Research on ellipsis has been prominent in theoretical linguistics for several decades. One question that has attracted considerable attention is how apparently incomplete syntax serves to convey semantically complete messages. Consider, for instance, that questions are typically answered using ellipsis, i.e., with phrases in place of sentences (e.g., A: Have you talked to Mary Beth? B: For about five seconds). Such a phrase with sentential semantics can be analyzed either as a stand-alone phrase or as a fragment left behind by a sentential structure that remains unpronounced. Both options have been pursued in theoretical syntax (e.g., Ginzburg and Sag 2000, Merchant 2001, 2004, Culicover and Jackendoff 2005), and both are supported by a range of empirical facts. In this talk, I will situate my research in this landscape, focusing on evidence in favor of nontransformational analyses, i.e., those that treat elliptical utterances as stand-alone phrases.