To determine the effect of counselling on condom use by prostitutes.Cohort study.Field-based study in The Gambia.Thirty-one (12 HIV-positive and 19 HIV-negative) prostitutes.Post-test HIV counselling.Levels of condom use.Overall, counselling had no effect on condom use.Scarce resources should be directed towards providing condoms in bars rather than counselling.HIV infection is much more prevalent in prostitution than in the general population of the Gambia (26.2 vs. 1.7%, respectively, in 1988). The short-and long-term effects of counseling on condom use by prostitutes was examined by determining condom use in 29 prostitutes, 11 HIV-positive (group 1) and 18 HIV-negative (group 2), before and 1 month after counseling and in 31 prostitutes, 12 HIV-positive group 3) and 19 HIV-negative (group 4), before and 2-5 month after counseling. HIV-positive women were told that reinfection would increase their chances of developing AIDS as well as lead to transmission of the virus. 3515 sexual contacts were recorded before counseling, 1252 after 1 month, and 4949 at 2-5 months later. Group 1 women reported an 89.3% rate of use before counseling; 1 month later, 9 maintained their rate, 1 increased use from 56 to 81%, and 1 decreased use from 31 to 26%. Group 2 women had 88% use before and, in 11 of the 12 women, 90% use after counseling; the 12th woman reduced use to 66% and was observed to increase her alcohol intake significantly. In both groups, 4 increased use and 3 reduced use. In Group 3, condom use declined insignificantly from 89.3 to 83.5%. 4 women maintained 90% use, 6 reduced use, and 2 increased it. Group 4 women decreased use insignificantly from 88% before to 84.4% after counseling. 8 women maintained original levels of use, 4 increased, and 7 decreased (1 from 90% to as low as 17%). In groups 3 and 4, 74% had 90% use precounseling; this fell to 69% at 1 month and 36% at 2-5 months after counseling, a significant decrease in usage. In all of the prostitutes, condom use increased in the first month after counseling and fell 2-5 months later. Counseling, therefore, failed to increase overall condom use. Possible reasons for this failure include the Senegalese backgrounds of many of the women (they were accustomed to health checks and free condoms); the low number of AIDS deaths in the Gambia, which keeps the disease a distant possibility instead of a pressing reality for the prostitutes; and the consideration of condom use as primarily a method to prevent contraction of the traditional sexually transmitted diseases. When resources are scarce, it may be more important to provide free condoms in bars frequented by prostitutes than to provide counseling.