Theory suggests temperamental reactivity [negative affectivity (NA)] and regulation [effortful control (EC)] predict variation in the development of emotion regulation (ER). However, few studies report such relations, particularly studies utilizing observational measures of children's ER behaviors in longitudinal designs. Using multilevel modeling, the present study tested whether (1) between‐person differences in mean levels of mother‐reported child NA and EC (aggregated across age) and (2) within‐person changes in NA and EC from the ages of 18 to 42 months predicted subsequent improvements in laboratory‐based observations of children's anger regulation from the ages of 24 to 48 months. As expected, mean level of EC (aggregated across age) predicted longer latency to anger; however, no other temperament variables predicted anger expression. Mean level of EC also predicted the latency to a child's use of one regulatory strategy, distraction. Finally, decreases in NA were associated with age‐related changes in how long children used distractions and how quickly they bid calmly to their mother. Implications for relations between temperament and anger regulation are discussed in terms of both conceptual and methodological issues.