The aim of this study was to assess the comparative importance of a broad array of background variables in examinees' performance on the General Test of the GRE using analytical methods designed to disentangle the confounding among the various characteristics. The experimental sample consisted of the 3,145 examinees who took the General Test at the October 1989 administration. The examinee background characteristics analyzed fell into two broad groups: 1) initial characteristics (sex, ethnicity, parental education, geographical region of residence, and age); and 2) intervening variables (college-related characteristics--major, research university, public- private institution, college selectivity, Ph.D. productivity ratio--and college-related performance-- undergraduate GPA). The General Test scores analyzed were Verbal Quantitative, and Analytical. Conclusions include: 1) the relationships (direct, indirect, and total associations) between the examinees' initial characteristics and their performance on the General Test were generally modest. Among these variables, parental education generally had the most consistent and strongest relationships; 2) the intervening variables partly mediated the associations for the two strongest initial characteristics of the examinees: parental education and sex; 3) the intervening variables generally had stronger and more consistent relationships with performance on the test than did the examinees' initial characteristics; and 4) the present results are broadly consistent with previous findings. It is further concluded that additional research needs to go beyond these social categories to the underlying psychologically relevant processes that produce the observed associations with the intervening variables and test performance, and that intervening variables of this kind need to be taken into account in future investigations of group differences in test performance.