This study examined the effects of a successive stimulus pairing procedure (respondent training) on formation of conditional discriminations and equivalence classes. Different training protocols (linear, many-to-one, one-to-many), and training and test arrangements (simultaneous, simple-to-complex) were used. A simultaneous protocol was used in Experiment 1. During training, adults were exposed to multiple random series of stimulus pairs. Stimuli of the same pair were presented one after the other (e.g., A1-->B1, C1-->B1, A2-->B2, C2-->B2, A3-->B3, C3-->B3). These series were followed by a match-to-sample test series involving symmetry probes (e.g., B-A, B-C) mixed with equivalence probes (A-C). Experiments 2 through 4 involved preschool children. Experiment 2 was a modified replication of Experiment 1 (Observing A-->B and C-->B. Testing A-B, C-B, A-C, and vice versa). Experiment 3 was the same except that a simple-to-complex protocol was used (e.g., training A-->B, testing A-B and B-A, training C-->B, testing B-C and C-B, and testing A-C and C-A). Experiment 4 was the same as Experiment 3 except that only symmetry and equivalence relations were tested (e.g., training A-->B, testing B-A, training C-->B, testing B-C, and testing C-A). Symmetry and equivalence were obtained most quickly with adults trained on simultaneous many-to-one protocols. With children, however, the simultaneous protocol was not effective. The simple-to-complex protocol produced much better results which were virtually the same for all training arrangements (linear, many-to-one, one-to-many).