Bayesian and classical approaches to classification of vocational data were compared using an educational data set from a longitudinal study of professional growth and development in organizations (P. Ruohotie et al., 1994). Data were from 2,430 workers in companies in Finland who completed a questionnaire with behavior and background statements. The main purpose of this study was to look for new possibilities in analyzing multiform vocational data starting from the level at which traditional linear methods become too complex to apply. After describing the data and the theory of professional growth, the paper discusses linear discrimination and its use in the social sciences. Bayesian modeling with the BAYDA software package is described. It is concluded that linear and nonlinear methods support each other depending on the subject of the study. The Bayesian approach, in the form of the BAYDA program, is still under rapid development, but it appears to provide a valuable tool for analysis. (Contains 10 figures, 21 tables, and 12 references.) (SLD)