The static surface tension (sigma(e)) of aqueous solutions of a fatty alcohol versus temperature exhibits a minimum. Thus, a temperature gradient which is created at the free surface of such a solution at temperatures higher than that of the minimum (sigma(e)) should induce a surface flow from the cold area to the hot one. This was indeed observed even at temperatures much lower than that of the minimum. A possible explanation is the evaporation of alcohol at the hot area where the surface tension increases, and its condensation on the cold area where the surface tension decreases, the alcohol being tranported by the gas phase. The thickness of the gas phase over the liquid could play a role in the observed flows. A device was built in which the thickness of the gas phase could be adjusted. Experiments performed with a 6.2 x 10(-3) m solution of n-heptanol indicate a lowering of the surface velocity where the thickness of the gas phase is reduced,