Several correlations have been proposed for prediction of three-phase relative permeability from two-phas~ data or from saturation/capillary-pressure relationships. These include the models of Stone, Hirasaki, Corey et al., Naar and Wygal, Land, Alema~, and Parker et al. This paper compares predictions of these models with predictions of two additional models (saturation-weighted interpolation and true-linear interpolation). The comparison is made using all the published three-phase experimental relative permeability data complete enough for application of the models.The comparison shows that the models are often not very good predictors of the experimental data. This points out a need for better relative permeability models in cases where three-phase flow may have a significant effect. In most cases, straightline interpolation or saturation-weighted interpolation between the permeabilities at the two-phase boundaries of the three-phase flow regiori provided a better fit of the experimental data than did the theoretically-based models.The paper also demonstrates the utility of comparing the effects of different relative permeability models before using them in a reservoir simulation. Sometimes the result of a simulation can be heavily biased by the choice of relative permeabili~y model used. j=l J J J J * p3j + P 6 .(s)where each of the saturations S. is normalized with respect to a maximum and minimurl! value (unspeci-