Evaluation of level of foreign language proficiency based on eye movement data

Abstract

Subject of study. Foreign language proficiency was evaluated using the optical recording of eye movements. Representation of unconscious cognitive processes in oculographic data during reading and correlation of eye movement parameters with the complexity of the text perceived by the reader were analyzed. Aim of study. The study was aimed at determining the correlation between the eye movement features of individuals reading texts in native and foreign languages, considering the peculiarities of the language and level of language proficiency. Method. The study comprised two stages. The total sample size was 63 Russian-speaking students, aged from 19 to 27 years, for whom English was a foreign language. The subjects’ level of proficiency in the Russian and English languages was estimated using the C-test method. The experimental task consisted of reading the English texts in the slides and answering the comprehension questions after a nine-point calibration procedure. Eye movements were recorded binocularly using an SMI-High Speed Tracker 1250 setup at a sampling rate of 500 Hz. The first stage was aimed at investigating eye movement features of native Russian speakers who are highly proficient in the English language while reading texts in both languages. In the second stage, the eye movement features of individuals reading texts in a foreign language at different levels of proficiency were investigated. Main results. The eye movement parameters of the individuals varied, depending on whether they were reading texts in their native language (Russian) or the foreign language (English), even at high proficiency in the foreign language. The eye movement patterns changed when they were reading texts in the foreign language (English) at different levels of proficiency. Practical significance. Based on the obtained data, the text complexity for the test subject can be evaluated. The text complexity evaluation using eye movement features enables algorithms to be designed for normalizing different texts in terms of complexity. The results of the study can be used when designing automated systems for evaluating foreign language proficiency level.