In pronunciation, segmental accuracy has been recognized as an important aspect in contributing to a non-native speaker's intelligibility. Nonetheless, there has been a lack of research focusing on the role of segmental errors in understanding extended discourse. Furthermore, previous research studies on intelligibility have largely been conducted in a controlled condition where a listener's cognitive process is more limited than in a real-life setting. In addition, proficiency level has not been considered as one of the factors contributing the intelligibility of non-native speech. This thesis uses a think-aloud methodology to investigate how native English speakers perceived how segmental errors contributed to reduced intelligibility of academic discourse produced by three Korean speakers with varying oral proficiency.;Five native American English listeners watched the teaching demonstrations performed by the three Korean speakers of English who had been rated as being at the advanced, intermediate and beginner levels or oral proficiency in English. While listening, the native speakers paused whenever they encountered a communicative breakdown and described the nature of the breakdown in understanding. Both quantitative and qualitative analyses were conducted. Listener results for the numbers of communicative breakdowns, and the numbers of the locations where communicative breakdowns were associated with segmental errors, were compared. The types of segmental errors identified by the listeners were analyzed to determine which errors appeared to impact intelligibility the most.;The findings showed that the listeners had communicative breakdowns for different reasons depending on proficiency levels. For the lower-level speakers, listeners stopped most often for segmental errors, whereas they stopped for non-phonological reasons as often as for segmental errors while listening to the advanced speaker. Also, not all segmental features were equally important for the listeners in this study. The consonants in syllable final position seemed to be important for understanding extended discourse on an academic topic. The comparison between the intermediate and beginner speakers suggested that not only was pronunciation important, but also that non-pronunciation related factors were important in being perceived as an effective speaker in an academic context. For nonnative speakers of English, and for Korean speakers of English in particular, a pedagogical approach which prioritizes certain segmental features depending on proficiency level is suggested.