A STUDY OF PSYCHOLOGICALLY OPTIMAL LEVEL OF ITEM DIFFICULTY

For the purpose of selecting items in a test, this study presented a v iewpoint of psychologically optimal difficulty level, as well as measu rement efficiency, of items. A paper-and-pencil test (P and P) compose d of hard, moderate and easy subtests was administered to 298 students at a university. A computerized adaptive test (CAT) was also administ ered to 79 students. The items of both tests were selected from Shiba' s Word Meaning Comprehension Test, for which the estimates of paramete rs of two-parameter item response model were available. The results of P & P research showed that the psychologically optimal success level would be such that the proportion of right answers is somewhere betwee n .75 and .85. A similar result was obtained from CAT research, where the proportion of about .8 might be desirable. Traditionally a success rate of .5 has been recommended in adaptive testing. In this study, h owever, it was suggested that the items of such level would be too har d psychologically for many examinees.