The acquisition of complex wh- questions in the L2 English of Canadian French and Bulgarian speakers: Medial wh-constructions, inversion phenomena, and avoidance strategies

This dissertation examines the development of long-distance wh- movement questions in the L2 English of (Canadian) French and Bulgarian speakers. The main phenomenon under investigation is medial wh- constructions (wh- scope marking and wh- copying). Such constructions are of particular interest because they are unattested in both the L1 and the L2 of the two learner populations; at the same time, they are licensed options in a number of other typologically distinct languages, of which the participants report no knowledge. As such, medial wh- constructions pose a learnability problem in L2 acquisition: how can a learner "know" something that is not supported by either the native language or the target input, but is attested in other languages? Two experiments, a written grammaticality judgment multiple-choice task and an oral elicited production task, were carried out with the two different learner populations and with English native speaker controls. The written experiments showed that medial wh- constructions co-exist and compete with the target English long-distance structure at the early and intermediate stages of acquisition of both the French and the Bulgarian speaking participants; at the advanced stages of acquisition, both populations showed evidence that medial wh- representations had been successfully eliminated from the interlanguage grammar, and the L2 data converged with that of the native speakers. In the oral elicitation experiments both the French and the Bulgarian speaking participants resorted to medial wh- and a number of other strategies aimed at avoiding long-distance wh- movement; I argue that such strategies are due to both the derivational complexity and the high processing load associated with long-distance wh- movement. The account developed to address the findings of the dissertation incorporates insights from both nativist and domain-general views on acquisition. The proposal is that L2 grammars have to be UG-constrained in order for the learnability problem to be resolved. In addition, the acquisition process has to be strongly driven by the input, allowing learners to make extensive use of a general probabilistic learning mechanism; this mechanism helps them to gradually eliminate the competing representations unsupported by the L2 input and to converge with the grammatical target. This approach is in principle applicable to both L1 and L2 acquisition and accounts for some relevant similarities between the two