Case-driven movement to T of non-nominative NPs
What is new about the data with respect to recent work on A-movement of non NOM NPs to a preverbal position is that the underlying complements in these examples are structurally Case-marked and optionally non-D-linked. This contrasts with the familiar paradigm from Icelandic in (1), where a lexically Case-marked complement may be displaced into Spec-TP.
(1)
In (1a), where the main verb assigns lexical (i.e., quirky) Case to its complement, the passive construction exhibits the standard raising of the internal argument to Spec-TP without the standard Case- or agreement-checking motivation. In the absence of such movement, as in (1b), an expletive is merged in Spec-TP. When neither raising nor merger of an expletive into Spec-TP occurs, the result is deviant as (1c) illustrates. Under an analysis in which there is no independent EPP requirement, the obligatory presence of preverbal material in (16a-b) must follow from some other feature of T. Boeckx (2000a) argues that the requirement for an overt Spec-TP in such instances follows instead from an enriched machinery for Case-checking, involving the presence of a “Generic Case” feature in T, which attracts the lexically Case-marked complement or an expletive. Under this view, EPP-motivated movement in (1a-b) can still be held to overlap with a Case-checking operation.
To be sure, the existence of lexically Case-marked complements that are displaced into subject position is among the thorniest problems involved in reducing the EPP to Case (or to some other feature of T). For Icelandic, the Generic-Case proposal and an independent EPP-analysis are roughly equivalent since they account for the same set of data. Note, however, that the generic Case proposal cannot be extended to account for the data in (6–8). The main empirical problem with applying Boeckx’s proposal to the Russian and Ukrainian examples is that the preverbal internal arguments in these accusative unaccusatives are structurally Case-marked, in contrast to the derived lexically Case-marked subject of the Icelandic passive in (1a). Under the Generic-Case checking mechanism for this movement, unaccusative V’s complement would be forced to check two distinct structural Cases (with the peculiar property of bearing a morphological reflex of the former only), thereby creating a chain that is doubly Case marked.
That the direct object of accusative unaccusatives bears structural (rather than lexical) ACC is demonstrated on the basis of the following facts. First, the direct object undergoes Genitive of Negation (GenNeg) and Partitive Genitive (GenPart) formation, processes that are well known to apply to structurally Case-marked elements only. Note the examples from Russian and Ukrainian in (2) and (3), respectively:
(2)
(3)
The examples in (2–3) are in contrast to lexically Case-marked NPs, which fail to undergo GenNeg, as shown in the Russian examples in (4):
(4)
Another argument for a structural Case analysis of the ACC complement of accusative unaccusatives is the “heterogeneous” NP-internal case-marking pattern for numerically quantified nominals. When the numeral bears structural Case, the remainder of the NP is obligatorily marked GEN. If the numeral bears lexical Case, the remainder of the NP will be marked “homogeneously” with the same lexical Case, as in (5).
(5)
Using these Case-marking patterns as a diagnostic for structural versus lexical Case, we see in (6–7) that the ACC complement of accusative unaccusatives does indeed bear structural Case. Under the theory that these predicates assign lexical ACC, the homogeneous Case-pattern would be falsely predicted to occur.
(6)
(7)
Note, finally, that if the ACC complement of accusative unaccusatives were lexically, rather than structurally, Case-marked, this would predict that two ACC NPs (one lexical (an Experiencer) and the other structural) should be able to cooccur in the same clause. Indeed, Sigurðsson (1992) notes this possibility for Icelandic:
(8)
Similar ACC-ACC predicates do not occur in Russian and Ukrainian (or elsewhere in Slavic) where both ACC NPs are arguments. Note, for example, the ungrammaticality of (9), falsely predicted to converge with two ACC NPs under the theory that one of the two is lexically Case-marked (that is, where the first ACC is an Experiencer, parallel to the Icelandic example in (8)):
(9)
Returning to the Icelandic paradigm in (1), note that the DAT-NP/EXPL alternation reflects an interpretive distinction that obscures the precise nature of the checking relation with T. The raised internal argument in (1a) is obligatorily interpreted as definite. Under a theory of Generic Case, it would follow that this interpretive effect is, in some sense, built in to the checking operation. This presents an additional empirical obstacle for applying such a Case-checking mechanism to the preverbal ACC NPs in Russian and Ukrainian accusative unaccusatives since these preverbal ACC NPs are optionally interpreted as non-generic indefinites (i.e., they are optionally non-D-linked).
To summarize, the preverbal ACC NP in Russian and Ukrainian accusative unaccusatives is structurally Case-marked and indefinite (under neutral discourse). We have shown in this section that this ACC NP does not enter into a Case (or agreement) relation with T. Furthermore, we have distinguished the Russian and Ukrainian accusative unaccusatives from the superficially similar lexically Case-marked passive construction in Icelandic, which, unlike the Slavic constructions, involves an obligatory definite interpretation for the raised nominal expression.
