Generative Grammar Theoretical Framework Conditions
The movement of whoi into the external COMP1 is out of a tensed sentence and therefore prohibited by (1).
(1)
The effect of the TSC is to make all categories in embedded finite clauses, with the exception of the embedded COMP (to be discussed below), inaccessible to rules that relate them to categories in matrix clauses. The TSC holds for rules of construal (e.g., the rule that assigns coindexing between a reflexive pronoun and its antecedent) as well as for movement rules (see Chomsky (1973, 241) and Chomsky (1976) for important discussions). Thus, in (2), the TSC can be construed as blocking the coindexing of the reflexive pronoun in the embedded finite clause with the matrix subject.
(2)
It is important to note that under this interpretation the TSC does not directly account for the ill-formedness of (2). Rather, it is the proper binding condition on bound anaphora that determines the ill-formedness of (2). (2) is excluded because it contains a “bound anaphor” that is not bound to an antecedent.
The account of (2) just given crucially depends on construing the TSC as a condition on the application of rules. Suppose though that coindexing by construal rules applies freely without any conditions (or even that indexing in the base is free, thus allowing coindexing in the base). In this instance, a derivation could yield (3) rather than (2).
(3)
(3) does not violate the conditions on proper binding because the reflexive pronoun is bound to an antecedent that c-commands it. Nonetheless, the binding relation expressed in (3) is not well-formed. This follows automatically if we interpret the TSC as a condition on proper binding rather than as a condition on the application of rules (cf. Chomsky (1976; 1977b)).
Interpreted as a condition on proper binding at the level of surface structure (technically, interpreted surface structure—i.e., after the application of construal rules), the TSC would be formulated as shown in (4):
(4)
Given the convention that “bound anaphors” must be properly bound, (4) accounts for the ill-formedness of (5).
(5)
Here too a condition on proper binding for lexical bound anaphora extends naturally to trace binding as well. Note that the analysis of the relation between a moved phrase (NP in all the cases considered thus far) and its trace as that of bound anaphora is an empirical issue and not a matter of principle or definition.
At this point we might ask whether there is any difference between interpreting the TSC as a condition on the application of rules, as in (1), and interpreting it as a condition on proper binding, as in (4). The answer is in the affirmative; but the relevant facts involve traces rather than lexical bound anaphors like reflexive pronouns.
Interpreted as a condition on proper binding, the TSC (i.e., (4)) automatically accounts for the ill-formedness of (6).
(6)
whoi in COMP1 binds a trace in a tensed sentence, S′ 2—thus violating proper binding under (4). (6) is a structure that occurs in the two derivations. Interpreted as a condition on proper binding, the TSC is all that is needed to account for the misgeneration of (5). However, if the TSC is interpreted as a condition on the application of rules, only the derivation is accounted for and so we need to stipulate the SCC (or TEP) in addition to account for the derivation. Thus, it does make a difference how the TSC is interpreted. Only under the proper binding interpretation is it possible to derive the empirical effect of the SCC and thereby eliminate what turns out to be a redundancy in the content of the theory of grammar. In the case of (5), then, the SCC is superfluous. In section 3 it will be shown how this analysis extends to all cases in which violations of the SCC result in misgeneration.
Our analysis illustrates how the empirical effects of the SCC can in part be derived from independently motivated conditions on proper binding. This results from an interpretation of the TSC as a condition on proper binding. Note that the Specified Subject Condition (Chomsky (1973; 1976)) is also susceptible to such an interpretation.
(7)
Formulated in this way, the SSC accounts for the ill-formedness of the following structures.
(8)
Here again, lexical bound anaphors and traces pattern together.
Taken in conjunction with one another, the TSC and SSC prohibit binding between all categories in matrix and embedded clauses, with the exception of binding that involves the subject of an embedded nonfinite clause or the embedded COMP. In effect, these conditions specify “opaque domains” with respect to binding. A domain is “opaque” if a category within it may not be bound to a category outside. Thus, the predicates of embedded clauses and the subjects of finite clauses constitute opaque domains with respect to categories in a matrix clause. Viewed in this way, the TSC and SSC define an “opacity principle” for anaphora in natural language.
As defined, the Opacity Principle permits binding between matrix categories and the subject of an embedded nonfinite clause as illustrated in (9).
(9)
Comparing (9) with (8), we see that in nonfinite clauses, only the predicates constitute opaque domains.
It is important to note that predicates of nonfinite clauses are not opaque in any absolute or global sense. Thus, consider (10):
(10)
According to a global interpretation of the Opacity Principle, the binding relation between Jilli in the matrix clause and the trace ti in the embedded predicate would not be permissible. Since (10) yields a grammatical sentence, this interpretation must be rejected. Instead, a local interpretation of the Opacity Principle is required. Towards this end, we observe that there is a crucial difference between the structures in (10) and (8c)—namely, no possible antecedent intervenes between Jilli and ti in (8c), whereas ti in the embedded subject intervenes between Jilli and the ti of the embedded predicate in (10). Thus, in (10), neither the binding between the matrix subject and the embedded subject nor the binding between the embedded subject and the embedded object violates the Opacity Principle. In each instance the binding relation is local, in that each antecedent/anaphor pair is adjacent with respect to any other possible antecedent or anaphor. Each pair constitutes a link in a chain of binding relations. A chain is well formed if and only if each link is well-formed. Given this rather informal sketch of a linking convention, the Opacity Principle can be said to designate the subject of nonfinite clauses as an “escape hatch.”
This analysis extends quite naturally to the embedded COMP. Thus, compare (11) and (12).
(11)
(12)
Structure (11) yields a grammatical sentence, Who did John say Fred saw?; whereas (12) yields an ungrammatical sentence, *Who did John know what frightened? Given the linking convention, the Opacity Principle predicts this difference. The binding relation between whoj and its trace in (12) violates the TSC and SSC. In (11), the link between the trace ti in S2 and the one in COMP2 does not violate the Opacity Principle. Since (11) yields a grammatical sentence, it must be assumed that the link between whoi in COMP1 and ti in COMP2 is also well-formed. That is, the Opacity Principle must be formulated in such a way that COMP is also an “escape hatch.” (See Chomsky (1973; 1977b) for a detailed discussion of this issue.)
Such a formulation is given in (13).
(13)
In order that COMP have the required escape hatch property, “finite clause” in (13i) must be construed as S rather than S′. Therefore, the domain of a subject and a finite clause are the same, i.e., S. The linking convention (13b) guarantees that only adjacent categories in a binding chain will be subject to opacity conditions on proper binding.
The Opacity Principle plays a central role in the derivation of the empirical effects of the SCC, as will be documented in the following section.
