The emotional reactivity hypothesis holds that, over the course of phylogeny, the selection of animals with less reactive temperaments supported the development of sophisticated social-cognitive skills in several species, including humans (Hare, 2007). In the ontogenetic human case, an emotional reactivity hypothesis predicts that children with less reactive temperaments will reach certain milestones in theory-of-mind (ToM) development more quickly. We examined relations between temperament and false-belief understanding in 102 preschool-age children from China and the United States. Temperament was measured via parental ratings of behavior as well as with physiological measures of children's reactivity (hypothalamic-pituitary-adrenocortical axis reactivity gauged via salivary cortisol). In accord with an emotional reactivity hypothesis, children with certain reactive temperaments--specifically, those who were more aggressive and those who were both socially withdrawn and physiologically reactive--evidenced poorer social-cognition. However, our findings also force amendment to the ontogenetic emotional reactivity hypothesis. For the majority of children in both countries, physiological reactivity predicted more advanced ToM, perhaps by facilitating social engagement and attention to social stimuli. Moreover, children who were withdrawn from social interaction yet nonreactive, especially Chinese children of this temperament, evidenced advanced ToM. Thus, some forms of social disengagement may foster social-cognitive development in certain sociocultural contexts. (Contains 3 tables and 1 figure.)