CONNECTED SPEECH AS A FACTOR OF LISTENING DIFFICULTIES IN DIGITAL DISCOURSE: A PILOT STUDY ON KAZAKHSTANI EFL LEARNERS

Abstract

The present study examines how Kazakhstani EFL learners with self-reported B1–C1 English proficiency perceive and decode connected speech in authentic digital discourse. Existing research relies mostly on scripted audio, therefore this paper uses unedited TikTok videos instead. Such content reflects real-world speech conditions more accurately. Sixteen third-year students of the “Foreign Philology” faculty transcribed six audio phrases which included different connected speech processes — elision, yod-coalescence, assimilation and lexical reduction. Additionally, each phrase was analyzed acoustically in Praat in order to check whether the process actually occurred in the signal, rather than being assumed from the transcript. Results showed that even advanced learners systematically had problems decoding a few phrases while still remaining confident in wrong answers. The most striking finding was that when several phonological processes overlapped in one phrase, the difficulty did not simply add up. Instead, it produced a completely different, non-existent word in the listener’s perception — and this was visible on the spectrogram itself. Perceptual failure turned out to depend on how deeply speech was phonologically reduced and on how familiar the listener was with the underlying word. By contrast, language proficiency level itself poorly explained these results. The study claims that university language programs need to bring authentic digital discourse into the classroom, in order to close the gap between what students are taught and what they actually hear in real life.