Description of contents: The evolution of division of labour: The file ‘Division of labour data.xls’ contains the results of our analytical and numerical equilibrium analysis (see supplementary information). This data was used to create figure 3 and similar data was used to create figure 5. We present all parameter combinations that were considered, which strategies were found to be stable for each combination, and the ESS values of p, q1 and q2 for each combination. The evolution of extreme specialisation: We include the dynamic, individual-based simulation codes that were used to generate figures 4a-4d. This code was written in Matlab R2015a. ‘Reproductive specialisation.m’ generates figure 4a, ‘Helper specialisation.m’ generates figure 4b, ‘Phenotypic_repulsion.m’ generates figure 4c and ‘Feedback_curve.m’ generates figure 4d. ‘DOLsimq1s.m’ and ‘DOLsimgflips.m’ are the simulation codes used in the above files. Uninvadability analysis: We include the dynamic, individual-based simulations codes that were used to show that an equilibrium is uninvadable by a particular mutant strain. This code was written in Matlab R2015a. 'Mutant_invasion_simulation.m' takes a known equilibrium and parameter combination and evaluates whether a given mutant strain can invade. This plots 95 percent confidence intervals of the proportion of mutant individuals in the population over time from repeated simulations. 'Mutant_invasion_simulation.m' calls 'multinvplot.m' which calls 'DOLsim11.m'. This kind of analysis was used to show that all of our equilibria are uninvadable ( supplementary information and in supplementary figures 2 and 3).