Abstract: Interventions occur when animals interfere in competitive interactions between two or more individuals. Interveners can alter the nature of the ongoing interaction by targeting one party (attacking, biting) and supporting the other. Three theories have been proposed to account for intervention behaviour: kin selection, reciprocity and direct benefits. The kin selection hypothesis predicts that interveners will selectively support relatives over nonrelatives; the reciprocity hypothesis predicts that when intervener [`]A' supports individual [`]B', later [`]B' will intervene and support [`]A'; and the direct benefits hypothesis predicts that target/support patterns should serve the immediate interests of the intervener. We tested the reciprocity and direct benefits hypotheses by exploring third-party interventions in play fighting among littermates of domestic dogs. Interveners in dyadic play did not preferentially target or support preferred playmates of the intervener. Interveners targeted the dog in the losing role at the time of the intervention, and they did not show reciprocity in support. Taken together, these last two findings suggest that littermates benefit directly and use interventions opportunistically to practise offence behaviours directed at littermates already behaving subordinately. Opportunities to practise targeting in a playful setting may help structure dominance relationships among littermates. Additionally, the tendency for puppies to do what the other is doing (target the dog in the losing role) may pave the way for synchronizing cooperative behaviours during group hunting and territorial defence. The types of behaviours used to intervene changed over development, but the outcome following an intervention remained stable.