The motivation for the syntactic cycle
The syntactic cycle was initially proposed for methodological reasons. Adopting a cyclic principle of rule application allowed for a fundamental simplification of the theory of grammar as formulated in the early 1960s. In particular, some powerful theoretical machinery such as generalized embedding transformations, transformational markers, and the as yet unformulated type 2 projection rules of Fodor-Katz semantic theory could be eliminated in favor of recursive base rules and the cycle. The result of adopting the cycle was a more highly structured theory of grammar that was also weaker in expressive power; specifically, it did not allow for the possibility that singulary transformations would have to apply to a matrix S before a constituent S could be embedded in it. (See Chomsky (1966) for a detailed discussion of this point.) However, there was no discussion then of the empirical motivation for the cycle; and in fact, it seems as if no empirical differences between the two theories were noticed at that point.
In the rather extensive literature on the cycle that has developed since its initial proposal, it has been claimed that the cycle is empirically motivated by grammatical sentences whose derivations require that some pair of rules applies in an A-B-A order— that is, cyclically. Rather than review this literature (and the equally extensive counterliterature claiming that these arguments for the cycle can be subverted), it will be sufficient to examine what is perhaps the strongest case of this sort in English, the interaction between the rules of NP Preposing and There Insertion.
Within the framework presented above, There Insertion consists of two independent operations, an NP movement and an insertion of there into an empty NP category (see Milsark (1974)) without affecting the index of that category (cf. Dresher and Hornstein (1979)). The former operation is a rightward movement of an NP around be, one realization of the rule “Move NP.” NP Preposing is yet another realization of the general rule. Given that the two NP movements cannot be distinguished from one another under our framework, an A-B-A derivation could only be Move NP—Insert there—Move NP.
Consider the following actual case. (1) gives a derivation for (2).
(1)
(2)
There were reported to have been tigers sighted in Spain.
Note that proper binding is observed in (1e). (1) provides an A-B-A derivation for (2) within the framework under consideration.
The derivational arguments for the cycle claim that the A-B-A ordering is not merely sufficient, but also unique in yielding the correct result. Clearly this is not true for (2), which could be derived by applying all the NP-movements before inserting there in the matrix subject. The derivation would be identical to (1) except for the substitution of (3) for (1d).
(3)
Recall that np in (1), as well as ti, represents an empty NP category, [NPi e], where the category bears an index i. Thus, replacing np with ti merely involves changing the index of np to i (or adding the index i if it is assumed that basegenerated categories do not have indices when they are empty).
Within the framework under consideration, it seems unlikely that insertion operations that do not affect the indexing of a category will be crucial for any movement operation. This leaves the possibility of an A-B-A interaction between “Move NP” and “Move wh-phrase.” Here too there will be nothing crucial about such an interaction. Given that once a wh-phrase moves into COMP it may only move into another COMP (see Chomsky (1977b)), NP Movements and Wh Movements will be essentially independent of one another.
That there will be no crucial A-B-A arguments for the cycle within the theoretical framework assumed here is exactly what should be expected. Given filters, the facts of derivation with respect to surface structures are of no empirical significance. Moreover, misgeneration (and not generation) is what constitutes empirical motivation for this or that analysis. Since the major movement rules assumed) apply without constraint, we can assume also that they will generate all the grammatical sentences as well as a considerable number of ungrammatical structures. Generation of grammatical sentences is no longer an issue; rather, it is misgeneration that has to be accounted for. From this it follows that only misgeneration resulting from violations of the SCC constitutes empirical motivation for the cycle.
