Using data collected from 218 street drug users in 11 cities in the United States, we examined the test-retest reliability of the alcohol-related items on the Risk Behavioral Assessment (RBA; National Institute on Drug Abuse, 1993), an instrument commonly used in drug abuse research. With a 48-hr retest interval, findings indicated excellent to good reliability for the following variables: age of first use, ever used alcohol, and days used alcohol in last 30 days. Items with fair to poor reliability were number of occasions used alcohol in last 30 days without injecting and number of times used alcohol immediately before or during sex. These findings suggest that self-report items on alcohol use from the RBA are generally reliable, particularly items that ask for more general, rather than specific, information.