Subjected to moisture variations under a long-time mechanical loading, wood shows a delayed behaviour, where the total response is not the simple accumulation of those corresponding to the mechanical loading and the zero-load swelling-shrinkage deformation. Indeed, one can observe supplementary couplings phenomena between the two types of loading. These phenomena are regrouped under the name of mechanosorptive effects. The aim of the proposal thesis was to realize a very meticulous experimental study that would bring new elements for trying to explain the observed mechanosorptive effects on the wood, and particularly about the structural level where the principal mechanisms would take place. The peculiarity of the study is related to the choice of the specimens with small thicknesses. A special experimental apparatus was developed to conduct mechanical tests on small specimens under constant and variable climatic conditions. The first part of this apparatus is a mechanical press, equipped with a command program permitting to set complex sorts of loading and unloading programs in force or displacement control. The second part is a hygrothermal cell, fixed on the clamps of the press, which was realized to control the climatic conditions around the specimen. The regulation of the temperature is controlled by a thermostatic bath. Creep-recovery tests in tension with cycles of humidity variations were realized. A first series of tests was done on initially dry and initially moist specimens. It permitted us to observe the general characteristics of the hygro-mechanical wood behaviour. On the next step, several groups of creep-recovery tests were done for the same stress level. The only varying parameter was the load-duration. The results obtained shown that the load-duration has no significant influence on the mechanosorptive deformation. Influence of both humidity cycles and loading-unloading cycles and of humidity cycles and step-wise unloading on mechanosorptive deformation has been studied. The results obtained let us suppose that the mechanosorptive deformation could be decomposed into two parts, related probably to two different mechanisms. Relaxation-blotting out tests under variable climate has been performed on initially dry and initially moist specimens, thanks to the experimental apparatus which was developed. A software was developed for the simulation of linear and non-linear elasto-visco-plastic behaviour of materials. It is based on the interna1 variables method. Its application to the behaviour of wood, under constant climate, is illustrated on an example based on a relaxation-blotting out test.