Research on transactive memory systems (TMSs) implicitly assumes that metaknowledge (i.e., the "knowledge of who knows what") is uniformly distributed among team members. Relaxing this assumption results in a more realistic notion of team cognition in which the distribution of metaknowledge can take different forms. Demonstrating the importance of this conceptual shift, we compare teams in which metaknowledge is concentrated within one central member (a centralized TMS structure) with teams in which metaknowledge is distributed evenly among the members (a decentralized TMS structure). We predicted that centralized metaknowledge can give teams a performance advantage over decentralized metaknowledge, because centralized metaknowledge can allow the central member to function as a catalyst for information exchange and integration. We proposed this catalyst effect to be contingent on the extent to which the distribution of task information among members poses high coordination demands to effectively integrate members' knowledge. In a laboratory team decision making experiment (N = 112) we found the predicted interaction effect between TMS structure and the distribution of task information. Furthermore, the experiment supported our hypotheses about the mediating role of the transactive retrieval process and the ensuing team information elaboration.