Sleep Deprivation and Time-on-Task Performance Decrement in the Rat Psychomotor Vigilance Task

Abstract

Study Objectives:

The rat psychomotor vigilance task (rPVT) was developed as a rodent analog of the human psychomotor vigilance task (hPVT). We examined whether rPVT performance displays time-on-task effects similar to those observed on the hPVT.

Design:

The rPVT requires rats to respond to a randomly presented light stimulus to obtain a water reward. Rats were water deprived for 22 h prior to each 30-min rPVT session to motivate performance. We analyzed rPVT performance over time on task and as a function of the response-stimulus interval, at baseline and after sleep deprivation.

Setting:

The study was conducted in an academic research vivarium.

Participants:

Male Long-Evans rats were trained to respond to a 0.5 sec stimulus light within 3 sec of stimulus onset. Complete data were available for n = 20 rats.

Interventions:

Rats performed the rPVT for 30 min at baseline and after 24 h total sleep deprivation by gentle handling.

Measurements and Results:

Compared to baseline, sleep deprived rats displayed increased performance lapses and premature responses, similar to hPVT lapses of attention and false starts. However, in contrast to hPVT performance, the time-on-task performance decrement was not significantly enhanced by sleep deprivation. Moreover, following sleep deprivation, rPVT response times were not consistently increased after short response-stimulus intervals.

Conclusions:

The rat psychomotor vigilance task manifests similarities to the human psychomotor vigilance task in global performance outcomes, but not in post-sleep deprivation effects of time on task and response-stimulus interval.

Citation:

Oonk M, Davis CJ, Krueger JM, Wisor JP, Van Dongen HPA. Sleep deprivation and time-on-task performance decrement in the rat psychomotor vigilance task. SLEEP 2015;38(3):445–451.

INTRODUCTION

Sleep deprivation in humans results in performance deficits within multiple cognitive domains, including attention, memory, and executive functions.^1,2 The neurobiological processes involved in these deficits remain poorly understood³; animal models of sleep deprivation and cognitive performance are needed to help elucidate them. The effects of sleep deprivation on animals in tasks involving mnemonic functions, including object recognition⁴ and spatial memory,^5–7 have been characterized. However, the most profound and operationally relevant effects of sleep deprivation on human cognitive performance involve deficits in sustained attention.⁸ A few animal performance tasks that capture sleep deprivation-induced attentional processes have been developed,^9–12 and here we investigate one such task.

In human studies, a gold standard performance task for investigating the effects of sleep deprivation on sustained attention is the human psychomotor vigilance task (hPVT).⁸ The hPVT is a simple reaction time task requiring subjects to respond as quickly as possible to a visual stimulus appearing at random intervals between 2 sec and 10 sec across a 10-min task duration.¹³ Conventionally, four categories of hPVT responses are distinguished¹⁴: normal responses, defined as response times (RTs) < 500 msec; lapses of attention, defined as RTs ≥ 500 msec; nonresponses or time-outs, defined as RTs reaching 30 sec (at which point a new stimulus presentation is initiated); and false starts or premature responses. Sleep deprivation causes an increase in lapses of attention, false starts, and, after prolonged sleep deprivation, nonresponses.¹⁵ Underlying these changes is a systematic skewing of the RT distribution, such that longer RTs are substantially increased.¹⁶

On the hPVT, RTs become longer and more variable with increasing task duration— this effect is referred to as the time-on-task effect. The time-on-task performance decrement is exacerbated by sleep deprivation.^15,17 It has been posited that the time-on-task effect and the sleep deprivation-induced performance decrements share neurobiological substrates.^18,19

In addition, RTs on the hPVT are longer when preceding response-stimulus intervals (RSIs) are shorter. The RSI effect appears to be independent of sleep deprivation and time-on-task effects; no significant interaction of the effect with sleep deprivation or time on task has been found.²⁰

Christie and colleagues^10,21 developed a rat analog of the hPVT, an operant task requiring rats to respond with a simple behavior to a brief stimulus presented at quasi-random intervals. The rat psychomotor vigilance task (rPVT) is believed to be a model of sustained attention, because continually focused attention to detect the stimuli is required in order to be rewarded. Specifically, after a 0.5-sec light stimulus, rats have a 2.5-sec window (hold period) to poke their nose into a water delivery port to receive a water reward. An omission, defined as a failure to respond within 3 sec of the onset of the light stimulus, results in a 10- sec time-out, during which no new stimuli are presented and the chamber house light is turned off. A premature response (i.e., false start), defined as a response during the intertrial interval (ITI), results in a 10-sec timeout as well (Figure 1). A response during a time-out is also counted as a premature response, but does not trigger an additional time-out. The total task duration of the rPVT is 30 min. To motivate performance on the rPVT, rats are water deprived for 22 h prior to each session.

Figure 1.

Flowchart of rat psychomotor vigilance task (rPVT) session timing and stimulus/response contingencies. At session onset, the house light turns on and the intertrial interval (ITI) starts, varying between 3 and 7 sec. At the end of the ITI, the stimulus light turns on for 0.5 sec followed by a 2.5-sec hold period—the animal has 3 sec to respond. When the rat nose-pokes, it receives a water reward, and after the hold period a new ITI starts. If the rat neglects to nose-poke during the hold period (omission), the house light turns off for 10 sec (time-out) and the ITI is reset. After 10 sec the house light turns back on and the ITI starts again. When the rat nose-pokes during the ITI (premature response), it also receives a 10-sec time-out. When the rat nose-pokes during the timeout, nothing happens (but the nose-poke is registered as a premature response).

Behavioral outcomes for the rPVT are analogous to those for the hPVT and include mean RT, number of correct responses, number of lapses, and number of premature responses. As with the hPVT, RTs are calculated as the time between light onset and the response (nose-poke in rats). Because of the large baseline variability in response speed between individual rats, lapses are not scored relative to a fixed threshold (e.g., RTs ≥ 500 ms as typically used in the hPVT), but are defined as trials with RTs at least twice the rat's average baseline RT.¹⁰ In earlier work on the rPVT, lapses were not included in the calculation of mean RT.¹⁰ Here we do include lapses in the calculation of mean RT in order to retain compatibility with the hPVT.

The number of correct responses on the rPVT includes all correct responses, including responses that are lapses, but not omissions. The number of lapses on the rPVT does include omissions (the number of omissions or nonresponses is not analyzed separately as is sometimes done for the hPVT). Premature responses on the rPVT, like false starts on the hPVT,¹⁵ may be interpreted as indices of motivated responding.¹⁰

In Fischer Norway rats, Christie et al. found a slowing of mean RTs and an increase in lapses after sleep deprivation,¹⁰ similar to the performance degradation observed on the hPVT after sleep loss.¹⁵ In the current rat sleep deprivation study, we extend the work of Christie et al.¹⁰ to Long-Evans rats and investigate the time-on-task and RSI effects on the rPVT.

MATERIALS AND METHODS

Animals

We used 24 male Long-Evans rats (Charles River, Wilmington, MA), 6 w of age at start of training. Rats were housed, in pairs, in an Association for Assessment and Accreditation of Laboratory Animal Care-approved vivarium at 21 ± 1°C. They were on a 12:12 h light:dark cycle and were fed ad libitum. Water was restricted to 2 h per day during and following rPVT bouts for the duration of the study (including training days). Body weight was monitored twice weekly; all rats displayed normal growth throughout the study. The study was approved by Washington State University's Institutional Animal Care and Use Committee and was consistent with National Institutes of Health guidelines.

Equipment

Operant chambers (model #80004, Lafayette Instruments, Lafayette, IN), equipped with a water delivery port with infrared beams for nose-poke detection, were used for the study. Two lights were simultaneously presented as the operant stimulus; one was located centrally above the water delivery port and the other inside the port, to ensure stimulus detection even if the rats were poking their nose into the port. A 20-μL water reward was delivered when a task-appropriate nose-poke occurred. ABET II software version 2.15 (Lafayette Instruments, Lafayette, IN) was used to run the operant schedules, collect data, and analyze responses.

Behavioral Training

Training sessions were 30 min in duration. Each rat was trained daily, at the same time each day, between ZT3 and ZT8. Training began with two acclimatization sessions during which the rats were placed in the operant chambers without stimulus or reward presentation. Following acclimatization, rats were trained to poke their nose into the water delivery port for a water reward. When rats nose-poked to obtain the water reward, a light was introduced as a discriminative stimulus signaling operant-dependent availability of the reward. Initially, a 30-sec light stimulus was used. After rats were able to obtain more than 90 correct responses and fewer than 25 omissions 3 days in a row, the stimulus duration was incrementally reduced from 30 sec to 15 sec, to 10 sec, to 5 sec, to 2.5 sec, to 1 sec, and finally to 0.5 sec. A random ITI, varying from 3 sec to 7 sec in steps of 1 sec, was introduced halfway through the training when the stimulus duration was 5 sec.

Advancement criteria for shorter stimulus durations described originally by Christie et al.¹⁰ were more than 100 correct responses and fewer than 20 omissions. Of our Long-Evans rats, 30% were unable to meet the criteria. We therefore adjusted the advancement criteria to more than 90 correct responses and fewer than 25 omissions. Three rats were excluded because they did not meet these training criteria. Rats included in the experimental sessions were all able to respond with a nose-poke to a 0.5-sec stimulus according to the advancement criteria.

Experimental Sessions

After training, rats were randomly divided into two groups, each of which underwent two experimental sessions using a crossover design. During the first experimental session, one group (n = 12) performed the rPVT after undisturbed, ad libitum sleep (baseline), whereas the other group (n = 9) underwent sleep deprivation for 24 h prior to rPVT testing. Rats were in their home cages during sleep deprivation. Novel objects and gentle handling were used to keep the rats awake. See the online supplement for an assessment of the effectiveness of our sleep deprivation method.

The second experimental session was conducted 1 w later with conditions reversed. The experimental sessions occurred at the same time as each rat's daily training sessions, between ZT3 and ZT8. In the week between the two experimental sessions, rats underwent daily rPVT maintenance sessions to sustain performance levels.

Rats were monitored with video cameras during experimental sessions. None of the rats fell asleep during the rPVT in the baseline condition, but seven rats fell asleep during the rPVT after sleep deprivation and as such showed large numbers of omissions. The rats exhibited, on average, 1.8 sleep episodes throughout the 30 min rPVT session, with an average duration of 3.1 min. Per Christie et al.,¹⁰ rats showing more than 50 omissions were excluded from rPVT data analyses; this occurred in only one rat (56 omissions). The other six rats that fell asleep during the rPVT made fewer than 45 omissions and data from these animals were retained. The final sample size was n = 20.

Data Analyses

Global rPVT performance based on mean RT, number of correct responses, number of lapses, number of premature responses, and total number of stimuli presented in each 30-min rPVT session were analyzed using mixed-effects analysis of variance (ANOVA) with condition (baseline versus sleep deprivation) as repeated factor, order of conditions as a covariate, and a random effect on the intercept.

To examine time-on-task effects, the mean RT data, number of correct responses, number of lapses, and number of premature responses were binned in 5-min segments. Segment time was included in the ANOVA as an additional repeated factor, and a condition by segment time interaction was assessed.

To examine the effect of RSI, the mean RT data were binned according to the preceding RSI into nine categories (< 5 sec, 5–6 sec, 6–7 sec, 7–8 sec, 8–9 sec, 9–11 sec, 11–16 sec, 16–31 sec, and ≥ 31 sec). RSI was defined two different ways: (1) the time from the last correct response to the next stimulus onset (RSI-CR), and (2) the time from the last response, whether correct or erroneous, to the next stimulus onset (RSI-LR). For analysis, RSI categories (rather than segment time) were included in the ANOVA as an additional repeated factor, and a condition by RSI interaction was assessed, for both RSI-CR and RSI-LR.

For mean RT, the data points of seven 5-min segments (distributed over five rats), 51 RSI-CR categories (distributed over 18 rats) and 27 RSI-LR categories (distributed over 14 rats) were missing because there were no correct responses in these segments or categories. Five rewarded responses with RT ≤ 10 ms were considered premature responses, and those data points were excluded from all analyses. Outliers were defined as data points more than three standard deviations removed from the mean. The mean RT of one 5-min segment was an outlier and was therefore excluded from all analyses. For lapses, the data points of three 5- min segments were outliers and were therefore excluded from the segment time and main analyses. For premature responses, the same applied for one 5-min segment. The excluded outliers represented 1.1 % of the overall data set.

RESULTS

Overall 30-min rPVT performance was degraded after 24 h of sleep deprivation compared to baseline, as shown in Figure 2. Although the difference in mean RT did not reach statistical significance (F_1,18 = 3.6, P = 0.075), the number of correct responses was significantly decreased (F_1,18 = 35.4, P < 0.001) and the numbers of lapses and premature responses were significantly increased after sleep deprivation (F_1,18 = 11.6, P = 0.003, and F_1,18 = 16.1, P < 0.001, respectively). The total number of stimuli presented (not shown) was also significantly decreased after sleep deprivation compared to baseline (94.2 versus 127.1, respectively; F_1,18 = 20.3, P < 0.001).

Figure 2.

Means and standard errors of overall 30 min rat psychomotor vigilance task performance in the baseline condition (white) and after 24-h sleep deprivation (gray). * P < 0.01. ** P < 0.001.

Time-on-task effects are shown in Figure 3. Mean RTs tended to increase over the 30- min sessions (F5,200 = 2.1, P = 0.065), but there was no significant condition by time interaction (F5,200 = 1.4, P = 0.22). The number of correct responses decreased over time on task after an initial increase (F5,208 = 5.8, P < 0.001), but again there was no significant condition by time interaction (F5,208 = 0.2, P = 0.95). Likewise, the number of lapses increased over time on task (F5,205 = 5.9, P < 0.001), but without significant condition by time interaction (F5,205 = 0.1, P = 0.99).

Figure 3.

Means and standard errors of rat psychomotor vigilance task performance over time on task during baseline (dotted line) and after 24-h sleep deprivation (solid line). The average number of premature responses includes responses made during time-outs; these responses did not result in additional time-outs.

The number of premature responses decreased over time on task (F5,207 = 5.6, P < 0.001). Moreover, there was a significant condition by time interaction (F5,207 = 2.6, P = 0.025). Whereas premature responses were relatively stable over time on task at baseline, they were initially increased and declined over time on task after 24-h sleep deprivation.

The effect of RSI on mean RT is shown in Figure 4. Mean RT increased with greater RSI-CR (F8,271 = 7.3, P < 0.001) and RSI-LR (F8,295 = 8.3, P < 0.001), although for RSI-CR there was an initial decrease (Figure 4, top panel). There was no significant condition by RSI interaction for RSI-CR (F8,271 = 0.7, P = 0.69) or RSI-LR (F8,295 = 0.2, P = 0.99). These analyses were repeated using only the RT data from the first six RSI categories (< 5 sec, 5–6 sec, 6–7 sec, 7–8 sec, 8–9 sec, 9–11 sec) in order to exclude the longer RSIs increased due to 10-sec time-outs. This confirmed that the initial decrease across RSICR categories (Figure 4, top panel) was statistically significant (F5,184 = 2.7, P = 0.024); the condition by RSI interaction for RSI-CR remained non-significant (F5,184 = 0.1, P = 0.98). However, for RSI-LR, the RSI effect was not significant across RSIs < 11 sec (Figure 4, bottom panel; effect of RSI: F5,199 = 0.9, P = 0.48; condition by RSI interaction: F5,199 = 0.5, P = 0.76).

Figure 4.

Means and standard errors of rat psychomotor vigilance task mean response times as a function of the preceding RSI during baseline (dotted line) and after 24 h sleep deprivation (solid line).

DISCUSSION

Performance on the rPVT was impaired after 24-h sleep deprivation as compared to baseline: the number of correct responses decreased and the numbers of lapses and premature responses increased (Figure 2). The observed sleep deprivation-induced impairments were comparable to those occurring in humans after sleep deprivation.¹⁵ They were also similar to deficits observed in earlier rat sleep deprivation studies using various sustained attention tasks.^9–12 However, an increase in premature responses after sleep deprivation, as observed here, has not been previously reported in rats.

The reduced number of correct responses after sleep deprivation could be interpreted as evidence of reduced motivation.¹⁰ However, because of the increased lapses and especially the increased premature responses (and associated 10-sec time-outs), the total number of stimuli presented was significantly reduced after sleep deprivation as well. The reduced number of correct responses thus cannot be interpreted independently of the other changes in performance on the rPVT. Moreover, on the hPVT, increased premature responses (false starts) after sleep deprivation have historically been interpreted as evidence of increased compensatory effort indicative of heightened motivation.¹⁵ The current data do not resolve this issue unambiguously, but an explanation of the findings solely in terms of reduced motivation seems unlikely.

Similar to hPVT data, baseline rPVT performance showed a decrement over time on task for correct responses (which decreased) and lapses (which increased). Yet, in contrast to hPVT data, mean RT was relatively unaffected and premature responses showed an unanticipated decrease over time on task. Moreover, sleep deprivation did not accelerate the performance decrement over time on task.¹¹ The only significant interaction of sleep deprivation with time on task in our study involved an initial increase and subsequent decrease of premature responses over time on task after sleep deprivation (Figure 3). In humans, premature responses (false starts) increase over time on task after sleep deprivation.¹⁵

In this study, the effect of RSI on mean RT was dominated by increasing mean RT for RSIs longer than approximately 11 sec (Figure 4). These longer RSIs, which occurred on the rPVT because of 10-sec time-outs, are not possible on the hPVT, so whether humans would show the increased mean RT with prolonged RSIs is not known. For RSIs shorter than approximately 11 sec, when RSIs were measured from the last correct response to the next stimulus onset (i.e., RSI-CR), an RSI effect of greater mean RT for the shortest RSIs was seen across RSIs—both at baseline and after sleep deprivation (Figure 4, top panel). This resembled the RSI effect documented for the hPVT, which is similarly independent of sleep loss.

On the hPVT, the effect of greater mean RT for shorter RSIs is only seen for RSIs below approximately 7 sec.^20,22 This difference with the rPVT may reflect fundamental differences between humans and rats in the neurobiological substrates of task performance, but it may also be related to differences between task characteristics of the hPVT and rPVT. It is believed that cognitive processing time needed for response preparation is involved in causing the RSI effect.²⁰ After the presentation of the 0.5-sec light stimulus, rats had a 2.5-sec hold period to poke their nose into the water delivery port and receive the water reward. During part or all of this hold period, rats may not have been preparing to respond to the next stimulus. The effective RSI range over which response preparation plays a role may have been longer in the rPVT for that reason.

Yet, on the rPVT, more than one response after a stimulus is not uncommon, and this raises the question of how to best define the RSI. Given the hypothesized role of response preparation in the RSI effect,²⁰ the most recent response preceding a stimulus would seem to be the most relevant. When RSIs were measured from the most recent response to the next stimulus onset whether or not that response was correct (i.e., RSI-LR), however, the hPVT-like RSI effect was no longer observed on the rPVT (Figure 4, bottom panel). The presence of a hPVT-like RSI effect on the rPVT thus remains unclear due to definitional ambiguity.

The differences between rPVT performance and characteristic hPVT performance may limit the generalizability of rPVT study findings (e.g., concerning underlying brain mechanisms) to humans. This is particularly relevant with regard to the time-on-task effect and its interaction with sleep deprivation. For example, according to a recent theory,^19,23–25 the time-on-task effect may be a consequence of local, use-dependent sleep in neuronal networks subserving cognitive processes involved intensively in the performance of a task. Per the theory, as the duration of the task progresses, the likelihood of local sleep and attendant performance degradation increases, and prior sleep loss enhances this effect by promoting sleep both locally and globally. However, our study did not show these phenomena in the rPVT as they have previously been documented for the hPVT. Therefore, before the rPVT can be used to examine the theory and the attendant brain mechanisms, methodological issues need to be considered.

Several methodological differences between the hPVT and the current rPVT paradigm are noteworthy. Rat performance was motivated by using water rewards following a period of water deprivation. The use of incentives during task performance may have reduced the effect of sleep deprivation on the time-on-task decrement, as has been reported for humans.²⁶ Also, across rPVT task duration, rats may have become increasingly satiated, thereby decreasing the thirst-induced motivation to perform. This could be reflected in the decrease in premature responses over time on task (Figure 3). However, it would remain to be explained then why this decrease was most prominent after sleep deprivation.

In the rPVT, after the 0.5-sec light stimulus, there was a 2.5-sec window (hold period) for the rats to obtain the water reward. Thus, time pressure would not seem to be as critical a factor for performance as in the hPVT. Also, performance errors resulted in 10-sec time-outs, which increased the RSI for the next stimulus and may thus have decreased the intensity of task performance. In addition to potentially affecting the RSI effect in the rPVT (Figure 4), these factors would have dampened the time-on-task effect if it is indeed, as theorized,¹⁹ a use-dependent phenomenon. Moreover, because prior sleep deprivation increases performance errors (Figure 2), the dampening of the time-on-task effect should have been stronger after sleep deprivation than at baseline. This could have counteracted any amplification of the time-on-task effect by sleep deprivation and, given that statistical power for interactions is relatively low,²⁷ could explain the lack of significant interactions in rPVT outcomes (Figure 3).

Substantial effects of condition on overall rPVT performance (Figure 2) would appear to rule out insufficient duration of sleep deprivation (24 h) and circadian timing of rPVT testing (between ZT3 and ZT8) as potential limitations of our study design (see also the online supplement). However, the method of sleep deprivation, i.e., gentle handling, may have influenced our results. Sleep deprivation has been reported to induce a manic state in rats,²⁸ and handling by a human investigator increases stress-like responses.²⁹ The initial increase in premature responses in the sleep deprivation condition could reflect hyperactivity introduced by the method of sleep deprivation.

Our findings indicate that the rPVT shows promise as an animal model for the hPVT, especially when data are aggregated over time on task and presented as overall performance outcomes. However, differences between observations in the present rPVT paradigm and well-documented characteristics of hPVT performance—in particular with regard to the time-on-task effect—need to be resolved to ascertain generalizability of rPVT study findings to humans. The unexpected temporal pattern of premature responses after sleep deprivation, the absence of an interaction between sleep deprivation and the time-on-task effect in correct responses and lapses, and the ambiguity in the RSI effect indicate the need to adjust several parameters of the rPVT to improve its experimental validity.

Some of the features of the rPVT paradigm may be inherent to operant conditioning. In ongoing research, we are investigating whether unconditioned, self-motivated sleep deprivation through intracranial self-stimulation is a viable solution. Animal studies are needed to elucidate the cellular and molecular mechanism mediating performance impairment due to sleep deprivation. Building on the pioneering work of Christie and colleagues,¹⁰ the further development of a rodent model of sleep loss-induced performance impairment that is analogous to a well-established human performance task such as the hPVT will therefore be worthwhile.

DISCLOSURE STATEMENT

This was not an industry supported study. This work was supported by NIH grant R21NS085605 (CJD), R01NS025378 (JMK) and an Elliot D. Weitzman, M.D. Research Grant from the Sleep Research Society Foundation (HVD). The authors have indicated no financial conflicts of interest.

SUPPLEMENTAL MATERIAL

To examine the effectiveness of the sleep deprivation method used in the rat psychomotor vigilance task (rPVT) experiment of our main study, we conducted a polysomnographic experiment using the same 24-h sleep deprivation methods. Electroencephalogram (EEG) and electromyogram (EMG) were recorded for the assessment of sleep/wake states. The study was approved by Washington State University's Institutional Animal Care and Use Committee and was consistent with National Institutes of Health guidelines.

Male Long-Evans rats (n = 8; 9–10 w old) were individually housed under standard vivarium conditions (21 ± 1°C; 12 h light:dark cycle), with ad libitum food and water. Rats were anesthetized using intramuscular ketamine-xylazine (87 and 13 mg/kg, respectively) and surgically implanted with EEG electrodes over the left frontal and parietal cortices, and over the cerebellum, which served as a ground. An EMG electrode was placed in the dorsal neck muscles. The electrodes were secured to the skull with dental cement. Rats were given a minimum of 7 days to recover from surgery.

Rats were placed in sound-attenuated chambers and allowed to habituate to EEG/EMG recording cables for 2 days prior to baseline recording. Baseline EEG and EMG signals were recorded for 24 h starting and ending at ZT6. The EEG was high-pass and low-pass filtered at 0.1 Hz and 100 Hz, respectively, and the EMG was high-pass and low-pass filtered at 30 Hz and 3 kHz, respectively.

After the 24-h baseline period, rats were sleep deprived for 24 h using novel objects and gentle handling— the same methods used for sleep deprivation in the rPVT experiment. Note that during the rPVT experiment, sleep deprivation started following the baseline rPVT sessions, which were run in series; the first rats starting at ZT3 and the last rats ended at ZT8. Here, ZT6 was chosen as a representative start time that allowed ZT alignment of the baseline, sleep deprivation, and recovery periods within and between rats and comparison with published literature.

EEG and EMG recordings continued throughout the 24-h sleep deprivation period and for the 24-h recovery period after sleep deprivation. The signals were sampled at 128 Hz and manually scored in 10-sec epochs as wakefulness, rapid eye movement sleep (REMS) or non-REMS (NREMS) using Sleep Sign Software (Kissei Comtec Co., Matsumoto, Japan). EEG slow wave activity (SWA; 0.5–4 Hz) was determined by fast Fourier transforms of Hanning window-filtered artifact-free NREMS epochs.

Percentage of time scored as wakefulness during the 24-h baseline and sleep deprivation periods was analyzed using mixed-effects analysis of variance (ANOVA) with a fixed effect for condition (baseline versus sleep deprivation). In addition, time scored as wakefulness during the 24-h sleep deprivation period was analyzed in 2-h time bins using mixed-effects ANOVA with a fixed effect for time. Time spent in NREMS and REMS during the 24-h baseline and recovery periods, as well as SWA, were analyzed in 2-h time bins using mixed-effects ANOVA with fixed effects for condition (baseline versus recovery), time and their interaction. All mixed-effects ANOVAs had a random effect on the intercept.

Wakefulness was significantly increased during the 24-h sleep deprivation period compared to the 24-h baseline period (92.2% vs. 51.1%; F_1,7 = 680.1, P < 0.001). Time awake gradually decreased across the sleep deprivation period, with a brief increase at the end of the dark phase followed by a further decrease (F11,77 = 11.3, P < 0.001; Figure S1).

Figure S1

Means and standard errors of time scored as wakefulness during the 24-h sleep deprivation period. Gray background indicates lights off.

During the first 12 h of the 24-h recovery period after sleep deprivation, compared to baseline, there were rebounds of both NREMS (condition by time interaction: F11,161 = 5.0, P < 0.001; Figure S2, top panel) and REMS durations (condition by time interaction: F11,161 = 9.8, P < 0.001; Figure S2, middle panel). Furthermore, SWA increased during the first 6 h of the recovery period (condition by time interaction: F11,161 = 23.6, P < 0.001; Figure S2, bottom panel). These rebounds are similar to those previously reported for rats after 24 h sleep deprivation using gentle handling.^1–4

Figure S2

Means and standard errors of nonrapid eye movement sleep (NREMS), rapid eye movement sleep (REMS) and normalized slow wave activity (SWA) during the 24-h baseline period (white circles) and the 24-h recovery period after 24-h sleep deprivation (black circles). Gray background indicates lights off.

In conclusion, these findings indicate that our sleep deprivation method resulted in considerably more wakefulness (92.2%) compared to baseline (51.1%). Moreover, homeostatic sleep pressure was evidenced by characteristic NREMS and REMS rebounds and NREMS EEG SWA responses observed during the subsequent recovery period. These polysomnography-based results corroborate the performance-based evidence of the rPVT experiment that our sleep deprivation method was effective.