(a) REM-triggered averaging of single-unit firing rate in wakefulness, separately for epochs tagged as ‘non-visual’ (open eyes in a dark room, green) or ‘visual’ (patient watching TV/DVD or interacting with people in a well-lit room, blue). s-REMs during REM sleep are superimposed for comparison (red). Top: mean waveform of EOG traces for ‘visual’ w-REMs (blue), ‘non-visual’ w-REMs (green), and s-REMs (red) could not reveal differences between conditions. Middle: average peri-REM-time-histogram (‘visual’: N=79 units and 1,250 REMs; ‘non-visual’: N=47 units and 140 REMs) expressed in percentage change of baseline. Statistical analysis matched for sample size (n=140 REMs in both conditions) revealed a significant decrease in firing rate prior to REM onsets for both ‘visual’ and ‘non-visual’ epochs, whereas a significant increase was observed following REM onsets only for ‘visual’ w-REMs (‘visual’: P<0.05, ‘non-visual’: P=0.33, Mann–Whitney U-test). Bottom: dERP time-locked to REMs onset (‘visual’: N=287 REMs in 70 channels, ‘non-visual’: N=287 REMs in 50 channels). Number of REMs were equated across conditions (n=287). Note the negative deflection between 200 and 500 ms for ‘visual’ REMs only. Error bars show s.d. across REMs. (b) Average EOG and intracranial dERPs in one participant when performing saccades in darkness with eyes closed (cyan, N=42), eyes open in darkness (dark blue, N=47), when performing saccades to pictures presented at the periphery (green, N=55) or when being presented with images directly at the center of fixation (magenta, N=59). Top: EOG traces corresponding to the four different experimental conditions. Note the major difference in shape between the eyes closed and eyes open conditions. Bottom: MTL dERPs time-locked to the onset of saccades/image presentation. Error bars show one s.d. Note that while evoked responses following image presentation and saccades appear similar, saccades in darkness reveal a much weaker or absent modulation.