Data acquisition

The motor activity of participants was continuously recorded with actigraphy (Micro Motionlogger Sleep Watch, Ambulatory Monitoring, Inc., Ardsley, NY) in order to control their sleep timing and duration; a given participant wore the same device during the whole 21-day period. The actigraphs collected data in 1-min intervals in two available modes: the zero-crossing mode (ZCM; i.e., frequency of activity signal crossing a zero threshold) and the proportional integrating measure (PIM; i.e., intensity or under the signal curve), see Fig 2A–2E. The EEG recordings were performed in the laboratory, whereas the actigraphy data were collected continuously both in the laboratory and in the natural environment.

A-B: a sample of the first two days (black) and three nights (red) of activity recording of a single subject in ZCM and PIM modes. C: density histogram of ZCM versus PIM activity (vertical and horizontal axes, respectively) collected from all subjects on all days (BASE, SR, and RCV) with the highest concentration of data around point (0, 0); the relation between ZCM and PIM is non-linear with ZCM saturating around 250. D-E: separate histograms of ZCM and PIM recordings, respectively; ZCM has two peaks, at 8 and 252, marked by red triangles corresponding to diurnal and nocturnal activity, while PIM is unimodal. F: rest and activity periods (red and black stripes) are defined as segments of activity continuously smaller or greater than a threshold (red dashed line); lengths of these segments are collected into duration distributions in G. G: complementary cumulative distributions of rest and activity durations in SR period of all subjects in log-log scale; γ is the exponent of power-law tail of rest CCDF (slope of the blue dashed line). Inset: γ exponents in baseline, sleep restriction and recovery conditions. H: γ exponents for each day of the experiment obtained from rest duration CCDF of all subjects.

Each day the subjects’ performance in terms of cognitive information processing was measured in a classic Stroop test. The subjects were asked to decide whether the name of the color matches with the ink in which it is written (congruent conditions) or not (incongruent conditions). So the more automated task (reading the word) interfered with the less automated task (naming the ink color) resulting the difficulty in inhibiting more automated process known as the Stroop effect. In total there were N_s = 432 randomly shuffled stimuli presented in 3 separate blocks (144 stimuli each) with a short break in between each block, as indicated in Fig 3. Half of the stimuli were congruent, and half were incongruent. The inter-stimulus interval between each stimulus was between 1500 ms and 3500 ms (in steps of 400 ms, with 2500 ms on average); the entire task lasted 22 minutes and 11 seconds on average.

Vertical dashed lines indicate pauses between blocks of the Stroop task. The curves represent means over subjects and days within a given condition. and the shaded regions 95% confidence intervals; The straight lines are linear fits; the RT slope is significantly higher in SR. Red triangles point to RTs dropping just after the pause, which happens in all conditions.

All participants underwent one training session before beginning of the experiment in order to avoid the learning effect.

Each day, participants’ brain activity was being monitored with an EEG (64 electrodes, 500 Hz sampling frequency; Geodesic Sensor Net, EGI System 300, USA), while they were in a resting state (for 8 minutes with open eyes, RSeO, followed by 8 minutes with closed eyes, RSeC), and next performing Stroop test. The EEG session for a given participant took place in the laboratory at the same time every day, in the morning (8–11 a.m.) or evening (6–9 p.m.) according to the participant’s diurnal preferences.