In order to investigate the impact of density on the classification behavior of particles in the static classifier of Vertical Spindle Mill, the sensitivity of overflow yield to the increase of air amount for narrowly sized pyrite, carborundum, quartz and coal samples were compared in a lab-scale classifier, respectively. Response surface methodology is used to analyze the combined effect of size and density on the classification. Wide size classification was also conducted and results show that both the yield and R90 of overflow increase with the decreasing of density, and the growth of air amount would also lead them to rise. The Whiten’s model was applied to illustrate the influence of density on the sharpness of classification, corrected cut size and fishhook effect. Results show that material with a lower density would have a higher fishhook effect parameter, classification sharpness and corrected cut size. The increase of air amount would result in a more evident fishhook effect for the high density material. Based on the Whiten’s model, a new classification efficiency model with the addition of particle density in various forms was established. This new model could describe the classification efficiency of materials with different density in the identical experiment conditions.