A Rasch Model for Detecting Learning While Solving an Intelligence Test

A dynamic extension of the Rasch model (Verhelst & Glas, 1993, 1995) is developed from a Bayesian point of view, and it is shown how this permits application of the model in a wide variety of test settings. In particular, the method allows for an adequate modeling of learning throughout a test, determining whether learning has occurred and whether individual differences in learning rate should be assumed. An example is provided in which the model is applied to a computer-administered intelligence test. A satisfactory fit of the model was found for these data. Results indicated that learning did occur, and that there might be individual differences in learning rate.