Chlamydia trachomatis is the most prevalent sexually transmitted bacterial pathogen. Nevertheless, selective, rather than universal, screening for chlamydia has been recommended, largely because testing is expensive and requires considerable technical expertise. A total of 1,348 women in four family planning clinics in San Francisco, California, were screened from March 1987 to January 1988 to identify criteria for selective screening. Of these, 9.2% had a positive chlamydia test using direct fluorescence. Logistic regression analysis identified five factors associated with infection: age less than 25 years, cervical friability, single marital status, a new sexual partner within the past 3 months, and lack of barrier contraceptive use. No single risk factor or combination of risk factors had both a high sensitivity and a high positive predictive value for infection. While screening all women who were unmarried would detect 93% of those with chlamydia, the positive predictive value of 10.7% was not much higher than the overall prevalence. Conversely, screening all women with cervical friability, which had a positive predictive value of 23.2%, would only detect 11% of those with chlamydia. On the basis of the authors' findings, selective screening should not be used in high prevalence populations in which all women are at risk and should be screened for chlamydia.