Kin selection is a powerful tool for understanding cooperation among individuals, yet its role as the sole explanation of cooperative societies has recently been challenged on empirical grounds. These studies suggest that direct benefits of cooperation are often overlooked, and that partner choice may be a widespread mechanism of cooperation. Female eider ducks (Somateria mollissima) may rear broods alone, or they may pool their broods and share brood-rearing. Females are philopatric, and it has been suggested that colonies may largely consist of related females, which could promote interactions among relatives. Alternatively, shared brood care could be random with respect to relatedness, either because brood amalgamations are accidental and nonadaptive, or through group augmentation, assuming that the fitness of all group members increases with group size. We tested these alternatives by measuring the relatedness of co-tending eider females in enduring coalitions with microsatellite markers. Females formed enduring brood-rearing coalitions with each other at random with respect to relatedness. However, based on previous data, partner choice is nonrandom and dependent on female body condition. We discuss potential mechanisms underlying eider communal brood-rearing decisions, which may be driven by the specific ecological conditions under which sociality has evolved in this species.