Infanticide was first observed in langurs nearly 50 years ago, and this rare phenomenon has been inferred to have either an evolutionarily adaptive function or to be a pathological and non-functional behavior. In this study, we report 5 male takeover events in one-male groups of white-headed langurs in the Nongguan Karst Hills, Guangxi, China from 1998 to 2006. We recorded 13 attacks on 9 infants by extra-group males or new resident males. During the male takeovers, all of the infants younger than 6 months (with an average age of 3.6 months [N = 11]) in the groups disappeared. The infant death rate during the 4.2 months after takeover by a new male was significantly higher than the infant death rate calculated for most of the year. Older infants that were still nursing (with an average age of 14.1 months [N = 7]) were often attacked and seriously wounded by the extra-group males or new resident males, but all of them survived. The interbirth intervals of females whose infants were assumed to be killed by males were significantly reduced relative to those of females in groups with stable male tenure (mean = 10 months vs 25 months). Our data suggest that males kill unrelated and unweaned infants during the takeover period to decrease the time until the infants’ mothers resume fertility. Thus, infanticide would support sexual selection theory in white-headed langurs. The data also show that infanticidal behavior was directed toward the infants, especially those who were still nursing. Female dispersal may function as a counter-strategy to avoid infanticide.