Understanding how social species differentially allocate reproduction within groups is a major problem in evolutionary biology. A key issue is whether relatedness among nestmates can be assessed by group members and then used to determine levels of skew. Several studies have shown a relationship between skew and relatedness, but this might indicate correlational rather than causal relationships. For example, if fecundity has a genetic component, then more closely related females might have similar reproductive outputs, and hence low skew, owing to genetic similarity in genes controlling reproduction, rather than relatedness per se. Here we show that in a facultatively social bee ovarian differentiation between queens and secondary reproductives, prior to egg-laying declines with increasing relatedness. This finding could not be explained by genetically determined similarities in fecundity, by variation in body size or age, or by selective execution of brood by workers. Our results indicate that females are able to assess pairwise relatedness, either directly or indirectly, and use this information to mediate ovarian development. This points a way forward for future developments in skew theory by throwing attention onto what has been largely treated as a 'black box': the ability of organisms to acquire and process the kinds of information needed for skew theories to operate.