Our first goal was to examine the relations among observational, physiological, and self-report measures of children's anger. Our second goal was to investigate whether these relations varied by reactive or proactive aggression. Children (272 second-grade boys and girls) participated in a procedure in which they lost a game and prize to a confederate who cheated. Skin conductance reactivity, heart rate reactivity, self-reports of anger, angry facial expressions, and angry nonverbal behaviors were measured for each turn of the game. We used multi-level regressions to calculate the relations among the 10 pairs of the five anger variables over the course of the game. Six of the 10 pairs of anger variables were positively related. These findings suggest that measuring children's anger using any one approach may not capture the full complexity of children's overall experience and expression of anger. Furthermore, three of the 10 relations were stronger at higher levels of reactive aggression, although none varied by proactive aggression. These findings suggest that reactive aggression is related to greater cohesiveness in the experience and expression of anger than is proactive aggression.