Computational fluid dynamics (CFD)-discrete element method (DEM) simulations are designed to model a pseudo-two-dimensional (2D) fluidized bed, in which bed thickness is minimal compared to height and length. Predicted bed behavior varies as the simulations are conducted on increasingly refined computational grids. Pseudo-2D simulation results, in which a single computational cell spans the bed thickness, are compared against fully-three-dimensional (3D) simulations results. Both pseudo-2D and fully-3D simulations exhibit high accuracy when sufficiently refined. Indicators of bed behavior, such as bed height, bed height fluctuation, bubble generation frequency, and segregation, do not appear to converge as the cell size is reduced. The Koch-Hill and Gidaspow drag laws are alternately employed in the simulations, resulting in different trends of results with computational grid refinement. Grid refinement studies are used to quantify the change in results with grid refinement for both three-dimensional, uniform refinement, and for two-dimensional refinement on pseudo-2D computational grids. Grid refinement study results indicate the total drag converges as the computational grid is refined, for both 3D and pseudo-2D approaches. The grid refinement study results are also used to distinguish the relatively grid-independent results using the Koch-Hill drag law from the highly grid-dependent Gidaspow drag law results. Computational cell size has a significant impact on CFD-DEM results for fluidized beds, but the grid refinement study method can be used to quantify the resulting numerical error.