In queenless ants, workers compete aggressively to be reproductives instead of sterile helpers. To limit the number of reproductives, either mating or egg-laying by mated workers can be regulated. In Gnamptogenys menadensis, all mated workers reproduce, and field data presented here indicate that mating is not random. We removed gamergates experimentally (n = 16 groups) to induce mating and gamergate replacement. Virgin workers engaged in agonistic interactions, allowing us to classify them into two groups: dominants performed antennal boxing and biting, while subordinates did not. However, a few subordinates cooperated to immobilize individual dominant workers. This effectively reduced the number of dominants because immobilized dominants became subordinates. Six to 12 wks after the removal of reproductives, between one and eight dominants per colony started to perform 'sexual calling' outside the nest entrance. Sixteen tout of 33) dominants mated with foreign males introduced in five experimental groups. None of the subordinate workers mated. The probability of mating is thus regulated by dominance interactions among workers and immobilizations initiated by infertile workers in this polygynous queenless ant.