The short answer: Implications for direct compositionality (and vice versa): Online Appendices

Language 92.2, June 2016 s1 THE SHORT ANSWER: IMPLICATIONS FOR DIRECT COMPOSITIONALITY (AND VICE VERSA): ONLINE APPENDICES PAULINE JACOBSON Brown University These appendices address two additional arguments that have been given for the SILENT LINGUISTIC MATERIAL hypothesis (SLM; see in particular Merchant 2004). These arguments purport to show that a movement analysis of the fragment answer is necessary, and this in turn supports SLM. For if there is movement of the fragment, there must be surrounding material out of which the fragment moved. The first argument (Appendix A) concerns preposition stranding and has played a key role in much literature (especially on SLUICING, but many of the same points are relevant to fragment answers). The second concerns islands (Appendix B). I show that both cases are unconvincing, for two reasons. First, the actual facts are unclear. Second, even if the purported generalizations were correct, SLM does not actually account for the full range of facts. As to the account of the facts under Qu-Ans, here too there remain open questions, so I am not putting these domains forward as arguments for Qu-Ans either. The point is simply that both domains need to be better understood before any conclusion can be drawn from them. At this point, however , they certainly do not provide evidence for SLM. APPENDIX A: THE P-STRANDING GENERALIZATION One of the most frequently cited arguments for an SLM analysis of sluicing—and which has also been used to argue for SLM in fragment answers—is the so-called PREPOSITION STRANDING (hereafter, Pstranding ) generalization (Merchant 2001, 2004). The argument is intended to support not just SLM PER SE, but also the movement-plus-deletion/silencing account of SLM. I elucidate the argument in the context of fragment answers by first considering the situation in English, a language that allows P-stranding in WHquestions and other fronting constructions. Thus both 1a and 1b are possible questions. (1) a. To whom did Carly reveal her secret? b. Who did Carly reveal her secret to? Focusing just on 1a for the moment, either of the answers in 2 is also possible. (2) a. To Mick. b. Mick. (Notice that both versions of 2 are also possible answers to 1b, a fact that I put aside for now but return to below.) The crux of the P-stranding argument is that languages that do not allow P-stranding (and hence do not allow questions analogous to the version in 1b) allow only answers with a preposition, analogous to the English case in 2a; fragment answers parallel to the English answer 2b are not allowed. The claim is that this follows if fragment answers require movement followed by deletion/silencing; 2b could not be derived since the movement required to front just a single NP (as opposed to a full PP) would be blocked. In other words, using English words for the sake of illustration, the non-P-stranding language would not allow the fragment answer in 2b because this would covertly have to be a structure like 3, which would be disallowed in that language (the notation ⊗ is used to mean that sentences corresponding to 3 are bad in the class of languages that do not allow P-stranding). s2 (3) ⊗Mick, [FP Shh [Carly revealed her secret to t]]. Merchant 2004 illustrates this generalization by citing English, Swedish, Norwegian, and Danish all as languages of the English variety (P-stranding allowed, and fragment answers like 2b allowed as answers to questions like 1a). By contrast, Greek, German, Yiddish, Czech, Russian, Bulgarian, and Hebrew all do not allow P-stranding in questions, and similarly do not allow fragments like 2b as answers to questions such as 1a. (Additional data regarding the corresponding argument for sluicing is provided in Merchant 2001.) However, it is not at all clear what conclusion to draw from this domain, since the issues are murky for at least three reasons. First, assume that the generalization is correct. Then it is once again hasty to jump to the conclusion that 2b must COME FROM a full fuller sentence of the form in 3. Once again, the two could be bad for the same reason. After all...