Objective: Analysing reliability and validity of the widely used 20 Questions of Gamblers Anonymous (GA20). Basic Research Design: GA20 and South Oaks Gambling Screen (SOGS) questionnaires administered under supervision and tested for validity and reliability measures. Participants: 127 problems gamblers (mean age 41.67, SD 12.58; 79% men; 5 years mean problem duration) who were to start or had recently started treatment at one of two self-help associations and a sample of 142 non-pathological or social gamblers who were paired with problem gamblers by age (mean age 38.5, SD 12.96) and sex (60% men). Results: Reliability was very high using Cronbach's alpha (0.94) and Spearman-Brown (0.95). Convergent validity, measured as correlation between GA20 and SOGS was very high at 0.94, which is to be expected due to the similarity between both screens. Construct validity, measured by factor analysis, revealed one important dimension which explained 50.6% of total variance. Two other dimensions explained much less variance (6% each). Only three questionnaire items do not significantly contribute to the first factor. Discriminatory validity as ability to differentiate between problem and social gamblers was measured comparing the answers of both groups. The differences in results were found to be markedly significant at p<0.0001. Diagnostic efficacy at the GA cut-off point of seven affirmative answers was also found to be very high, with sensitivity (percentage of true problem gamblers indicated) at 98.4% and specificity (percentage of non-problem gamblers identified) at 99.3%. Conclusions: Comparing reliability and validity with that of SOGS, GA20 is no doubt as good as the best clinical and diagnostic instrument proposed at present. It is reliable enough to be used as a diagnostic instrument, produces less false positives than SOGS and is easy to administer. It shares several critical points with SOGS, like having affirmative answers point in the same direction on all items and not being time-period specific. Comparison to other assessment instruments is difficult due to lack of psychometric studies of them. Recommendation: More attention should be paid to the qualitative information provided by self-help associations.