| Summary: | 碩士 === 國立中央大學 === 學習與教學研究所 === 103 === While collocations have received substantial attention in second language research, the colligational details surrounding many of them have been neglected though they cause problems for many learners long after they have mastered the collocation. Colligation is a patterning more complex than collocation. For example, we should use spend time doing (not to do) something while we can use take time to do something. This thesis is an investigation of two data-driven approaches to learning English collocation-specific colligations.
The two data-driven approaches to learning these patterns differed with respect to whether the learner’s query results came in the form of concordance lines containing a key word in syntagmatic sequences and requiring the learner to detect collocations and colligational patterns from these exemplars or came in the form of patterns listed for the learner that showed both syntagmatic and paradigmatic relations which were in turn linked to concordance lines exemplifying each pattern separately. BYU-BNC was used for the first approach and StringNet for the second. Eighty-two students in college were divided into three groups, including one control group and two experimental groups: BYU-BNC Group in an example-driven approach and StringNet Group in a pattern-driven approach. Participants underwent a training session and an experiment session, as well as a pre-test, an immediate post-test, and a delayed post-test. A questionnaire containing ten Likert-scaled items solicited data from the experimental groups on three areas: (1) ease of use, (2) effect on learning, and (3) willingness to use and two open-ended questions on participants’ elicited perceptions of the two learning tools.
Based on the results reported in this thesis, the two different approaches caused different effects on learning collocation-specific colligation. The StringNet Group showed significantly greater gains than the Control Group in the immediate post-test and significantly greater gains than both the BYU-BNC Group and the Control Group in the delayed post-test. These results revealed that the syntagmatic and paradigmatic dimensions of a learning tool are both important in assisting learners to acquire collocation-specific colligations, and the colligational knowledge is likely to be so subtle that it was hard to attain the improvement immediately. On the other hand, both the qualitative and quantitative questionnaire show that StringNet was perceived by that group’s participants as more helpful, effective, and simpler to use compared with the BYU-BNC group’s perceptions of that tool. All in all, the findings suggest that it might be better for teachers to teach colligation using a pattern-driven approach with the DDL tools which show both the syntagmatic and paradigmatic dimensions.
|
|---|
