In two experiments, children aged 3, 4 and 5 years (N = 61) were given conflicting information about the names and functions of novel objects by two informants, one a familiar teacher, the other an unfamiliar teacher. On pre-test trials, all three age groups invested more trust in the familiar teacher. They preferred to ask for information and to endorse the information that she supplied. In a subsequent phase, children watched as the two teachers differed in the accuracy with which they named a set of familiar objects. Half the children saw the familiar teacher name the objects accurately and the unfamiliar teacher name them inaccurately. The remaining half saw the reverse arrangement. In post-test trials, the selective trust initially displayed by 3-year-olds was minimally affected by this intervening experience of differential accuracy. By contrast, the selective trust of 4- and 5-year-olds was affected. If the familiar teacher had been the more accurate, selective trust in her was intensified. If, on the other hand, the familiar teacher had been the less accurate, it was undermined, particularly among 5-year-olds. Thus, by 4 years of age, children trust familiar informants but moderate that trust depending on the informants' recent history of accuracy or inaccuracy.