摘要：My point here is that the strong generative capacity of Kayne’s (1994) classical version
of the Antisymmetry Theory (AT) is greater than usually claimed. Thus, AT is not as
restrictive as it seems at first. In and of itself, this is neither good nor bad. It is an
empirical matter whether the additional types of structures apparently generated by AT
correspond to representations of natural language sentences. However, since those were
initially thought to be blocked by the mechanisms of AT, and initially thought not to
exist, it is worth showing that, unless AT is modified accordingly, such structures are
indeed predicted to be well-formed. In what follows, I presuppose full knowledge of
AT from the reader, and I adopt the AT metalanguage to analyze each case. Given AT,
the following types of configuration are supposed to be blocked by the Linear
Correspondence Axiom (LCA): (i) n-ary branching (n > 2); (ii) heads adjoined to nonheads;
(iii) non-heads adjoined to heads; (iv) multiple specifiers; and (v) multiple
adjunction to heads. But look at (1-4):
