Prior research indicated that computer-mediated glosses had an overall medium effect on second language reading comprehension. This study investigated the effects of computer-mediated glosses on reading comprehension. It also investigated the predictive power index of the e-glosses use with regard to reading comprehension performance, from a stepwise multiple regression analysis. A quasi-experimental method, involving the nonequivalent, pretest-posttest design was employed for this study. Participants were randomly assigned to an experimental group (n = 17) and a control group (n = 21). Participants in the experimental group were provided with a computer-mediated dictionary condition, doing their readings with e-glosses available as hypertext links to unfamiliar lexicon. The control group accessed the same reading texts in textbooks without any additional dictionary or glossary explanations available to them. ANCOVA analysis showed that the instructional method integrating e-glosses in computer-based instruction effected significant improvement in the experimental participants' reading comprehension. Furthermore, Stepwise Multiple Regression Analysis suggests that the use of hypertext-rich e-glosses can be predictive of improved reading comprehension performance. Pedagogical implications of the study are discussed at the end.