The application of the Monte Carlo method to the solution of deep-penetration radiation transport problems requires the use of “importance sampling.” A systematic approach to obtaining an importance function is to calculate the solution of the inhomogeneous adjoint transport equation (using the Monte Carlo estimator of the answer of interest as the source term) and to use this adjoint flux (or value function) as the importance function. The adjoint flux is calculated for simplified geometries using one-dimensional discrete ordinates methods. In three-dimensional deep-penetration Monte Carlo calculations the alteration of both the transport and the collision kernel is desirable. The exponential transform is quite useful for altering the transport kernel. However, selection from the altered collision kernel is much more difficult. The approach taken here is to introduce an angular grid with 30 discrete directions fixed in the laboratory coordinate system, along which particles are required to travel. After determining appropriate scattering probabilities and values of the importance function for each of the discrete directions, the selection of the outgoing direction and, hence, energy from the resulting discrete distribution is easily performed. The effects of the discrete angular grid and the capability of angular-biased Monte Carlo have been investigated for neutron transport by comparison with standard Monte Carlo and discrete ordinates calculations, experiment, and exact analytic solutions for several configurations. In all cases the discrete grid alone (no angular biasing) was observed to have no significant effect on the results. Monte Carlo calculations were performed utilizing the exponential transform, nonleakage, source energy biasing, Russian roulette, and splitting plus the angular biasing. The results of these calculations illustrate the general usefulness of this discrete grid approach to angular biasing in several ways. First, meaningful results were obtained with angular biasing at much greater distances from the source than were practically possible with the earlier biasing techniques. The answers, variances, and computer times were all on the same order or better than those obtained with the earlier biasing techniques. Finally, this method utilizing the discrete grid to incorporate angular biasing requires very little human interaction once the adjoint configuration is selected.