Perturbation viscometry is a recently developed method which measures the logarithmic gradient of the viscosity-composition curve for gas mixtures using a variant of capillary-tube viscometry. A gas mixture flowing through a capillary has its composition perturbed by the addition of a small flow of gas, normally one of the pure components of the gas mixture. Two pressure changes, the first due to the change in flowrate and second due to the change in viscosity are seen at the capillary. The logarithmic viscosity gradient is calculated from the ratio of these two pressure changes. Integration of logarithmic viscosity gradients measured for the full composition range yields the mixture viscosity relative to the viscosity of either component of the gas mixture. This method is attractive because, for measurements of equal precision, integration of the gradients is potentially much more precise than conventional methods that measure the absolute viscosities directly. Integration of the logarithmic viscosity gradients could be accomplished either by application of the trapezium rule or numerical integration of a polynomial fitted to the data. Here a method with better theoretical foundations is presented. The method fits a differential form of the Sutherland equation to the logarithmic viscosity gradients. The fitted parameters of the equation are then used to generate the viscosity ratios by substitution into the normal form of the Sutherland equation. The quality of fit obtained may be used to test of the consistency of the experimental data. Experimental data gathered for the mixtures argon-nitrogen, helium-argon and helium-nitrogen at 98.5°C are analysed using this procedure. The internal consistency of the data is checked by direct comparison of calculated values for the gradients generated from the fitted Sutherland parameters with actual experimental data. The calculated relative viscosities are compared with extant data and theoretical predictions and show good agreement. Also the calculated Sutherland parameters are well within the scatter of the parameters obtained from alternative sources.