Using an instrument based on a national contest in France determining researchers’ location, we find evidence of peer effects in academia, when focusing on precise groups of senders (producing the spillovers) and receivers (benefiting from the spillovers), defined based on field of specialisation, gender and age. These peer effects are shown to exist even outside formal co-authorship relationships. Furthermore, the match between the characteristics of senders and receivers plays a critical role. In particular, men benefit a lot from peer effects provided by men, while all other types of gender combinations produce spillovers twice as small.