Abstract
Two conceptions of observational learning were compared using a domino selection task that allowed subjects to follow a modeled response rule, to match particular numerical instances, or both. 80 college students showed strong rule-adoption and identification but very little mimicry of specific instance values. Live modeling produced stronger rule-adoption than did modeling a target through presentation of completed arrays, but live modeling and modeling a target led to comparable rule-description. There was greater mimicry of simple than more complex modeling sequences, but no difference between live and target-modeling conditions. The results supported a symbolic concept-attainment view of observational learning phenomena rather than an operant response-marching explanation.