The relationships between examinee background characteristics and performance on the GRE General Test were appraised by a structural equation modeling analysis. The examinees' initial characteristics (sex, ethnicity, parental education, geographic region, and age) had modest relationships with their test performance. Of these, parental education had the most consistent and strongest association. Sex also had an appreciable association, but it was limited to the quantitative score. College-related characteristics (college major and the institution's public vs. private control, Carnegie classification, selectivity, and Ph.D. productivity) and undergraduate grade-point average (GPA) generally had stronger and more pervasive relationships with test performance than did the examinees' initial characteristics, not only mediating the associations of the examinees' initial characteristics with test performance but also making independent contributions in their own right. The associations were especially strong for school quality (a composite of public vs. private control, selectivity, and Ph.D. productivity), college major, and undergraduate GPA.