REM-triggered averaging of multiple signals of interest in wakefulness (left) and REM sleep (middle), as well as during controlled visual stimulation (right). Intracranial recordings (c–e) focused on MTL regions. (a) Mean (± s.e.m.) waveform of EOG traces (light colours: amplitude; dark colours, velocity). (b) Mean scalp ERPs computed at Pz. (c) Mean depth EEG potentials (dERP), (d) mean peri-REM time histogram for ‘biphasic’ MTL neurons in wakefulness (n=79 units, n=16,738 REMs), REM sleep (n=47 units, n=4,510 REMs), and visual stimulation (n=196 units, n=23,248 trials). Note the reduction in firing rate (−400 to 0 ms) and increased activity (150–550 ms) relative to REM onsets, and similar increased activity upon visual stimulation (right). Yellow shading highlights a time window in which negative components in the dERPs are associated with increase neuronal activity in all three conditions. Bars illustrate statistical deviances from baseline corrected for multiple comparisons (Methods). (e) Representative examples of activity in individual neurons (black inset: spike waveform) around REM onsets in wake and sleep (same neuron from parahippocampal gyrus) and visual stimulation (different neuron but from same brain region). Peri-REM-time histograms were computed on 120 ms bins (100 ms overlap). Mean firing rate of each neuron is marked with a dashed horizontal line, grey bars above illustrate statistically significant deviations from baseline (t-test, P<0.05). Error bars denote the s.e.m. across patients (a,b), dEEG channels (c) or REMs (d,e).