Analogy items from the Scholastic Aptitude Test (SAT) were evaluated for differential performance by black and white examinees. Black and white examinees were first matched for overall SAT-V scores prior to conducting item analyses. A content and psycholinguistic analysis of 220 disclosed SAT analogy items from 11 test forms was performed. Regression analyses indicate that black examinees consistently perform differentially better than matched white examinees on the hard analogy items. However, for easy items, particularly those that involve "science" content, white examinees appear consistently to perform differentially better than matched black examinees. In addition, semantic relationships dealing with part/whole relationships in the item stem also contributed negatively to black examinee percent correct responses. Three variables (item difficulty, science content, part/whole relationship) together account for 30% of the variance between the two ethnic (black and white) groups. Of these three significant predictors, two are semantic (part/whole and science content) while the third (item difficulty) reflects a non-semantic factor. Several hypotheses are advanced to explain these findings. Appendix A lists scoring categories, and appendix B lists the variable names and presents a table of intercorrelations of 39 variables. (Contains 2 tables, and 18 references.) (Author/SLD)