An automatic rat sleep classification system is described which records and analyses bioelectrical signals from 32 rats over extended periods of time. At present this system is used routinely for the screening of drug effects on sleep. The analysis is based on 3 signals, the parieto-occipital EEG, nuchal EMG and a movement indicator signal. The on-line analysis is done per epoch of 2 sec and involves power spectral analysis of the EEG and rectification and integration of the EMG and movement signals. The automatic sleep staging into 6 stages (active and quiet waking; quiet, deep, pre-REM and REM sleep) is performed off-line. Parameters derived from a discriminant analysis of visually scored tracings of individual rats constitute the basis for the automatic scoring procedure. The movement index is used to discriminate between active and quiet waking, while the use of the EMG level improves the separation of waking and REM sleep. After the construction of hypnograms from these computer scorings a set of parameters can be extracted which characterizes the sleep-waking behavior of each individual rat. These parameters are then used to compare statistically the 2-4 treatment groups which make up each experiment of 32 rats. Experimental validation of the system is reported in an accompanying paper.