A 21-item measure of gambling abstinence self-efficacy (GASS) was developed. A principal component analysis of 101 pathological gamblers supported the use of a total score that showed good internal and retest reliability as well as four subscales: 1) winning/external situations; 2) negative emotions; 3) positive mood/testing/urges; 4) social factors. The total and subscales showed moderate relationships with single item ratings of confidence to abstain from gambling and weak or non-significance relationships with demographic and gambling-related variables. The total score and three of the subscales showed evidence of predictive validity for gamblers not currently involved with treatment. Higher self-efficacy was related to fewer days of gambling over a 12-month period. These results provide preliminary support for the reliability and validity of the GASS.