THE GERMAN INTELLIGENCE-TEST I-S-T-70 - A NALYSIS AND EVALUATION

The Intelligence Structure Test I-S-T 70 Amthauer (1970) was analysed in a study with 996 applicants for training programs in various skille d trades (school diploma: Hauptschule N = 205; Realschule N = 394; Gym nasium N = 397). Reliabilities (split-half and Cronbach's alpha), vali dities (correlations with CFT 2, dictation, simple arithmetic, dexteri ty test) and item characteristics were calculated. Factor analyses of the subtests were conducted and the dependence of the test scores on f orm (A and B), sex and education was checked. Forms A and B of the sub test word choice were not parallel. The reliability for form A was una cceptably low. Item analyses showed, especially in this subtest and pa rticularly in form A, many items with very low, sometimes even negativ e, item-test correlations. Distractor analyses of items with particula rly low item-test correlations showed that in some cases a distractor was preferred and correlated higher with the test than the ''right'' a nswer. Subgroups of items from the cube-tasks defined by different pro cessing strategies (''Surface cubes'' and ''Three-dimensional cubes'') showed differential validities. The ''Three-dimensional cubes'' prove d to be the better items. The I-S-T 70 correlates relatively high with the CFT and low with the dexterity test, as expected. Multiple regres sions showed that the subtests have differential validities and that a combination of subtests often correlates higher with an external test than the overall test. The factor analyses of the subtests in combina tion with the reliabilities speak for restrictions of profile analyses . In sum, the test must be revised, but is worth revising. Suggestions are made for a revision and the user is advised how to deal with the present shortcommings.