Thermogravimetric data on the devolatilization rate of beech wood are re-examined with the aim of incorporating the effects of high heating rates (up to 108 K min(-1)) in the global kinetics. The mechanism consisting of three independent parallel reactions, first-order in the amount of volatiles released from pseudo-components with chief contributions from hemicellulose, cellulose and lignin, is considered first. It is found that the set of activation energies estimated by Gronli et al. [M.G. Gronli, G. Varhegyi, C. Di Blasi, Ind. Eng. Chem. Res. 41 (2002) 4201-4208] (100, 236 and 46 U mol(-1), respectively) for one slow heating rate results in very high deviations between predicted and measured rate curves. The agreement is significantly improved by a new set of data consisting of activation energies of 147, 193 and 181 U mol(-1), respectively. In this case, the overlap is reduced between the reaction rates of the three pseudo-components whose chemical composition is also modified. In particular, instead of a slow decomposition rate over a broad range of temperatures, the activity of the third reaction is mainly explicated along the high-temperature (tail) region of the weight loss curves. The performances of more simplified mechanisms are also evaluated. One-step mechanisms, using literature values for the kinetic constants, produce large errors on either the conversion time (activation energy of 103 W mol(-1)) or the maximum devolatilization rate (activation energy of 149 U mol(-1)). On the other hand, these parameters are well predicted by two parallel reactions, with activation energies of 147 and 149 kJ mol(-1). (c) 2005 Elsevier B.V. All rights reserved.