Preschool children and adults received training on three sets of successive discriminations: (1)A1-R1,A2-R2, (2)B1-R1,B2-R2, and (3)A1-R3,A2-R4. Then they received tests assessing derived stimulus-response relations (B1-R3, B2-R4) and stimulus-stimulus relations (e.g., A1-B1,A2-B2), Four training protocols were used. The protocols differed with regard to the order in which the sets were trained: Many-To-One(1-2-3), One-To-Many-1 (1-3-2), One-To-Many-2 (3-1-2), and One-To-One (3-2-1 or 2-3-1). The adults displayed class-consistent B-R and A-B performances over all conditions. The children displayed class consistent B-R performances more often in Many-To-One and One-To-Many than in One-To-One. Their A-B performances were highly consistent with the trained A-R and tested B-R performances. Present findings are consistent with the stimulus equivalence account rather than with the mediated response account. (C) 2001 by Academic Press.