From the fall of 1981 through June 1984, more than 850 disabled examinees took special administration of the Graduate Record Examinations (GRE). Grade point averages were obtained on 278 disabled students; 236 had enough complete data to be included in the study. Disabled students earned lower mean GRE scores than their nonhandicapped counterparts but, except for visually impaired students, they earned overall graduate grade point averages very close to those of nonhandicapped students. Differences in the GRE performance of blind students compared to students with other visual disabilities raised questions for further research. The predictive validity of the GRE scores obtained in nonstandard administrations was estimated with empirical Bayes procedures. The differences between actual and predicted grade point averages were more negative for disabled students. The correlations between predicted and actual scores were more modest than results for nonhandicapped students. In addition a distinctive pattern was observed for disabled students: higher predicted five-year averages (FYA) scores were accompanied by increased overprediction. However, considerable caution was recommended in interpreting the results of the study because of the lack of information on nonhandicapped students in the same departments as disabled students. Appendices include a description of the use of Empirical Bayes methods and an annotated bibliography of reports from "Studies of Admissions Testing and Handicapped People." (Author/JAZ)