Response thresholds alone cannot explain empirical patterns of division of labor in social insects