Surface roughness is one of the most important characteristics in machining processes. This study presents probabilistic models to estimate surface roughness experimentally in multi-pass horizontal surface grinding operations from three independent distributions such as the initial surface roughness distributions of workpiece, the distributions of the wheel radius, and the distributions of distances between major active grains. To specify the model characteristics from surface roughness measurements, either the probability satisfying a given surface roughness or the range of surface roughness satisfying a given probability have been estimated while grinding conditions are fixed. Finally, the relationship between grinding conditions satisfying surface roughness range under a given probability can be established.