The influence of question format on item responses is examined using split ballot techniques and sex-role stereotypes in a junior college sample. Two formats were used: direct, in which questions asked only about women, and comparative, in which items explicitly compared women with men. Significant differences by format occurred on seven out of eight stereotypes. Students rated women more favorably when direct rather than comparative steretoype items were used. However, relationships between stereotypes and other variables were similar, regardless of format. Explanations for these results and implications for attitude research are discussed. (Author)