This study investigated the intelligibility of Chinese and American graduate students to their Indian, Chinese, Korean, and American peers. A psycholinguistic word-recognition-in-noise study investigated the effects on intelligibility of speakers’ L1 and segmental pronunciation accuracy and how this varied by listeners’ L1 and word familiarity. Participants included 6 male graduate students (Chinese & American) as speakers and 72 male and female graduate students (Indian, Chinese, Korean, & American) as listeners. Since there was no speech corpus publicly available that would help answer the research questions, the researcher first compiled the Buckeye GTA Corpus, which includes L1 and L2 speech recordings of Chinese, Indian, Korean, and American university students. A series of logistic regression mixed-effects models revealed that speaker L1, listener L1, and word familiarity were significant predictors of intelligibility. Speaker segmental accuracy did not significantly predict intelligibility at the “TA-certified” proficiency level. Consequently, in the academic setting, attention should be paid to increasing both international and American listeners’ discipline-specific vocabulary. In addition, both international and American university students should receive the linguistic perception training necessary to accommodate the range of accent diversity that has become a reality of academic settings today.