Investigation of grammatical and lexical errors in three different level Turkish EFL learners' spoken corpus

The importance of learner corpus studies is widely recognized in the fields of second language acquisition (SLA) and corpus linguistics since learner corpora offer valuable suggestions for second language instruction (Granger, 2009). Despite the importance of the speaking skill in English as a Foreign Language (EFL) contexts, the scarcity of time-series and longitudinal studies exploring learners' spoken interlanguage continues to pose problems to teaching and learning English. In accordance with this, the present study aimed to identify the grammatical and lexical errors in three different level Turkish EFL learners' spoken performance, track the progress of those errors over time, and offer instructional suggestions. For these purposes, the corpus approach (Leech, Myers and Thomas, 1995) was used and corpus-assisted error analysis was conducted using AntConc 3.5.8 (Anthony, 2019), a very popular corpus analysis software. In order to compile a specialized learner corpus, 105 participants who were studying at the English language preparatory program intended for English-major students were grouped into A2, B1, and B2 levels (according to Common European Framework of Reference) on the basis of their scores obtained from Oxford Quick Placement Test (OQPT). Afterwards, 105 participants in the spring term and 29 of these 105 learners who continued their studies in the summer term were audio-recorded, and oral data recordings (25 hours of recording) were then transcribed in order to create the specialized corpus called Yeditepe Spoken Corpus of Learner English (YESCOLE). YESCOLE had four sub-corpora: YESCOLE-A2 (A2 level learners' corpus), YESCOLE-B1 (B1 level learners' corpus), YESCOLE-B2 (B2 level learners' corpus), and YESCOLE-LONG (longitudinal corpus of A2, B1, and B2 level learners). After identifying the utterances in each sub-corpus as the basic unit of analysis, a linguistic error categorization scheme was prepared on the basis of the literature, and grammatical errors were analyzed with respect to syntactic, morphological and morpho-syntactic errors, while lexical errors were analyzed in terms of lexico-semantic and semantico-syntactic errors. In line with the error categorization, the transcribed data were error-tagged manually and made ready for analysis in AntConc corpus analysis software. The findings derived from the spoken corpus showed that young-adult Turkish learners of English in the research context committed more grammatical errors than lexical errors. The most common errors found in the spoken corpus were omission of articles and prepositions in the noun phrase under syntactic category. Besides, the participants at all levels had difficulty in plurality and third person singular –s errors under morpho-syntactic error category. Furthermore, YESCOLE-A2 sub-corpus showed that A2 learners had more errors related to verb-phrase and word order than B1 and B2 learners. Although significant differences were not found in terms of the percentages of errors across A2, B1, and B2 levels, developmental progress of errors across these levels also revealed improvement, backsliding, and stabilization patterns of errors. In addition, u-shaped development, which indicates backsliding and improvement of errors over time, was observed in lexical and morpho-syntactic errors at A2 level. Thus, the findings of this dissertation offer valuable teaching implications for university-level A2, B1, and B2 levels in Turkish EFL context as well as suggestions for further research on learner spoken corpora for researchers.