Neither fragments nor ellipsis *

Jason Merchant (2004, and Chap. 3, this volume) proposes to account for all speech acts performed with “fragments,” whether in discourse-initial position or otherwise, by appealing to syntactic ellipsis. Though his proposal is insightful, I offer empirical and methodological considerations against it. Empirical problems include: (a) His alleged “elliptical sentences” do not embed the way they should; (b) in some cases where Merchant requires fronting to take place, it is blocked – either by an island (e.g., in English) or because nonsubject fronting is not allowed in the language in question (e.g., in Malagasy); and (c) his “limited ellipsis” strategy, allowing do it and this is __ to be licensed in discourse-initial position, is not general enough. The methodological problem is that, his protests to the contrary, Merchant’s view multiplies hidden structure without necessity.