The Syntactic Cycle
In its initial formulation—essentially as in (1) (cf. Chomsky (1966))—the cycle is a general principle that determines, in part, the order of application of transformations with respect to syntactic domains in phrase markers.
In a derivation, for all syntactic domains α in a phrase marker, a linear sequence of transformations applies to a domain αi before applying to αj, where αj contains αi.
“α” denotes the set of categories that constitute cyclic domains.1Under (1), rules apply to successively larger cyclic domains until the entire phrase marker has been processed. This mode of application is commonly referred to as “bottom-to-top.”
A sharper notion of the cycle is provided by the Strict Cycle Condition (Chomsky (1973, 243)):
Strict Cycle Condition (SCC)
No rule can apply to a domain dominated by a cyclic node A in such a way as to affect solely a proper subdomain of A dominated by a node B which is also a cyclic node.
The SCC ensures that once a cycle has been passed in a derivation, it is inaccessible to any rule that does not analyze it as a subdomain—i.e., by making crucial reference to some constant term in the matrix domain. For example, suppose that, given a complex structure (3), a rule Ri applying on the S1 cycle transformed the structure of S2 in such a way that a rule Rj could then apply solely within S2, where the structural description of Rj was not met before the application of Ri.
In this derivation Ri feeds Rj, and the feeding relationship holds between a cyclic domain and its cyclic subdomain. The SCC, but not the cyclic principle in (1), rules out such derivations.
As a concrete illustration, consider the following derivation (4), in which Wh Movement violates the SCC (but not (1)) and results in misgeneration (i.e., an ungrammatical output).
Specifically, the movement that maps (4c) onto (4d) violates the SCC. The movement of who from the embedded COMP to the matrix COMP on the S′ 1 cycle feeds the movement of what in S′ 2 into the embedded COMP. The latter movement cannot apply when the embedded COMP is filled. When it does apply, the application crucially involves only terms within the cyclic subdomain S′ 2. (4) gives a partial derivation of (5).
(5) constitutes the crucial case for the SCC in previous discussions (see Chomsky (1973)). In section 3, many other cases will be discussed.
The SCC is not merely an ancillary principle to (1)—which is usually conceived of as “the cyclic principle.” Rather, the SCC subsumes this notion of cycle (see also Lasnik and Kupin (1977)). Clearly if rules may not apply solely within cyclic subdomains on any given cycle, then the only point in a derivation where they may legally apply solely within those subdomains is on the cycle of the sub-domain. Given the SCC, a rule that can apply to the most deeply embedded cyclic domain of a phrase marker must apply on the cycle of that domain or not at all. Therefore, a stipulation about the order in which subdomains of a phrase marker must be operated on by rules (e.g., (1)) is superfluous, since this ordering (i.e., bottom-to-top) follows from the SCC alone.
In fact, without the SCC or its equivalent, it is questionable whether there is a coherent notion of cycle at all. That is, without the SCC it is possible to construct derivations involving rule orderings that wildly violate bottom-to-top ordering but are consistent with (1). Given a structure of n cyclic domains where some optional rule may apply to each domain, it would be allowed under (1) to refrain from exercising the option to apply these rules until the last cycle. If the structural descriptions of these optional rules are still met in the various cyclic subdomains of the last cycle, then the option to apply them may be exercised. In this case, the rules could apply in every possible order—only one of which is bottom-to-top. Clearly (1) is not sufficient to ensure that only cyclic (bottom-to-top) ordering is permissible. In marked contrast, the SCC permits only the cyclic ordering.
The fact that the SCC reduces the class of possible derivations to only those that are cyclic has no empirical significance in the hypothetical case cited above because the output of the noncyclic derivations will be identical to that of the cyclic derivation. The interaction of the SCC and the two possible derivations of (6) provides a concrete illustration.
Both derivations involve NP Preposing in S′ 1 and S′ 2. The SCC prohibits the derivation in which the rule applies to S′ 1 before it applies to S′ 2. Yet because the outcome is the well formed (6) in any case, the exclusion of one possible derivation of (6) does not alter the strong or weak generative capacity of the grammar. Reductions of this sort are therefore without empirical consequence with respect to the class of possible languages a grammar generates.
In contrast to (6), there are cases in which the SCC excludes derivations that result in misgeneration. (4) above is one example. Such cases provide the empirical motivation for the SCC. Ill-formed strings whose derivations violate the SCC (e.g., (5)) constitute the empirical content of the SCC.
In the next section, (5) will be considered as paradigmatic of SCC violations. It will be demonstrated that the empirical effect of the SCC can be derived within trace theory from an independently motivated condition on traces. This condition makes no reference to notions like “cyclic domain” or “stage of the cycle.”
