This experimental research investigated the behaviour of a concrete wall, with a local zone very permeable to air, subjected to accident conditions. The high level of permeability in the concrete was obtained for one specimen with a porous concrete having connected pores (intrinsic permeability of 10 −16 m 2). The aim of this work was to study comparatively, in the laboratory, the permeability of a non-cracked concrete wall under two conditions. A cylindrical specimen 1.3 m thick was used. No appreciable stresses were applied on the concrete. The first condition was at ambient temperature, under an increasing pressure (up to 0.42 MPa) of the air applied on one face of the specimen, the other one being at atmospheric pressure. The second condition was an accident scenario with simultaneous effects of temperature and gas (a mix of air and steam) pressure applied on one face, the other one remaining at atmospheric pressure and temperature. During the test, the lateral face of the cylindrical specimen was thermally isolated and made leak tight. So, a uni-dimensional experimental analysis was performed. The accident conditions consisted of a rise from ambient conditions to a temperature of 141 °C and a relative pressure of 0.42 MPa (steam pressure of 0.377 MPa and air pressure of 0.043 MPa) for 3 h, the maximum values remaining constant for several days. Thermocouples, pressure taps and moisture gauges were implanted in the specimens at the moment of casting, to provide local information about the inner wall under the simultaneous effects of temperature and steam pressure. Outside the specimen, the chamber enclosed a condenser with a humidity meter and thermocouple, so it was possible to quantify separately the saturated airflow with a flow-metre and the condensed water. During the permeability test at ambient temperature, the field of pressure was affected by the variation of local permeability of the concrete. Furthermore, some movement of interstitial water inside the specimen induced an increase of the predicted outflow, no liquid phase appeared. For the natural porous concrete, during accident testing, the field of pressure advanced slowly, the outflow progressively increased to obtain a maximum at 70 h, after this time on the one hand the airflow rate decreased and became equal to zero; on the other hand, the liquid flow rate was constant, at the end of the test, about 30 cm of the specimen was water saturated. A uni-dimensional numerical analysis was performed. The THM model (non-saturated porous media thermo-hydro-mechanic) included in Code_Aster ® was used. Two fluid phases can be considered in the material: a liquid phase (water) and a gas phase (dry air plus vapour) with the liquid/vapour phase changes. The thermodynamic aspects dealt with open systems framework, with temperature, capillary pressure and gas total pressure as variables. Observing the considered conditions (progressive saturation of the wall), the shape of sorption isotherm and permeabilities had an important influence on the results. The numerical results were in good agreement with experimental results, on the phenomenology and on flow rate through the wall. This study gave some indicators about the relations between airflow during a permeability test and gas (air + steam) flows during accident testing.