Abstract
This study aims to explore the potential of oral reading tasks to establish learners' proficiency when compiling learner corpora. Informed by research on oral reading fluency, we selected a text containing a variety of linguistic features and submitted it to 68 English learners in Taiwan, who were interviewed for the construction of a large spoken corpus of L2 English across proficiency levels.
Their proficiency was rated by trained Common European Framework of Reference (CEFR; Council of Europe, 2001, 2018, 2020) raters and ranged from A1 to B2. The performances in the reading passage were analyzed for reading rate and reading disfluencies.
The relationship between reading measures and language scores was analyzed using spine plots, revealing a strong association between reading rate and language level. The number of disfluencies did not show a significant association with language level when all disfluencies were counted together.
However, when different types of disfluencies were treated separately, false starts were found to be associated with language level (even though the relationship was less clear than the one reported between reading rate and language level). The study demonstrates that including a carefully selected reading passage among the tasks when compiling spoken learner corpora may be an efficient way of collecting data relating to learner performance in speech.