Satellite infra-red images taken from April-May to September-October often show persistent ribbons of cool water along the European continental slope, from West Ireland to South-West Brittany (Pingree and Mardell, 1981; Le Tareau et al., 1983; Mazé, 1983; Mazé et al., 1986). The cooling increases northwestward and follows almost exactly the top of the shelf break. This effect appears to be connected with the increase in intensity of the baroclinic internal tide field. A global model is constructed to simulate the cooling process above the edge by interaction between internal tide waves and mixing. First, a three layer model simulates the generation and propagation of nonlinear waves induced over the shelf break as a result of the propagation of the barotropic tide. Then, mixing processes associated with wind action on the sea surface, tidal friction on the sea floor, and thermocline instability when the local Richardson number becomes less than ¼ are taken into account. Numerical runs on a transect perpendicular to the shelf break show the formation of a narrow spot of cool water above the edge when wind forcing is applied. This cool spot is advected by tidal currents from the slope to the shelf and conversely. Just above the shelf break, in the area of the internal tide generation, the internal mixing is larger than that taking place on the slope and on the shelf, leading to a thickening of the thermocline. On the other hand, a gust of wind has a very different effect on surface cooling during neap tides or spring tides. Thus, when the upper mixed layer is shallow, it appears that, during spring tides, a low intensity gust of wind is able to cool surface waters, whereas the same wind during neap tides induces no cooling. Evidence of such characteristics have been observed in shelf break areas. In particular data collected during the Ondine 85 cruise in the Bay of Biscay show internal tides whose shape and amplitude can be compared with the model results. Shelf break cooling does not appear clearly in the temperature data. However, weak entrainment in the surface layer may be expected from nutrient data at spring tides (P. Le Corre, personal communication). The absence of a significant shelf break cooling can be explained by the presence of a rather deep seasonal thermocline (40 m depth in average) and weak winds during the experiment leading to low entrainment rates at spring tide on one hand, and by a positive heat budget for the ocean across the sea surface on the other hand. Therefore, a quasi-balance may be expected between cooling due to wind induced entrainment and warming due to a net surface heat flux positive in average over diurnal cycles during the experiment.