To determine the reliability of serial naloxone challenges, five heroin addicts stabilized on methadone were given 0.2 mg naloxone intravenously on three consecutive days. Two-factor ANOVA revealed that the between-subjects effect accounted for at least 63% of the variance in each of the dependent measures, and the effect of challenge number for at most 22%. Withdrawal tended to be slightly more severe for the first challenge. Intraclass correlation coefficients were calculated for the area-under-the-curve of the change from baseline: subject-rated (.74) and observer-rated (.71) withdrawal, pulse (.88), systolic blood pressure (.58), diastolic blood pressure (.85), and skin temperature (.63).