The learnability of A-bar chains
This paper discusses the learnability of long <i>wh</i>-questions. Taking longitudinal data from child Dutch an example, I will show that the acquisition of long <i>wh</i>-movement has been thoroughly prepared in previous acquisition steps. Each step defines a local relation that is preserved in the next acquisition step. The long <i>wh</i>-questions appear first with an intermediate pronoun in the speech of the child. My data show that the intermediate pronoun relies on the (previously acquired) relative paradigm which appears in Dutch as an apparently irregular mixture of <i>d</i>-pronouns and <i>w</i>-pronouns. The present view on the learnability of A-bar chains will lead to the following conclusions: (i) Long <i>wh</i>-movement can (hence must) be successive cyclic; (ii) Pied-piping follows from the preservation of licensing contexts only; (iii) Syntactic islands need not to be “learned”. They follow from a non-overlap of local movement domains.