The research agenda for DSM‐5 emphasizes the implementation of dimensional trait models into the classification of personality disorders (PDs). However, because assessment psychologists may still want to recover the traditional DSM‐IV categories, researchers developed a count technique that uses sums of selected Five‐Factor Model facets to assess the DSM‐IV PDs. The presented study examined the convergent and divergent validity of different linear combinations of trait facets to describe specific DSM‐IV PDs in a heterogeneous clinical sample (N = 155) with sufficient prevalence of all PDs, using semi‐structured interviews to obtain all diagnostic information, and comparing alternative counts from five different sources for each PD. The results show that none of the schizotypal, antisocial, and dependent counts succeeded in combining good convergent with adequate divergent validity. However, the original counts could be optimized for five of the seven remaining PDs by using alternative Five‐Factor Model prototypes. The diagnostic and taxonomic implications of these findings are discussed. Copyright © 2012 John Wiley & Sons, Ltd.