The main metrological characteristics of a French version of the 49-item addiction research center inventory (ARCI) were evaluated using data collected in three controlled studies in healthy subjects. An analysis of variance showed no study effect, so the three studies were pooled. The test-retest reliability coefficients after placebo evaluated by a Spearman rank correlation test were 0.64 (P < 0.0001) for subscale A, 0.49 (P < 0.0001) for subscale BG, 0.55 (P < 0.0001) for MBG, 0.58 (P < 0.0001) for PCAG and 0.27 for LSD (not significant). Using the same test, the test-retest reliability coefficients after amphetamine were 0.73 (P < 0.0001) for subscale A, 0.61 (P < 0.0001) for subscale BG, 0.71 (P < 0.0001) for MBG, 0.46 (P < 0.0001) for PCAG and 0.66 for LSD (P < 0.0001). In order to assess the predictive validity of the translated questionnaire, areas under curves were calculated from the ROC diagrams for the three scores, amphetamine (A), benzedrine group (BG) and morphine benzedrine group (MBG). Two criteria validity were used: the desire to take amphetamine another time and the discrimination of the allocated treatment (amphetamine or placebo). The calculated areas under curves indicated a good capacity of prediction of the three ARCI subscales (A, BG, MBG) for both criteria.