Lexical Case phenomena
The analysis of Case phenomena presented in this paper is based on a theory of Case structure in generative grammar that has been developed over the past decade (see Chomsky 1986a for some general discussion and Chomsky 1981 for more technical details of the basic theory). Central to this theory of Case is the Case Filter (1), which plays a crucial role in determining the distribution of lexical (that is, phonetically realized) versus nonlexical NPs (trace and PRO) in sentences.
The Case Filter designates as ill formed any lexical NP that is not marked for Case. Under the well-motivated assumption that the infinitival subject position in such constructions is not marked for Case, the Case Filter excludes infinitival indirect questions with lexical subjects as in (2a) (in contrast to (2b), where the infinitival has a nonlexical subject—in this instance PRO—and to (2c), where the indirect question is a finite clause).
Under this analysis, it must follow that the Case Filter does not apply to D-Structure representations. If it did, then raising constructions such as (3a) would be ruled out at D Structure where the lexical subject is in an infinitival subject position, illustrated in (3b).
(3c) shows that a lexical subject cannot occur in the infinitival subject position of a raising construction at S-Structure, in contrast to the subject position of a finite complement (for example, (3d)). (3a) demonstrates that the Case Filter applies after movement transformations.
If Case assignment applies after movement transformations and the Case Filter applies to syntactic representations derived via movement rules, there must be a mechanism for assigning Case to an NP that occurs in a Case-marked position at D-Structure and is moved to a position at S-Structure that is not Case-marked, as with the moved wh-phrase in (2). For the analysis of (2), we assume that the wh-phrase what inherits a Case marking via the trace that it binds in the complement object position. Thus, this mechanism for Case assignment in such representations derived by movement transformations can be construed as Case inheritance via trace binding.
Though Case inheritance holds generally for movements from a Case-marked D-Structure grammatical function position (for instance, object position) to an S-Structure position that is not a grammatical function position (for instance, the specifier position of CP), it does not seem to apply in constructions where movement is between two grammatical function positions. In (3a), for example, where movement occurs between two grammatical function positions (matrix subject and complement subject), the moved NP is assigned Case by virtue of moving into a Case-marked position from a position that is not Case-marked. This assumes that a verb like seem does not assign objective Case to the complement subject position—in contrast to a verb like expect, which does, as illustrated in (4).
The difference between seem and expect that is thought to account for the difference in Case-marking possibilities is that seem, in contrast expect, does not assign a semantic function (or θ-role) to its subject. Given this correlation between θ-role assignment and Case-marking possibilities, it is assumed that a verb that does not assign a θ-role to its subject may not Case-mark an NP that it governs. This correlation generalizes to passive constructions as well, where the passive predicate does not assign a θ-role to its subject.
Thus, even though the passive predicate governs the infinitival complement subject, it does not assign objective Case to this NP, in contrast to the corresponding active predicate in (4). Thus, (5a) constitutes a Case Filter violation. In (5b) the D-Structure complement subject has moved into matrix subject position where it is marked for nominative Case in the normal fashion. This correlation between the inability of a verb to assign a θ-role to a subject and its inability to assign Case to an NP it governs is generally referred to in the literature as Burzio’s generalization (see Chomskey 1986:139–141 and Burzio 1986: sec. 3.1). It is standardly assumed that this failure of Case assignment results from a mechanism of “Case absorption,” which is induced by passive morphology for passive predicates.4 As we will show, this assumption requires some revision when we consider the fuller range of Case phenomena, which includes lexial Case—that is, Case marking that is determined as a lexical property of certain heads (for instance, V and P) in some languages, as opposed to Case marking determined solely in terms of syntactic configuration (henceforth configurational Case).
The Case-theoretic analysis of (2)–(5) given above, essentially the standard analysis, rests on several assumptions that we would like to examine in some detail in this and the following sections. Let us suppose that Case is assigned as an index to the maximal phrasal projection of N (designated as NP, in contrast to nonmaximal phrasal projections, which will be designated as N*). At this point two basic questions arise about the process of Case assignment: (i) What is the formulation of the rule of case assignment? and (ii) Where does this rule apply in relation to other rules of grammar (in other words, where is Case assignment located in the organization of a grammar)?
The formulation of the rule (or rules) of Case assignment crucially affects the interpretation of the Case Filter. Suppose the rule of Case assignment is stated in the optimally simple form (6).
If (6) is interpreted as an optional rule, then its particular behavior with respect to various constructions (when it must apply versus when it cannot apply) will be determined by general principles of grammar. For example, when the rule fails to apply to a lexical NP in a canonical Case-marked position (for instance, subject of a finite clause), then the resulting representation violates the Case Filter. When Case is assigned to a lexical NP in a syntactic position that is not licensed for Case (for instance, infinitival subject position in indirect questions, as in (2)), then the resulting representation violates the general principle of proper Case licensing stated in (7).
A Case index will be properly assigned where it is governed by an appropriate element (for instance, accusative Case governed by V and nominative Case governed by agreement). (7) is independently needed to exclude instances where (6) assigns the wrong Case index to an NP (for instance, assigning accusative Case to the subject of a finite clause). Thus, under the formulation of Case assignment as (6), the explanation for why indirect infinitival questions in English may not have lexical subjects has two parts, one of which involves the Case Filter and the other, a principle of Proper Assignment.
An alternative to (6) would be a set of specific rules, each of which assigns a particular Case to an NP in a particular configuration. This is assumed in the standard analysis, where the lexical subject of an infinitival indirect question would never be assigned a Case and would therefore always constitute a Case Filter violation. This solution conflates Case assignment with Case licensing, which, we will argue, need to be distinguished for the analysis of lexical Case phenomena.
A third alternative would be to consider (6) as an obligatory rule in Universal Grammar (UG). Thus, every NP in a given phrase marker will be assigned Case. Under this analysis the Case Filter is essentially useless. To account for the “Case Filter effects” under the standard analysis, the principle of Case licensing would be restricted to phonetically realized NPs. Note that this is necessary in any event since PRO as well as lexical NPs will be Case-marked if (6) is obligatory, and presumably Case-marked empty categories are not subject to any particular Case principle.
