The average estimates of a group of individuals are generally better than the estimates of the individuals alone, a phenomenon commonly referred to as the wisdom of crowds. This has been shown to work for many types of simple tasks, but has generally been performed on subjects that do not communicate with one another. We report group aggregation performance for more complex tasks, involving reconstructing the order of time-based and magnitude-based series of items from memory. In half of these tasks, subjects receive the previous subjects final ordering in a serial fashion. The aggregate for communicating subjects is better than that for independent subjects. We introduce a Bayesian version of a Thurstonian model to show how each subject combines their individual, private knowledge with the previous individuals ordering. The model also shows that individuals can produce estimates in the shared information condition that are better for aggregating than independent estimates.