Lack of progress in reducing linewidth roughness of lithographic features has led to investigations of the use of postlithography process smoothing techniques. However, it remains unclear whether such postprocessing will sufficiently reduce the detrimental effects of feature roughness. Thus, there is a need to understand the efficacy of postprocessing on not just roughness reduction, but on the negative device impacts of roughness. We derive model equations of how roughness impacts lithographic performance and incorporates smoothing using postprocessing. These models clearly show that postprocess smoothing works best by increasing the correlation length. Increasing the correlation length is very effective at reducing high-frequency roughness that impacts within-feature variations but is less effective at reducing low-frequency roughness that impacts feature-to-feature variations. It seems that postprocess smoothing is not a substitute for reducing the initial roughness of resist features.