Beyond lexical and syntactical complexities, the pragmatic meanings of English modal verbs in discourse can be recognizably difficult (Coates 1983, Hinkel, 2001). The use of modal verbs is further complicated by the fact that a choice of a modal verb does not always permit a single pragmatic interpretation. Although the previous studies have contributed to an understanding of the rhetorical functions common to L2 writing can be achieved (Hinkel, 1995c, 199b, Hyland 1998, Maynard 1993), most of them have focused on the frequency count of modal verbs used in L2 writing. This study has thus two phases of empirical research with one sample group of 50 Korean EFL students enrolled at a MA programme in a single university: The first phase is a corpus-based analysis of two different students' academic essay samples: the corpus of the target group was taken from the Korean EFL university students (113,459 tokens) and the reference corpus was constructed from 50 L1 writers' academic essay samples extracted from a corpus of British Academic Written English (BAWE)(124,589 tokens). Secondly, a series of recall interviews with 15 Korean MA students out of the sample group were undertaken to shed light on the corpus data. The ultimate goal of this study is to examine specific patterns in their use of modal verbs and diagnose problematic areas of Korean EFL learners in written academic discourse. It thus compares and contrasts the frequency rate of a range of modal verbs classified in the there groups of modal verbs from Hinkel (2001)'s taxonomy. A detailed contextual analysis identifies the functions that the writers use modal verbs to perform in terms of epistemic and root meanings. The recall interviews were designed to provide richer and supportive data by seeking their thought patterns and thus explore how far their L1 transfer or developmental factors can affect the learners' choice of modal verbs. The results from the yield several important data. first, a range of modal verbs including semi-modals occur more frequently in the Korean student corpus than in L1 students' corpus. Although there are differences in the frequency rate among the different semantic classes, this does correspond to the results from several other studies (Hinkel 1995, 1999) which revealed that the overall frequency of a list of modals expressing either possibility and ability or obligation and necessity in Korean, Japanese and Korean texts contained significantly was higher than are native speaker student ones.