Slow-wave sleep during a brief nap is related to reduced cognitive deficits during sleep deprivation

Abstract

Sleeping for a short period (i.e. napping) may help mitigate impairments in cognitive processing caused by sleep deprivation, but there is limited research on effects of brief naps in particular. Here, we tested the effect of a brief nap opportunity (30- or 60-min) during a period of sleep deprivation on two cognitive processes with broad scope, placekeeping and vigilant attention. In the evening, participants (N = 280) completed a placekeeping task (UNRAVEL) and a vigilant attention task (Psychomotor Vigilance Task [PVT]) and were randomly assigned to either stay awake overnight or sleep at home. Sleep-deprived participants were randomly assigned to receive either no nap opportunity, a 30-min opportunity, or a 60-min opportunity. Participants who napped were set up with polysomnography. The next morning, sleep participants returned, and all participants completed UNRAVEL and the PVT. Sleep deprivation impaired performance on both tasks, but nap opportunity did not reduce the impairment, suggesting that naps longer than those tested may be necessary to cause group differences. However, in participants who napped, more time spent in slow-wave sleep (SWS) was associated with reduced performance deficits on both tasks, effects we interpret in terms of the role of SWS in alleviating sleep pressure and facilitating memory consolidation.

Statement of Significance.

This research asks whether brief naps can mitigate cognitive impairments caused by sleep deprivation. Brief naps are a promising intervention but have not been widely studied. We find that with a large sample, 30- and 60-min nap opportunities are not long enough to improve criterion performance, establishing a potential boundary condition on nap duration. However, we also find that duration of slow-wave sleep during naps predicts morning performance. The cognitive processes we assess (placekeeping and vigilant attention) play a supporting role in performance of a wide range of tasks, suggesting generality of the results. Other methodological strengths include a large sample, baseline assessments to control for individual differences, and a separate sleep control group to confirm sleep-deprivation effects.

Introduction

Sleep loss poses a significant economic burden, costing the United States up to $411 billion annually [1], and has played a role in a number of catastrophic accidents, including navy shipwrecks [2] and nuclear incidents [3]. These effects are perhaps not surprising, given that sleep deprivation impairs a wide range of cognitive processes. Two important cognitive processes that underly many types of performance are vigilant attention and placekeeping [4–16]. Vigilant attention is the ability to maintain consistent attention over time and is arguably the most widely studied process used to characterize cognitive effects of sleep deprivation. Placekeeping is the ability to complete a series of steps in a prescribed order, without repetitions or omissions, in the face of distractions and interruptions [17]. Placekeeping is less studied than vigilant attention but also has broad scope, playing a role in procedural performance [17], multitasking [18], problem solving [19], tasks that measure fluid intelligence [20, 21], and other cognitive domains that require systematic linear thinking. Placekeeping also represents an independent measure of effects of sleep deprivation, which acts on placekeeping through a direct causal path that does not include vigilant attention [15, 16], and unlike vigilant attention is subject to learning effects that could reveal effects of memory consolidation on task performance [16]. Thus, vigilant attention and placekeeping are suitable testbeds for studying effects of sleep deprivation, in that each supports performance in diverse task environments and the two are complementary in relevant ways.

Given that the effects of sleep deprivation on cognitive performance are robust and consequential, a broadly relevant research goal is to find ways to mitigate them. One approach is through pharmacological intervention with stimulants such as caffeine. However, caffeine appears to have limited scope, reducing impairments on vigilance and reaction time measures, but producing little to no benefit for more complex and diverse higher-order processes [22–25], including placekeeping [14].

An intervention that may have broader scope is napping. There are at least two mechanisms through which short periods of sleep could affect cognitive performance. First, napping alleviates the pressure to sleep from the homeostatic sleep drive, which is considered a major contributing factor to failures of attention [26, 27]. The homeostatic sleep drive is an internal biochemical propensity to maintain balance between sleep and wake. As wakefulness accumulates, the pressure to sleep increases [28–32] and creates instability in the wake state and difficulty in maintaining consistent attention [26, 33–35]. Slow-wave sleep (SWS), which occurs during non-REM (NREM) stage 3, is particularly important for reducing sleep pressure, evidenced by the fact that, as the duration of wakefulness and sleep pressure accumulates, the time spent in SWS increases [36, 37].

A second mechanism through which napping could reduce cognitive deficits caused by sleep deprivation is by stabilizing new memories. SWS has been linked to cortical reorganization, a process important for memory consolidation [38–42]. Synaptic changes associated with learning during wakefulness induce local changes in slow-wave activity that correlate with post-sleep performance improvement [43]. Thus, the slow oscillations generated during SWS are important for synaptic stability and plasticity required for memory organization, including enhancing, stabilizing, and integrating memories into neural networks [39, 44]. Stable long-term memory representations are crucial for any task that requires learning, including placekeeping [45]. Thus, a nap may help offset cognitive deficits by providing an opportunity for sleep-dependent consolidation processes to stabilize the memory representations necessary for task performance.

Napping has shown promise as an intervention to mitigate impairments caused by sleep deprivation. Longer naps, spanning 2–4 h, reduce deficits in alertness and vigilant attention [46–50] as well as some aspects of higher-order cognition such as working memory [48, 51, 52] and logical reasoning [53]. Brief naps of an hour or less, although less frequently studied, have also been found to reduce deficits in alertness and vigilant attention [54–56], and one study found that 10- and 30-min naps both helped to maintain performance on an associative learning task compared with no nap [55]. Thus, there is some evidence that napping during a period otherwise devoid of sleep may be an effective intervention for a diverse range of processes, including attention and higher-order processes. However, most existing research used naps spanning several hours, which are not always feasible or practical. Research into brief naps of an hour or less is much more limited, particularly for higher-order cognitive processes.

In the current study, we address two questions concerning the effect of brief naps on cognitive performance after sleep deprivation. First, we assess whether brief naps are an effective intervention for reducing the negative effects of sleep deprivation on vigilant attention and placekeeping. We hypothesized that a brief bout of sleep would mitigate performance deficits on both tasks. Second, we assess whether specific kinds of sleep during naps are beneficial for maintaining performance. Given that SWS is important for reducing sleep pressure and promoting consolidation of complex learning, we hypothesized that SWS would be related to the ability to maintain cognitive performance during sleep deprivation, specifically on a task requiring memory consolidation.

An overview of the study design is as follows. In the evening, participants performed a placekeeping task (UNRAVEL [17]) and a vigilant attention task (PVT [57]) under double-blind conditions, as a baseline assessment that allowed us to control for stable individual differences in task performance. Afterwards, participants were randomly assigned to stay awake overnight in the laboratory (the sleep-deprived group) or go home and sleep for the night (the sleep group). Between 04:00 and 06:00 am, sleep-deprived participants received a 0-, 30-, or 60-min nap opportunity. Participants assigned either a 30- or 60-min nap opportunity were set up with polysomnography to assess sleep quality and architecture. The following morning, all participants again completed UNRAVEL and the PVT. The inclusion of a sleep control group that completed the tasks at the same timepoints as the sleep-deprived group allowed us to rule out confounding effects of time-of-day or practice.

Methods

Participants

Participants were native English-speaking undergraduate students at Michigan State University. Participants were eligible for participation if they had never been diagnosed with a memory or sleep disorder, were not color blind, did not have a strong time-of-day preference (scores of 42–58 on the Morningness–Eveningness Questionnaire [58]), and did not have any major sleep disturbances (scores of 0–10 on the sleep disturbance section of the Pittsburgh Sleep Quality Index [59]). Participants slept a minimum of 6 h the night before the first day of the study and woke up on the first day of the study by 09:00 am. Participants were instructed to refrain from napping prior to arrival at the laboratory and to not consume caffeine, alcohol, or other drugs for 24 h prior to the study and for the duration of the study, with an exception for contraceptives. Compliance was assessed at screening with self-report. Participants who failed to follow criteria did not participate in the study.

Of an initial sample of 334 participants, 15 were excluded for missing data, 25 for failing to meet an accuracy criterion during the evening session of UNRAVEL (described in the Materials section), 9 for attrition, 4 for not obtaining any sleep during the nap opportunity, and 1 for noncompliance with instructions. After these exclusions, the final sample contained 280 participants (sleep n = 106, sleep-deprived n = 174). Demographic information was missing for five participants. Among the remaining 275 participants, 178 were female, the age range was 18 to 26 years old (M = 18.91 SD = 1.19), and the average education level was 12.82 years (SD = 1.11). The various groups in our design (described in the Procedure section) were balanced in terms of age, gender, and education level.¹ Most participants self-identified as White (77.1%), with an additional 8.7% identifying as Black or African American, 5.5% as Asian or Pacific Islander, 5.5% as Multiracial, 2.2% as Hispanic or Latino, and 1.1% as Other.

Participants recorded their sleep habits for five nights prior to the first day of the study. Summary information from the sleep diaries is reported in Table 1. On average, participants slept between 7.5 and 8 h each night, including the night before the first day of the study. The sleep-deprived and sleep groups reported sleeping similar amounts.

Table 1.

Sleep diaries for the five nights prior to the first day of the study

	Average sleep time, 5 nights prior to study	Total sleep time, night before study	Time to bed, night before study	Time of awakening, first day of study
Sleep	7 h 54 min (57 min)	7 h 45 min (1 h 1 min)	00:04 (1 h 12 min)	08:16 (53 min)
Sleep-deprived	7 h 44 min (47 min)	7 h 34 min (55 min)	00:07 (1 h 1 min)	08:08 (57 min)
Difference	t(259) = 1.59, p = 0.11	t(259) = 1.50, p = 0.14	t(259) = 0.37, p = 0.72	t(259) = 1.18, p = 0.24

Note. N = 261; 19 participants had incomplete or missing sleep diaries. Standard deviation in parenthesis.

Procedure

Participants were recruited for a study on sleep deprivation and napping with the knowledge that they would either remain awake overnight in our laboratory, with or without a nap opportunity, or go home and sleep for the night. They arrived at 22:00 pm for the evening baseline session in which they completed sleepiness and mood assessments (see the supplemental online material [SOM] for task descriptions and analyses), UNRAVEL, PVT, and other cognitive assessments that were part of another study. Completion of all tasks took approximately 2 h. Afterwards, participants were randomly assigned to either remain awake overnight in the laboratory (the sleep-deprived group) or to go home and sleep for the night (the sleep group). Researchers and participants were blind to condition until all evening testing was completed. Participants assigned to the sleep group were then given a Charge 2 activity monitor (Fitbit Inc., San Francisco, CA) to track their sleep at night and were given a ride home. Participants returned the monitors when they returned for the morning session.

Participants assigned to the sleep-deprived group remained in the laboratory overnight. These participants were additionally assigned to receive either a 0-min (n = 54), 30-min (n = 56), or 60-min (n = 64) nap opportunity. On any given night, either all participants were allowed a nap (30-min and 60-min conditions) or none were (0-min condition). On nights where participants were allowed a nap, nap opportunity duration (30-min or 60-min) was randomly assigned. We blocked nights based on whether or not participants were allowed a nap to avoid potential negative mood states if participants who were not allowed to nap were paired with participants who were. Nights with and without naps were randomly conducted throughout the year.

The start time of the nap opportunity was restricted to occur between the hours of 04:00 and 06:00 am. This time frame was chosen so that there would be sufficient time between awakening and morning tasks (which began at 08:30 am) to reduce negative effects of sleep inertia on task performance. In addition, this time frame roughly coincides with the circadian nadir when participants would likely be at their sleepiest and, thus, fall asleep faster. Variability in nap timing within the 04:00 to 06:00 am window had no effects on morning performance.²

Sleep-deprived participants remained awake overnight in the laboratory, except during the nap opportunity, and were monitored continuously by two trained research assistants. Participants were permitted to read, do homework, watch TV or movies, play board/card games, or engage in other quiet activities, but were not permitted to engage in any activities that would activate the autonomic nervous system. Participants were permitted to consume any food or beverage that did not contain caffeine or alcohol. Every hour (01:00, 02:00, 03:00, 04:00, 05:00, 06:00, 07:00, and 08:00 am) sleep-deprived participants completed sleepiness and mood assessments (discussed in the SOM). Participants were sleep-deprived for approximately 24 h before starting the morning tasks (with the exception of the nap).

At 08:30 am on the second day of the study, sleep participants returned to the laboratory and all participants completed the morning session, which included mood and sleepiness assessments (reported in the SOM), UNRAVEL, PVT, and other cognitive tasks associated with another study. The morning session lasted approximately 1.5 h, after which sleep-deprived participants were given a ride home.

Materials

UNRAVEL

We used the UNRAVEL task [17, 45] to measure placekeeping. UNRAVEL is an acronym in which each letter refers to a step that participants perform in the order specified by the order of the letters within the acronym. Each letter, or step, identifies a different two-alternative forced-choice decision rule that participants must apply to a randomly generated stimulus. Figure 1 shows two sample stimuli and the seven decision rules. The stimulus contains no information about what step to perform and any rule can apply to any stimulus, so participants must remember where they are in the sequence. Participants perform the steps in a loop, returning to the “U” step after completing the “L” step.

Figure 1.

Above: Example of two randomly generated stimuli from the UNRAVEL task. Below: The UNRAVEL rules that correspond to each step (letter) in the UNRAVEL acronym, and the correct keyboard responses for each rule based on the two stimuli above. The bolded letters represent the possible response options for each rule. Figure adapted with permission from Altmann, Trafton, and Hambrick [¹⁷].

Performance is periodically interrupted by a typing task. During the interruption, two randomized strings of letters appear on the computer display, one string at a time, and the participant must type each string correctly into a box. Each string comprises the 14 possible UNRAVEL responses (Figure 1), so as to generate some interference with memory for the step performed before the interruption. Interruptions last about 20 s. After an interruption, the participant must remember where they were in the UNRAVEL sequence prior to the interruption in order to correctly resume their place. There were ten interruptions per block. A session of UNRAVEL took about 35 min to complete and consisted of four test blocks, each block containing 66 trials on average (SD = 12), where a trial comprises one performed step.

During the evening session, participants received instructions on how to perform the task and completed a practice phase to familiarize them with the procedure. If participants made 15 or more errors during practice, they completed the practice phase a second time to help ensure that they understood the task. If participants made fewer than 15 errors, they progressed to the test blocks of UNRAVEL. During the morning session, participants completed a brief practice to remind them of the rules and procedure before completing the test blocks.

The behavior of interest is the rate of placekeeping errors, which are steps performed out of order. Placekeeping errors can be detected because every rule has unique response options, so from any response we can determine which step the participant intended to perform. Placekeeping errors are coded with respect to the step performed on the previous trial. For example, if steps “N,” “R,” “V,” and “E” are performed in succession, “V” would be an error because the “A” step was skipped, but “E” would not be an error because it correctly follows “V.”

There are two types of placekeeping errors, which we measure and analyze separately because they occur at substantially different frequencies. Post-interruption errors occur on the trial immediately following an interruption, whereas non-interruption errors occur on a trial following another trial. The two error types measure nearly the same set of cognitive operations but, due to the interruption, post-interruption errors reflect a more substantial burden on memory to keep task-relevant representations active and, as a result, occur more frequently than non-interruption errors. The rate of each error type (post-interruption or non-interruption) was computed by dividing the number of errors of that type by the number of trials on which that error type could occur.

Errors in applying the decision-rules can also occur (e.g. the participant responds “red” when the correct response was “yellow”; see Figure 1). These errors are infrequent and are reported in the SOM.

A correct trial is one on which there is neither a placekeeping error nor a decision-rule error. If fewer than 70% of trials in a block were correct, the participant was instructed to be more accurate at the end of the block. We excluded participants from all analyses whose accuracy was below 70% on two or more blocks (out of four) during the evening session because we could not be sure they understood the task.

Psychomotor Vigilance Task (PVT)

We used the PVT [57] to measure visual vigilant attention. Participants monitored a blank computer screen for the appearance of a red circle and were instructed to make a mouse click as quickly as possible when the circle appeared. The mouse click caused the circle to disappear and triggered feedback on reaction time which remained on the screen for 500 ms. The circle appeared at random intervals between 1 and 10 s. The task lasted 10 min. The outcome of interest was the lapse rate, which we calculated by dividing the number of lapses (reaction times greater than 500 ms) by the number of trials, where a trial comprises one appearance of the circle.

Polysomnography

Participants assigned to the 30- or 60-min nap opportunity were set up with polysomnography (PSG). Scalp measurements were taken in accordance with the international 10–20 system and a total of 14 electrodes were applied. Electroencephalographic (EEG) electrodes were placed at: F3, F4, C3, C4, O1, O2, and FpZ (ground). Additionally, there were two EOG electrodes (one on each eye) and three EMG electrodes on the chin. Finally, reference electrodes (M1 and M2) were placed on the mastoids. Data were recorded continuously using the Philips Respironics Alice 6 LDx Diagnostic sleep system (Koninklijke Philips N.V., Eindhoven, Netherlands) using Sleepware G3 software.

Dimensions of interest included sleep quality, sleep architecture, and spindle activity during the nap opportunity. To assess sleep quality, we collected total sleep time, sleep latency (duration from the start of the nap opportunity to sleep onset), and wake after sleep onset (WASO). To assess sleep architecture, each 30-s epoch was visually scored into one of five states based on standard practices from the American Academy of Sleep Medicine manual version 2.5 [60]: Wake, NREM stage 1 (N1), NREM stage 2 (N2), NREM stage 3 (N3), or rapid eye movement (REM). Two trained coders scored each participant’s data. The average agreement for scored sleep stages between the two coders was high (92%) and we therefore used one coder’s scoring for the analyses.

To assess spindle activity, the EEG data were loaded into MATLAB (MATLAB, R2019b) using the EEGLAB toolbox [61]. Sleep spindles were then coded automatically, following a procedure adapted from Wamsley and colleagues [62]. Electrode C3 (referenced to M2) was used to identify spindles. Spindle detection was based on the following standard criteria: duration greater than or equal to 0.5 s, frequency between 11–16 Hz, and within an NREM epoch. Spindle count was defined as the total number of NREM spindles during the nap opportunity and spindle density was calculated by dividing spindle count by the amount of time in NREM sleep.

Actigraphy

After all evening baseline testing was completed, participants who were randomly assigned to the Sleep group were given a Fitbit Charge 2 wrist actigraphy monitor (Fitbit Inc., San Francisco, CA) to estimate their sleep during the night between the evening and morning sessions. For each participant, we collected information on the amount of time in bed, total sleep time, number of awakenings, and time spent awake based on Fitbit’s algorithm. We calculated sleep efficiency by dividing total sleep time by time in bed.

Statistical analyses

We conducted three sets of analyses for each of three measures, the measures being the PVT lapse rate, the UNRAVEL post-interruption error rate, and the UNRAVEL non-interruption error rate. To stabilize variance in the measures, which are all rates, we applied the arcsine-root transformation prior to analysis. However, for clarity, we report descriptive statistics on the original scale, as rates back-transformed from the mean transformed rates estimated by analyses of covariance (ANCOVA), as described below.

The first set of analyses examined effects of the experimental manipulations—sleep deprivation and nap opportunity—on morning performance. We conducted ANCOVAs with evening performance on the measure of interest (the lapse rate or one of the two error rates) as the covariate. To examine effects of sleep deprivation, we conducted an ANCOVA that included group (sleep, sleep-deprived) as a between-subjects factor and collapsed across nap opportunity subgroup. To examine effects of nap opportunity within the sleep-deprived group, we conducted a separate ANCOVA, on that group, that included nap opportunity subgroup (0-, 30-, 60-min) as a between-subjects factor. We conducted pairwise comparisons to probe any significant main effects of nap opportunity on performance, again using evening performance as a covariate. In the SOM, we report raw rate data for the morning and evening sessions, as well as raw lapse counts for the PVT for comparison with existing studies that report counts instead of rates.

The second set of analyses asked whether sleep architecture during the nap was related to morning performance. We collapsed across the 30- and 60-min nap opportunity subgroups and performed separate stepwise regressions for the three morning performance measures (the lapse rate and the two error rates). First, we tested for multicollinearity among sleep variables by examining their first-order correlation matrix, adopting a criterion of r = 0.50 (a moderately strong correlation) as a basis for omitting variables from the set of predictors. Then, for each regression, we entered evening performance to control for stable individual differences in performance, and the sleep variables, which were sleep latency (min), N1 (min), N2 (min), N3 (min), REM (min), WASO (min), and NREM spindle count. The sleep variables were then backwards eliminated from the regression sequentially until no variables met the criteria to be eliminated (IBM SPSS version 25.0). We chose backwards elimination because it starts with all of the sleep variables together in the model, which helps to reveal potential combinations of sleep characteristics related to performance.

The third set of analyses asked whether total sleep time for participants in the sleep group was related to morning performance. We performed separate hierarchical regressions for the three morning performance measures. In the first step of each regression, we entered evening performance on that measure to control for individual differences in performance. In the second step, we entered total sleep time.

Results

Effects of sleep deprivation and nap opportunity

Results of comparing the sleep-deprived group (collapsing across nap subgroups) with the sleep group are plotted in Figure 2 and reported in Table 2. The sleep-deprived group had a higher lapse rate than the sleep group, F(1, 277) = 56.23, p < 0.001, η _p² = 0.169, a higher post-interruption error rate than the sleep group, F(1, 277) = 15.96, p = 0.001, η_p² = 0.054, and a higher non-interruption error rate than the sleep group, F(1, 277) = 8.96, p = 0.003, η_p² = 0.031. These results confirm that there was an impairment due to sleep deprivation that naps could potentially mitigate.

Figure 2.

Morning performance measures for the sleep and sleep-deprived groups. Values are back-transformed from the mean transformed lapse and error rates estimated by analysis of covariance using evening performance as a covariate. Error bars are 95% confidence intervals.

Table 2.

Morning PVT lapse rate and morning UNRAVEL placekeeping error rates

	Lapse rate	Post-interruption error rate	Non-interruption error rate
Sleep (n = 106)	0.068 [0.054, 0.083]	0.148 [0.116, 0.182]	0.014 [0.006, 0.025]
Sleep-deprived (n = 174)	0.155 [0.138, 0.171]	0.242 [0.212, 0.272]	0.038 [0.027, 0.050]

Note. Values are back-transformed from the mean transformed lapse and error rates estimated by analysis of covariance. 95% confidence intervals in brackets.

Results of comparing the different nap subgroups (0-, 30-, and 60-min) within the sleep-deprived group are plotted in Figure 3 and reported in Table 3. Lapse rates did not differ significantly, F(2, 170) = 2.16, p = 0.119, η_p² = 0.025. However, post-interruption error rates did differ significantly, F(2, 170) = 3.92, p = 0.022, η_p² = 0.044. Specifically, the 0-min subgroup performed marginally better than the 30-min subgroup, F(1, 107) = 3.32, p = 0.071, η_p² = 0.030, and the 30-min subgroup performed significantly worse than the 60-min subgroup, F(1, 117) = 7.56, p = 0.007, η_p² = 0.061. The 0-min subgroup did not differ from the 60-min subgroup, F < 1. Non-interruption error rates did not differ significantly across nap subgroups, F(2, 170) = 0.235, p = 0.098, η_p² = 0.027.

Figure 3.

Morning performance measures for the sleep-deprived group, separated by nap opportunity subgroup. Values are back-transformed from the mean transformed lapse and error rates estimated by analysis of covariance using evening performance as a covariate. Error bars are 95% confidence intervals.

Table 3.

Morning PVT lapse rate and morning UNRAVEL placekeeping error rates, separated by nap opportunity within the sleep-deprived group

	Lapse rate	Post-interruption error rate	Non-interruption error rate
0-min (n = 54)	0.136 [0.107, 0.168]	0.214 [0.161, 0.273]	0.038 [0.019, 0.065]
30-min (n = 56)	0.171 [0.140, 0.204]	0.298 [0.237, 0.360]	0.052 [0.029, 0.081]
60-min (n = 64)	0.129 [0.103, 0.157]	0.188 [0.142, 0.240]	0.020 [0.008, 0.038]

Note. Values are back-transformed from the mean transformed lapse and error rates estimated by analysis of covariance. 95% confidence intervals in brackets.

These results show the expected negative effects of sleep deprivation on cognitive performance, but no benefit of brief naps. Indeed, there was some evidence that 30-min naps may have had a negative effect on cognitive performance, as measured by post-interruption errors, but this effect would have to replicate before we considered it interpretable.

Relationship between sleep quality and architecture and morning performance

Although the brief nap opportunities we tested did not improve morning performance, the quality or architecture of sleep during naps may still have been related to morning performance. We examined this possibility using stepwise regressions for each morning performance measure with sleep variables as predictors, collapsing over nap subgroup (30-min, 60-min). Descriptive and inferential statistics for the sleep variables separated by nap subgroup are reported in Table 4. The correlation matrix for sleep variables collapsed over nap subgroup is reported in Table 5. We omitted two sleep variables as predictors based on multicollinearity, using r = 0.5 as a cutoff for highly correlated variables. We omitted total sleep time because it was highly correlated with minutes in N3, r = 0.843, p < 0.001, retaining minutes in N3 because total sleep time redundantly measures minutes in the other sleep stages. We omitted NREM spindle density because it was highly correlated with NREM spindle count, r = 0.749, p < 0.001, retaining NREM spindle count because, like the other predictors, it is a total rather than a rate.

Table 4.

Sleep architecture for the 30- and 60-min nap opportunity subgroups

	30-min	60-min	t	p
Total sleep time (min)	24.64 (5.93)	53.22 (7.62)	22.69	< 0.001
Sleep latency (min)	4.86 (4.89)	4.81 (5.29)	−0.05	0.96
N1 (min)	4.29 (3.82)	4.75 (3.61)	0.67	0.50
N2 (min)	9.78 (3.44)	16.98 (8.34)	6.03	< 0.001
N3 (min)	10.09 (7.10)	30.96 (13.17)	10.59	< 0.001
REM (min)	0.48 (1.73)	0.52 (2.22)	0.11	0.91
WASO (min)	1.14 (2.07)	1.95 (4.00)	1.35	0.18
NREM spindle count	12.16 (8.56)	27.89 (17.89)	6.00	< 0.001
NREM spindle density	0.53 (0.37)	0.54 (0.34)	0.15	0.88

Note. n = 120. Standard deviation in parentheses. Bold, p < 0.05.

Table 5.

Correlation matrix for sleep variables for the 30- and 60-min nap opportunity subgroups

	1	2	3	4	5	6	7	8	9
1. Total sleep time (min)	1
2. Sleep latency (min)	−0.356**	1
3. N1 (min)	−0.039	0.098	1
4. N2 (min)	0.469**	−0.078	0.073	1
5. N3 (min)	0.843**	−0.380**	−0.337**	−0.005	1
6. REM (min)	−0.050	0.124	0.087	−0.093	−0.162	1
7. WASO (min)	−0.154	0.181*	0.283**	0.118	−0.308**	0.116	1
8. NREM spindle count	0.495**	−0.176	0.069	0.404**	0.307**	0.000	−0.002	1
9. NREM spindle density	−0.072	0.108	0.107	0.155	−0.190*	0.081	0.116	0.749**	1

Note. n = 120. Values are Pearson’s r. *, p < 0.05. **, p < 0.01.

The regression results are reported in Table 6. For the morning lapse rate, the final regression model included significant predictors of evening lapse rate, minutes in N3, and sleep latency. More time spent in N3 and a longer sleep latency were associated with a lower morning lapse rate. For the morning post-interruption error rate, the final model included significant predictors of evening post-interruption error rate, minutes in N3, and WASO. More time spent in N3 and more WASO were associated with a lower morning post-interruption error rate. However, four participants had WASO > 3 standard deviations above the mean; when these participants were removed from the final model, the effect of WASO fell below significance, B = −0.022, SE_B = 0.014, β = −0.133, t = −1.62, p = 0.108. Finally, for the morning non-interruption error rate, the final model included the evening non-interruption error rate as a non-significant predictor and minutes in N3 as a marginally significant predictor. More time spent in N3 was marginally associated with a lower morning non-interruption error rate. Variance inflation factor scores in all models were below 2, indicating that multicollinearity was not a problem.

Table 6.

Final regression models for morning performance measures, collapsed across the 30- and 60-min nap opportunities

	B	SE B	β	t	p
Lapse rate
Evening lapse rate	0.927	0.121	0.583	7.69	< 0.001
N3 (min)	−0.002	0.001	−0.179	−2.21	0.029
Sleep latency (min)	−0.008	0.003	−0.207	−2.53	0.013
Post-interruption error rate
Evening post-interruption error rate	0.809	0.120	0.518	6.77	< 0.001
N3 (min)	−0.004	0.002	−0.185	−2.30	0.023
WASO (min)	−0.018	0.007	−0.199	−2.47	0.015
Non-interruption error rate
Evening non-interruption error rate	0.315	0.223	0.128	1.41	0.161
N3 (min)	−0.002	0.001	−0.153	−1.69	0.094

Note. n = 120.

The regression results indicate that amount of SWS (minutes in N3) was related to morning performance for both our tasks, and that sleep latency and WASO also predicted morning performance but less consistently.

Relationship between actigraphy in the sleep group and morning performance

Actigraphy monitors were used to estimate nighttime sleep between the evening and morning sessions in the sleep group. The results from the monitors are reported in Table 7. We limit our analysis to the relationship between total sleep time and morning performance because actigraphy monitors correlate well with PSG on this measure but less reliably on other measures [63]. Total sleep time was significantly related to the morning lapse rate, B = −0.001, SE_B < 0.001, β = −0.28, t = −3.50, p = 0.001, indicating that participants who slept more between sessions made fewer lapses in the morning. Total sleep time was not related to the post-interruption error rate, B < 0.001, SE_B < 0.001, β = 0.01, t = 0.08, p = 0.94, or the non-interruption error rate, −0.001 < B < 0, SE_B < 0.001, β = −0.02, t = −0.21, p = 0.83.

Table 7.

Sleep characteristics recorded from actigraphy monitors for the night between the evening and morning sessions in the sleep group

Total sleep time	Time spent in bed	Time spent awake	Number of awakenings	Sleep efficiency
5 h 57 min (52 min)	6 h 17 min (52 min)	20 min (11 min)	1.73 (4.84)	94.66% (3.03%)

Note. N = 82; 24 participants were missing actigraphy data due to technical failure or experimenter error. Standard deviation in parenthesis. Sleep efficiency is calculated by dividing the total sleep time by the time spent in bed for each participant.

Discussion

Sleep deprivation impairs a diverse set of cognitive processes, but mitigating these impairments has proven challenging, particularly for higher-order processes like placekeeping. Napping interventions may benefit a variety of processes, potentially due to the multifaceted effects of SWS on cognitive functioning. Brief naps in particular are appealing because they are more feasible and practical than longer naps. However, brief naps have attracted less research than longer naps, and no previous studies have tested the effects of naps on placekeeping.

Using a large sample, we found that brief naps, occurring in 30- or 60-min windows toward the end of a 24 h period of sleep deprivation, did not mitigate the impairments in vigilant attention or placekeeping caused by sleep deprivation. Thus, naps longer than 60-min may be necessary to mitigate such impairments. This conclusion may apply quite broadly, as vigilant attention and placekeeping support performance in a wide range of task environments.

However, duration may not be the only factor that influences nap effectiveness. Other factors may include the time lag between nap and performance, and the nature of other tasks performed during this lag. Prior work has found that, compared with no nap, a 10-min nap opportunity was effective at mitigating impairments in the PVT caused by sleep deprivation [55] and a 60-min nap opportunity remediated performance on a simple reaction time task [54]. However, in the study involving the PVT [55], the lag between awakening from the nap and PVT performance was 47 m or less. In the current study, the lag was about 2.5 h, with time of awakening from a nap averaging approximately 06:00 am and the morning assessments beginning at 08:30 am. The substantially longer lag in our study may explain the contrasting findings. However, our participants also performed UNRAVEL immediately before they performed the PVT. UNRAVEL is a cognitively demanding task and may have caused fatigue or decreased motivation that diminished the benefits of the nap on vigilant attention. A brief nap opportunity may benefit vigilant attention, regardless of lag since the nap, if no other intervening tasks are performed [54]. In future work it would be useful to examine the relationship between lag since napping and activity during the lag to assess how the two factors might interact.

We did find that sleep factors that were related to sleep pressure predicted morning performance. More SWS (minutes in N3) predicted better morning performance on both of our tasks. Sleep-deprived participants spent most of their allotted nap time in N3 (M = 48%, SD = 25%), and every 10-min increase in N3 reduced post-interruption errors by about 4% and lapses and non-interruption errors by about 2%. SWS has a reciprocal relationship with sleep pressure, such that sleep pressure builds up across wakefulness and is subsequently alleviated with time spent in SWS [36, 37, 64]. That is, SWS both reflects sleep pressure and alters the amount of sleep pressure. Given this reciprocal relationship, our data suggest that participants with more sleep debt obtained both more SWS during naps and more performance benefits from SWS during naps. These results are consistent with the view that buildup of sleep pressure is one of the primary causes of lapses on the PVT [26, 33–35] and may suggest that buildup of sleep pressure also increases placekeeping error rates (but see below).

Sleep latency and WASO were also related to morning performance but less consistently. Longer sleep latency predicted better performance on the PVT and more WASO predicted better performance on UNRAVEL as measured by post-interruption errors. Sleep latency measures physiological sleep propensity [65] and thus is another measure of sleep pressure. Sleep latency in our data was quite brief (M = 4.81 min, SD = 5.10 min), as would be expected for sleep-deprived individuals experiencing high sleep pressure. WASO is a measure of sleep fragmentation and, although more WASO often predicts worse performance [66, 67], in our data more WASO predicted better performance. One explanation for our finding is that, in a sample of sleep-deprived but otherwise young and healthy participants, more WASO may primarily reflect lower sleep pressure causing a lower propensity to remain asleep. This interpretation should be treated with some caution, as there were four participants with WASO scores more than 3 standard deviations above the mean, and when these participants were removed, the effect of WASO fell below significance. Nonetheless, in context of the SWS findings we reported above, our sleep latency and WASO data suggest that people experiencing less sleep pressure may be more resilient to effects of sleep deprivation and less likely to benefit from a nap.

For the UNRAVEL task, SWS may also have reduced placekeeping errors by facilitating memory consolidation. SWS plays a role in the consolidation and stabilization of declarative memory [38–42], and these benefits can occur within the timeframe of a nap [68–71]. Consolidation may be especially relevant to maintaining learning of new and complex tasks, such as UNRAVEL. Indeed, participants show improved performance on UNRAVEL across the evening session [16], and the newly formed memories driving this improvement may be subject to consolidation processes. In contrast, the PVT is known to be resistant to learning effects [47, 72] and may not benefit from consolidation during SWS. Thus, different mechanisms may have linked SWS to reduced deficits in vigilant attention and placekeeping, a possibility that will be important to explore in future work.

Although this study had several strengths, such as a large sample, baseline assessments of performance, a sleep control group, and polysomnographic recordings of sleep during the nap opportunities, it also had limitations. Sleep participants slept an average of about 6 h the night between the evening and morning sessions and may therefore have been experiencing mild sleep restriction when completing the morning tasks. If this were the case, then we may have underestimated the effect of sleep deprivation in our analyses of the behavioral data. In addition, although we assessed sleep between sessions in the sleep group objectively via wrist actigraphy, we did not use polysomnography, which is the gold standard for recording sleep. Lastly, compliance with the screening criteria of refraining from napping and from consuming caffeine, alcohol, or other drugs for 24-h prior to the study was assessed via self-report rather than through objective means. Although we have no reason to believe that dishonest reporting differed in a systematic way between groups, we cannot rule out the possibility that prior napping or drug use affected our results.

In conclusion, a brief nap of 30 or 60 min did not significantly mitigate deficits in morning vigilant attention and placekeeping performance caused by sleep deprivation. However, sleep quality and architecture during naps were related to morning performance, with time spent in SWS related to morning performance on both tasks. The role of time spent in SWS implicates sleep pressure as an important factor in morning performance but may also point to consolidated learning of the UNRAVEL task as a factor in reducing the placekeeping error rate. Several other factors may play a role in the effectiveness of nap interventions. Future research should explore how nap duration, the timing of a nap in relation to task performance, and circadian placement of a nap might interact to optimize the effectiveness of a nap on cognitive performance. In the interim, our results highlight the difficulty of mitigating the cognitive deficits caused by sleep deprivation and the need to exercise caution when performing vigilance and procedural tasks under conditions of sleep deprivation when there is a premium on attentive and accurate performance.

Funding

This research was supported by grants N000141612841 and N000142012739 from the Office of Naval Research to the second and third authors.

The first author was supported by grant T32 HL082610 from the National Institute of Health.

Disclosure Statement

Financial disclosure: The authors report no conflicts of interest.

Nonfinancial disclosure: The authors report no conflicts of interest.

The data underlying this article will be shared on reasonable request to the corresponding author.