Participants learned about a structure without explicit explanation, either by using relational schema induction (G. S. Halford, J. D. Bain, M. T. Maybery, & G. Andrews, 1998), which requires completion of instances of the structure, or by memorizing instances. Emergence of structured knowledge over trials was assessed by ability to map the structure to an isomorph and thereby to generate novel instances of the structure (the generativity test). Schema induction was found to be an effective way to induce acquisition of the structure, and the generativity test was a sensitive measure of structure acquisition, as indicated by converging measures. Memorization produced only low levels of structured knowledge. A conception of structured knowledge acquisition based on relational schema induction is proposed.