Three major personality prototypes were derived on the basis of a Big Five instrument (NEO-PI) by means of replicated cluster analysis in two Spanish samples (a sample from the general population and a student sample). The replicability of the three prototypes within and their consistency between the two samples were evaluated. In addition, subtypes were analysed in a similar way. Finally, the relation between prototype assignment and level of education was examined in the sample from the general population. Within-study replicability was satisfactory only for the student sample. Comparison with the results for a similar instrument (NEO-PI-R) applied to a German sample showed satisfactory consistency only for the student sample. Discussion centres on the strong sample dependency of the results. Copyright © 2002 John Wiley & Sons, Ltd.