Several mechanical relaxation phenomena are observed on glasses. The characteristics of these phenomena vary from one type of glass to another but some features seem to be common to all solids in the vitreous state. Hence, results about: 1) oxide glasses; 2) organic glasses; 3) vitreous semi-conductors; 4) metallic glasses are recalled. Then, it is shown, that there are a relaxation process about the glass transition T g (primary or α relaxation) and several relaxation processes more or less differenciated at lower temperature (secondary relaxations): if one is restricted to the temperature range not too lower than T g , a peak or an anomaly in the spectrum internal friction versus temperature more or less resolved from the increase of damping due to the α relaxation, is often observed. This feature seems to be associated to delayed elasticity. We propose that such a behavior corresponds to a second relaxation phenomenon (β relaxation) which is, as α relaxation, characteristic of the vitreous state. Most models describing such a behaviour remain very phenomenological. Hence, it seems interesting to propose an interpretation following more physical basis: after a short review about ideas previously put forward and concerning molecular mobility in vitreous solids near T g , a new description of elementary processes of deformation in glass induced by an applied stress, is presented. Such a model takes again into account the assumption about localized processes (high free volume sites) but structural relaxation is considered in order to explain that internal stresses associated to such processes can be relaxed, thus allowing new processes to occur. Then the deformation of vitreous solids near T g might be essentially controlled by recovery phenomena. From such a model of deformation, dynamic modulus (internal friction and modulus effects) can be calculated and the results are discussed by comparison to experimental data Bilan des resultats concernant les proprietes micromecaniques des solides vitreux (oxydes, semi-conducteurs, composes organiques, alliages metalliques), qui permet d'illustrer l'existence des relaxations α (viscoplasticite) et β (viscoelasticite). Discussion des modeles utilises jusqu'ici pour interpreter ces relaxations. Proposition d'un nouveau modele reprenant l'idee que ces processus sont localises (sites a volume libre eleve), mais en tenant compte de l'effet de la relaxation structurale qui supprimerait les contraintes internes associees a ces processus, qui pourraient alors etre reactives