Viscosity is an important parameter to assess heavy oil and bitumen upgrading operations, such as bitumen dilution to meet pipeline viscosity specification limits. In conceptual design studies that involve blending of high- and low-viscosity materials, the experimental measurement of viscosity is impractical; therefore, such studies employ mixing rules to estimate mixture viscosity. Mixing rules for conceptual design evaluations where limited or no information apart from the viscosity and density of the bitumen and solvents is available were of interest. This study determined which viscosity mixing rules could be used with such limited input, what the predictive errors were, whether the performance of those mixing rules were measurably affected by changes in chemical composition, and if mixing rules could be applied at the <10 wt % solvent concentration in bitumen. Binary mixtures of 1-10 wt % of six different solvents (1-methylnaphthalene, decahydronaphthalene, 1,2,3,4-tetrahydronaphthalene, butylcyclohexane, butylbenzene, and n-decane) and Athabasca bitumen were prepared, and their viscosity, density, and refractive index were measured at 303, 313, and 333 K (30, 40, and 60 degrees C) and atmospheric pressure. The performance of mixing rules in predicting these properties for the binary mixtures at low solvent concentration was evaluated. It was found that the simple mixing rule, log [log (nu(m) + 0.7)] = Sigma(wi) log [log (nu(i) + 0.7)], and that by Miadonye et al. (Petrol. Sci. Technol. 2000, 18, 1-14) consistently gave the better viscosity estimation with an absolute average relative deviation (AARD) of around 30%. Within this uncertainty, there was no evidence indicating that the mixing rules could not be used for viscosity prediction of bitumen-solvent mixtures at low dilution levels. The mixing rules appeared not to be affected by the chemical nature of the solvent, and if it had an effect, it was of the same order or less than the uncertainty of viscosity prediction.