The uncertainty in computed species concentrations resulting from measurement imprecision in reaction rate components is investigated in two tropospheric photochemical models. In this study, which extends Thompson and Stewart <1991>, we perform statistical analysis on reaction rate coefficients to focus on two aspects of model uncertainty: (1) the change in the magnitude of concentration uncertainty as temperature varies throughout a model grid and (2) the difference resulting from selection of mean, as opposed to median, rates for model calculations. Reaction rates are treated as realizations of random variables having statistical properties given by component terms and their imprecisions. These assumptions lead to expressions for probability distributions of bimolecular rates and for the high- and low-pressure limits for termolecular and thermolytic processes. They also imply that bimolecular rates and high- and low-pressure limits used in photochemical models correspond to median values taken from a lognormal distribution. We derive analytic expressions for mean values, which are always larger than the medians, an intrinsic property of lognormal variables. We suggest that comparison of species concentrations computed using median and mean rates can provide some measure of the effect of rate imprecision in multidimensional models. We also find that the temperature dependence derived for the imprecision in bimolecular rates differs from that given in standard references, with our expression giving a smaller increase in imprecision as temperatures deviate from 298 K. Results are illustrated using a box and a one-dimensional (1-D) model.