An integrated thermal-hydraulic feedback module has previously been developed for the Monte Carlo transport solver MC21. The module incorporates a flexible input format that allows the user to describe heat transfer and coolant flow paths within the geometric model at any level of spatial detail desired. The effect that the varying levels of spatial homogenization of thermal regions has on the accuracy of the Monte Carlo simulations is examined in this study. Six thermal feedback mappings are constructed from the same geometric model of the Calvert Cliffs core. The spatial homogenization of the thermal regions is varied, giving each scheme a different level of detail, and the adequacy of the spatial homogenization is determined based on the eigenvalue produced by each Monte Carlo calculation. The purpose of these numerical experiments is to determine the level of detail necessary to accurately capture the thermal feedback effect on reactivity. Several different core models are considered: axial flow only, axial and lateral flow, asymmetry due to control rod insertion, and fuel heating (temperature-dependent cross sections). The thermal results generated by the MC21 thermal feedback module are consistent with expectations. Based on the numerical experiments conducted, it is concluded that the amount of spatial detail necessary to accurately capture the feedback effect on reactivity is relatively small. Homogenization at the assembly level for the Calvert Cliffs pressurized water reactor model results in a power defect similar to that calculated with individual pin cells modeled as explicit thermal regions.