In the non-isothermal approach to the determination of kinetic constants, the mathematical analysis of the a-time curve is based on the assumption of a rate equation, which in some cases, is not consistent with the isothermal curves. For example, during the reduction of U3O8 by hydrogen, we have shown from isothermal and isobaric experiments that a model of nucleation and anisotropic growth was able to fit the experimental rate curves. This makes the analysis of non-isothermal experiments very complex. A first attempt to solve this problem is presented here, in which the effect of a sudden change in the reaction temperature during an experiment is analysed. Taking into account the differences due to the sudden change in the calculation of the rate, a good agreement was found between the calculated and experimental d alpha /dt vs. a curves.