As with most ellipsis mechanisms, theoretical accounts of stripping face significant challenges. The greatest challenge is to come up with a coherent explanation of the stripped material. The insight that the remnant of stripping is always a constituent is straightforward. The difficulties arise when one attempts to discern which individual constituents can and cannot be a remnant. For instance, why are the remnants in the following cases disallowed?: :: *Larry has read the text, and he has enjoyed it. - Failed attempt at stripping :: *Susan promised to read the page, and she promised to copy it. - Failed attempt at stripping :: *An instructor flubbed the problem, and a student flubbed the problem too. - Failed attempt at stripping The remnants in these examples are constituents in
constituency grammars (
phrase structure grammars), since every individual word is by definition a constituent in constituency grammars. These words are not, however, constituents in
dependency grammars, since they dominate other (elided) material. The constituency vs. dependency distinction is therefore one avenue that one might pursue to locate an explanation of such cases. If one chooses a constituency-based grammar, however, then the explanation might draw attention to the distinction between projection levels (see
X-bar theory): the remnant must qualify as a maximal projection (as opposed to an intermediate or minimal projection). Another avenue to explore for an explanation is to focus on the elided material (as opposed to on the remnant). In most cases, the elided material cannot be characterized as a constituent. It can, however, be characterized as a
catena. The following dependency grammar trees illustrate this explanation in terms of catenae. The elided material is indicated with a lighter font shade: :: The elided word combinations form chains (catenae), that is, the elided words are linked together by dependencies in the vertical dimension. The word combinations
do you want and he
said...that he knew it are catenae. These two examples must be compared to the following two: :: These sentences are bad, and one can explain their badness by acknowledging the status of the elided words as non-catenae; the elided words are not entirely linked together in the vertical dimension. The object pronoun
it in the a-tree and the indefinite article
an in the
b-tree are not linked directly to the other elided material. This observation may explain why these attempts at stripping fail. The elided material should qualify as a catena. ==See also==