This study analyzes sexual-preference profiles in a sample of 420 sexual aggressors who exhibited a valid profile during their initial phallometric assessment. The sexual stimuli used in the process were audiotapes describing sexual-offending scenarios. Two types of sexual stimuli sets were used: one developed for sexual aggressors against women and a second developed for sexual aggressors against children. Penile responses were recorded during stimulus presentation using a mercury-in-rubber strain gauge. Classification analyses (hierarchical and K-means clustering combination) were conducted separately for three groups of sexual aggressors: (a) sexual aggressors against children (n = 253), (b) sexual aggressors against women (n = 138), and (c) mixed sexual aggressors (n = 29). The sexual aggressors against children exhibited four penile-response profiles, the sexual aggressors against women two penile-response profiles, and the mixed sexual aggressors only one penile-response profile. In addition, analyses carried out on randomly split subsamples established that the generated profiles for sexual aggressors against children and sexual aggressors against women were stable.