According to the Motivated Information Processing in Groups (MIP-G) model, groups should perform ambiguous (non-ambiguous) tasks better when they have high (low) epistemic motivation and concomitant tendencies to engage in systematic (heuristic) information processing and exchange. The authors tested this prediction in an experiment with four-person groups performing a complex and dynamic decision making task. Group confidence was measured after extensive training and prior to actual group decision-making. Task ambiguity was manipulated. Results showed that when task ambiguity was low, group confidence indeed benefits decision quality and group performance. But when task ambiguity was high, group confidence hurt decision quality and group performance. Implications for theory and practice are discussed. Copyright © 2010 John Wiley & Sons, Ltd.