The tensile strength of water measured dynamically is less than that measured statically. It is claimed that this anomaly, which has been known for a long time, can be accounted for if one assumes that the transition layer at the free surface of a liquid contains a region for which the effective velocity of sound is very small. This region should be a source of cavitation nuclei, thus lowering the tensile strength. The hydrodynamics of reflection of pressure pulses by such a region are investigated and the imperfect reflection of pressure pulses by the free surface observed by Sedgewick and Trevena (1976) can be explained on the same hypothesis.