This paper argues that (1) as measures of emotional empathy in children are refined, there is increasingly strong support for the hypothesis that emotional empathy can serve both as an inhibitor of antisocial behavior and as a motivator of prosocial behavior, but that (2) since the relation between children's empathy and their behavior is confounded by a number of factors the measures require further refinement. Research is described which investigates the responses of children in grades one, three and five to violent and non-violent events shown in slides and on television. No sigificant evidence was found that children's assessed level of empathy was related to their aggressive or altruistic behavior toward peers. It is argued that the criterion measures of altruism and aggression are not responsible for this result, but that an aspect of the empathy measure, children's reports of their own affect, is not reliable. When a measure not dependent on children's self-reports was used, the relation between empathy and behavior appeared dependent upon (1) empathic responses to specific affects and the relevance of specific empathic emotions to the criterion behaviors, (2) situational factors in the empathy eliciting and criterion contexts, and (3) the age and sex of the children. It is suggested that both younger and older children are aroused by violent episodes, but that the latter are more likely also to empathize with the victim. A short description of the Feshbach and Roe measure of empathy and five tables of data are appended. (Author/RH)