Divorce Licensing: Separate Criteria for Predicate and Clausal Ellipsis

Current theories of ellipsis share the assumption that ellipsis is licensed by the presence of a licensing feature, [E], on a functional node in syntax (Lobeck 1995, Merchant 1999, 2004). This dissertation provides evidence from both Mainstream American English (MAE) and African American English (AAE) that the functional heads that license ellipsis must be phonologically overt, which is unexpected under current theories. AAE is particularly important to establish this generalization due to the fact that, although the pronunciation of auxiliary be and the possessive ’s morpheme is typically optional preceding a full predicate, new experimental evidence reported in this dissertation shows that this optionality disappears in elliptical contexts. This shows that predicate ellipsis can only be licensed in an Agree relation established between a phonologically realized functional morpheme and the lexical phrase it c-commands. I argue that the functional morphemes that license predicate ellipsis must be overt because the [E] feature is not present in syntax prior to the insertion of lexical items bearing this feature. Phonologically reduced, contracted and zero-marked forms are not appropriate licensors because only the full form of a lexical item bears the licensing feature, in accordance with the theory of Structural Deficiency (Cardinaletti and Starke 1994). Beyond predicate ellipsis in AAE and MAE, this new Agree analysis also accounts for crosslinguistic ellipsis phenomena such as Verb Stranding Verb Phrase Ellipsis in languages like Hebrew, Irish, Swahili and Portuguese. Ultimately, the analysis proposed here entails that predicate ellipsis, in which overtness is required, and clausal ellipsis, wherein the head said to license ellipsis is necessarily silent, are subject to different licensing conditions.