A sample of borstal boys was selected (56). All had an assessment of mental state and ratings of sociopathy and aggressiveness performed, using structured interview techniques. A group of RAMC apprentices matched for age, sex and social class, was also examined (56). Eight-channel bipolar scalp EEG recordings were taken, 4 channels (T4–T6, P4–O2, T3–T5, P3–O1) being recorded on analogue tape. The tape-recorded signals were filtered off line through a low pass filter, multiplexed and digitized on to magnetic tape. Power spectral analysis was carried out on this data, using the University of London CDC 60 600 computer. Three 30-second epochs were analysed: eyes closed, eyes open, and eyes closed. No significant EEG differences were found between the borstal boys and the control sample. Neither visual inspection nor power spectral analysis revealed any EEG differences between the borstal boys and controls. Correlation coefficients between the power at each frequency from all four channels, the violence and sociopathy ratings and the time spent in institutional care in months for each boy were computed. These showed that violence contributes not more than 1 per cent and sociopathy not more than 5 per cent to the variance of the EEG power. Duration of stay in the institution accounted for as much as 20 per cent of the variance of the EEG power with significant correlations at 7 Hz (positive) and 11 Hz (negative). Duration of stay is thus a variable that must be controlled for during the investigation of subjects in penal institutions and long-stay hospitals.