movement out of adjunct clauses in russian: evidence from semantics

Although in most familiar cases sentential adjuncts are islands, syntactic evidence for overt movement out of adjunct clauses in Russian has been around for a while. The present paper makes a case for covert (LF) movement out of conditionals. The data is the same as Ivlieva (2011) used to motivate her ‘laziness’ theory of pronouns bound by universal antecedents. The analysis proposed here exploits the possibility of movement out of conditionals (manifest in overt syntax elsewhere), in particular that of Acrossthe-Board extraction out of both the protasis and the apodosis.

Although in most familiar cases sentential adjuncts are islands, syntactic evidence for overt movement out of adjunct clauses in Russian has been around for a while.The present paper makes a case for covert (LF) movement out of conditionals.The data is the same as Ivlieva (2011) used to motivate her 'laziness' theory of pronouns bound by universal antecedents.The analysis proposed here exploits the possibility of movement out of conditionals (manifest in overt syntax elsewhere), in particular that of Acrossthe-Board extraction out of both the protasis and the apodosis.
[1] i n t r o d u c t i o n Finite adjunct clauses are generally believed to be strong islands for syntactic movement (Szabolcsi & den Dikken 2003, 214), i.e. no constituent may cross the border of an adjunct while moving: (1) *[Which college] i did you become a linguist [after you entered t i ]?
There are several known sorts of exceptions from this constraint.First, several languages (Japanese, Korean, Malayalam) have been reported to be exempted from the restriction (Yoshida 2006).Second, English rationale clauses are not subject to it (Culicover 1997, 253), as Culicover's (2) witnesses.Besides that, Barker (2012) provides numerous examples where the binding relation does not observe c-command, including (3) and the like (see his § 3.7), although on his account those are no exceptions as his proposed replacement for the c-command requirement rests on the notion of 'the order of evaluation'. (2) Who i did you go to Harvard [in order to work with t i ]?
(3) After unthreading each i screw, but before removing it i , make sure to hold the screw in place while separating the screw from the driver.
In what follows, I will survey other exceptional cases, mostly known data from Russian.Purely syntactic arguments will, however, serve a somewhat ancillary role, as the main body of evidence to be dealt with below is of semantic nature: LF daniel tiskin movement out of an alleged island will be a likely explanation for the perceived semantics.
The rest of the paper is organised in the following way.Section [2] introduces the cases of overt extraction from adjuncts in Russian.Section [3] is dedicated specifically to extraction out of conditionals.In Section [4] I finally turn to Ivlieva's examples,which are presented in [4.1] and analysed in terms of covert ATB movement in [4.2].A brief conclusion follows.
[2] r u s s i a n a n d t h e c o n s t r a i n t s o n e x t r a c t i o n In case of conjunction islands, where it is not typically possible to extract from one of the conjuncts (but see the discussion below), the option for extraction is Across-the-Board (ATB) movement, i.e. simultaneous extraction from all conjuncts which results in a single higher copy binding two or more traces (de Vries to appear).This strategy is readily available in Russian ( 4), but alongside with it there exist cases of 'semantically asymmetric coordination' (like in English, Lakoff 1986) where some of the conjuncts simply lack the target of extraction ( 5 For a survey on extraction from adjuncts in English, see Chaves (2013, 293).In addition, Yoshida (2006) shows the possibility of extraction from if -islands in Japanese, and Etxepare (2013, 489) provides an analogous example for Spanish, claiming that the conditional has to be indicative, as opposed to subjunctive, for the extraction to be possible.In the case of Russian, this does not seem to be a firm restriction (i), and in addition extraction is possible with infinitives (ii): Besides ATB-based analyses, in Williams (1990) one finds an approach to the phenomenon of parasitic gaps (pgs), as exemplified by Williams' (11), via ATB movement.(This approach was later criticised by Postal (1993), largely on the basis of pgs being more restricted in distribution than ATB movement.)As long as ATB, as it is usually conceived, requires coordination, Williams proposed to view conjunctiveness as a continuum based on syntactic as well as semantic considerations.
The more 'conjunctive' the structure is, the easier it is to extract ATB.
(11) Who i would you warn t i before striking pg i ?
On the other hand, Ivlieva (2007) proposed to extend the notion of a pg onto cases where no overt movement has taken place (12).Ivlieva assumes the covert A ′movement of the gap's antecedent into the Topic position and hints to the possibility of cross-linguistic variation in the overtness of movement operations.
[4] Vicente (2016) interprets the following example, due to Andreas Kathol, as a case of ATB movement without coordination: (i) This is the book i that everyone who reads i raves about i .
OSLa [4] ' u n i v e r s a l l a z i n e s s ' a s c o v e r t e x t r a c t i o n Having surveyed the possibilities for overt extraction from conditional clauses in Russian, we may now turn to a semantic puzzle discovered by Natalia Ivlieva (as presented in Ivlieva 2011).She has noted that a pronoun in the main clause of a conditional may be anaphoric on a universal antecedent in the conditional clause.I will show that Ivlieva's account is not unproblematic and propose another view which puts Ivlieva's cases in line with the data presented above.
[4.1] Unbound anaphora in conditionals Consider Ivlieva's example, given in ( 13).The pronoun in the second clause is anaphoric on the DP každyj 'everyone' in the first clause, although universal antecedents have not been previously recognised as suitable antecedents for unbound anaphora 5 (except for telescoping; see fn. 10).Moreover, the perceived truth conditions for one reading of ( 13) are stronger than expected on the assumption that if -clauses are islands for LF movement.What (13) says on that reading is, roughly, 'whoever minds his own business will be more useful and, in particular, if it so happens that everyone minds his own business, then everyone will be more useful'; in other words, the relative scope is ∀ > if.
( 'If each i of them married one of the sisters, his i brother would become his i brother-in-law' [5] Cf.: 'Generally <...> it is concluded that the universal quantifier shares <...> the characteristic of being externally static' (Groenendijk & Stokhof 1991, 51). [38] daniel tiskin Ivlieva assumes that Quantifier Raising (QR) is clause-bound, but even granted that it is not, QR together with binding into the main clause gives only the stronger reading.Thus Ivlieva's own account is spelled out in terms of pronouns of laziness (Nouwen submitted, i.a.), i.e. pronouns that 'go proxy' for their antecedents, so that the meaning of the sentence may be recovered by substituting a copy of the antecedent for the pronoun. 6Given this, ( 13) is regarded as a shorthand for (15) Esli každyj budet zanimat'sja svoim delom, každyj prinesët bol'še pol'zy.'If everyone minds his own business, everyone will be more useful'

[4.2] Arguments for LF movement
There are at least two problems for Ivlieva's analysis.First, the laziness approach lacks generality.This becomes manifest once we switch from universal antecedents to those with different quantificational force (as suggested in Tiskin 2015).Besides každyj, vse 'everyone' and mnogie 'many' also give rise to 'lazy' interpretations, although in this case the anaphor should be plural. 7Ivlieva's account, if extended to cover (16), would predict that oni should be interpreted as if a token of mnogie occupied its place.However, this yields incorrect truth conditions, according to which it is sufficient for the truth of ( 16) that some large subset of 'them' use Mac computers in case some (maybe other) large subset lives now.Meanwhile, the two readings (16) actually has are as follows: a. 'if some large subset i lived now, then that i subset would use...' b. 'for some large subset: if any i of those lived now, (s)he i would use...' [6] In addition to this, one has to account for the stronger reading.Ivlieva does this by manipulating situation indices (Barwise & Perry 1983) on predicates: a situation may comprise only a fragment of a world, thus even in a world where not everyone minds their own business there is a situation whose domain of individuals is exhausted by those who do mind; what (13) says is that in such situations every individual in the domain is being more useful.[7] An anonymous reviewer points out the similarity between ( 16) and cases of intersentential unbound anaphora such as (i).Here the pronoun they refers to the (maximal) set picked out by its antecedent.Plural anaphora is possible with universal antecedents, too.This weakens my first objection to Ivlieva, as two distinct analyses may be needed after all, depending on the number feature of the anaphor.
(i) [Many students] i in my class passed the exam.They i solved all the problems.
OSLa volume 9(1), 2017 movement out of adjunct clauses in russian [39] Ivlieva's reading differs from (a) in that no anaphoric link is established between the subset mentioned in the conditional clause and that of the main clause, 8 so that it is incorrectly predicted that any large subset would do, not necessarily the one mentioned in the conditional clause.This problem would not arise under a binding-based approach.
The second problem is that without anaphora, as in ( 17), only the scoping 'if > ∀' is possible.This is not explained by any straightforward account based on the idea of laziness, as on such views it is unclear why the presence or absence of replaceable material in the second clause may have any effect on the interpretation of the material in the first clause.Although Ivlieva (2007) developed an analysis of (12)-like pgs in terms of covert movement, she does not consider the possibility of such an analysis for (13).Nevertheless, stipulating ATB movement for the strong reading does help, provided that a way is found to assimilate if -clauses to cases of coordination (as e.g. in Williams 1990).Then it would be possible to say that the quantifier moves ATB at LF 9 out of both clauses if it occurs in both but cannot move if it only occurs in one.The weak reading will then have to be accounted for in terms of dynamic binding (Groenendijk & Stokhof 1991;Kamp & Reyle 1993), which will have to be extended onto universal quantification as well.The apparent problem tied to this analysis is how to explain that an overt pronoun is found in the second clause, which should [8] Note that a distinctive feature of pronouns of laziness is their ability to support non-coreference: (ii) This year the president is a Democrat.Next year, he will be a Republican.(Nouwen submitted) [9] Despite the existing arguments against covert ATB movement (see e.g.Bošković & Franks 2000), there is (albeit scarce) evidence for it, summarised in de Vries (to appear, § 2.2.7).
daniel tiskin not be expected of ATB movement, especially if it is covert. 10Regarding the latter problem, a reviewer has pointed out that the ATB analysis of (13) finds additional support in Bobaljik's (2002) architecture of interfaces.According to Bobaljik, who builds upon the copy theory of movement (Chomsky 1995, i.a.), either LF or PF have the choice of which copy to interpret and make their choices independently of each other.Thus, in (13), after the covert ATB movement of každyj, the upper copy is interpreted but the two lower copies are pronounced; an additional mechanism, perhaps close to that behind pronouns of laziness, replaces the second lower copy with a pronoun (see also (3)).( 18 However, if we go for an ATB analysis for (13) but allow for asymmetric extraction in the case of ( 9) and its kin, we run into a predicament.The reason is that free extraction out of adjuncts would result in the scoping '∀ > if ' being predicted for (17) as well.A reviewer notes that one would have to formulate constraints on overt and covert movement differently in order to get this right.
[5] c o n c l u s i o n The wide availability of overt extraction from adjunct clauses in Russian suggests that the same mechanism may be at play in the cases where covert movement is semantically detectable.Following Bobaljik's proposal to endow PF with some amount of freedom as to which copy to pronounce, I have suggested that Ivlieva's cases of binding by universal antecedents from within a conditional clause be accounted for by means of covert ATB movement of the antecedent.The account dovetails with Ivlieva's treatment of pg-like configurations in Russian, but problems for such an account listed above are not negligible.a c k n o w l e d g e m e n t s Thanks to the audience of the GRAMNORUS workshop in Oslo (April 2016) and to the two anonymous reviewers for OSLa, who provided thorough comments and important suggestions.The author is fully responsible for the remaining errors. [10] Another instance of unbound anaphora to universal antecedents is telescoping (i), for which Keshet (2007) proposed a solution based on its semantics.In (i) anaphora is possible because the second sentence elaborates on the situation introduced in the first one, which allows Keshet to assume that the 'discourse continuity' between the two sentences in (i) makes them parts of a single overall structure where each degree candidate is able to bind the anaphor.
(i) Each i degree candidate walked to the stage.He i took his diploma from the dean... (Roberts 1989) Although Ivlieva shows that her 'universal laziness' is licensed by 'eo ipso' semantics, i.e. by the fact that the situation of minding one's own business is itself the situation of being more useful, the analogy with telescoping is incomplete, as 'universal laziness' is hardly possible cross-sententially.(I am grateful to a reviewer for drawing my attention to it.) of them lived now, they would be definitely using Mac computers' describes what he understands well enough, a brilliant documentation will result' (Google) ).In the room where I slept, there was one bed such that whoever lay down on it to sleep fell ill with fever...' (RNC) It is also that Petya Farber is a dreamer such that what(ever) he sees in his night dreams he would tell out loud in the daylight' (Google) 'This is the tea made you but didn't finish drinking'[2]Should any of the types (or both) be assimilated to correlatives, the extraction analysis would not have to be abandoned, as extraction out of correlatives is attested in Russian as well.It may target either the relative clause (i) or the correlate (ii), or both, in which case the differences in case-marking force the use of a resumptive pronoun (iii).
And in the middle of the flower, there is hidden a number s.t.if you add 2 to it, the result will be 10'(Google)What is perhaps more surprising, movement out of adjuncts may take place from both clauses.4Ifcases such as (10) are treated as exemplifying ATB extraction, they will violate the usually assumed symmetry requirement for ATB (but see an overview of other sorts of asymmetries in de Vries to appear, 2.2): '...unforgettable hand-made English, which you will never forget if you once heard it' (RNC) To isolate the weaker reading, Ivlieva provides an intricate scenario.Suppose there are two brothers and two sisters.Then the main clause of (14) is true only if two marriages happen; one is insufficient.Therefore, the reading 'whoever marries...' is excluded, and only the weak reading remains.
use'If everyone i minds his own business, he i will be more useful' Ivlieva claims that (13) has two readings.The other reading is weaker and amounts to the 'if > ∀' pattern.