Despite its common use, the term“craving”is controversial. Moreover, its measurement can be problematic, since craving is purely subjective and is highly influenced by setting and drug availability. Thus, one might expect patients in protective settings such as hospitals to experience little craving, despite potentially being at high risk to relapse upon discharge. In this study of 73 patients hospitalized for cocaine dependence, we developed a series of five questions to measure different aspects of cocaine craving: 1) current intensity, 2) intensity during the previous 24 hours, 3) frequency, 4) responsiveness to drug-related conditioned stimuli, and S) imagined likelihood of use if in a setting with access to drugs. We hypothesized that some hospitalized patients with low craving levels might believe their risk of relapse to be high in another setting. We then examined the relationship between inpatient craving levels and 3-month treatment outcome. Results revealed a high level of internal consistency, with Cronbach's alpha scores of. 82 to. 94 on daily measures; factor analysis revealed unidimensionality; and each of the five items showed significant decreases in craving over time. The question regarding likelihood of use in another setting consistently scored higher than other questions on current craving (p&.001). Predictive validity was assessed by examining 3-month outcome. Among the 37 patients who participated in a follow-up study, there were no statistically significant relationships between 3-month cocaine use outcome and any item on the craving scale. Neither inpatient craving levels nor predictions regarding future use correlated with actual 3-month outcome.