Most of the available information on the effects of impulse noise on hearing is derived from temporary threshold shift (TTS2) measurements performed 2 min after a single exposure to small-weapon noises. TTS is known to recover as a linear function of the logarithm of time when it is induced by a continuous noise of moderate intensity. Following the exposure to impulse noise, several investigators have reported individual exceptions to the log-time relation, e.g. increases in TTS during the first hour of recovery. These authors observed a 'rebound recovery function' for most of the exposed men, and they conclude that this phenomenon '... has implications for the use of TTS in the construction of damage risk criteria for hazardous noise exposure ..., a single measure, such as the widely used TTS2 may not be an adequate index of the magnitude of the TTS'. In order to thoroughly investigate in man the existence of 'delayed' TTS following the exposure to actual weapon noises, the 'French Committee on Weapon Noises' carried out the following study. Three groups of soldiers (28 subjects) wearing no hearing protection were exposed in the free field over 2 days to impulse noises produced by a rifle. Békésy audiograms were obtained from each subject just before the exposure, and at 5 min, 1 h and 4 h after exposure. All audiometric tests were carried out even when no TTS was observable in the first postexposure audiogram. A significant number of subjects showed a 'delayed TTS' and/or 'rebound recovery'. The maximum TTS was observed at 1 h after exposure, but the observation of a delayed recovery and a rebound recovery indicate that audiometric tests should be performed in all cases at least up to 4 h after the exposure. More detailed work is necessary to establish what changes may be necessary in the present damage risk criteria for impulse noises of a very high level.