The goal of the present study was to examine how young children form impressions of peers with implications for understanding and ameliorating peer rejection. We focused on four questions: whether young children make dispositional attributions from observed behavior; how children combine complex behavioral information into a coherent evaluation; how other traits are inferred; and how modifiable are existing impressions. The participants were 51 kindergarteners, 53 second graders, and 104 college students. All subjects watched videoclips of a same-sex child actor performing three mean or kind behaviors, three smart or not smart behaviors, and three shy or not shy behaviors. Two of the three behavioral dimensions had the same valence given by a valence condition while the third dimension had opposite valence. Some subjects also had prior expectancies of one trait of predominant valence. We hypothesized that in the expectancy condition, only the expected trait will be represented in memory, while all three traits will be represented if there is no expectancy. After providing ratings of how much they liked the actor and how various traits described him or her, two more video clips of the same actor were presented. These new behaviors were incongruent in valence with the expected trait or, in the no-expectancy condition, with one of the traits with predominant valence. Ratings of liking and the same personality traits were then collected again. The results showed that expectancy did not suppress the encoding of the unexpected traits but it made the ratings of the expected trait more extreme. Importantly, all age groups made dispositional attributions of all three presented traits that accurately reflected the valence of the observed behaviors, although 5-6 year-olds gave less negative ratings to traits with negative valence. The global evaluation of the actor was predicted by all presented trait ratings, and was evaluatively consistent with the inferred trait ratings. The presentation of incongruent behaviors resulted in substantial modifications of all impressions that had predominant valence. Kindergarteners modified their ratings the least, especially if their impressions were positive. Implications for understanding peer rejection are discussed.