A Corpus-based Error Analysis of Spoken English Produced by University Students

Abstract

The present study developed a corpus of spoken English produced by Korean EFL learners to identify its characteristics, types of errors and collocational errors. The spoken corpus was compiled of interviews with 134 Korean university students. Descriptive statistics were computed, and errors were identified and then classified accordingly using a qualitative approach. The study identified a total of 33,306 tokens and 2,308 types of words in the spoken corpus. In examining the frequency of all the individual words (types) that made up the corpus, it was discovered that almost 90% of the word families belonged to the first 1,000 most frequent English words native speakers use. With respect to error analysis, omission was the most frequent type of error, with misformation in second place and addition in third. In terms of collocational errors, G8-D (verb + preposition + object) and L1 (verb + noun/pronoun) were the most frequent types of grammatical and lexical collocation errors in the spoken corpus. Some implications are drawn concerning types of errors, including collocational errors, with respect to improving English learning and teaching as well as materials’ development.