I this thesis, a proposed computerized task-based speaking test (PTBST) was designed based on the communicative language ability model proposed by Bachman (1990), and its effect on the predictive power of the academic English proficiency exam (EPE) administered at the Middle East Technical University, Northern Cyprus Campus (METU NCC) as an English proficiency screening criterion was examined. Two important constructs for academic success were operationalized in PTBST not covered by EPE: the ability to synthesize written an aural stimuli into spoken responses and presenting the resultant synthesis orally in a comprehensible and fluent fashion. Correlation, regression, and factor analyses were conducted to explore the predictive power of PTBST and EPE as well as to examine the construct exclusiveness of these two tests. The results showed a better correlation of PTBST with the GPAs of non-engineering students compared to METU EPE. PTBST also showed to have a potential to exhibit the same results for engineering students. Moreover, an exploratory factor analysis showed that the PTBST measures a different construct compared to METU EPE, which justifies adding PTBST to it. Also, the results of this study corroborated the correlations reported for EPE and GPA in METU Ankara. In addition, two distinct types of synthesizing were found and showed a different pattern of correlation with engineering and non-engineering students. Finally, it was found that it might be a better practice to report proficiency tests with different weightings of its sections, giving higher weightings to the sections more relevant to engineering vs. nonengineering disciplines. Key words: English for academic purposes, language proficiency tests, academic achievement prediction studies, task-based speaking tests