Schlenker, Philippe. 2003. "Clausal Equations (A Note on the Connectivity Problem)" (not the final version; the final version appeared in  Natural Language and Linguistic Theory 21: 157-214, 2003)

[Full paper in pdf]  

Abstract:  According to a variety of tests, What John likes is himself displays the same c-command relations as John likes himself. But none of these relations appears to hold on the surface; this is the 'connectivity problem'. Revisionists maintain that the problematic examples are identity sentences with no hidden structure, but that none of our c-command tests is infallible. Conservatives claim that our c-command tests are reliable, but that the clause John likes himself is indeed present at some level of representation. Siding with the Conservatives, we follow Ross's original insight and suggest that connectivity sentences equate a concealed question with an elided answer: [What John likes]= [John likes himself].  New arguments are given for each component of the analysis, and it is shown that connectivity effects are obviated when the elements that are equated are referential rather than clausal. The correct truth-conditions are derived from the semantics of identity, together with Groenendijk & Stokhof's semantics for questions. The analysis is then extended to cases of DP connectivity, such as His worry is himself, by suggesting that semantically dyadic nouns have an additional argument position, yielding the representation: [?x his [worry x]]=[his [worry himself]]. Finally it is shown that recent objections based on 'anti-connectivity' effects misfire, because the same facts hold of question-answer pairs, as is expected on the present approach.