As a consequence of the analysis proposed here, it appears that an element in COMP can have one of two things happen to it at S-structure. It may be deleted, provided that it is
$\left[ {{*{20}c} \alpha & {WH} \\ {\alpha i} & {} \\ } \right]$\left[ {\begin{array}{*{20}c} \alpha & {WH} \\ {\alpha i} & {} \\ \end{array} } \right]
, or it may fuse with another element, provided that one of the two elements is
$\left[ {{*{20}c} \alpha & {WH} \\ {\alpha i} & {} \\ } \right]$\left[ {\begin{array}{*{20}c} \alpha & {WH} \\ {\alpha i} & {} \\ \end{array} } \right]
. In the latter event, an
$\left[ {{*{20}c} \alpha & {WH} \\ {\alpha i} & {} \\ } \right]$\left[ {\begin{array}{*{20}c} \alpha & {WH} \\ {\alpha i} & {} \\ \end{array} } \right]
element also dictates the resultant form of fusion, with the consequences for the proper government of subject traces discussed above.
While taking its inspiration from Pesetsky's rule of special deletion in COMP, fusion is distinct from it in applying also to Wh-questions and relative constructions where the resultant form in COMP is a Wh-phrase. Thus, three constructions, rather than two, involve fusion: Wh-questions, Wh-/that relatives, and in some languages and dialects, that-e constructions. To summarize in somewhat rough terms, a subject gap is licensed when a Wh-phrase may potentially appear in pre-subject position. When this is the case, either the [`(X)]\bar X
syntax. In retaining or acquiring an index, it permits appropriate proper government of subject position in both Wh-questions and relatives. In particular, it explains how a complementizer which is otherwise not indexed comes to behave like an indexed Wh-phrase. Further, FIC-T explains how for some languages and dialects the [–WH] complementizer may act as an indexed element and license a subject gap when the Wh-phrase corresponding to the gap is in a more remote position.
This analysis has a number of advantages. For one thing, it allows a direct characterization of varying levels of acceptability among the constructions discussed above. It also separates the sometimes-accepted
that-e constructions from the much more rarely accepted
whether-e constructions. Second, it offers a concrete and general mechanism for explaining the apparent acquisition of head status by a Wh-phrase moved into a COMP already containing a natural
[ *20c + WH + COMP ]\left[ {\begin{array}{*{20}c} { + WH} \hfill \\ { + COMP} \hfill \\ \end{array} } \right]
head. Fusion thus offers an explicit account of how compliance with the requirement that a [+WH] COMP must have a [+WH] head takes place, while still allowing the [+WH] and [–WH] complementizers parallel status in phrase structure. Third, for English, fusion offers an account of what the resultant forms can be. In particular, it explains how Wh-phrases interact with complementizers and thus offers a more general account of the licensing of subject gaps while avoiding the complication of making
that subject relatives exceptions (to the
that-e filter) or requiring special rules which do no other work.
Thanks are due many people for various contributions to this work in its present and earlier forms, including the anonymous readers and editors for their thoughtful and detailed commentary. I am also especially indebted to Wieke Benjamin, Greg Carlson, Carol de St. Victor, Richard Hudson, Greg Iverson, S. Jay Keyser, Tom Larsen, Richard O'Gorman, Michel Raby, John Ter Haar, Rosemarie Whitney, and the students of Language and Society, for their generous contributions to this work. Any errors are my own.