Coordination, ATB and ellipsis

"Shared constituents" such as what did in (3) are then to be handled in terms of ellipsis. In short, I advocate a return to the "large-conjunct-plus-reduction" conception of coordinations associated with early generative grammar (Chomsky 1957) although with reduction understood as "ellipsis" of terminals (PF-material) and not as deletion of syntactic structure. If (1/2) is replaced by (5b.), conjunctions uniting clauses can be seen as "discourse connectives", heads that take root clauses as their arguments. This conception fits most naturally with the idea mat coordinate structures should not be treated as symmetrical, n-ary branching, and multi-headed exceptions to XMheory (in particular, to the binary branching hypothesis of Kayne (1984), and the antisymmetry hypothesis of Kayne (1993)). However, it may be necessary to allow for "embedded" coordination I n coordination seems a pretty clear case, motivated on syntactic grounds. But I see no reason for assuming conjuncts can be any smaller than complete "extended projections" in the sense of Grimshaw (1991). If correct men the rationale for assuming ATB-rule application disappears. Further argumentation builds on the following points: