Determined personality prototypes in a population-based sample, and validated them in a sample of prison inmates. Cluster analyses were performed using personality traits based on the Big-Five model in two studies. A representative sample of 1,908 German adults aged 18-96 years and a sample of 256 prisoners (aged 15-35 years) completed the German version of the NEO-Five Factor Inventory (NEO-FFI). A two-step cluster procedure, using the Ward followed by the k-means method, yielded five clusters in the population sample: resilient, overcontrolled, undercontrolled, confident, and reserved personality types. The first three were comparable to previously identified temperament-based prototypes. The five-cluster solution was confirmed in the inmate sample, but only when clusters were determined using an approach based on discriminant functions derived from the representative sample. This procedure reduces within-cluster heterogeneity by excluding individuals with unclear trait configurations (residuals) from further analysis. These algorithm-based prototypes were related to demographic variables, sentence length, drug use, family environment, and childhood delinquency. In contrast, the sample-based approach yielded neither the expected personality prototypes nor their expected pattern of relationships with sociodemographic variables. A five-prototype conception of personality is therefore proposed. Furthermore, the algorithm-based approach is concluded to reproduce more valid personality prototypes than the sample-inherent derivation.