A Bayesian approach for finding latent classes in data is discussed. The approach uses finite mixture models to describe the underlying structure in the data and demonstrate that the possibility of using full joint probability models raises interesting new prospects for exploratory data analysis. The concepts and methods discussed are illustrated with a case study using a dataset from a recent educational study on how teachers evaluate teaching concerning ethical awareness. The Bayesian classification approach has been implemented for the personal computer under the Linux operating system. It presents an appealing addition to the standard toolbox for exploratory data analysis of educational data. (Contains 4 figures and 21 references.) (Author/SLD)