For humans and other social species, social status matters: it determines who wins access to contested resources, territory and mates [1–11]. Human infants are sensitive to dominance status cues [12,13]. They expect conflicts to be won by larger individuals [14], those with more allies [15], and those with a history of winning [16–18]. But being sensitive to status cues is not enough; individuals must also use status information when deciding whom to approach and whom to avoid [19]. In many non-human species, low-status individuals avoid high-status individuals, and in so doing avoid the threat of aggression [20–23]. In these species, high-status individuals commit random acts of aggression toward subordinates [23] and even commit infanticide [24–26]. However, for less reactively aggressive species [27,28], high-status individuals may be good coalition partners. This is especially true for humans, where high-status individuals can provide guidance, protection and knowledge to subordinates [2,29,30]. Indeed, human adults [31–33], human toddlers [34], and adult bonobos [35] prefer high-status individuals to low-status ones. Here we present 6 experiments testing whether 10- to 16-month-old human infants choose high-or low-status individuals—specifically, winners or yielders in zero-sum conflicts—and find that infants choose puppets who yield. Intriguingly, toddlers just six months older choose the winners of such conflicts [34]. This suggests that although humans start out like many other species, avoiding high-status others, we shift in toddlerhood to approaching high-status individuals, consistent with the idea that for humans, high-status individuals can provide benefits to low-status ones.