On a Lexical Parameter in the Government-Binding Theory
The ‘Government-Binding’ theory of Chomsky’s recent work provides the outlines of general theories of abstract Case, and of binding. Implicitly, the theories make a variety of parameters available within universal grammar. It will be the purpose of the present paper to explore the nature of certain of these parameters with the goal of determining how it is possible for the language learner to ‘fix’ them. In the course of the discussion, it will become evident that markedness is crucially involved.
Unable to display preview. Download preview PDF.
- 1.Even with S deletion, there must be limitations on exceptional case marking, as is evident from the behavior of seem — *It seems John to be intelligent’ — even though `J seems [e to be intelligent]’ indicates that S has deleted (or the Empty Category Principle would be violated). Thus, seem must be specified as not assigning Case.Google Scholar
- 2.Interestingly, binding theory does not require that PRO be bound here, as it has no governing category. Since it is clear that PRO must pick up its reference from they rather than being `arbitrary’ in reference, an additional stipulation on structures of this sort is needed.Google Scholar
- 3.Note that contrary to what one might expect, this example does not also violate principle (2c) of the binding theory. Since each other here is not governed, it has no governing category, and hence no binding requirement.Google Scholar
- 4.Presumably S deletion will be inapplicable when for is present as the complementizer for the want class. In such cases as (a) (a) I want very much [S for [s you to leave]] for must assign Case to you, and, by implication, govern it. (b) *I want very much [S for [s PRO to leave]] bound anaphors other than PRO should be grammatical in these complement subject positions, since the governing category will be the larger S. (Recall that S is not a possible governing category.) Unfortunately, the relevant data, as in (c), is rather unclear. (c) ?We want very much [r for [s each other to win]]Google Scholar
- 5.A coindexed pronoun will also be impossible when for is present as in fn. 4. Bill, wants very much [r for Is him, to win]] Again, the governing category of him is the larger S (even though the governor is for.) Him is then bound in its governing category violating (2B).Google Scholar
- 6.Vietnamese is uninflected, yet the central differences between finite and infinitive clauses that are apparent in English show up in this language as well. Apparently, the relevant notions are quite abstract. See Dang (n.d.) for examples and discussion.Google Scholar
- 7.An alternative, and perhaps more natural, constraint is the Functional Uniqueness principle of Freidin (1978: 536, fn. 25). This principle prohibits a lexical NP from filling more than one argument position in the logical form of a sentence. It would straightforwardly exclude (9a, b).Google Scholar
- 8.A possible alternative is suggested by certain observations in Anderson (1979). In effect, Anderson suggests that a preposed NP within an NP must bear a very specific semantic relation to the head noun. In the NPs considered above, there is of course no semantic relation. Why NPs should have this strong requirement that Ss lack is not clear. Further, as Anderson notes, the way in which the observation is instantiated in her theory would not, in fact, extend to (12, 13).Google Scholar
- 9.Chomsky (class lectures 1980) has re-examined these facts involving NP preposing and suggested a different approach. He argues that S deletion is impossible for want, as it is for try. The difference is that complementizer for is available as an option with want and can assign Case (prior to complementizer deletion). If for is not selected in the base, (19) will be out directly by ECP. If for is selected, there is a problem. (a) can be ruled out by the for — to filter of Chomsky and Lasnik (1977): (a) *J is liked [g for Is e to do that]] But if for is selected and later deleted, the filter is not violated. Further, the deleted for would be present in LF, apparently saving (a) from the effects of ECP. It must apparently be stipulated under this approach that although complementizer for counts as a governor for Case-assignment, it does not count as a governor for ECP. The subject trace would then be ungoverned. A possible alternative, suggested by observations of Anderson and Dang (1979), is that both S and S are bounding nodes for subjacency when COMP is filled.Google Scholar
- 10.Note that Chomsky’s alternative to S deletion for want-class verbs in English (see fn. 9) is not available in Vietnamese No complementizer ever shows up in the complement of moun. Google Scholar