To estimate the effects of night work on the human cognitive function, P300 event related potentials (ERPs) evoked with an auditory "oddball" paradigm were recorded for 17 nurses (mean age 27.4 +/- 6.1 years; range 21-41 years) for 3 epochs; after a night of work, after a day of work and on a holiday, and for 12 elderly security guards (mean age 62.8 +/- 2.2 years; range 60-67 years) for 2 epochs; after a night of work and on a holiday. The Stress Arousal Check List (SACL) was used for all the subjects just prior to the P300 ERP recording to determine the extent of stress and the arousal grades. Fourteen of the nurses were in their twenties (mean age 24.9 +/- 2.6 years; range 21-29 years), and 3 (39.0 +/- 2.6) were older (36, 40 and 41 years). The 14 nurses were classed as the young group. All the security guards were classed as the elderly group. In the young group, the stress grade scores increased significantly (P < 0.05) and the arousal grades decreased significantly (P < 0.01) after night work as compared to the holiday values. Although statistically not significant, the amplitude of the P300 component tended to decrease after night work, whereas the latency was very stable for these 2 epochs. The P300 latencies of the 3 older nurses were as stable as those of the young group, but their amplitudes were significantly reduced after night work as compared with the holiday amplitudes (P < 0.05). In contrast, the 12 elderly security guards showed no statistically significant changes in the scores for the stress and arousal grades between the 2 epochs, after night work and holiday. Latency prolongation however, was statistically significant (P < 0.01) after night work. The P300 amplitudes for many of the elderly security guards also tended to decrease after night work, but were not statistically significant. The P300 amplitude is considered to reflect the amount of attentional resources and the latency to reflect the time needed for the cognitive process, indicating that the elderly security guards experiences slowing of the cognitive process in night work. Our results suggest that the effect of night work on the cognitive function is greater for elderly than for young workers. We conclude that P300 can be used to evaluate changes in the human cognitive function produced by night or rotating shift work and that the results provide useful information with which to plan shift schedules on the basis of worker age.