When deciding whether to trust someone's claims, how do children combine - over multiple interactions - information about that person's general behavioral tendencies (traits) with that person's ongoing (and changing) rate of providing accurate claims? Children aged 4-8 played 11 rounds of a find-the-sticker game. For each round, an informant looked into two cups and made a claim about which cup held a sticker. Children guessed the sticker's location and the sticker's actual location was revealed. Prior to the game, children received information that the informant was either honest or dishonest. In Study 1 (N = 201, 105 female, 96 male), the informant provided inaccurate information on the first five trials and then provided accurate information for the remaining trials (55% overall accuracy). In Study 2 (N = 144, 89 female, 55 male), the informant produced a less predictable pattern of (in)accuracy, but remained 55% accurate overall. The trait information children initially received about the informant's honesty strongly influenced their epistemic trust when they lacked additional information about the informant's reliability (the earliest trials). When children's first-hand experiences with the informant prevented them from making strong predictions about the informant's future behavior, only children approximately 7 years and older utilized trait information to guide their epistemic trust. These results demonstrate some similarities in children's causal reasoning about the physical world and their social reasoning. The results also demonstrate developmental patterns in how children weigh different types of social information at different junctures in social interaction.