A Conditional Inference Tree Analysis of Can and May in Korean EFL Learners’ Writings

This paper applied an analysis using conditional inference trees to the corpus data of native and non-native speakers and examined the behaviors of some of linguistic factors in the data sets. For this purpose, the corpus data in Lee and Yu (2017) were taken. In the corpus data, all the sentences with two modal verbs can and may were extracted from two corpora (the ICE-USA corpus and the Korean component of the TOEFL11 corpus) and twenty linguistic factors were annotated. This paper applied a conditional inference tree analysis to the annotated corpus data so that it can be examined how differently some of linguistic factors behaved in two different types of data sets. Through the analysis, the following findings were observed: (i) the factors Sense, Mood, AnimType, and SubjMorph were highly ranked in the (American) native speakers’ corpus, whereas the factors Sense, AnimType, VerbSem, and SubjMorph were highly ranked in Korean EFL learners’ one, (ii) although the factor Sense was located at the top of the conditional inference trees, their behaviors were different in the two types of corpora, and (iii) while the use of three types of Senses was fairly balanced in the native speakers’ English, that of the Korean EFL learners was biased. The analysis results imply that slightly different types of criteria are involved in the determination of choice between can and may in (American) native speakers and Korean EFL learners.