Numerous theoretical studies have shown that information aggregation through voting is fragile. We consider a model of information aggregation with vote-contingent payoffs and generically characterize voting behavior in large committees. We use this characterization to identify the set of vote-contingent payoffs that lead to a unique outcome that robustly aggregates information. Generally, it is not sufficient to simply reward agents for matching their vote to the true state of the world. Instead, robust and unique information aggregation can be achieved with vote-contingent payoffs whose size varies depending on which option the committee chooses, and whether the committee decision is correct.