A reversal design was employed for the analysis of transfer of fear and avoidance through equivalence classes. Two 5-member equivalence classes (A1-B1-C1-D1-E1 and A2-B2-C2-D2-E2) were established. Then B1 and C1 were paired with shock (CS+) and served as SD s in avoidance training (B2 and C2 were trained as CS-/S∆ s for avoidance). Further avoidance training followed with D1 and E1 (as SD s) and D2 and E2 (as S∆ s), with the first presentation of each of these stimuli serving as the first transfer test. Afterwards, aversive conditioning contingencies were reversed: B2 and D2 were paired with shock and trained as SD s for avoidance, B1 and D1 were presented without shock (CS-/S∆ s). Transfer was tested again with C1, E1, C2 and E2. This reversal was implemented to allow for the within-subject replication of transfer effects upon changes in the function of only a subset of each class's elements. Avoidance (key presses) and conditioned fear (skin conductance and heart rate) were simultaneously measured. Results show a clear transfer effect for avoidance, with between- and within-subject replications. For physiological measures, transfer effects in the first test could only be imputed on the basis of group-based inferential statistical analysis. Evidence for between-subject replication was weaker, with only a limited proportion of participants meeting the individual criterion for transfer.