Healthy volunteers were recruited via local and online advertisements. None of the subjects reported a history of sleep disorder, psychiatric or neurological diseases, or current intake of psychotropic medication. All subjects were required to participate in two EEG recordings (one ignored and one attended oddball condition, see below) with an interval of 7 days between the recordings. These two recordings were performed in a pseudorandom order. Not all subjects participated in the second session due to lack of compliance or availability, leaving 45 subjects in the ignored and 49 subjects in the attended condition. Within these participants, those who exhibited insufficient arousal variability during the 2-h recording (i.e. too much EEG-vigilance stage A1; n = 6 in the ignored and n = 10 in the attended condition) were further excluded. As result, the final sample consisted of 39 subjects in the ignored (22 females, age = 23.90 ± 3.93) and 39 in the attended condition (24 females, age = 24.46 ± 4.44), respectively. The study was approved by the local ethics committee of the University of Leipzig (075-13-11032013). Each subject gave written informed consent prior to the first recording. All subjects received 20€ or course credits (psychology students) for their participation.
The 2-h EEG recordings began between 1 and 4 p.m. in a light-dimmed and sound attenuated room. The temperature in the booth was maintained around 25 °C at the beginning of each recording. For each individual, the time of assessment was the same in both sessions. During the EEG recording, subjects lay comfortably on a lounge chair while standard (500 Hz) and deviant (1000 Hz) tone were presented in an oddball sequence with stimuli probabilities of 80 and 20% respectively. In the ignored condition, subjects were instructed to ignore the tones, while in the attended condition they performed a simple cognitive task such as pressing a button to target stimuli. At the beginning of each recording, the body position was changed from upright to laid-back. During the recording, subjects were instructed to close their eyes, relax and not fight against an urge to sleep. When subjects did fall asleep, they were woken up after 5 min and asked to answer a common question (e.g. today’s date) before they were allowed to continue the task. This process was repeated until the end of the experiment in order to acquire enough data from each arousal state.
EEG-recording and EEG-vigilance staging
The EEG was recorded at 1000 Hz with Ag/AgCl electrodes and DC amplifiers (QuickAmp; Brain Products GmbH, Gilching, Germany) from 31 sites (Fp1, Fp2, F3, F4, F7, F8, Fz, FC1, FC2, FC5, FC6, C3, C4, T7, T8, Cz, FT9, FT10, CP5, CP6, TP9, TP10, P3, P4, P7, P8, Pz, O1, O2, PO9, PO10) according to the extended international 10–20 system using EasyCap (EASYCAP Brain Products GmbH, Gilching, Germany), and referenced against common average. Impedance of each electrode was kept below 10 kΩ. Bipolar electrodes were placed laterally to the left and right eyes to monitor horizontal eye movements and above and below the right eye to monitor vertical eye movements.
EEG data were analyzed using BrainVision Analyzer 2.1 software (Brain Products GmbH, Gilching, Germany). First, the EEG raw data were pre-processed according to standard operating procedures (see VIGALL manual  or refer to Additional file 1). After that, all 1-s EEG-segments were classified into seven different EEG-vigilance stages using VIGALL 2.1 (available at http://research.uni-leipzig.de/vigall/).
To assess the R–R intervals of HR (in ms), an electrocardiogram (ECG) was recorded at a 1000 Hz sampling rate using a bipolar channel of the QuickAmp amplifier. Electrodes were placed on both forearms. R-peaks were marked using the CB correction module of BrainVision Analyzer (Brain Products GmbH, Gilching, Germany). The results were visually checked and corrected if necessary.
To assess SCL (in µSiemens), a bipolar channel of the QuickAmp amplifier was used with a constant voltage of 0.5 V (GSR module, Brain Products GmbH, Gilching, Germany). Two Ag/AgCl electrodes (with an overall diameter of 13 mm) were placed at the thenar and hypothenar of the non-dominant hand. A low pass filter of 1 Hz was applied to exclude phasic components of the electrodermal activity due to stimuli presentation (in both conditions) and response (only in the attended condition).
Segments identified as artifacts in the EEG were also marked as artifacts in the ECG and SCL channels. Only artifact-free segments were used in further analyses. VIGALL also provides calculations for R–R intervals and SCL values: R–R interval was computed as the mean of the R–R intervals across three consecutive artifact-free 1-s segments. The HR was calculated for each segment (indexed by 60,000/R–R intervals in ms). SCL value was computed as mean of all data points in each 1-s segment.
In order to account for the considerable degree of variability in SCL raw values between subjects, for each subject SCL values were z-transformed against the mean and standard deviation over 2 h when assessing overall SCL differences between EEG-vigilance stages. SCL values were also z-transformed against the mean and standard deviation in each corresponding time block for each subject when differences within each time block were examined.
A minimum criterion of 10 epochs for each EEG-vigilance stage was set in order to obtain reliable HR and SCL values. Subjects with an insufficient number of epochs were excluded from the comparisons of respective stages. This step resulted in different sample sizes for each EEG-vigilance stage. Some stages, such as A1 and B1, which were frequent, contained more subjects, whereas others, especially A3 and C, which rarely occurred, had fewer subjects (see Additional file 2).
To analyze differences in ANS activity between EEG-vigilance stages across the entire 2 h and within each time block, paired sample (within subject) t tests were used. The different sample sizes in the EEG-vigilance stages precluded adequate stage comparisons (due to listwise deletion) by repeated measures analyses of variance (rmANOVAs). Hence, comparisons were only made for pairs with a sufficient sample size (n > 10). All statistical analyses were conducted using IBM SPSS Statistics version 20 (IBM, Armonk, NY, USA).
The time-on-task effect on HR and SCL was analyzed with rmANOVAs. In these analyses, the arousal stage was kept constant by restricting the analyses to a respective EEG-vigilance stage across the four consecutive time blocks (min 1–30, min 31–60, min 61–90, min 91–120). However, these analyses could only be performed in the EEG-vigilance stages A1, A2, B1 and B2/3 in the ignored condition and in stages A1, B1 and B2/3 in the attended condition because the sample sizes (n ≤ 10) across all four blocks were insufficient in the remaining stages. When significant main effects were present, post hoc tests for multiple comparisons were conducted with adjustments for significance level using the Bonferroni method (p < 0.0125). When analyzing the time-on-task effect on SCL we did not z-transform the data for two reasons. First, the time course and percentage of low EEG-vigilance stages in each time block may have increased with time (see Additional file 2), possibly resulting in a smaller z-score in low stages in earlier versus later time blocks. This could have led to an artificial time effect. Second, because we used rmANOVAs to examine within-subject effects over time, the inter-individual variations likely had little influence on the results.