"Awareness-Reinforcement-Composition" Approach In The Production Of English Learners' Lexical Chunks

The study on lexical chunks has attracted the attention of researchers and instructors in the linguistic field since the beginning of the last century. So far, history has witnessed a rich field of studies on this phenomenon. This paper studied the application of the lexical approach, the "Awareness-Reinforcement-Composition" approach, in the author's words, in non-English majors' English classes, and examined its effect on the output of students' lexical chunks with the help of a new method to extract lexical chunks from the corpora which is named "kfNgram-deletion" method by the author. During the experiment which lasted for three units, over a period of two months, students at two English proficiency levels of the experimental groups were instructed in the"Awareness-Reinforcement-Composition" approach, while those of the control groups were not told the concept of lexical chunks but asked to write compositions when a unit was finished. The compositions were submitted, processed and put into a corpus. Lexical chunks of different lengths were obtained with the method of "kfNgram-deletion" and data were analyzed from three perspectives: in general terms, structural terms and functional terms.Through quantitative and qualitative analyses, the study has proved that "kfNgram-deletion" method is more suitable than the traditional methods to extract lexical chunks. There are a great number of overlaps and re-accountancy in the traditional ways of extracting lexical chunks. The lexical chunks obtained in the traditional ways are difficult to classify and analyze. On the other hand, lexical chunks obtained through the "kfNgran-deletion" method do not overlap and are easy to classify and analyze. This method is more suitable for this study and the author strongly recommends that the method be tried in other studies.What is more, the "Awareness-Reinforcement-Composition" approach does have some effect on the output of students' lexical chunks. Generally speaking, students from the experimental groups use far more types of lexical chunks than those in the control groups, no matter what English language proficiency the subjects have and how lengthy the lexical chunks are. Also students from the experimental groups can write longer lexical chunks than those in the control groups, and subjects of higher English language proficiency level do better than those of the lower ones. As the experiment proceeds, their performance on lexical chunks gets better. Structurally speaking, clause constituents take the largest percentage of all lexical chunks, followed by the percentage of full clauses, and incomplete phrases. But experimental groups produce about 15% to 18% of full clauses and 75% of clause constituents, while the control groups produce fewer full clauses (about 10%) and they rely more on clause constituents (about 80%). Taking full clauses into consideration, compared with the control groups, the experimental groups use more full clauses and their full clauses are much longer than that of the control groups. It reveals that students of experimental groups have raised their awareness of big lexical chunks and are more likely to produce lexical chunks as big as possible. Functionally speaking, students from the experimental groups do better than those in the control groups. They not only produce more time/place chunks, epistemic chunks, and focus chunks but also produce longer topic-specific ones. But neither of the groups do well in the production of quantity chunks. At the same time, we get similar results of both groups on the production of framing chunks. It indicates that students have already acquired this function of lexical chunks before the experiment. On the whole, the language proficiency level does not contribute much to the production of all kinds of lexical chunks.