Abstract : Thermal oxidation was often studied by conventional methods: thermogravimetric measurement and SEM observations. The use of AE has not yet been widely developed in the field of thermal oxidation. The present work aims to use AE to monitor on-line oxidation and detect physical phenomena occurring at high temperature. It focuses on two groups of materials: susceptible to catastrophic oxidation (Ti, Zy-4) and passive (Cr, Ni). This work first presents the oxidation of pure Ti at 900°C under 150 mbar of O2, as a model material. It also discusses the influence of temperature, atmosphere and surface condition on acoustic behaviour. Oxidation of Zy-4, pure Cr and Ni in pure oxygen and/or under water vapour is then studied. Elastic waves from oxidation are converted into AE signals via a sensor, which is placed outside the furnace by an alumina wave guide. A correlation between AE measures and classical results was highlighted. AE technique was able to detect catastrophic oxidation and identify failure modes of the oxide. Beyond the AE discriminated parameters, we observed two distinct populations: the so-called normal population is attributed to background noise and relaxation of stresses in the oxide associated with micro-mechanic creep, probably located at grain boundaries. AE velocity may then depend on stress level and/or their relaxation. Post-transition population is associated to oxide failure leading to breakaway oxidation. According to the transition energy level, the oxide failure modes could be identified. However, AE did not give conclusive results on passive materials. Recorded signals are mainly caused by background noise.