Research on the acquisition of L2 collocations has examined different instructional conditions under which these can be best learnt. However, none has investigated the efficacy of a corpus-assisted contrastive analysis and translation approach on learning non-congruent collocations despite the evident theoretical and pedagogical grounds in support of the integration. The current study investigates this integration and argues for its implementation.
Research during the past 20 years has indicated that natural language consists mostly of formulaic sequences or chunks (Erman & Warren, 2000). The mastery of formulaic sequences including collocations is believed to be an important component of successful language learning (Wray, 2000). Non-congruent collocations (collocations that do not have L1 equivalents) are evidently difficult to acquire (Peters, 2016), produce (Nesselhauf, 2005) and process by EFL learners (Wolter and Gyllstad, 2013). Very few experimental studies (e.g., Laufer & Girsai, 2008; Chan & Liou, 2005) have examined the instructional conditions under which non-congruent collocations can be best learnt, however. These studies argued for the efficacy of a contrastive approach (Laufer & Girsai, 2008) which entails interlingual comparisons with learners' L1 and for a pedagogical integration of corpus resources to develop collocational knowledge. Nonetheless, no empirical study has attempted to investigate the value of both approaches on the acquisition of non-congruent collocations. This study attempts to fill this gap. It examines 129 EFL Saudi learners as they learn 30 non-congruent adjective-noun collocations under three instructional conditions (non-corpus-assisted contrastive analysis and translation (CAT), the corpus-assisted CAT, and the corpus-assisted non-CAT) in addition to the control, no exposure, condition. After a pre-test of passive and active recall, the participants were exposed to the target collocations in six teaching sessions according to their experimental condition. Then, an immediate post-test and a three-week delayed post-test were administered to all. The results showed that the corpus-assisted CAT group significantly outperformed the other two groups on all the tests.
