This work compares the performance among objective space partitioning with adaptive ε-ranking, subvector dominance assignment, and epsilon dominance assignment methods that have been recently proposed for many-objective optimization. These three methods enhance selection using different strategies to recalculate the primary or secondary ranking of solutions and have been implemented using the framework of NSGA-II. The first method focuses on the primary ranking of solutions by partitioning the objective space into lower dimensional subspaces and re-ranking solutions within each subspace using an adaptive epsilon-ranking procedure. On the other hand, the latter two methods focus on the secondary ranking of solutions, replacing crowding distance with a substitute assignment distance. As test problems, we use scalable MNK-Landscapes with 4 ‹ M ‹ 10 objectives, N=100 bits, varying the number of epistatic interactions per bit K in the range 0 ‹ K ‹ 50.