ABSTRACT
In previous work 1 summarized in the Introduction to this book, the Level Embedding (LE) regime derives a number of correlations called the Locality-Reconstructivity-Target correlations (LRT) in Williams (2003). The locality correlations are of the form “the higher a position in F-structure that a movement targets, the larger the embedded clause it can extract something out of.” The locality correlations are not without content—among them is the prediction that the complement of subject-raising verbs must be at most a TP—but they do not seem to cover the full range of classic island effects, the kinds that fall under subjacency. In fact, subjacent movement is impossible in the model of Williams (2003), so in fact I have owed a general account of islands since then. This is it.
