Lexical versus configurational Case in Russian
This section initiates an investigation of lexical Case phenomena that we hope to show has significant consequences for the theory of grammar as it has been formulated in the standard work on generative grammar (see Chomsky 1981, 1986). We begin with a discussion of lexical versus configurational Case in Russian as a striking illustration of the different syntactic properties of these two types of Case. In the following discussion configurational Case designates a Case marking that is licensed solely in terms of a canonical syntactic configuration (for instance, accusative Case on an NP governed by V or nominative Case on an NP governed by the agreement element). Lexical case designates a Case marking on an NP that is associated with a particular lexical head and that differs from the canonical configurational Case that would otherwise be assigned to the NP that bears the lexical Case.
In Russian the lexical versus configurational Case distinction shows up clearly with respect to the verbal object.
(1)
In both (1a) and (1bi) the verb governs its NP object. The verbs poceloval and pomog differ in that the latter requires that its object occur in the dative Case—thus, (1bii), where the object is Case-marked accusative, is ill formed even though the NP occurs in a syntactic configuration that licenses accusative Case. Note that in Russian Case is also morphologically realized on lexical modifiers of N (such as determiners and adjectives). This can be accounted for if Case marking of all such lexical elements of the NP results from propagation of the Case index on NP to the lexical constituents in the government domain of the lexical head N.
Russian exhibits a striking difference between lexical and configurational Case with respect to nouns modified by certain numerals.
(2)
In the “quantified” noun constructions in (2), when the NP is marked for configurational Case, the numeral manifests the appropriate configurational Case marking, whereas the remainder of the NP is obligatorily marked in the genitive. In contrast, a lexically Case marked NP shows no such “Case splitting,” as illustrated in (2bi) versus (2bii). (For further discussion of these constructions, see Freidin and Babby 1984 and Babby 1987).
The failure of Case splitting in lexically Case-marked NPs can be accounted for by assuming that lexical Case marking involves a head-to-head relation (like selection). Thus, the verb pomoc’ ‘help’ has a lexical property that imposes dative Case on the head of its NP object. This can be stated as a syntactic feature+_____DAT that is interpreted like a selectional feature (for instance+_____[+animate] for the verb frighten) and specifies a head-to-head relation between a verb and its object. Satisfaction of lexical properties where they exist in languages seems to be obligatory and to take precedence over structural properties where the latter conflict with the former. This follows if (3) is a principle of grammar.
(3)
Given that lexical Case properties must be satisfied, it follows from our assumption that Case is assigned to a maximal phrasal projection that there can be no Case splitting in lexically Case-marked NPs. In order for a lexical Case index to propagate from NP to its head, all projections of N will be lexically Case-marked and presumably all lexical constituents governed by N as well.
