Three groups of subjects received training in stimulus equivalence classes with a respondent-type training procedure. Stimulus contingency was positive for the first group, zero for the second, and negative for the third. Afterwards, all participants were tested with a matching-to-sample task for four baseline relations (A1-B1, A2-B2, C1-B1, and C2-B2), four symmetry relations (B1-A1, B2-A2, B1-C1, and C2-B2), and four combined symmetry and transitivity relations (A1-C1, A2-C2, C1-A1, and C2-A2). After this, they were retrained and retested twice. Explicit reinforcement was not programmed during the training or testing phases. The number of subjects who learned the baseline relations and showed the emergence of derived relations was higher in the positive contingency group than in the zero or negative contingency groups, although in all cases at least one retraining phase was required in order to reach the criteria. This finding contributes to the literature suggesting that stimulus pairing is the fundamental variable underlying emergent stimulus functions.