Effect of grammaticality judgements on the relationship between metalinguistic awareness and second language proficiency.

This study investigated the relationship between metalinguistic awareness (MLA) and L2 proficiency of second language learners from an experiential (communicative) and an analytic (grammar) approach to learning French. The model tested (Bialystok & Ryan, 1985) posits that linguistic and metalinguistic proficiency are composed of two components--'analysed knowledge' and 'control' over that knowledge. All tasks (e.g., conversation, literacy, metalinguistic) require some degree of both components. Since the development of the components is the result of different learning experiences, learners may master aspects of proficiency related to an increase in analysed knowledge but not master ones which stem from an increase in control, or vice versa. This is problematic in that it is conceivable that learners may have a very different mastery of the language when task demands change. Advanced university level French L2 learners (n = 64) completed a French Test of Proficiency and a written and oral grammaticality judgement test. The grammaticality judgement task required subjects to judge sentences for grammaticality, to identify errors, to correct errors and to provide the rule which had been violated. The first level of analysis looked at the performance of the entire sample on the grammaticality judgement tests and on the test of L2 proficiency. Then further analyses were carried out in light of the demands tasks made upon the two components and with subjects grouped according to learning methods; that is whether they had been exposed to a communicative or grammar approach to L2 learning. Results show that there is a positive relationship between MLA and L2 proficiency when analyses are run for the entire sample, but that once learners are grouped according to learning methods, that positive relationship no longer exists for communicative-approach learners. Tasks which made similar demands did not always correlate. Significant differences were found in learners' performance depending upon whether the tasks required high levels of analysed knowledge, control, or both and depending upon the type of learning method. Implications for the model and for L2 learning conclude the study