Recent developments in dichotomous factor analysis based on multidimensional item response models (Bock and Aitkin, 1981; Muthen, 1978) provide an effective method for exploring the dimensionality of questionnaire items. Implemented in the TESTFACT program, this "full information" item factor analysis accounts not only for the pairwise joint frequencies of correct/incorrect responses, but also for additional information in higher order joint frequencies in the sample of dichotomously scored items. This paper illustrates this method's utility by analyzing a questionnaire on affective outcomes of schooling using the expected a posteriori (EAP) method of estimating ability scores. The 40 item questionnaire with four subscales (Punctuality, Honesty, Cooperation, and Curiosity) was developed to measure a set of potential outcomes of schooling's "latent curriculum". A stepwise full-information item factor analysis was performed on data from 700 elementary school students which identified three factors: interpersonal relations, active participation, and studious attitude. The advantages of using EAP scores over the use of raw scores to simplify the interpretation of multivariate analysis of variance where there are fewer concepts to discuss are illustrated. The full information factor analysis is strongly recommended for construct validation in the initial preparation stage of item construction for psychological measuremnt. (BS)