Dominance interactions determine reproductive status in many animal societies, including many cooperatively breeding vertebrates and eusocial Hymenoptera without queen-worker dimorphism. Typically, the dominant individual monopolises reproduction, and subordinates behave like helpers. In Dinoponera queenless ants, workers are totipotent females and can potentially reproduce, yet only the top-ranking worker actually reproduces. Individual workers ranked immediately below the dominant breeder worker (gamergate) are hopeful reproductives. Whether or not a worker benefits from joining the hierarchy of high-ranking workers depends on the trade-off between the probability of becoming dominant and reproducing directly, and the colony-level cost of an additional lazy high ranker. Inclusive fitness models predict that the length of the dominance hierarchy depends on relatedness, colony size, and the linearity of the hierarchy. Here, we test the effect of colony size by comparing hierarchy length among three species that differ in colony size (Dinoponera australis: median=14 workers, quartiles=10 and 19 workers; D. gigantea: median=41, quartiles=33 and 74; D. quadriceps: median=78, quartiles=55 and 90). Although difficulties in defining where the hierarchy ends hamper comparisons, the results are in broad agreement with the predictions. Hierarchies are close to the predicted lengths and are longer in species with larger colonies (one, three and three workers in the three species in order from smallest to largest colony vs two, three and four predicted). These conclusions are further supported by determining Kokko and Lindström's λ index of skew, which is smaller (i.e. characteristic of a longer hierarchy) in species with larger colonies.