The acquisition of parameterized grammars

In this paper, I look at some problems that arise if second language learners are assumed to acquire the L2 on the basis of unconstrained hypothesis testing, since this crucially involves the assumption of the availability of negative evidence and does not seem to be sufficiently 'fine-tuned' for the learner to acquire certain aspects of the L2 grammar. Instead, it is proposed that L2 acquisition, like L1, may involve the prior knowledge of a number of highly contrained principles of Universal Grammar which restrict the number of options available to the learner. In some cases, these principles may be subject to parametric variation, so that they do not work in identical fashion in L1 and L2, allowing for the possibility of transfer of the L1 parameter setting. Taking subjacency as a case in point, a pilot study is reported in which native speakers of French and Spanish learning English as a second language were tested for their judgements on the bounding status of S in the L2 grammar. In English S is a bounding node, whereas in French and Spanish it is not. It was found that subjects were less accurate in their judgements on the bounding status of S in English than native-speaking controls, with a number of subjects consistently transferring the L1 parameter. In most cases, subjects showed improvement with increasing level of ESL proficiency. The majority of subjects adopted a consistent position with respect to the question of the bounding status of S, suggesting that they were not indulging in unconstrained hypothesis testing.