We assessed the validity of self-reported sex and condom use by comparing self-reports with prostate-specific antigen (PSA) detection in a prospective study of 210 female sex workers in Mombasa, Kenya. Participants were interviewed on recent sexual behaviours at baseline and 12-month follow-up visits. At both visits, a trained nurse instructed participants to self-swab to collect vaginal fluid specimens, which were tested for PSA using enzyme-linked immunosorbent assay (ELISA). Eleven percent of samples (n = 329) from women reporting no unprotected sex for the prior 48 hours tested positive for PSA. The proportions of women with this type of discordant self-reported and biological data did not differ between the enrolment and 12-month visit (odds ratio [OR] 1.1; 95% confidence interval [CI] 0.99, 1.2). The study found evidence that participants failed to report recent unprotected sex. Furthermore, because PSA begins to clear immediately after exposure, our measures of misreported semen exposure likely are underestimations.