BruckDorothy Department of Psychology (S089), Victoria University, PO Box 14428, MCMC Melbourne, Victoria 8001, Australia. Fax: +61 3 9365 2218.
Sleep inertia, the performance impairment that occurs immediately after awakening, has not been studied previously in relation to decision-making performance. Twelve subjects were monitored in the sleep laboratory for one night and twice awoken by a fire alarm (slow wave sleep, SWS and REM sleep). Decision making was measured over 10 3-min trials using the ‘Fire Chief’ computer task under conditions of baseline, SWS and REM arousal. The most important finding was that sleep inertia reduces decision-making performance for at least 30 min with the greatest impairments (in terms of both performance and subjective ratings) being found within 3 min after abrupt nocturnal awakening. Decision-making performance was as little as 51% of optimum (i.e. baseline) during these first few minutes. However, after 30 min, performance may still be as much as 20% below optimum. The initial effects of sleep inertia during the first 9 min are significantly greater after SWS arousal than after REM arousal, but this difference is not sustained. Decision-making performance after REM arousal showed more variability than after SWS arousal. Subjects reported being significantly sleepier and less clear-headed following both SWS and REM awakenings compared with baseline and this was sustained across the full 30 min. In order to generalize this finding to real-life situations, further research is required on the effects of continuous noise, emotional arousal and physical activity on the severity and duration of sleep inertia.
Sleep inertia can be described as a rising trend in alertness, reflecting a wake-up effect. This occurs despite the fact that the sleep from which a person has awoken may have fully dissipated their sleep need ( Folkard and Akerstedt 1992). The waking up process is gradual, demonstrating that the recently asleep brain of a person experiencing sleep inertia is not necessarily sleep starved but is struggling to move from a sleep state toward full alertness ( Balkin and Badia 1988).
While sleep inertia is typically modest and short lived in normal, non-sleep-deprived individuals ( Dinges 1990), there is no agreement about its actual duration. Dinges (1990) suggested that it can endure for 1–20 min. Wilkinson and Stretton's (1971) nocturnal study found that sleep inertia endured for up to 15 min. In contrast, Dinges et al. (1981) found that it only lasted for 1–5 min. Akerstedt et al. (1989) suggested that although sleep inertia is normally rapidly dissipated, it appears that it may be extended under certain circumstances. For instance, it can be more severe if arousal from sleep occurs during the first half of the night or during slow wave sleep (SWS) ( Dinges and Kribbs 1991).
The significance of the sleep inertia effect is such that it has also been incorporated into a number of theoretical models. Folkard and Akerstedt (1992) have devised a three-process model of the regulation of alertness and sleepiness, which aims to predict alertness ratings at any given point in time and on any given sleep/wake schedule. The model incorporates Process W, which is initiated at awakening, and refers to the ‘waking up’ effect. The Achermann and Borbley (1992) model proposes that homeostatic and circadian processes, as well as a sleep inertia component, account for experimental data of alertness ratings.
Balkin and Badia (1988) conducted a study on the relationship between sleep inertia and sleepiness and found no conclusive evidence suggesting that sleep inertia is qualitatively different from typical sleepiness. This is consistent with the suggestion that sleepiness is not necessarily a response to sleep need, but rather, may reflect the incomplete disengagement of sleep processes ( Pivik 1991).
The potential impact of sleep inertia on decision making has significant implications as people are called upon to make complex decisions soon after awakening in a variety of field settings. This includes emergency workers and people awoken by a disaster, e.g. fire and flood. Decision making is a complex cognitive process. It involves the accurate search and appraisal of information, understanding the options and choosing the best from the available alternatives. The latter involves evaluating probable consequences of an action. During this process the person continues to seek new information and to re-evaluate old information, and when the required level of confidence is reached a final decision is made ( Festinger 1964).
The effect and duration of impaired decision making due to sleep inertia is unknown. As there is no consensus on the duration of sleep inertia, which may vary depending on the type of task, it was not possible to be definitive in predicting how long the effect would last in relation to decision making. Consideration of the studies discussed earlier suggest that a period of 30 min should span the complete possible duration of sleep inertia.
This study represents the first step in evaluating possible impairment in decision-making performance due to sleep inertia. Specifically, the study investigates the time required to achieve waking levels of decision-making performance, and documents subjective ratings of sleepiness and clear-headedness during the postulated sleep inertia period. It also compares the level of performance decrement, subjective sleepiness and clear-headedness when awakening from SWS compared with REM sleep.
Twelve subjects (three males, nine females), aged between 18 and 30 y (mean age 22.3 y, SD 3.6 y) participated in this study. They were acquaintances of the researchers and students from Victoria University who volunteered in response to an internal advertisement. They reported themselves to be in good health and normal sleepers. All subjects received $10 (Australian) to cover transport expenses to the university sleep laboratory.
A Grass model 7 polygraph/paper chart recorder was used to record relevant polysomnographic measures. The alarm to arouse sleepers was a recording of a standard smoke detector fire alarm presented at a volume of 75 dB at the pillow. Subjects slept in a single bedroom adjacent to the sleep-recording apparatus.
The performance measure used was the ‘Fire Chief’ decision-making task ( Omodei and Wearing 1993a, 1993b, 1995), which was installed on an IBM clone 386 computer in the bedroom. This is an interactive computer administered task which has been programmed to capture the essential attributes of real-life decision-making situations. Subjects are required to assume the role of a chief fire officer responsible for fire control in a specified area comprising various types of landscape elements and fire appliances (trucks and helicopters). The subjects are able to control the spread of the simulated fires, which move according to the wind direction, by dispatching firefighting appliances to drop water on them. Refilling of the appliances is required at local dams.
More specifically, a relatively simple ‘Fire Chief’ forest scenario was used comprising trees, five scattered dams, 10 houses, one helicopter and one truck available for use. Two developed fires were specified for the start of each trial. One spot fire was specified in the middle of the trial. All the events, including the changes in wind strength and direction and number of fires, were pre-set. Hence, fires were programmed to develop under conditions of fixed changes in both strength and wind direction and subjects were warned that the wind strength and direction would change at various times throughout the trials. All 10 trials (each of 3 min duration) were programmed to have the same characteristics to ensure as much consistency as possible across the trials. However, in each trial the houses and dams were positioned differently so that they appeared different. The task involves easy-to-remember and easy-to-use mouse and keyboard procedures.
The Karolinska Sleepiness Scale (KSS) was used to obtain a subjective measure of sleepiness. It is a 9-point scale which ranges from ‘extremely alert’ (1) to ‘extremely sleepy, fighting sleep’ (9). KSS ratings have been shown to consistently relate to polysomnographic variables ( Akerstedt and Gillberg 1990) and performance tasks ( Gillberg, et al. 1994 ).
The ‘Clear-headed’ rating scale was an unvalidated scale constructed for this study to determine how clear-headed subjects felt after arousal at particular points in time. It consists of a 5-point scale that ranges from ‘extremely clear-headed’ (1) to ‘not at all clear-headed’ (5).
Each subject spent one night in the sleep laboratory at Victoria University. Subjects arrived at the sleep laboratory several hours before their usual bed time. Subjects were randomly assigned to either a morning baseline group or an evening baseline group. All subjects were first given instructions and a demonstration of the ‘Fire Chief’ decision-making task, which was followed by a 1-h session on the task to ensure familiarity and control for practice effects. Subjects were then prepared for the polysomnograph recordings. This procedure involved the standard attachment of 13 surface electrodes to the face and scalp to record brain waves, eye movements and muscle tension. ( Rechtschaffen and Kales 1968). Half the subjects (the evening baseline group) were then required to complete the 10 3-min baseline trials of the ‘Fire Chief’ task and, at intervals of 6 min, the KSS and the clear-headed rating were completed. A break of at least 1 h was placed between the practice and baseline trials. This was to ensure that the first trial during baseline would not be experienced as a continuation of the practice trials. This was important as first trials on ‘Fire Chief’ are likely to show a ‘warm-up’ decrement (Omodei, personal communication). The trials presented for familiarization were different to those used for experimentation.
In order to avoid possible systematic bias arising from the baseline trials always preceding the sleep inertia trials, the other half of the subjects were assigned to a morning baseline group. They performed their baseline trials in the morning after breakfast and prior to leaving. Any confounding circadian effects arising from morning vs. evening baseline trials would be averaged out when the group was compared with their performance under experimental conditions. (The existence of morning vs. evening differences would also be tested.) All subjects were fully aware of the protocol and were instructed to do as well as possible on all trials of the task. They were all given the opportunity to listen to the alarm prior to going to sleep.
All arousals were sudden. For each subject, the alarm was activated twice during the night; once during SWS (Stage 4 where possible and if not, Stage 3) and once during REM sleep. The order of waking from these two stages was counterbalanced across subjects. This was to avoid any order effect, which could have occurred if, for example, the REM awakening always occurred after SWS awakening. It was also decided not to awaken subjects too soon after sleep onset. One rationale for this was to have the awakenings most closely coincide with the time of night when most residential fire fatalities occur (00.00 to 04.00) ( Jones 1983). The other reason was to avoid the first SWS cycle as this was likely to show the greatest sleep inertia effect. Rather than find a maximum effect, this study was interested in determining the size of an effect that was potentially more generalizable across the 00.00 to 04.00 period. (Results from this study have been incorporated into a human behaviour fire risk model being developed at Victoria University.) Accordingly, all subjects were allowed to sleep through the SWS of their first cycle. Subjects in the ‘REM first’ condition were awoken in their first REM period and then during the SWS of their second cycle. Subjects in the ‘SWS first’ condition were awoken during the SWS of their second cycle and then their second REM period. All subjects were first awoken during the period 85–154 min after sleep onset. The second awakening was always within 2 min of first entering the desired sleep stage. In practice this second awakening was 10–96 min after re-entering sleep from their first experimental awakening. While most subjects fell asleep quickly after the first awakening, two subjects took over 90 min to re-enter sleep. The clock time of awakenings varied from 00.24 to 05.15am.
Stage scoring was carried out according to the method of Rechtschaffen and Kales (1968). When the electrophysiological recordings indicated that the subject had entered the desired sleep stage for a minimum of 1.5 min, the alarm was activated by the researcher. When the subject reached full EEG wakefulness, the researcher turned off the alarm and then entered the bedroom and turned on a very dim light. The subject was immediately required to fill out the KSS and clear-headed rating, which only required a few seconds. These responses were further required after 6, 12, 18, 24 and 30 min. The computer was moved to the subject so it was unnecessary for them to leave the bed. Each subject was required to perform 10 3-min trials on the decision-making task sitting up in bed. Their performance score was given to them after each 3-min trial. When the session was completed the subject was allowed to return to sleep until the second awakening. After the second awakening, subjects were allowed to sleep undisturbed until morning.
The experiment was approved by the Victoria University Human Experimentation Ethics Committee.
With regard to SWS arousals, it was possible to awaken two-thirds of the subjects from Stage 4, and the remaining subjects had to be woken from Stage 3. An independent t-test was conducted to compare the average experimental performance scores of the Stage 4 awakenings (n=8) vs. the Stage 3 awakenings (n=4) and no significant difference in decision-making performance was found between these two groups. In addition, both baseline and sleep inertia average performance scores for the subjects who performed their baseline in the evening were compared (using independent t-tests) with the scores of the morning baseline group and no differences were found. Visual inspection of all the data used in the above analyses also indicated no systematic differences between groups.
To analyse the performance and subjective rating dependent variables a series of ANOVAs were used which all involved three conditions (baseline, SWS and REM) by time (10 trials covering 30 min) with repeated measures on both factors. Alpha was set at 0.05. Given that the sleep inertia literature allowed the development of a number of expectations it was decided to use a priori planned contrasts to investigate any main effects found. This data analysis procedure was applied to decision-making performance, KSS ratings and clear-headed ratings over both the full 30-min period and a shorter period. The second, shorter period (first 9 min for performance, 12 min for ratings) was included to ensure that Type 2 errors were not being made arising from the reduced power of analysing the longer period. Given the diversity of duration for sleep inertia reported, the choice for the shorter period was somewhat arbitrary.
Where the ANOVA yielded significant effects, the planned contrast comparisons were made to address the following questions:1 Does baseline differ from scores following arousal? A decrement in performance/subjective ratings after arousal was hypothesized.2 Is sleep inertia following SWS arousal greater than from REM sleep arousal? While the literature is not unanimous on this point, several key studies suggest that this question would be answered in the affirmative (see Introduction).3 Does sleep inertia exert its main effects in the first 10 min or so and then become dissipated? It was hypothesized that sleep inertia effects would be greater initially.
Overall decision-making performance scores were automatically generated by the ‘Fire Chief’ program for each 3-min trial. Each score corresponds to the value of the landscape segments saved, that is, the percentage of trees and houses that are still unburned at the end of the trial. It is important to understand that in the ‘Fire Chief’ program a performance score (percentage saved) is generated from each trial even if no decisions are made. These were subtracted from performance scores in order to determine the true amount actively saved during each trial by subjects. For example, if subjects made no attempt to make decisions during Trial 1 (0–3 min), they would still score 36.32%, and so this figure was subtracted from their baseline, SWS and REM arousal mean scores obtained during Trial 1. Across the 10 trials the percentage subtracted varied from 33.33 to 36.59%. (These adjustments were also applied to the preliminary analyses.)
Table 1 shows the means and SD of performance scores in the three conditions for the 10 trials. Figure 1 allows a visual analysis of the main trends and differences for decision-making performance.
Table 1. . Means, SD, decrement and relative performance capability (%) for decision-making performance comparing baseline (B), SWS arousal and REM sleep arousal (n=12)
It was found, using a repeated measures ANOVA, that there was a significant condition effect (F =7.50, P=0.01) but no significant time effect (F =6.86, P=0.07) or interaction effect. A Helmert planned contrast on the condition effect considered whether the baseline scores were different to the scores after SWS and REM arousals (averaged). A significant difference was found (F =15.78, P=0.002). Consideration of the descriptive statistics in Table 1 indicates that the subjects performed better in the baseline condition than in the arousal conditions. This Helmert planned contrast also tested whether performance following arousal from REM sleep was different to performance following SWS arousal and no significant difference was found.
As described in the data analysis section a similar ANOVA was also conducted on a shorter period to determine whether performance differences were evident within the reduced time sample. Thus the first three trials were analysed (9 min). A significant condition effect (F =13.08, P=0.002) and time effect (F =18.77, P=0.000) were found but there was no significant interaction effect. The Helmert planned contrast on the condition effect found that baseline performance scores differed significantly from the average of REM and SWS scores (F =28.48, P=0.000) and that performance after SWS arousal differed from REM arousal (F =10.34, P=0.008). Table 1 and Figure 1 indicate that baseline performance was better than performance upon arousal and that REM arousal performance was better than SWS arousal performance. A difference planned contrast was performed to determine the nature of the changes across time. It was found that trial 1 (0–3 min) differed from trial 2 (F =29.32, P=0.000) and that trial 3 differed from the average of trials 1 and 2 (F =7.08, P=0.022). Mean values show that performance in trial 1 was clearly reduced. (A summary of all key results can be found in Table 3.)
Table 3. . Summary of ANOVA results in terms of where key differences were found ( Note: for the condition effect B=Baseline, SWS=following arousal from SWS, REM=sleep arousal ratings
Relative performance capability – decision making
In order to obtain quantitative information about the comparative performance at baseline and following arousal several simple calculations were made. For each time block, within both SWS arousal and REM arousal, comparisons were made with baseline performance to obtain a measure of the relative performance capability (RPC) according to the following formula:RPC percentage=100 –[(PB – PA)/PB ×100].where: PB=performance at baseline for a trial, and PA=performance following arousal for a trial (calculated separately for SWS and REM).
Results of these calculations are shown in Table 1 and indicate that in the first 3 min following SWS arousal subjects were performing at only 50.6% of their baseline performance. In contrast, following REM arousal subjects were performing at 65.2% of their baseline in the first 3 min. While there is some variability across the time blocks, after 12 min arousal from either stage produced RPCs of ≈80% or better.
Subjective sleepiness (KSS)
KSS ratings every 6 min over the 30-min performance period were subjected to ANOVA analysis (as described in the data analysis section), with the means and SD shown in Table 2. A significant condition effect was found (F =24.69, P=0.000) but no significant time or interaction effect was noted. A Helmert planned contrast found that baseline ratings differed significantly from the average of SWS and REM arousal ratings (F =40.24, P=0.000), while REM arousal ratings did not differ significantly from SWS arousal. Subjects rated themselves as much more alert in the baseline condition compared with the arousal conditions.
Table 2. . Means (SD) for Karolinska Sleepiness Scale ratings 9KSS, (1–9 where 1=alert) and Clear-headed ratings (1–5, where 1=extremely clear-headed) comparing baseline with SWS arousal ratings and REM sleep arousal ratings (n=12)
A different pattern of results, however, emerged when only the first 12 min of KSS ratings were submitted to analysis. Both a significant condition (F =24.49, P=0.000) and time (F =10.91, P=0.003) effect were found using ANOVA analyses as described previously. No interaction effect was found. As for the full 30-min analysis, Helmert planned contrasts found that baseline ratings differed significantly from the arousal ratings (F =53.61, P=0.000) but in this analysis REM arousal ratings did not differ from SWS arousal ratings. The time effect was further explored using difference contrasts and the rating at 6 min was found to differ significantly from the one given immediately upon sitting up in bed when awoken (0 min; F =8.49, P=0.041) but the rating at 12 min did not differ from the average of the ratings at 0 and 6 min.
As with the KSS, ratings of clear-headedness were taken every 6 min between trials of the decision-making task. The pattern of significant ANOVA analyses was the same as for the KSS. For the 30-min analysis a condition effect (F =13.74, P=0.001) but no time effect or interaction effect was yielded. Helmert planned contrasts found that baseline ratings differed from those taken during the night upon arousal (F =24.00, P=0.000) but REM arousal ratings did not differ from SWS arousal ratings. For the analysis of the first 12 min, both a significant condition (F =23.55, P=0.000) and time (F =10.91, P=0.003) effect were found using ANOVA analyses. No interaction effect was found. Helmert planned contrasts found that baseline ratings differed significantly from the arousal ratings (F =24.00, P=0.000) but REM arousal ratings did not differ from SWS arousal ratings. The time effect was investigated using difference contrasts and the rating at 6 min was found to differ significantly from the one given at 0 min (F =45.70, P=0.000), but the rating at 12 min did not differ from the average of the ratings at 0 and 6 min. Mean values are in Table 2.
Intra-individual correlations were performed on the dependent variables (performance, subjective sleepiness and clear-headedness) to detect any relationships. There were no significant correlations among the performance scores and their corresponding sleepiness and clear-headed ratings. However, the sleepiness ratings correlated consistently with their corresponding clear-headed ratings (five of the six coefficients were above 0.63, P< 0.05).
Summary of results
For an overview of the key significant results from the series of ANOVAs conducted, see Table 3.
The most important finding from this study is that sleep inertia reduces decision-making performance for at least 30 min. This is longer than the sleep inertia literature reports to date, possibly arising from the greater complexity of the current task. Previous research using pathologically sleepy populations has shown that complex tasks are the most sensitive to sleepiness ( Hood and Bruck 1997). The current findings are consistent with the theoretical models mentioned previously that incorporate the postulated sleep inertia component. For instance, Folkard and Akerstedt's (1992)‘process W’ of the three process model of the regulation of alertness and sleepiness is further supported. The results also support the notion that sleep inertia is a period of confusion and decreased alertness, which consequently impairs the essential cognitive abilities of vigilance and alertness, necessary for sound and rational decision making ( Janis and Mann 1979). This finding has real-life implications given that many decisions in emergency situations (e.g. awakening to fire, instructions given by doctors, emergency repair workers, military personnel) are made within 30 min after arousal from sleep.
Consistent with the findings of persisting sleep inertia in performance, are the subjective rating scales where subjects reported themselves as sleepier and less clear-headed, compared with baseline, for at least 30 min after both SWS and REM awakening.
By far the greatest impairments (in terms both of performance and subjective ratings) were found within the first 3 min after abrupt nocturnal awakening. Decision-making performance was as little as 51% of optimum (i.e. baseline) during these first few minutes. However, after 30 min performance may still be as much as 20% below optimum (see Table 1 for the relative performance capabilities).
Table 1 and visual inspection of Figure 1 suggest that decision-making performance after REM arousal showed more variability than after SWS arousal. Perhaps these findings suggest that following REM sleep, subjects are able to arouse themselves for a brief period of time (i.e. 3–9 min post-arousal) to improve performance, whereas subjects may be less able to do so after SWS arousal. Consistent with the performance improvement in the 3–9 min interval is the indication that subjects’ subjective sleepiness ratings improved markedly at the 6 min time point after REM arousal. Sleep inertia is said to be a period of great attentional lability with lapses ( Tassi et al. 1992 ) and it has been argued that subjects can exert additional effort for brief periods of time to overcome performance deficits ( Dinges and Kribbs 1991). A highly stimulating task will help foster the motivation needed for selected periods, but cannot guarantee that the impairment in performance produced by sleepiness will be completely overridden indefinitely. Horne (1988) proposes that, within some range of sleepiness, performance deficits can be overridden by the input of motivational or attentional resources. It is conceivable that subjects were able to temporarily arouse themselves more easily following REM sleep awakenings compared with SWS, since REM sleep is characterized by higher brain reactivity and excitability than SWS ( Dujardin et al. 1989 ).
The initial effects of sleep inertia during the first 9 min are significantly greater after SWS arousal than after REM arousal, but this difference is not sustained beyond the first 9 min. This finding of a difference related to pre-awakening sleep stage is consistent with most of the literature (see Introduction) but to our knowledge, this is the first to show differential effects over time. It is possible that a stronger SWS sleep inertia effect (and hence a sustained performance difference compared with REM arousal) may have been maintained if the SWS arousal had been from the first sleep cycle rather than from the second.
In several analyses ( Table 3) a significant time effect was noted. With regard to decision-making performance the statistical differences found would appear to be largely a result of the large initial decrement (trial 1) that was found across all conditions (baseline, SWS and REM sleep) and is a result of the Fire Chief decision-making task having a ‘warm up effect’. It cannot be argued that this specifically arises from sleep inertia because no interaction effect was found, with performance on trial 1 in the baseline condition also clearly being reduced. Similarly a time effect was found in both the sleepiness and clear-headed subjective ratings, whereby the rating immediately upon awakening was significantly more sleepy and less clear-headed than the subsequent 6-min rating, but no interaction effect reached statistical significance.
Correlational analyses suggest that there were no significant relationships between any of the decision-making scores and the corresponding sleepiness ratings, or between the decision-making scores and the corresponding clear-headed ratings. However, in the present study most of the subjective sleepiness ratings did significantly correlate with the clear-headed ratings. As the clear-headed scale was exploratory, this latter finding raises the question of whether clear-headedness is a part, or subjectively the same as, sleepiness. While it may be too early to be certain from this study, the present results provide no evidence of the two variables being independent constructs.
One potential confounding factor, consistently found in the sleep inertia literature, is that it is unclear how one differentiates sleep inertia from the circadian influences on performance. This study attempted to minimize confounding effects due to circadian factors by counterbalancing the order of SWS and REM awakenings across subjects (given that both occurred on the one night) and also by having the baseline scores being a combination of morning and evening scores (this was also required to avoid a systematic bias related to experience with the performance task). Nevertheless, circadian factors do exert their influence when we compare baseline performance with sleep inertia performance. As in all sleep inertia studies (except those using naps or sleep deprivation designs) a circadian effect is nested within the sleep inertia effect and thus the sleep inertia decrement found represents the sum of both a circadian middle-of-the-night decrement and a sleep inertia effect.
Work in this area, like that of arousal thresholds, should ideally control systematically for a variety of potential confounding variables including time of night, length of prior sleep, amount of prior sleep stages and duration of prior wakefulness. While it would be impossible to simultaneously control for many of these factors, a greater understanding of the effects of such variables may resolve the current inconsistencies in the literature, especially in relation to sleep stage of awakening and duration of sleep inertia given similar tasks ( Stones 1977; Bonnet 1983; Dinges et al. 1985; Mullington and Broughton 1994 ).
There is anecdotal evidence that some individuals consistently experience a strong sleep inertia effect while others feel they are able to become alert and function well relatively quickly. Visual inspection of the raw data of this study suggested that two of the 12 subjects showed particularly strong, consistent and enduring sleep inertia effects and it would be interesting to pursue further the examination of which variables (e.g. circadian, personality) may be associated with such individual differences.
This study did not investigate the effects of noise on sleep inertia as the alarm that aroused subjects was turned off immediately. Tassi et al. (1992) found that an intense continuous noise produced a total abolishment of sleep inertia after an early nap, yet noise had no effect on a later nap. This demonstrates that the effects of noise on sleep inertia requires further investigation. Furthermore, the present study required the subject to stay in bed, with only a very dim light on. It is acknowledged that activity and perhaps bright light, may also abolish or decrease sleep inertia, and this requires further research.
Although much remains to be determined about the variables influencing decision-making performance upon abrupt awakening, this study is the first step in documenting the effects of sleep inertia on decision-making performance; an area of research with important implications for understanding psychological functioning in emergency situations.
The assistance of Dr Mary Omedei with helpful advice at various stages of this study was much appreciated. Thanks also to Dr Bernadette Hood for useful comments on the manuscript.
Accepted in revised form 18 December 1998; received 14 May 1997