Any endothermic or exothermic reaction is accompanied by self-cooling or self-heating. In reacting systems in which heat transfer is controlled by conduction, non-uniform temperature-position profiles are established. Examples of this situation are the exothermic decomposition of gaseous diethyl peroxide and the endothermic decomposition of nitrosyl chloride at low pressures (when convection is unimportant). In kinetic studies, allowance must be made for the non-uniform temperature to derive accurate isothermal velocity constants and Arrhenius parameters. In the present paper, the necessary corrections have been derived for a reactant in the steady state whose reaction rate varies exponentially with temperature and in which the temperature excess varies from point to point, being zero at the boundary (Frank-Kamenetskii's conditions). The geometries considered are the slab, cylinder and sphere. The temperature gradient at the surface in the steady state ($\Gamma $) occupies a key position, and this is exploited to find the correction factor required to convert 'observed' rate constants to isothermal conditions, and thence to correct 'observed' activation energies and pre-exponential factors. The correction factor is found to be simply related to Frank-Kamenetskii's $\delta $ (a dimensionless measure of heat-release rate). A similar analysis is given for systems hotter or cooler than their surroundings but uniform in temperature--such as well stirred fluid systems or small solid crystals (Semenov's conditions). In these circumstances, systems of arbitrary geometry may be studied, and no approximation need be made to the Arrhenius function. For either type of boundary condition, uncorrected activation energies are overestimates in exothermic reactions and underestimates in endothermic reactions. Explicit relations are derived for making corrections. Boundary conditions intermediate between the two extremes investigated can also be treated though the resulting expressions are more cumbersome. In an appendix, an alternative 'experimental' approach is made to the elimination of errors from measured reaction velocities. This approach identifies the measured velocities with a temperature intermediate between those at centre and surface. The optimum choice, which weights the central and surface temperatures in the ratios 2:1 (slab), 1:1 (cylinder) and 2:3 (sphere), gives exactly correct results for the cylinder and acceptable precision for the slab and sphere even to within 5 K of the explosion limit. Other correction methods are also discussed.