This study investigated the effect of diagram formats on performance on geometry items in order to determine whether certain examinees are affected by different item formats and whether such differences arise from the different intellectual demands made by these formats. Thirty-two experimental, multiple-choice geometry items were administered in two pretest versions; one version did and one version did not provide a relevant diagram. A minimal amount of editing of the item stem was sometimes required for the version without diagrams in order to provide the relevant information formerly provided by the diagram. Each experimental unit was administered as part of the American College Test Assessment pretest procedures to approximately 2,900 examinees in randomly equivalent samples; only one pretest unit was administered to any one examinee. The data were analyzed using log linear analysis. No significant differences between males and females were found either in item difficulty by item type or in item discrimination by item type. Traditional assumptions about gender differences in mathematics may need reassessment. Six data tables and eight sample test items are included. (Author/TJH)