Expletive elements to, eto and sententional arguments in Russian

I discuss the syntax of Russian expletive elements to, and eto and focus on constructions with a non-agreeing nominal predicative subcategorizing for a dative subject (Dative-Predicative-Structures, DPS) and a sententional argument - bare that-clause (čto P-clause) or headed that-clause (to, čto P-clause), cf. Mne stydno, [čto P] ‘I am ashamed that P’, mne jasno, [čto P] ‘It is clear to me that P’, mne protivno, [čto P] ‘it is disgusting to me that P’. Data and previous research. Unstressed inanimate non-referential elements eto ‘this one’, ‘it’, ‘that’, and to ‘that one’, ‘it’, ‘that’ are morphologically ambiguous between Nom. Sg. N. and Acc.Sg. N., cf. eto mne jasno, mne jasno, to, čto P, and eto jasno, čto P co-occur with DPS predicatives. Since neither anticipatory eto nor correlative to are obligatory in the presence of a dative subject or a sententional argument, they are not recognized as subjects by authors who analyze DPS as a structure without agreement. However, Sperber (1972) and Apresian (1985) claim that those DPS predicatives that have a sententional argument (CP or IP) or combine with non-referential eto, to select sententional arguments eto, to as surface subjects. Zimmerling (2009; 2012) observes that a combination eto + dative subject + sententional argument is impossible (*eto mne jasno, [čto P] ‘It1is clear to me [that P]’) and argues that expletive eto alternates in the surface subject position with dative and sententional subjects. Licensing of to, čto P clauses relates to factivity, according to Padučeva (2004) and Zalizniak (2006), who however do not filter out correlative to in oblique cases and in the phrases with prepositional government (cf. On raduetsa tomu, [čto P], lit. ‘He rejoices to [that P]’, On uveren v tom, [čto P], lit. ‘He is sure in that [that P]).

I argue that to and eto have different syntax. Expletive unstressed eto is a matrix clause element. It alternates in the subject position with dative subjects and with čto P clauses, does not form a constituent with the CP it antecedes. It has a propensity to fronting and often takes distant position to the CP. Expletive unstressed to belongs to the complement clause, forms a constituent with the CP and cannot be separated from it. It does not alternate with dative subjects and does not require fronting. Bare čto P arguments are licensed by a number of DPS predicatives which lack agreeing correlates (cf. X-u stydno, vse ravno, čto P; no agreeing adjectives *stydnyj, *vse ravnyj are possible). A number of DPS predicatives have agreeing correlates (cf. X-u protivno to, čto P; protivnyj-Adj), which license a dative-nominative structure with a nominalization like tot fakt ‘that fakt’ in the surface subject position (X-u protiven tot fakt [čto P]), but the corresponding predicatives do not license to, čto P clauses: *X-u protivno to, čto P. These facts confirm that DPS predicatives only have default agreement pattern (3 p. sg; 3 p. sg. n in the past tense) and undermine Sperber-Apresian’s claim that correlative to in DPS is matrix clause subject imposing standard number-and-gender agreement. I show that licensing of to, čto P clauses splits DPS predicatives in two groups: all DPS predicatives which license headed to, čto P clauses license bare čto P clauses, but the reverse is false. Licensing of to, čto P clauses is triggered by a feature related to factivity – ability of projecting event structure with a dedicated sub-event. DPS predicatives from stydno ‘X is ashamed’, vse ravno ‘it is all the same for X’, protivno ‘disgusting’ group only license bare čto P-clauses, but not to, čto P clauses (*X-u protivno to, čto P). These predicatives are strictly non-factive and denote intentional states, events where no sub-event can be singled-out. DPS predicatives from the X-u jasno ‘it clear to X’, X-u bezrazlično ‘does not matter’, X-u neizvestno ‘It is unknown to X’ group both license bare and headed čto P clauses, cf. X-u neizvestno, [čto P]~ X-u neizvestno [to, [čto P]], since they can both denote intentional states and project structure with a dedicated sub-event. The diagnostics for factivity/event structure should be based solely on the uses of [to, [čto P]] in direct cases (structural nominative and accusative), since the correlative to in oblique cases is a matrix clause element that gets its case locally from the matrix verb or preposition.

Comments. 

 For the details see the paper Zimmmerling (2014).

 Later contributions: Zimmerling (2019, 2021), check the section 'Related works'.

 In this 2014 conference talk given at the SLS 9 Conference in Seattle 2014, I do not stick to the notion of DP and to not treat headed that clauses as DPs. But I have nothing against this analysis, cf. later accounts of Lyutikova, Kniazev and Bondarenko. Do it if you like.