Teaching collocations through data-driven learning: Comparison of two approaches

The present study aimed to investigate the effects of two data-driven collocation learning approaches (Corpus Consultancy and Practice on English-Turkish Parallel Texts) on participants' receptive and productive collocational knowledge of form, use and meaning. The study employed a quantitative research design; the data were collected through a vocabulary size test, a vocabulary knowledge scale, and a receptive and productive knowledge tests with a total number of 43 participants (N 14 in the Web-based group, N 16 in the Parallel Texts Group, and N 13 in the Control group). The Corpus Group received training on using a corpus to find and induce the meaning of the target collocations (10 adjective-noun and 10 verb-noun) through the COCA corpus. On the other hand, the Parallel Texts Group studied a small corpus consisting of English extracts containing target collocations taken from COCA corpus side by side with their L1 translations. The Control group, however, was expected to find the meanings of the collocations by resorting online dictionaries. The perceptions of the participants on both experimental approaches were also elicited via a structured survey consisting of open-ended questions. The results showed that The Corpus Group outperformed the Control Group both in both receptive and productive tests, while the Parallel Texts Group's scores remained to be in between in most cases. The participants in the Corpus Group and the Parallel Texts Group shared their perceived benefits and drawbacks of the approaches.