ABSTRACT. The pyrolysis kinetics of Norwegian spruce and birch wood was studied to obtain information on the kinetics of torrefaction. Thermogravimetry (TGA) was employed with nine different heating programs, including linear, stepwise, modulated and constant reaction rate (CRR) experiments. The 18 experiments on the two feedstocks were evaluated simultaneously by the 2 method of least squares. Part of the kinetic parameters could be assumed common for both woods without a considerable worsening of the fit quality. This process results in better defined parameters and emphasizes the similarities between the woods. Three pseudocomponents were assumed. Two of them were described by distributed activation energy models (DAEM), while the decomposition of the cellulose pseudocomponent was described by a self-accelerating kinetics. In another approach all the three pseudocomponents were described by n-order reactions. Both approaches resulted in nearly the same fit quality but the physical meaning of the model based on three n-order reactions was found to be problematic. The reliability of the models was tested by checking how well the experiments with higher heating rates can be described by the kinetic parameters obtained from the evaluation of a narrower subset of 10 experiments with slower heating. A table of data was calculated that may provide guidance about the extent of devolatilization at various temperatureresidence time values during wood torrefaction.