The validity of using the homogenization approximation in a lattice end-leakage calculation was studied in a series of Monte Carlo scattering order analysis experiments. A method for using the homogenization approximation in Monte Carlo end-leakage calculations is described. The analysis indicated that, even with hydrogen moderation, a treatment of all collision sequences of fifth order or less in a faithful mock-up of the lattice was required to describe the end-leakage fraction accurately. In the case of nonhydrogenous moderators it was necessary to consider longer sequences. When all collision sequences of 10th order or less were treated in a faithful mock-up of the lattice, 80% of the total end-escape fraction was accounted for in a rigorous manner and a good estimate of the end-escape spectrum was obtained down to 25 kev. Escape fraction and spectrum estimates based on numerical integration over all scattering sequences of second order or less in a faithful lattice mock-up are shown to be misleading. The two essential factors which caused the lattice and homogeneous model results to differ were the smaller mean free path and larger absorption cross section of the homogeneous model and the directional character of the mean free path and absorption cross section in the lattice. As a result, longer collision sequences are required, on the average, to produce escape in the homogeneous model than in the lattice.