Syntax literature reports that resumptive pronouns (RPs) ameliorate island violations, but much psycholinguistics literature has found RPs to be no more acceptable than straightforwardly island-violating gaps, even though island production tasks consistently elicit RPs. However, psycholinguistic studies have typically compared RP and illicit gap conditions indirectly. We posit that RP island amelioration in comprehension is undetectable when participants cannot compare alternative sentences, and thus that the apparent production/comprehension split arises from methodological differences between perception and production experiments.
We present six experiments crossing three island types in two tasks (full-sentence forced choice and forced-choice fill-in-the-blank), manipulating gap location (island vs. nonisland). We find that RPs are preferred in islands and gaps in nonislands (p = .0001). This suggests that RPs do ameliorate island violations and that the production/comprehension split is a methodological artifact.