BACKGROUND CHARACTERISTICS OF EXAMINEES SHOWING UNUSUAL TEST BEHAVIOR ON THE GRADUATE RECORD EXAMINATIONS

We ordinarily expect item difficulty to be related to errors on tests; examinees generally tend to make errors on more difficult items and to answer easier items correctly. However, some examinees miss easy items and get more difficult items correct. The extent to which correct and incorrect responses are predicted by the difficulty of test items has been quantified in various ways. One method, originated by Sato (1975) and modified by Harnisch and Linn (1981), was used in the present study of item level data from the Graduate Record Examinations General Test. The modified Sato caution index was found to be of low reliability in these data and to be generally unrelated to a variety of background variables, although ethnic group showed a small but significant relation to the index. In these data the modified caution index showed a curvilinear relation to total test score, with examinees who showed very high or very low scores having higher index values than those in the middle range of test scores. Indexes calculated on the three sections of the test were uncorrelated with each other. Finally, the index did not moderate the relationship between test scores and self‐reported grades, which it should have done if it indeed indicates how well the test measures the intended construct for any individual. The conclusion reached is that the modified caution index adds little information that would be of value in interpreting GRE test scores.