Sexual preferences of 38 rapists were assessed phallometrically with and wi
thout a semantic tracking task in a counterbalanced design. Four categories
of audiotaped vignettes describing neutral interactions, consenting sex, r
ape, and nonsexual violence were employed as stimuli. In the semantic track
ing task, participants were instructed to press one button when violent eve
nts were described in the vignette and another when sexual activities were
described. Phallometric assessment with the semantic task better discrimina
ted between rapists and non-sex-offender participants (from an earlier stud
y) than the same assessment without the task. Among four rapists who had pr
evious experience with phallometric testing, there was a very strong correl
ation between deviance scores and tracking accuracy. Results suggest that t
he semantic task may improve discriminant validity, particularly among sex
offenders who have had previous experience with phallometric assessment.