The purpose of this study was to investigate whether group performance could be significantly improved by forming groups with members heterogeneous in information-processing preferences, as measured by the Myers-Briggs Type Indicator, and by providing those groups with instructions’ to facilitate consensus on the group’s solution to a complex, multistage decision task. Comparisons of four performance measures among undergraduate student groups (N = 38) differing in composition (homogeneous vs. heterogeneous) and mode of consensual instruction (instructed vs. not instructed) were conducted through a series of statistical procedures. The performance measures, obtained from individual and group solutions to NASA’s "Lost on the Moon" task, were: a) group scores, b) utilization of average member resources, c) utilization of the groups’ best member’s resources, and da) the proportion of groups achieving an assembly effect. The results of the analyses supported only those hypotheses predicting that groups instructed in consensus would demonstrate better performance on each of the measures. The hypotheses predicting composition main effects for the performance measures were not supported, nor were hypotheses predicting significant interaction effects. Implications of the findings and suggestions for further research are offered. / Ph. D.