With the exception of anxiety scores, perceived neighborhood safety was the most important predictor of depression scores in the model used, which included well-established individual-level predictors of depression, such as gender, history, of sexual abuse, and alcohol dependence. Significantly higher depression scores were found for women and for those with histories of childhood sexual abuse and physical abuse. Additional research is needed in order to clarify the link between perceived neighborhood safety and depression. Using street outreach methods, the sample of 443 out-of-treatment African-American crack users was recruited in Raleigh, NC between August 2000 and August 2002. In order to be eligible for the study, potential recruits were required to indicate at the initial contact that they were interested in substance abuse treatment and had the intention to reduce or stop substance use within the next 6 months. Other eligibility criteria were self-identification as African-American, being 18 years old or older, no formal treatment for substance abuse within the past 90 days, either having a positive urine test for crack cocaine or self-reporting its use on at least 13 of the past 90 days, and using crack more often than injecting-drugs. Perceived neighborhood safety was assessed with the following yes-no items: feel safe sleeping at night, walk in the neighborhood at night, and walk in the neighborhood in the daytime. In order to measure depression, five yes-no items were adapted from the depression scale of the Global Appraisal of Individual Needs. 6 tables, 63 references, and appended table of measures