MA. Pesetsky, David. 1995. Zero Syntax: Experiencers and cascades. Cambridge, MA: MIT Press. Potashnik, Joseph. 2012. Emission verbs. In The Theta System: Argument structure at the interface, ed. by Martin Everaert, Marijana Marelj, and Tal Siloni, 251–278. Oxford: Oxford University Press. Reinhart, Tanya. 2002. The Theta System: An overview. Theoretical Linguistics 28:229–290. Reinhart, Tanya. 2016. Concepts, syntax, and their interface: The Theta System, ed. by Martin Everaert, Marijana Marelj, and Eric Reuland. Cambridge, MA: MIT Press. Reinhart, Tanya, and Tal Siloni. 2005. The lexicon-syntax parameter: Reflexivization and other arity opera- tions. Linguistic Inquiry 36:389–436. Rivero, Maria Luisa, and Ana Arregui. 2012. Building involuntary states in Slavic. In Telicity, change, and state: A cross-categorial view of event structure, ed. by Violeta Demonte and Louise McNally, 300–332. Oxford: Oxford University Press. Rivero, Maria Luisa, and Milena Milojevic´-Sheppard. 2003. Indefinite reflexive clitics in Slavic: Polish and Slovenian. Natural Language and Linguistic Theory 21:89–155. Roeper, Thomas. 1987. Implicit arguments and the head-complement relation. Linguistic Inquiry 18: 267–310. RuÅi?kova, Eva. 1971. Pseudoreflexives in Slovak. In OSU working papers in linguistics 10, 205–211. https://kb.osu.edu/handle/1811/81807. Schoorlemmer, Maaike. 1993. Dative subjects in Russian. In Formal Approaches to Slavic Linguistics: The Ann Arbor Meeting, ed. by Jindrˇich Toman, 129–172. Ann Arbor: Michigan Slavic Publications. Schoorlemmer, Maaike. 2004. Syntactic unaccusativity in Russian. In The unaccusativity puzzle, ed. by Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert, 207–242. Oxford: Oxford Univer- sity Press. Shetreet, Einat, and Naama Friedmann. 2012. Stretched, jumped, and fell: An fMRI investigation of reflexive verbs and other intransitives. NeuroImage 60:1800–1806.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 THE DATIVE DISPOSITIONAL CONSTRUCTION IN RUSSIAN 279

Shetreet, Einat, Naama Friedmann, and Uri Hadar. 2009. The neural correlates of linguistic distinctions: Unaccusative and unergative verbs. Journal of Cognitive Neuroscience 22:2306–2315. Shetreet, Einat, Naama Friedmann, and Uri Hadar. 2010. Cortical representation of verbs with optional complements: The theoretical contribution of fMRI. Human Brain Mapping 31:770–785. Siloni, Tal. 2002. Active lexicon. Theoretical Linguistics 28:383–400. Siloni, Tal. 2012. Reciprocal verbs and symmetry. Natural Language and Linguistic Theory 30:261–320. Siloni, Tal, Julia Horvath, Hadar Klunver, and Ken Wexler. 2018. Idiom storage and the lexicon. Journal of Linguistics 54:189–215. Slobodchikoff, Tatyana. 2008. The argument structure of the dative desiderative in Slavic: Bosnian and Russian. In Proceedings of the Sixth International Conference Formal Approaches to South Slavic and Balkan Languages, ed. by Marko Tadic´, Mila Dimitrova-Vulchanova, and Svetla Koeva, 109– 114. https://www.academia.edu/3227266/TOWARDS_ENCODING_EVENT_STRUCTURE_IN_ WORDNET. Williams, Edwin. 1985. PRO and subject of NP. Natural Language and Linguistic Theory 3:297–315. Yip, Moira. 1998. Identity avoidance in phonology and morphology. In Morphology and its relation to phonology and syntax, ed. by Steven G. Lapointe, Diane K. Brentari, and Patrick M. Farrell, 216–246. Stanford, CA: CSLI Publications. Zeldowicz, Gennadij. 2011. On Russian dative reflexive constructions: Accidental or compositional? Studies in Polish Linguistics 6:153–171.

(Kim) The Computational Lexicology and Terminology Lab Vrije Universiteit Amsterdam De Boelelaan 1105 1081 HV Amsterdam The Netherlands [email protected]

(Siloni) Department of Linguistics Tel Aviv University 69978 Tel Aviv Israel [email protected]

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/ling_a_00341 by guest on 28 September 2021