We find, just as Nunes (1998) does, that in a
derivational theory, the only basic syntactic relation is c-command. A distinction
is made between representational c-command and
derivational c-command. X only c-commands elements dominated by Y when X is merged with Y in derivation (Epstein et al 1998:10). Derivational c-command suffices to explain the determination of order more easily than Kayne (1994) whose Linear Correspondence Axiom has been proven to be unworkable by Chomsky (1995a) as it is based on X' theory. Chomsky (1995a) continues to use
intermediate projection while adhering to Uniformity Condition (UC). Nunes (1998:163) points out that UC and Chomsky's (1995a:322) stipulation are against minimalism.
If we adopt Kayne's (1994) LCA to determine the order of categories in (1), we may find conflicting results. That is, DP asymmetrically c-commands "buy" and "it", and dominates "the" and "man". According to LCA, "the" and "man" should precede "buy" and "it". On the other hand, if intermediate projection is allowed to enter c-command relations, then LCA should require "buy" and "it" to precede "the" and "man", because T' asymmetrically c-commands "the" and "man", and dominates "buy" and "it". The controversial results disappear if intermediate projection is not allowed to enter c-command relations or LCA is revised. The adoption of derivational c-command makes intermediate projection and Kayne's LCA dispensable. If there's no intermediate projection and the projection of an existing category (syntactic object) doesn't c-command a newly merged category, then derivational c-command can be redefined as (3).
Suppose X and Y are non-terminals, x and y are terminals. X dominates x and Y dominates y. If X asymmetrically c-commands Y, then x precedes y.
(3) Derivational c-command (revised)
A c-commands B and its elements iff A is merged later than B in syntactic operations. (Dai 2002) Bare Phrase Structure Theory, as Chomsky(1995b) claims, abandons X' levels but maintains the use of Xo and XP. The same item may be either Xo or XP as there's no distinction between items and heads of their projections. If that's the case, we are justified in assuming HP iteration, further projection of HP with a new category merged or a moved category remerged. In contrast, Chomsky's persistent use of X', Xo, XP in DbP(1999) causes confusion in explaining changing syntactic relations in derivation.
With X' levels, distinction between Xo and XP, and phrase structure theory abandoned, we can assume multiple specifiers of XP (VP, v*P, TP) for feature checking and merging operations. If our approach is on the right track, then we can remove intermediate projection completely from derivation. What is projected and mapped to LF is always HP, that is VP, v*P, AgroP, TP and CP. In this sense, we have reached the goal of eliminating phrase structure theory, as suggested in Chomsky (1995a). Elimination of phrase structure theory means the end of representational theories and the birth of a purely derivational theory as we propose.