The purpose of this study is to examine the use of words and fixed expressions by intermediate-level EFL learners. It is widely recognized that learner corpora, such as ICLE (International Corpus of Learner English) or NICE (Nagoya Interlanguage Corpus of English), are used to reveal the characteristics of learner interlanguage. In SLA research, contrastive interlanguage analysis (CIA), proposed by Granger (1998), was performed to find a wide range of the features of language learners, including the overuse and underuse of words or phrases. CIA mainly focuses on the NS/NNS comparison and the NNS/NNS comparison. There were many cross-sectional learner corpus research studies, but few studies on longitudinal design have been completed so far. In the present paper, I would like to analyze the corpus, which consists of longitudinal writing data compiled by 93 intermediate-level EFL learners.

To conduct the present study, two corpora were compiled; one is the corpus of writing in which the topic was “a cellular phone" by intermediate learners (1EC: 1st Essay Corpus), while the other is the corpus of writing in which the topic was the “early introduction of English education" by the same learners (2EC: 2nd Essay Corpus). Using the N-gram model, the collocational phrases were extracted from each corpus, and I analyzed the similarities and differences between the two corpora.

The results showed that the expressions extracted from 1EC and 2EC have one similarity: the learners frequently used “topic-related" phrases (e.g., mobile phone(s), English education). On the contrary, there are two contrasting properties between 1EC and 2EC. In 2EC, the number of preposition types (e.g., in, to, of, at) increased. In addition, the learners frequently used phrases such as “I think" or “I think that" in 1EC, but they did not use these phrases as much in 2EC. In conclusion, this study clarifies that the collocations used by the learners were varied because of the influence of the topic or the instructions in the classroom.