We employ social network data from 25 randomly sampled voluntary associations to understand the factors associated with accurate perceptions of the political preferences of fellow group members. We build upon research in communication, social psychology, and social networks to identify relevant predictors. We analyze relationships at the dyadic level, but we also consider the aggregated accuracy of perceptions by ego of alters (“perceptiveness”) and the aggregated accuracy of perception by alters of ego (“explicitness”) regarding political candidate preferences using a multilevel modeling approach. We find relatively low levels of accuracy on average, and in general the variables that predict perceptiveness are not the same variables that predict explicitness. However, there is a consistent and strong link between the frequency of communication (viewed as an indicator of network tie strength) and accuracy both at the dyadic and aggregate levels. However, this relationship is highly contingent on the homophily of political preferences within the group.