Objectives. Measures of acquired immunodeficiency syndrome (AIDS)-related knowledge that have good psychometric properties are needed to evaluate the impact of educational interventions, particularly among impoverished populations. Few measures that satisfy these requirements are available. The authors describe the psychometric characteristics and composition of a subset of 21 AIDS knowledge items from the questionnaire used by the National Center for Health Statistics. Methods. The study was conducted with a convenience sample of women and their significant others or close friends in nine homeless shelters and 11 residential drug recovery programs in the skid row area of Los Angeles from 1991 to 1993. Trained nurses and outreach workers collected data from 486 predominately African-American adults before delivery of well-established, culturally sensitive AIDS educational intervention. A subset of 334 participants were re-interviewed 6 months later. Results. A total scale score was calculated from a one-factor solution, and two subscale scores were formed based on a two-factor solution. One subscale was composed of primarily cognitive items, whereas the second subscale contained mainly transmission-related items. The internal consistency reliability of the overall scale was 0.89; the two subscales had alphas of 0.80 and 0.92. The measures have face validity, and support for convergent validity was found. Further, even though AIDS knowledge in the sample was relatively high, all three measures showed marked improvement in the subsample of respondents who were re-interviewed after receiving AIDS-related educational information. Conclusions. These measures show promise for assessing AIDS knowledge and the efficacy of AIDS education programs in vulnerable populations. The findings reported here, however, are preliminary. Test-retest reliability needs to be examined, and additional studies are needed with samples that are more culturally diverse and include socially isolated individuals. (Authors)