It is generally assumed that the properties of wood against fatigue are good, but little is known about the properties of adhesively bonded wood, which represents today most of the wood-based products. Lap-shear samples glued with three common wood adhesives [two ductile one-component polyurethane (1C-PUR) systems and one brittle phenol resorcinol formaldehyde (PRF) adhesive] were tested under cyclical loads at three different climates [20°C, 35% - 50% - 85% relative humidity (RH)]. For the analysis of data, an empirical model based on reaction kinetics was developed. In addition, a probabilistic model was used to estimate the endurance limit and the expected run-out lifetime. Both models were combined to accurately model fatigue at high and low relative stress intensity. It was shown that ductile 1C-PUR adhesives perform better than the brittle adhesive system under dry climates (35%-50% RH). However, for higher RH, the brittle PRF adhesive showed better performance, most probably due to a better wood-adhesive adhesion in wet climate. An average endurance limit for tensile shear stresses between 20% and 48% of the mean tensile shear strength (TSS) was estimated for the tested adhesives. It was shown that the model parameters could be linked to fundamental physical constants through the reaction kinetics approach; however, further research is needed to correlate these parameters to specimen-specific quantities.