This study examined the prevalence of unprotected sex, other sexual risk behaviours, and factors associated with unprotected sex among men and women recently diagnosed with HIV in South Africa. One hundred and forty-nine outpatients (44 males and 105 females) were assessed, of whom 101 were sexually active at least 6 months prior to study entry. Subjects were asked about sexual risk behaviours with reference to their most recent sexual encounter. Logistic regression analysis was employed to determine the predictors of condom use, with independent variables selected from five general categories: (1) sociodemographic characteristics; (2) situational characteristics regarding sexual intercourse (i.e. alcohol or drugs used before intercourse); (3) clinical diagnoses; (4) negative life events; and (5) coping styles. Fifty-five patients (19 males and 36 females), representing 54.4% of those sexually active in the 6 months preceding the study, had not used a condom during the most recent intercourse. Compared with those who used condoms, participants who did not significantly reported shorter duration of HIV infection (t = -2.7, p < 0.001), have a current partner (chi2 = 3.98, p = 0.005), and lack knowledge of their partner's HIV status (chi2 = 4.78, p = 0.004). Also they were significantly more likely to engage in denial (t = 3.2, p < 0.002) and to use substances (t = 1.98, p < 0.05) as a means of coping. Logistic regression showed that shorter duration of illness (odds ratio (OR) = 1.2, 95% confidence interval (CI) = 1.01-1.41) and coping styles characterized by denial (OR = 0.6, 95% CI = 0.45-0.96) were significantly associated with unprotected sex. These data suggest the need for interventions to further reduce sexual risk behaviours in HIV-positive patients in South Africa.