PurposeThe idea that criminal behavior is a function of the offender's personality, also called the Homology hypothesis, has a long history in forensic psychology and criminology. This assumption, however, has been decried as lacking empirical support. In spite of much social concern relative to sexual offenses, there is virtually no research looking at the stability of offending pattern in sex offenders of adult women. This paper aims to fill some of the gaps.Design/methodology/approachLatent structure analyses were conducted on a secondary dataset including 145 serial rapists. A cross‐sectional, discrete time‐series design was used including a sequence of three offenses.FindingsModerate support was found for the three main assumptions underlying the Homology hypothesis. Offenses tended to share a relatively similar underlying structure, with the victimology and aggression components being more prominent than the sexual dimension. The three primary profiles identified, labeled “Passive”, “Stranger‐aggressive”, and “Antisocial”, were found to be about 50 percent stable across the sequence. Finally, the presence of significant dysfunction in the family of origin predicted membership in the “Antisocial” class, as well as increased the specificity and stability of this profile. The presence of early maladjustment was not related to any of the states.Research limitations/implicationsBased on the results, it is proposed that future research incorporates contextual‐environmental elements in order to increase the validity of the findings.Originality/valueThis study represents a unique attempt at documenting patterns of stability and variations across incidents of rape, using an institutional sample. Furthermore, it illustrates the use and potential benefits of latent structure models in criminological research.