Studies of sex differences in test performance have identified the content of test items, the context in which they are set, and the constructs purportedly underlying the item as important variables influencing performance. Test content often reflects experiences that traditional social roles have discouraged women from exploring. Often, the item pool in an occupational area dominated by men tends to be based on male experiences or to use male referents almost exclusively. The same is true for roles in which females predominate, but the domain is narrower than that for men. Mathematics and science items presented in a context more familiar to men than to women have been found to be more difficult for women. In examining item content, it is essential to identify content that is irrelevant to the intended construct but that might elicit different responses to items from different groups. Further research is needed on the differences between judgmental and statistical methods of item bias detection and on the types of bias detected by different methods. (Author/GDC)