(DOC) The evaluative dative reflexive

Evaluative reflexions: evaluative dative reflexive in South East Serbo-Croatian Boban Arsenijević, Universitat Pompeu Fabra, Barcelona b.arsenijevic<at>gmail.com 1. Introduction The aim of this paper is to present the relevant data and propose an analysis for the expression that I refer to as the Evaluative Dative Reflexive (EDR), as it is used in the South-Eastern dialects of Serbo-Croatian (SESC). This expression belongs to the class of non-selected datives, and is characterized by the semantic effects of a positive evaluation of the underlying eventuality, a low value on a certain scale (small quantity of the eventuality, its low informational relevance etc.), and some additional restrictions introduced in section 3. I argue for an analysis at the syntax-semantics interface, in which the crucial component of this dative is that it relates to the evaluative mood, specifying that the subject of the clause is also the subject of evaluation. I show how this analysis, which focuses on specifying the evaluative aspect of the semantic contribution of EDR derives other semantic effects as well as the syntactic properties of the expression. The paper is organized as follows. In section 2, I present the class of non-selected datives – dative constituents which are not specified as typical participants in the kind of eventualities denoted by the verb. Section 3 introduces data from SESC about a special type of non-selected datives – the Evaluative Dative Reflexive. Section 4 presents an analysis that I propose for this expression, and section 5 shows how the analysis copes with the properties presented in section 3. Section 6 concludes. 2. Non-selected datives Dative case, both cross-linguistically and in Serbo-Croatian (S-C), a dialect of which is in the center of interest of this paper, shows a tendency to appear on both selected and non-selected constituents. It is selected when the meaning of the verb (the type of eventualities it denotes) somehow introduces the presence of a participant that it expresses, and it is non-selected when no such relation with the verb can be observed. A typical use of the former type is to mark the recipient (0a) or the direction of some movement (0b), while a typical use of the latter is to mark the benefactive/malfactive meaning (0c, d). (0) a. Jovan daje knjigu Mariji. S-C J.Nom gives book.Acc M.Dat ‘Jovan gives/is giving a/the book to Marija.’ b. Jovan ide lekaru. J.Nom goes doctor.Dat ‘Jovan is going to (see) the doctor.’ c. Jovan je otvorio vrata Mariji. J.Nom Aux opened door.Acc M.Dat ‘Jovan opened the door for Marija.’ d. Jovan je Mariji postavio zamku. J.Nom Aux M.Dat set trap.Acc ‘J set a trap for Marija.’ 1 Other non-selected datives include the possessive dative (0a), the ethical dative (0b), the dative of interest (0c) and others (Horn 2008, Al Zahre & Boneh 2010). (0) a. Marija mu je videla sestru. S-C M.Nom he.DatCl Aux seen sister.Acc ‘Marija saw his sister.’ b. Kako si mi? how Aux.2Sg me.Dat lit. ‘How are you to/for me?’ c. Ja ti onda krenem na put. I you.DatCl then start_going on trip.Acc ‘And then I took a trip (an information for which you are interested).’ All these traditional classes of non-selected datives are closely related to each other, and often hard to differentiate. This is already obvious in (0b and c), where the ethical dative could be seen as a pragmatically colored version of the benefactive (the speaker presents the well-being of the hearer as something that she is concerned about, hence benefactive/malfactive, i.e. the speaker takes the information he is conveying to directly bear on the hearer, again as a kind of benefactive/malfactive role in this respect). Moreover, most examples of benefactive and malfactive datives involve the meaning of possession and vice versa, as illustrated in (0). (0) a. Jovan je Mariji slomio olovku. J.Nom Aux M.Dat broken pen.Acc ‘Jovan broke Marija’s pen (at her dammage).’ b. Marija je Jovanu namestila zglob. M.Nom Aux J.Dat set joint.Acc ‘Marija relocated Jovan’s joint (for his benefit).’ Yet another type of non-selected datives is the Evaluative Dative Reflexive (EDR), as I call it in this paper, also known as the Coreferential Dative (CD, Berman 1982, Al Zahre & Boneh 2010), Reflexive Dative (Borer 2005), Personal Dative (Horn 2008) etc. (for arguments that EDR belongs to non-selected datives, see especially Horn 2008). I decide to introduce a new term for the same element to an already long list, because I think that none of the existing terms points to an exclusive property of the phenomenon observed and hence is a potential source of confusion. It is coreferential and reflexive indeed, but indirect objects and benefactives can also be coreferential (with another nominal expression in the clause, or even just the subject) and reflexive; it involves a ‘personal attitude’, but this is not so straightforwardly read of the term Personal Dative and so on. As will be argued in this paper, its evaluative contribution is its core characteristic, and together with its reflexivity it isolates it from all other datives. While rare or totally absent in the standard S-C, EDR is highly frequent in some of its dialects, especially in the South-Eastern dialects of S-C (SESC). (From here on, unless otherwise specified, all examples are from this group of dialects, which behave uniformly in respect of the construction under discussion; most of the examples are from the dialect of the city of Niš, with around four hundred thousand inhabitants, which makes nearly half of the speakers of SESC.) (0) a. Jovan si sedi i gleda si film. J.Nom Refl.Dat sits andwatches Refl.Dat movie.Acc ‘Jovan’s sitting and watching a movie [+EDR effects].’1 1 To specify the additional meanning contributed by the EDR,, ahich is not contained in the English translation, I add ‘[+EDR effects] to each translation, rather than trying to paraphrase the effects. 2 b. Dao sam si još jedan ispit. given Aux.1Sg Refl.Dat more one exam ‘I took one more exam [+EDR effects].’ It is in most relevant respects similar to the EDR in most other languages in which this phenomenon exists, as it is described in, among others, Horn (2008), for English, and Al Zahre & Boneh (2010) for Syrian Arabic and Modern Hebrew. Throughout the paper, I consider EDR in different languages essentially the same phenomenon, but with certain variation. In some cases, the differences may as well be a matter of (in)precise observations or descriptions, the risk of which is great due to the sociolinguistic and pragmatic aspects playing a significant role in the use and interpretation of these expressions. Compared to the situation I other languages, SESC shows a greater variety of verbs (or verb classes) that appear in the EDR construction. This pattern of variation is not exceptional, as it resembles the patterns in other domains, such as that of the classes of verbs used as inaccusatives, as presented in Sorace (2000). While it is closely related to some of them, none of the traditional classes of non- selected datives may fully accommodate EDRs, as it both escapes certain properties of each of them, and has properties of its own which are not typical for any of these classes. The following section lists its typical properties, some of which are already described in the literature, and some of which are either observed for the first time, or characteristic only of the EDR as it is used in SESC. 3. Properties of EDR As Horn (2008) describes it, EDR does not bear on the truth conditions of the sentence it appears in. The sentence is fully grammatical and semantically sound without the EDR. Yet, with the reflexive, its meaning is somewhat different. The difference, i.e. the contribution of the reflexive, can be informally described through a number of components. Horn (2008:172) provides the following: (0) a. they always co-occur with a quantified (patient/theme) direct object. b. they cannot be separated from the verb that precedes and case-marks them. c. they are most frequent/natural with monosyllabic “down-home” type verbs (e.g. buy, get, build, shoot, get, catch, write, hire, cook). d. they lack any external (PP) pronominal counterpart. e. they can occur in positions where a true indirect object is ruled out and can co- occur with (rather than substituting for) overt dative/indirect object. f. they are weak pronouns (Cardinaletti & Starke 1996, 1999; Bresnan 2001) and cannot be stressed or conjoined. g. they have no full NP counterpart. h. there’s no consistent thematic role for these elements. i. most speakers have no absolute restriction against 3rd person pronominals but some exhibit a residual person-based asymmetry: 1st > 2nd > 3rd j. they are non-arguments coreferring with the subject. k. they do not combine well with negated verbs. In what follows, I give a description of a number of properties of EDRs in SESC. Some of the properties given here (e.g. (0b, d, f, g, i, j) are trivial or non-applicable in SESC because it uses a reflexive and not a pronoun. Moreover, as will be shown, EDRs in SESC do not share all the properties of their English counterpart described by Horn – for instance, they can occur with any verb (against (0a, c, k)). The list in (0) summarizes the properties of EDR in SESC presented in this section (as specified in the following 3 discussion, most of them have been observed for other languages where EDR is in use, by other authors). (0) a. they are always realized by a reflexive. b. the subject they are bound by has to involve some kind of intentionality. c. the eventuality in the respective clause is positively evaluated by the subject. d. they do not combine well with focal elements with an evaluative interpretation. e. information conveyed by the respective sentence implied to be of low relevance. f. the subject binding the EDR must be topical. g. the subject binding the EDR must be referential. f. they resist distribution over plural subjects. I first describe the semantic and pragmatic ones, and then those that are (also) related to the syntax of this expression. Unlike its English counterparts described by Horn, but on a par with its counterparts in most other languages that display this phenomenon, EDR in SESC is always realized as a reflexive, i.e. it is always anaphoric and bound by the clausal subject. Again on a par with its counterparts in other languages (Borer & Grodzinsky 1986), EDR in SESC cannot be stressed. And since pronominal elements that cannot take stress in SESC are realized as clitics – EDR in SESC s always a clitic. In other words, there is only one lexical item that realizes EDR in SESC: the dative form of the reflexive clitic: si. Al Zahre & Boneh’s (2010:9) observe that EDR comes with an interpretation “involving some intentionality attributed to the referent of the subject DP”. Even when the subject is the undergoer or some other participant, its involvement still assumes some kind of intentionality, awareness of the underlying eventuality.2 (0) a. Pera si zna odgovor. Pera Refl.Dat knows answer ‘Pera is having the answer ready [+EDR effects]’ not the more literal: ‘Pera knows the answer [+EDR effects]’ b. Grana (#si) plovi rekom. branch Refl.Dat sail river.Inst ‘The branch is floating down the river [+EDR effects].’ [this sentence improves if the branch is personified] Horn (2008:181) talks about the following interpretive effect of the use of EDRs: “the speaker assumes that the action expressed has or would have a positive effect on the subject, typically satisfying the subject’s perceived intention or goals”. I argue that this is actually an evaluative effect: the subject POSITIVELY EVALUATES the eventuality, or more precisely the meaning of the entire PolP (the eventuality located with respect to the reference time and assigned a polarity). Information with a negative bias is pragmatically degraded, often infelicitous with the EDR.3 2 Borer & Grodzinsky (1986) argue that EDRs may only be used with a verb that has an external argument, and Al Zahre & Boneh (2010) show that this is actually not correct. It may be that Borer and Grodzinsky were mislead exactly by the component of control as described here, which is similar with the control typical of certain external arguments, but is rather at the epistemic level than at the level of event- participants. To illustrate the difference, let me note that, as Al Zahre & Boneh show, in Modern Hebrew, even stative and non-verbal predicates, including the individual level ones, can be used with an EDR, as long as the subject holds some rather vague intentional control over them (though it is clearly not a controller in the sense of agency). The same holds in SESC. 3 As pointed out by an anonymous reviewer, the evaluational restriction is another property that can be observed on other types of expressions as well. For instance, epithets tend to carry a negative evaluation, even though positively evaluated uses are not fully excluded. The source of the restriction is probably pragmatic in both cases, even though it is linked to different syntactic and semantic phenomena. 4 (0) a. Upao sam (#si) u rupu. fallen Aux.1Sg Refl.Dat in hole ‘I fell into a hole.’ b. Pera (#si) slomio nogu. P.Nom Refl.Dat broken leg ‘Pera had his leg broken.’ Both these examples are fine in contexts in which their subjects consider it good for themselves to fall into a hole and broke a leg, respectively. Nevertheless, I provide arguments that this is rather a tendency, with a significant number of exceptions. Not all negatively biased examples sound so bad with an EDR (see section 5 for examples and a discussion). To further illustrate the evaluative effect, observe the examples in (0). When the sentences are pronounced with a strong stress on mnogo ‘much’, they get an interpretation of a strong subjective evaluation of a high degree to which the newspaper is judged as interesting. Intensifiers of this type do not combine with EDRs. The reason is that the EDR requires an evaluative interpretation, but the evaluative interpretation is already reserved for the focal intensifier.4 (0) a. Čitam (#si) MNOGO interesantne novine! read Refl.Dat much interesting newspaper ‘I’m reading SUCH an interesting newspaper.’ b. Pera (#si) video MNOGO dobar sat! P Refl.Dat saw much good watch ‘Pera saw SUCH a good watch.’ [the sentence improves with a possessive/benefactive reading of the reflxive] There is one more interpretive effect of EDR that involves attitude and is common for its use in all languages in which it is found. The speaker uttering the sentence containing the EDR considers the information conveyed by it to bear LITTLE RELEVANCE for the collocutors. In SESC, sentences containing EDR have a kind of marginal, parenthetical intonation. They cannot be exclamative, and cannot involve a strong focal stress on the clausal polarity or one of the arguments. (0) a. Dobio sam si sat. /no strong focal stress/ gotten Aux1Sg Refl.Dat watch ‘I got a watch [+EDR effects]’ b. #DOBIO SAM SI SAT! /strong focal stress on the sentence/ gotten Aux1Sg Refl.Dat watch c. #Dobio sam si SAT. /contrastive focal stress/ gotten Aux1Sg Refl.Dat watch In other words, a sentence involving EDR cannot be felicitously used in a context where it brings in the information that the hearer is wondering about. Its content is always somehow digressive – beside the main line of exchange of information. As it is the speaker’s attitude that the information is not relevant for the hearer, the hearer may well show interest for the information conveyed, and this is exactly what usually happens, and what probably is one prominent purpose of EDR – to smuggle into the discourse information that might be pragmatically (socially) inadequate, by attributing it a lower 4 More precisely, the problem is that the EDR takes one subject of evaluation (theclausal subject) and the evaluative focal element another one (the speaker), which yields a conflict – see sections 4 and 5 for an explanation. 5 degree of relevance (usually simply because the speaker may be judged egocentric for introducing such information without a ground in the current set of topical issues). Al Zahre & Boneh (2010) report that in Syrian Arabic, EDRs can only be used if the eventuality involved is modified to have a SMALL QUANTITY (by adverbs and quantifiers such as šway ‘little’, kam ‘several’, or numerals marking a relatively small quantity). Also in other languages that display instances of EDR, there is a similar tendency, though not necessarily with an overt marking. This is probably just a consequence of the marginal informational relevance of the contents of sentences involving EDRs. Observe the example in (0), where it is obvious that the quantification introduces only pragmatic information, without real quantity entailments. (0) A: Šta si radeo juče? what Aux2Sg done yesterday ‘What did you do yesterday?’ B: Malo sam si jeo palačinke. little Aux1Sg Refl.Dat eaten pancakes ‘I ate pancakes [+EDR effects]’ A: Kolko si pojeo? how_muchAux2Sg eaten_up ‘How much/many did you have?’ B: Uuu, mnogo, najmanje 20. many, fewest 20 ‘Huh, a lot, at least 20.’ Finally, as an illustration of the pragmatic nature of this type of quantification, observe a typical exchange in the beginning of a colloquial conversation in SESC in (0). (0) A: Šta se radi? what Refl.Acc does ‘What’s up?’/’What’s going on?’/’What are you doing?’ B: Evo malo. 5 Prtl.Deic little lit. “Here, a little.” – semantically empty, pragmatic content only The response does not say what the person is doing, it only says that it is a little of whatever it is. There seems to be a pragmatic convention, the mechanisms of which should be looked for in sociolinguistics, that whatever one is saying about oneself, it is by default of little relevance in the ongoing discourse, and one of the ways to express this is by a low degree quantification. This property interacts with the level of aspect. In SESC (and generally in S-C), only sentences involving imperfective verb forms can include an explicit quantification with an adverb such as malo ‘little’. With perfective verbs, this yields unacceptability. (0) a. Malo sam gledao TV. little Aux1Sg watched TV ‘I watched the TV for a little while.’ b. */# Malo sam pročitao novine. little Aux1Sg read_out newspaper ~‘I read the newspaper for a little while.’ 5 Evo is a deictic particle, similar in meaning to the italian ecco. It refers to something that the speaker is showing or presenting, and can be used ostensively, but also in a more abstract way, to refer to some discourse-prominent object, event, temporal interval or spatial location. 6 However, when the sentence involves an EDR (or even when it does not, but the respective reading of malo ‘little’ is suggested by the context or by the intonation), THE USE OF MALO ‘LITTLE’ no more depends on the aspectual properties of the verb. In such cases, malo ‘little’ clearly specifies the degree of relevance of the information conveyed. (0) a. Malo sam si gledao TV. little Aux1Sg refl.Dat watched TV ‘I watched the TV a little [+EDR effects].’ b. Malo sam si pročitao novine. little Aux1Sg Refl.Dat read_out newspaper ‘I read the newspaper a little [+EDR effects].’ These facts can be interpreted in two ways. One would be to conclude that EDR has nothing to do with aspect, as the quantification that is related to its meaning is at the level of informational relevance rather than at the level of aspect. The other is to conclude that the use of EDR shifts the aspectual status of the verb to imperfective, thus making it compatible with the adverbs of the type of malo ‘little’. Again, only a precise description of the aspectual restrictions imposed by EDRs could lead to an answer to this question, and as I propose an analysis that is orthogonal to the issues of aspect (although certain interactions are obvious), I do not devote a more elaborate discussion to this problem. Sentences involving an EDR come with restrictions on their information structure. The SUBJECT MUST BE TOPICAL This is already illustrated in (0), where the subject is a first person pro which must be topical. Topic on other constituents yields unacceptability, as illustrated in (0) and (0).6 (0) A: Znaš onaj sati što smo videli? know.2Sg that watch Comp Aux1Pl seen ‘You know that watch that we saw?’ B: Znam. Pera (#si) gai kupio!7 know.1Sg Pera Refl.Dat it.AccCl bought ‘I know. Pera bought it! (0) a. On si pije kafu. /(contrastive) topic on the subject/ he Refl.Dat drinks coffee ‘He is drinking coffee [+EDR effects].’ b. pro Pije si kafu. /topic on the subject, which is dropped/ pro drinks Refl.Dat coffee. ‘He is drinking coffee [+EDR effects].’ c. Kafu (#si) pije (on). /topical object/ coffee Refl.Dat drinks he ‘Coffee, he drinks’ Al Zahre & Boneh find this requirement for a topical subject so striking that they “speculate that the CD [their term for EDR] is actually related to a topic in an A'- position, itself associated with a thematic-argument.” (pg. 17). 6 The type of topic in all these cases is the aboutness topic, which preferrably coincides with the familiarity topic, in terms of Reinhart (1981). 7 In this paper, I assume the EDR use for each dative reflexive used in the examples. Sometimes another interpretation is available and yields different acceptability judgments (e.g. in this case, the example improves with a recipient interpretation of the dative reflexive clitic), but these other readings are ignored. 7 As observed by Al Zahre & Boneh, the SUBJECT of a clause involving an EDR must BE REFERENTIAL, i.e. cannot be arbitrary or generic. This is confirmed in SESC, as shown in literal translations of their Syrian Arabic and Modern Hebrew examples. The sentence in (0), on the intended interpretation with an arbitrary subject, cannot combine with an EDR. The EDR is fine if pro is interpreted as referential. (0) proARB popravili (#si) klimu. EDRs seem to RESIST DISTRIBUTION OVER SUBJECTS and probably even DISTRIBUTION IN GENERAL. They involve an effect of singularity: whatever interpretation is derived for the eventuality and its participants, the EDR subjects it to one evaluation. More precisely, they never combine with distributed or iterative referential eventualities. The example in (0a) has only the interpretation in which that the speaker each time answered correctly displays the positive evaluation, lower informational relevance, and other EDR effects. The reading in which for every relevant time, that the speaker answered correctly displayed the positive evaluation, lower informational relevance, and other EDR effects – is not available. Similar holds of the example in (0b), where only the collective interpretation is available for the conjoined constituents (as a group, they are solving crosswords). The distributive interpretation, where each is solving crosswords, and each of these facts is evaluated as positive and marginally relevant – is out. In section 4, I speculate that this is a consequence of the tight relation between the evaluative effects and the discourse update. (0) a. Ja sam si svaki put tačno odgovorio. I Aux1Sg Refl.Dat each time correct answered ‘[EDR effects](I answered correctly each time.)’ *‘Each time, [EDR effects](I answered correctly).’ b. Pera i Mika si rešavaju ukrštene reči. P andM Refl.Dat solve crossed words ‘[EDR effects](Pera and Mika are solving crosswords.)’ *‘[EDR effects](Pera is solving crosswords) and [EDR effects](Mika is solving crosswords)’ While other authors do not make similar observations on the use of EDR in other languages, there are some observations which I consider related to this one. In particular, Borer & Grodzinsky (1986) note that EDRs (Reflexive Datives in their terminology) cannot be coordinated, Al Zahre & Boneh show that they cannot be associated with a group of coordinated verbs, and Borer (2005) argues that they only occur with atelic interpretations (but see Al Zahre & Boneh, who show that telicity does not play a role in the use of EDRs in Modern Hebrew). While the ban on coordination of EDRs cannot be tested in SESC, where EDR is a clitic and clitics generally do not coordinate, its ban in languages in which otherwise it would be expected, as well as the ban on coordination of eventualities, needs an explanation. As for the aspectual facts, the empirical situation is still quite unclear (see Al Zahre & Boneh), but it would not be a surprise if it eventually turned out that they also derive from the singularity restriction on the evaluation time. Horn observes that in English, EDRs do not go well with NEGATION. Although certain examples are attested, they are all licensed by what Horn refers to as a syntagmatic priming (the expression involving an EDR is literally repeated in the negated sentence, from a sentence from the preceding discourse). Horn’s evidence is based on the quantitative data about antonym verbs as well as examples involving explicit negation, rather than on the unacceptability of negated examples, and I refer the reader to Horn’s paper for the precise data. 8 In SESC, EDRs are not (so) sensitive to negation, as long as it does not affect the positive evaluation of the sentential content by the subject. Examples of the type in (0) are well-formed, although some speakers find some of them slightly degraded. (0) a. Ja si danas ne odo na poso. I Refl.Dat today not went on work ‘I didn’t go to work today [+EDR effects].’ b. Pera si ni ogrebotinu ne dobi. P refl.Dat n_even scratch not got ‘Pera wasn’t even scratched [+EDR effects].’ 4. The analysis Previous accounts of the EDR were mostly based on the data from Modern Hebrew (Berman 1982, Borer & Grodzinsky 1986), with the exception of Al Zahre & Boneh (2010), who compare the Modern Hebrew data with those from Syrian Arabic, and Horn (2008), who uses the data from English. These analyses form two groups, one which tries to explain the effects of EDR in terms of conventional implicature (Horn, and to some extent Al Zahre & Boneh), and one which links the EDR to the aspectual and argument structure (all other accounts). An exception is the work of Al Zahre and Boneh, the aim of which is not to offer an analysis, but rather to systematize the data, adding important new observations and generalizations, and correcting some old ones which turned out to be wrong. They show that at least the accounts focusing on the aspectual and argument structure are based on incomplete data, and actually fail to explain the facts. Finally, the analysis in this paper is similar in spirit to the analysis of Boneh & Nash (2010), who target French, and who also argue that the dative element brings in a sense of affectedness, and that its attachment site is crucial for its interpretation. From the information available about their analysis, main differences are that they assume the subject in EDR constructions to be the agent, which is not necessarily the case in SESC, and their analysis generates EDRs no higher than TP, while in the present analysis it goes to MoodP, or even ForceP; as will be presented soon, the present analysis also goes into more detail about the mechanics deriving the effects observed. For reasons of space, I do not discuss the earlier accounts in more detail, but refer the reader instead to Al Zahre & Boneh who provide a detailed discussion. In deriving the analysis of EDR in SESC, and probably also more universally, I depart from the analysis proposed in Boneh and Nash (2010). Boneh and Nash argue convincingly that EDR (CDC in their terminology) is generated somewhere at the level of TP, i.e. higher than vP and lower than CP. In this section, I first offer some additional arguments for the same claim, and then propose a more precise location for EDRs: the projection of the evaluative mood (Cinque 1999’s MoodevaluativeP). In section 5 I show how the particular syntactic and semantic properties of EDR presented in sections 3 and 4 follow from generating EDR in this position. As shown in section 3, EDR has several pragmatic effects, among which the marking of a low informational relevance. This effect is often employed to serve certain social relations between the speaker and the hearer, such as the introduction of oneself as a topic. In this way, EDR is similar to the interested speaker dative (which is often classified together with the ethical dative), expressed in S-C by a first person clitic in dative case and to the interested hearer dative, and to the interested hearer dative (IHD), expressed by a second person dative clitic. IHD thus refers to the hearer, marking that the information conveyed, or the process of conveying it, is of a high relevance for the 9 hearer, or even takes place for her benefit. The entire narration that (0) is part of is somehow dedicated to the hearer. (0) Pera ti onda dohvati jabuku i baci je kroz prozor. P you.Dat.Cl then catches apple andthrows it through window ‘[FYI] Pera then catches the apple and throws it through the window.’ In SESC, in addition to the hearer-oriented relevance, it also marks a special social relation between the speaker and the hearer – they are in a close, almost intimate relation (a typical gesture that comes with the IHD is putting an arm on the hearer’s arm or shoulder). Note, in illustration, that IHD cannot take the form of a polite second person. (0) #Pera Vam dohvati jabuku i baci je kroz prozor. P you_polite.Dat.Cl catches apple andthrows it through window ‘[FYI] Pera catches the apple and throws it through the window.’ Next to the relevance and socially related effects, EDR and IHD are also similar in being restricted to the expression by only one lexical item. EDR is always lexically realized as a dative reflexive, and IHD as a dative second person clitic. Strong pronouns, lexical nominals or any other items are excluded. Assuming that IHD is expressing a set of features in the projection responsible for the performative force of the clause (ForceP), marking thus that the force of the clause goes in the hearer’s direction, we may speculate that EDR expresses a similar type of features, but specifying the direction of the clausal subject. This parallel explains (a) the restriction to clitics: clitics are light elements without lexical content, realizing bundles of functional features, and that is the reason why they have the capacity to realize not only arguments, but also contents of higher functional heads, such as Force and (b) the restriction on the use of EDRs that the subject of the clause must be topical: only a topical subject may bind a reflexive generated in a position as high as ForceP. In semantic terms, while IHD marks that the force of the clause is directed towards the hearer, and hence of a special relevance for her, the orientation towards the subject, especially when the subject is in first person, marks that the information is relevant to the speaker, hence at least by implicature not particularly relevant for the hearer. There are, however, also certain facts that indicate that these two elements are neither syntactically nor semantically identical. First, IHDs and EDRs may occur in the same sentence. (0) Mika ti si ustane i ode. M you.DatCl Refl.Dat gets_up andleaves ‘[FYI] Mika gets up and leaves [+EDR effect].’ If they were generated in the same position, and contributing semantics of the same type – they should never co-occur in a sentence, unless coordinated. Second, while EDR frequently appears in embedded clauses, IHD never embeds. (0) Mika kaže da (*ti) (si) on lepo ustane i ode. M says comp you.DatCl Refl.Dat he nice gets_up andleaves ‘Mika says that [*FYI] he simply gets up and leaves [+EDR effect].’ Since embedded clauses lack a fully specified force, this minimal pair suggests that IHD is indeed force-related, but that EDR does not sit in ForceP, but rather in some lower functional projection, as indicated by examples (0), and (0) which shows that the IHD cannot surface after the EDR. 10 (0) Mika (ti) si (*ti) ustane i ode. M you.DatCl Refl.Dat you.DatCl gets_up andleaves ‘[FYI] Mika gets up and leaves [+EDR effect].’ The lower bound can be determined by another type of non-selected dative – the benefactive. Let us assume with Pylkkanen (2002) that benefactives are generated somewhere at the level of vP, and observe (0). (0) Ček da si mu otvorim vrata. wait Comp Refl.Dat he.DatCl open door ‘Wait till I open the door for him [+ EDR effect].’ Finally, one sentence can have all three types of datives above, but necessarily in the strict order: IHD<EDR<Benefactive. (0) Ja ti si mu otvorim vrata... I you.Dat.Cl Refl.Dat he.DatCl open door ‘[FYI] I open the door for him [+ EDR effect].’ This tells us that indeed, as argued by boneh and Nash (2010), EDR is base generated somewhere between CP (more precisely, ForceP) and vP, i.e. within the IP layer (which I assume, with Rizzi 1997, Cinque 1999 and a lot of subsequent work, to be constituted by the specification of the components such as polarity, mood, tense and outer aspect, i.e. different types of MoodP, PolP, TP, AspP and perhaps other related projections). The asymmetries are not really unexpected, considering that the description of EDR involved specifying a number of different properties, some of which have nothing to do with force, while IHD seems to be restricted to the force-related effects. Moreover, the possibility to combine IHD and EDR in the same sentence, without a conflict, indicates that the relevance-related effect of one of them at the very least has to be cancellable, as else in some cases the same sentence would be specified for contributing information that is at the same time relevant and irrelevant for the hearer. I propose, based on the semantics of EDRs, that, in a finer analysis of the IP layer, the right projection to generate them is Mood0evaluative (which specifies the evaluative aspect of the mood, see Cinque 1999, and especially Liu 2007 for an implementation). The dative reflexive in this head introduces a variable that is bound by a c-commanding nominative subject, and specifies the subject of evaluation (i.e. the one who evaluates the semantic contents of the structure in the complement). The direct interpretive effect is that the predicate of the eventuality is evaluated by the clausal subject. Other evaluators determined overtly or in the discourse are allowed, as long as the subject is also involved, and as long as the evaluation by the subject is exempted from any more general evaluation. We arrive at the analysis in (0a), illustrated on an example in (0b). (0) a. [Subji [ForceP [IHD] ... [MoodPeval [EDR]i [vP [v] [ApplP [Benefactive]]]]] b. Ja ti si mu otvorim vrata... I you.Dat.Cl Refl.Dat he.DatCl open door ‘[FYI] I open the door for him [+ EDR effect].’ [TopicP Ja [ForceP [ti] ... [MoodP [si] [PolP ... [vP Ja [v] [ApplP [mu] [otvorim vrata]]]]]]] This, I argue, is the semantic entailment of the EDR: The eventuality is evaluated, and it is evaluated by the clausal subject. Other semantic and pragmatic effects, in particular: that this evaluation is positive, the lower degree of relevance of the information, the small quantity of the eventuality, the intentionality of the subject, and the ban on distributive interpretations all derive from this core property. 11 5. Explaining (and reexamining) the properties of EDR The positive interpretation of the eventuality is only an implicature of specifying that the subject is also a subject of evaluation, and of giving the subject’s evaluation a special status in this way. In over 70% of examples excerpted using Google (of nearly 300 examples), the subject is also the agent, and the controller of the evaluated eventuality. It is natural that if the evaluator holds control over the evaluated eventuality, she evaluates the eventuality as positive for herself (else she would have stopped it, or controledly push it in another direction). Sentences with EDRs involving other types of subjects inherit this generalized positive evaluation interpretation from those which make a majority (note that the asymmetry in frequency is so strong that Borer & Grodzinsky 1986 had made a generalization that only eventualities with an external argument license the use of EDR; Al Zahre & Boneh 2010 showed that this generalization was wrong). In any case, the positive evaluation of the eventuality is not an absolute condition for the use of EDRs, as shown in the examples in (0), where those in (0b-f are found via Google), and (0a) is from a live conversation. (0) a. Pao sam si ispit, sa’ ću vi’m šta ću radim. fallen Aux1Sg Refl.Dat exam now I’ll see what to do ‘I failed the exam, now I’ll see what to do [+EDR effects].’ b. Nego, nema veze, teb si ide to na savest... but n_have conection you.Dat Refl.Dat goes that on conscience ‘But, never mind, it goes to your conscience [+EDR effects]...’ c. Kome se ne svidja nek si ide!! who.Dat Refl.Acc not like let Refl.Dat go.3Sg ‘Those who don’t like it may leave [+EDR effects].’ d. Iz ovaj li moj dom da si idem? from this Q_Part mine home Comp Refl.Dat go.1Sg ‘Is it this home of mine that I have to leave? [+EDR effects]’ e. ipak džabe Raka uči, on si ide kući... nevertheless in_vain R learns he Refl.Dat goes home ‘Nevertheless, Raka’s learning is in vain, he’s going home [+EDR effects]...’ f. Nek si ide život... let Refl.Dat goes life lit. ‘Let go the life [+EDR effects]...’ ‘Let my life be sacrificed (for a contextually given reward) [+EDR effects]’ (used when a great sacrifice has to be made for a rather hedonist reason) In fact, almost one third of the examples excerpted from the internet can be judged more or less negative for the subject. In all these cases, however, there is a sense of the negative facts being a) evaluated and b) ‘none of the business of anyone else than the subject’. This is actually a rather general effect that comes with EDRs, and which is also observed (though not as a general one, but as one of the pragmatic options) by Al Zahre & Boneh (2010: 25) as an ‘isolation effect’. This effect naturally follows from the present analysis, more precisely from the specification of the subject as a (singled out, hence isolated) subject of evaluation I argue that exactly this effect of isolation of the clausal subject as a subject of evaluation is what triggers the effect of a lower informational relevance of the sentence containing an EDR. As the subject is taken aside, as an evaluator, from the other relevant subjects of evaluation (most importantly the interlocutors as the default 12 evaluators), the information conveyed receives a special status in the discourse. It carries an evaluation which is not (necessarily) shared by the speaker and/or the hearer, and therefore less directly updates the discourse (i.e. does not update it with an evaluation of the facts, but with an information about the subject’s evaluation of the facts). Moreover, the MoodPevaluative projection is closely related to the projection introducing evidentiality specification, MoodPevidential (see Liu 2007 for empirical evidence). Unless the evidential status of the sentence is explicitly specified, this brings about an effect of attributing to the subject also the role of the source of information. In itself, this gives a weak lower relevance effect, which can be made stronger when the pragmatic conditions are fulfilled, i.e. when the speaker intends to make a stronger effect of this kind. One way to overtly mark this status is the use of adverbs and quantifiers introducing the values from the lower sections of a scale. This is illustrated in (0), where the sentence with the adverbial malo ‘little’ even sounds more natural than the one without it, although both are fully acceptable. (0) a. Malo sam si jeo palačinke. little Aux1Sg Refl.Dat eaten pancakes ‘I ate pancakes a little [+EDR effects].’ b. Jeo sam si palačinke. eaten Aux1Sg Refl.Dat pancakes ‘I ate pancakes [+EDR effects].’ The sentence in (0a) is more appropriate when it is a digression, or introduces a new topic into the discourse (which gives it a stronger low relevance effect). The one in (0b) is more natural as an answer to the question ‘What were you doing (at the relevant time)?’ – i.e. in a context in which it has some informational relevance (answers a question), which it degrades by its own implicature of low relevance. In languages like Syrian Arabic, where EDR necessarily requires the presence of an adverb or quantifier of this kind (Al Zahre & Boneh 2010), EDR has probably become part of a construction specialized to express meanings with a strong low relevance effect. The component of intentionality of the subject that binds the EDR (Al Zahre & Boneh 2010:9) directly follows from the specification of the subject as an (isolated) evaluator: trivially, intentionality is a necessary property of any possible subject of evaluation – it is a prerequisite of evaluation. The very fact that the subject binds a variable that represents the evaluator forces an interpretation of the subject which attributes it intentionality, even when the actual referent normally does not display intentionality, as illustrated by the example in (0b), repeated here as (0). (0) Grana (#si) plovi rekom. branch Refl.Dat sail river.Inst ‘The branch is floating down the river (#[+EDR effects]).’ [this sentence improves if the branch is personified] EDR effects do not distribute over subjects or times, as already discussed in section 3 in relation to examples in (0), repeated here as (0). Several other restrictions, such as the ban on coordination both of EDRs and of the respective eventualities (Borer and Grodzinsky 1986), as well as the referentiality restrictions described in Al Zahre & Boneh (2010) subsume under this one. I speculate that the discourse status of evaluative attitudes is responsible for these effects. More precisely, (the discourse-representation of) the set of beliefs of the subject has to be updated by the evaluation introduced by EDR, and this effect cannot distribute (similar to the fact that force does not distribute). 13 (0) a. Ja sam si svaki put tačno odgovorio. I Aux1Sg Refl.Dat each time correct answered ‘I answered correctly each time [+EDR effects].’ b. Pera i Mika si rešavaju ukrštene reči. P andM Relf.Dat solve crossed words ‘Pera and Mika are solving crosswords [+EDR effects].’8 This is also part of the explanation of a final property of EDRs discussed in this section, namely that the subject that binds the EDR has to be topical. The direct discourse update of clauses involving an EDR (e.g. in the sense of file-card semantics of heim 1982) are in the domain of the beliefs of the subject. If any other event participant were topical, that should also be the referent directly targeted by the discourse update, and the semantic contribution of EDR could not be realized. The restrictions of EDRs being realized by a reflexive is not an effect, but rather a core property of EDRs. Hence, rather than explaining why it holds, one could speculate on the ways in which it lead to the independent establishing of this element in mutually unrelated languages. The reasons are rather pragmatic than a matter of grammar. As Horn (2008) indicates, there was probably a fortunate match between a niche in the set of frequent pragmatic patterns and the entailments of the particular syntactic structure with a referent with the evaluative capacity being both the evaluator and the controller of an event. This guaranteed a high frequency of the configuration, leading to its establishing as a construction, with the set of restrictions less directly related to its entailment varying across languages. There is also a syntactic condition that is fulfilled by the topical interpretation of the subject. In order to bind the dative reflexive, the subject needs to be higher than MoodPevaluative. Assuming that the subject normally moves only as high as TP, which is lower than MoodPevaluative (Cinque 1999), the subject must move higher up to get into a c-command relation with the EDR. This is achieved by its movement to TopicP. 6. Conclusion In this paper, I presented the data about the Evaluative Dative Reflexive as it is used in South-Eastern Serbo-Croatian. It shares most of its properties with its counterparts from Modern Hebrew, Syrian Arabic, French or dialects of English, but also differs from them in some interesting ways. I proposed an analysis according to which the semantic contribution of EDR is in specifying that the subject of the clause is a subject of evaluation of the underlying eventuality. Syntactically, it is generated in MoodPevaluative, where it realizes its semantic contribution, and is bound by the subject which moves up to TopicP. I showed how this analysis accounts for other presented properties of the EDR: the positive nature of the evaluation, the intentionality and the topical and collective interpretation of the subject, the low relevance of the information conveyed and the reflexive nature of the EDR. The analysis should also apply to EDRs in other languages – possibly with small modifications in the domains of variation (e.g. the obligatory use of a small quantity adverb in Palestinian Arabic as in Al Zahre & Boneh 2010), hopefully motivated by the different setting of syntactic parameters in these languages. 8 Note in further support of the weak low informational relevance effects, that this sentence cannot combine with adverbs such as malo ‘little’, and only in a very remote sense displays the low relevance effect. 14 Ackowledgments I am indebted to the editors of this volume, and the organizers of the workshop at which it was initiated for providing a great venue for the research on the aspects of variation in the use of datives, and to two anonymous reviewers for comments which have helped me to improve the paper considerably. Funding from the following grants is gratefully acknowledged: ‘The Origins of Truth’ (NWO 360-20-150), ‘Natural language ontology and the semantic representation of abstract objects’ (MICINN, FFI2010-15006 and JCI- 2008-2699). References: Al-Zahre, Nisrine & Nora Boneh. 2010. “Coreferential Dative Constructions in Syrian Arabic and Modern Hebrew.” MS. Université Paris VIII/ The Hebrew University of Jerusalem. Arsenijević, Boban. 2009. “{Relative {Conditional {Correlative clauses}}}.” In Rajesh Bhatt & Aniko Liptak (Eds.), Correlatives crosslinguistically, 131-156. John Benjamins. Berman, Ruth. 1982. “Dative marking of the affectee role: data from Modern Hebrew.“ Hebrew Annual Review 6, pp. 35-59. Boneh, Nora & Lea Nash. 2010. “Getting high.” Paper presented at the Colloquium on Generative Grammar, Barcelona, March 18-20th 2010. Borer, Hagit. 2005. Structuring Sense. Oxford University Press. Borer, Hagit. & Yosef Grodzinsky. 1986. “Syntactic Cliticization and Lexical Cliticization: The Case of Hebrew Dative Clitics.” In Hagit Borer (ed.) Syntax and Semantics 19, 175-215. New York: Academic Press. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford: OUP. Heim, Irene. 1982. The semantics of definite and indefinite Noun Phrases. Doctoral dissertation, University of Massachusetts, Amherst, Mass. Horn, Laurence R. 2008. “’I love me some him’: The landscape of non-argument datives.” In Bonami, O. & P. Cabredo-Hofherr (eds.), Empirical Issues in Syntax and Semantics 7, pp. 169-192. Liu, Chen-Sheng Luther. 2007. “The V-qilai Evaluative Construction in Chinese”, UST Working Papers in Linguistics 3: 43-62. Pylkkänen, Liina. 2002. Introducing arguments. Doctoral dissertation, Massachusetts Institute of Technology, Cambridge, Mass. Reinhart, Tanya. 1981. Pragmatics and linguistics. An analysis of sentence topics. Philosophica 27: 53–94. Rizzi, Luigi. 1997. “The fine structure of the left periphery’, in Lilliane Haegeman (ed.), Elements of Grammar, Kluwer, Dordrecht, 281–337. 15 Sorace, Antonella. 2000. “Gradients in auxiliary selection with intransitive verbs.” Language 76 (4), pp. 859–890. 16