wakefulness on behavior and neurophysiologic measures made during
performance of well-practiced working-memory tasks. Data...
employed in studies of mental workload. At three occasions over the
extended-wakefulness testing session, participants als...
changes in the incidence of slow eye movements, a bipolar derivation
was computed between a pair of electrodes placed at t...
correlation between these two error types over participants and test
intervals (Spearman rho = 0.69, P < .001).
Impact of ...
effect reflects some type of transient beginning-of-session difference in
the overall state of arousal. To evaluate this p...
latency, and response-variability measures. The participants and test
intervals that displayed the largest reduction in LP...
The Impact of Extended Wakefulness on the Phasic ERP
Careful analyses of behavior measures during sleep deprivation have
the participants and then tested by application to the remaining participant.
Figure 10 summarizes the results of the cros...
working memory capacity?! Intelligence 1990;14:389-433
17. Gevins A, Smith ME. Neurophysiological measures of working
ations in vigilance and the EEG spectrum. Electroencephalogr Clin
Neurophysiol 1983;55:108-13
61. Oken BS, Salinsky M. Ale...
Upcoming SlideShare
Loading in...5

The Impact of Moderate Sleep Loss on Neurophysiologic Signals ...


Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

The Impact of Moderate Sleep Loss on Neurophysiologic Signals ...

  1. 1. INTRODUCTION THIS PAPER EXAMINES THE IMPACT OF MODERATE SLEEP LOSS ON BEHAVIOR AND NEUROPHYSIOLOGIC MEASURES RELATED TO WORKING-MEMORY TASK PERFORMANCE. In children and adolescents, sleep loss tends to be associated with impaired performance on neuropsychologic tests and lower levels of scholastic achievement.1-3 Equivalent patterns of degraded performance on the job or on controlled tests have been observed in working adults experienc- ing interruption of normal sleep patterns.4-7 Not surprisingly then, sleep loss is frequently implicated in major workplace accidents.8,9 Many such accidents likely result from the fact that, although rote activities might be performed adequately even after moderate sleep loss, subtle changes in a drowsy individual’s cognitive ability might nonetheless diminish his or her capacity to respond optimally when circumstances arise that require rapid evaluation and thoughtful action. The ability to perform competently in such attention-demanding cir- cumstances depends in part on the integrity of working memory. Work- ing memory refers to the limited capacity to hold and manipulate infor- mation in mind for several seconds in the context of cognitive activity. In a sense, working memory is an outcome of the ability to control and sustain attention and to strategically focus it on a particular mental rep- resentation in the face of distracting influences.10,11 This ability plays an important role in comprehension, reasoning, planning, and learning.12 Indeed, the use of active mental representations to guide performance appears critical to behavior flexibility,13,14 and measures of this ability tend to be positively correlated with performance on psychometric tests of cognitive ability and other indexes of scholastic aptitude.15-17 Metabolic studies suggest that working memory and the neural mech- anisms of attention control involve cortical circuits linking regions of prefrontal cortex with posterior association cortexes.18-22 Activation of these circuits also occurs during reasoning and problem solving.23-25 Such activation can be detected in measurements of neuroelectric activ- ity recorded at the scalp. More specifically, a sustained-task-imposed change in working-memory requirements tends to produce characteris- tic tonic changes in the amplitude of attention-related spectral features of the ongoing electroencephalogram (EEG)26-28 that are punctuated by more phasic changes in components of the stimulus-locked, transient event-related potentials (ERP).29-31 Under normal conditions, neurophysiologic signals modulated by task-imposed variations in working-memory demands tend to be very stable. For example, in a recent study in which the test-retest reliability of task-related EEG spectral features was examined in well-practiced subjects, correlations of r > .9 were found between two test sessions with a 1-week lag.32 Despite this apparent stability when other factors are held constant, there is evidence that task-related neurophysiologic sig- nals are highly sensitive to stressors that impose some form of mild, transient cognitive impairment. For example, recent studies have demonstrated that working-memory-sensitive neurophysiologic signals vary in conjunction with low doses of alcohol33-35; diphenhydramine34; marijuana36; and prescription medications frequently used to treat neu- rologic37 and psychiatric38 disorders. Results of this type suggest that the cortical networks responsible for generating working-memory-sensitive neurophysiologic signals might also be disrupted in conjunction with moderate sleep loss. To evaluate this possibility, the current study examined the impact of extended Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 56 The Impact of Moderate Sleep Loss on Neurophysiologic Signals during Working- Memory Task Performance VARIATIONS IN SLEEP AND PERFORMANCE Michael E. Smith, PhD; Linda K. McEvoy, PhD; Alan Gevins, D.Sc San Francisco Brain Research Institute and SAM Technology, San Francisco, California Study Objectives: This study examined how sleep loss affects neuro- physiologic signals related to attention and working memory. Design: Subjective sleepiness, resting-state electroencephalogram, and behavior and electroencephalogram during performance of working-mem- ory tasks were recorded in a within-subject, repeated-measures design. Setting: Data collection occurred in a computerized laboratory setting. Participants: Sixteen healthy adults (mean age, 26 years; 8 female) Interventions: Data from alert daytime baseline tests were compared with data from tests during a late-night, extended-wakefulness session that spanned up to 21 hours of sleep deprivation. Measurements and Results: Alertness measured both subjectively and electrophysiologically decreased monotonically with increasing sleep deprivation. A lack of alertness-related changes in electroencephalo- graphic measures of the overall mental effort exerted during task execu- tion indicated that participants attempted to maintain high levels of perfor- mance throughout the late-night tests. Despite such continued effort, responses became slower, more variable, and more error prone within 1 hour after participants’ normal time of sleep onset. This behavior failure was accompanied by significant degradation of event-related brain poten- tials related to the transient focusing of attention. Conclusions: Moderate sleep loss compromises the function of neural circuits critical to subsecond attention allocation during working-memory tasks, even when an effort is made to maintain wakefulness and perfor- mance. Multivariate analyses indicate that combinations of working-mem- ory-related behavior and neurophysiologic measures can be sensitive enough to permit reliable detection of such effects of sleep loss in individ- uals. Similar methods might prove useful for assessment of functional alertness in patients with sleep disorders. Key Words: Fatigue, alertness, cognition, attention, working memory, electroencephalogram, event-related potentials Disclosure Statement This research was sponsored by the National Aeronautics and Space Administration, the National Institute of Mental Health, the National Institute of Neurological Disorders and Stroke, and The Air Force Office of Scientific Research. Submitted for publication February 2002 Accepted for publication June 2002 Address correspondence to: Michael E. Smith, Ph.D., SFBRI , 425 Bush Street, 5th Floor, San Francisco, CA 94108; Tel: 415. 837.1699; Fax: 415. 274.9574; E-mail:
  2. 2. wakefulness on behavior and neurophysiologic measures made during performance of well-practiced working-memory tasks. Data from a sam- ple of healthy young adult subjects were compared between an average alert daytime baseline test interval and 5 nighttime test intervals that occurred over the course of an extended-wakefulness session in which the participants performed tasks throughout a night of sleep deprivation. METHODS Subjects Sixteen healthy adults (21-32 years; mean age, 26 years; 8 females) participated. Most were full-time college students and as such they tend- ed to maintain fairly irregular sleep-wake schedules. All were non-smok- ers, social drinkers (1-10 drinks per week), and mild to moderate caf- feine consumers (1-3 cups of coffee per day). None were taking any cen- trally acting medications during the course of the study, and none had been previously diagnosed with any psychiatric or neurologic problems or sleep disorders. All participation was fully informed and voluntary and conformed to all institution and government guidelines for the pro- tection of human subjects. Participants were remunerated on an hourly basis for their efforts. To minimize attrition, a significant bonus was also awarded to all participants upon completion of all test sessions. Finally, to increase motivation to task performance, extra bonus payments were awarded in a competitive fashion to the two participants who had the best overall accuracy across all test sessions. Tasks The focus of the present paper is on the effects of sleep deprivation during working-memory tasks. Participants performed two difficulty levels of a continuous performance, n-back–style working-memory-task versions of which we have employed in prior EEG studies.17,26-30,32-34 Similar n-back tasks have been adopted in many other laboratories as a means to activate functional networks of working memory in a con- trolled fashion. Such applications have spanned conventional behavior studies,39 other electrophysiologic studies,40,41 studies of the effects of magnetic fields on cognitive function,42,43 and functional neuroimaging studies employing positron emission tomography or functional magnet- ic imaging methods.44-48 In the versions of this task employed in the present study (Figure 1), participants were required to compare the spatial location of the current stimulus with that of one presented previously. Briefly, single capital-let- ter stimuli, drawn randomly from a set of 12, were presented for 200 mil- liseconds once every 4.5 seconds on a computer monitor. At 1.3 seconds prior to stimulus onset, a warning cue (a small x) appeared in the center of the screen for 200 milliseconds. The letter stimulus occurred 1.3 sec- onds after the cue in 1 of 12 possible locations on the monitor. The iden- tity of the letter and its spatial position varied randomly from trial to trial. A small fixation dot was continuously present at the center of the screen. In an easy, low-working-memory-load version of the task, subjects were required to match the position of the current stimulus with the posi- tion of the very first stimulus in the block. In a difficult, high-working- memory-load version of the task, participants compared the current stim- ulus with that presented two trials previously. In this version, they were required to remember two positions (and their sequential order) for the duration of 2 trials (9 seconds) and to update that information on each subsequent trial. In both versions of the task, stimuli were presented in blocks of 53 trials (the first 3 trials were warm-up trials and were dis- carded from analysis). Matches occurred randomly on 50% of the trials. Participants were instructed to respond as quickly and as accurately as possible. Procedures Each subject participated in a total of 6 experimental sessions. The first session was a practice session in which subjects learned to perform the working-memory tasks to an asymptotic level. All subjects then par- ticipated in 5 subsequent sessions, separated by at least 1 week. Four ses- sions involved recording from participants before and after they had ingested alcohol, caffeine, antihistamine, or placebo. These sessions occurred in a counter-balanced order, and the data concerning drug effects are described in a separate report.34 A fifth experimental session was an extended-wakefulness manipulation that began in the evening and lasted until 6:00 AM the following morning. This extended-wakeful- ness session is the focus of the current report. However, since each of the four drug sessions involved a baseline recording prior to drug adminis- tration (around 12:00 PM on each day), the average of these alert, day- time, baseline sessions was used as a comparison point for the data col- lected in the extended-wakefulness session. In the week prior to the extended-wakefulness session, participants were required to keep a sleep diary indicating (among other things) the time on each day they went to sleep and the time they awoke. In the overnight extended-wakefulness session, subjects arrived at the laboratory at approximately 8:30 PM, were given a warm-up block of the working-memory task and other tasks, and were prepared for the EEG recording. They then participated in five 40-minute recording intervals spaced throughout the night. The first interval occurred on average at 11:00 PM, the second at 12:30 AM, the third at 1:30 AM, the fourth at 3:30 AM, and the fifth at 5:00 AM. The internal structure of each interval was the same as that used during the daytime baseline sessions. Within each interval, participants completed the Karolinska49 Subjective Sleepiness Rating Scale and had their EEG recorded while they performed two blocks each of the easy and difficult versions of the working-memory task. At each interval, EEG data were also recorded while the partici- pants rested quietly for 90 seconds with their eyes open and closed. Such resting-state data were analyzed in order to obtain objective neurophys- iologic measures of changes in alertness. The resting conditions, the subjective measures, and the four blocks of the working-memory task took approximately 20 minutes to complete during each testing interval. In the periods between these recording blocks, the participants performed other repetitive computer tasks to ensure continued wakefulness and to help induce mental fatigue. Thus, at each of the test intervals, participants also performed a 10-minute ver- sion of the Psychomotor Vigilance Task, which is a simple reaction-time task that has been frequently used in sleep deprivation studies,50-52 and a 5-minute version of the Multi-Attribute Task Battery,53 which is a com- puter-based multitasking environment that simulates some of the activi- ties a pilot might be required to perform and that has often been Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 57 Figure 1—Schematic diagram of the low memory load and high memory load versions of the n-back working-memory task. Every 4.5 seconds, one of 12 possible capital-letter stimuli appeared in one of 12 possible locations on a computer monitor. In the easy version of the task, subjects compared the location of the current stimulus, regardless of its identity as a letter, with the location of the very first stim- ulus in the block of 50 trials. In the difficult version of the task, subjects compared the location of the cur- rent stimulus with the location of the stimulus presented two trials previously.
  3. 3. employed in studies of mental workload. At three occasions over the extended-wakefulness testing session, participants also performed a 20- minute version of a simple visual “oddball” discriminative judgment vigilance task. During periods when the participants were not actively involved in one of the formal tasks, they were allowed to read, play video games, or surf the Internet, but no napping was permitted. EEG Recording and Preprocessing The EEG was continuously recorded from 28 scalp electrodes using a linked-mastoids reference. The electrooculogram was recorded from electrodes placed above and below one eye and at the outer canthus of each eye. Physiologic signals were band-pass filtered at 0.01 hertz to 100 hertz and sampled at 256 hertz. Data were digitally filtered offline with a zero-phase-shift, 0.5 hertz, high-pass IIR filter. Automated artifact detection was followed by application of adaptive eye-contaminant removal filters.54 The data were then visually inspected, and data seg- ments containing possible residual artifacts were eliminated from subse- quent analyses. To obtain power spectral estimates, Fast Fourier trans- forms were applied, and periodograms were computed on 50% over- lapped, 512-sample Hanning windows for all artifact-free trials and averaged over all data within a condition. Average spectra were convert- ed to dB power by normalizing them with a log10 transform. The ERP were computed by averaging together all of the artifact-free epochs of the stimulus-locked EEG for correctly classified trials within each con- dition for each subject. These epochs began 0.2 seconds before stimulus onset and extended to 1.0 second afterward. The ERP were low-pass fil- tered at 20 hertz prior to subsequent analyses. Analyses To examine how individual behavior and physiologic features varied over the extended test session, univariate repeated measures analyses of variance were used to compare data from the 12:00 PM baseline to data recorded across the different test intervals of the extended-wakefulness session. The experiment-related sources of variance in each of the indi- vidually measured parameters were assessed in univariate task (easy vs. difficult) or resting condition (eyes-open vs. eyes-closed) by test interval repeated measures ANOVA. Changes in mental function related to extended wakefulness would thus be expected to manifest as a main effect of test interval or as an interaction involving test interval. The Greenhouse-Geisser correction to degrees-of-freedom was employed to correct for any violations of the sphericity assumption in analyses involving repeated measures. In such cases, the reported p values corre- spond to the corrected degrees of freedom. RESULTS The sleep-diary data indicated that over the week prior to the extend- ed-wakefulness session, the participants typically fell asleep between 11:30 PM and 1:00 AM and received about 7 hours (range, 6-9 hours) of sleep on an average night. On the day of the extended-wakefulness ses- sion, participants reported awakening between 6:00 AM and 10:30 AM (average, 8:00 AM) after having obtained an average of 7 hours of sleep. Thus, the first test (11:00 PM) during the extended-wakefulness session occurred, on average, after 15 hours of wakefulness, and the final test session at 5:00 AM took place after about 21 hours (range, 19.5 to 23 hours) of wakefulness. No caffeinated beverages were consumed during the extended-wake- fulness session itself. Twelve of the participants consumed no caffeine after noon of the day of the extended-wakefulness session. The other 4 participants had last consumed caffeinated beverages in the late after- noon. None consumed a total of more than three caffeinated beverages on the day of the extended-wakefulness session. Impact of Extended Wakefulness on Subjective and EEG Measures of Sleepiness Subjectively, participants reported that they felt most alert in the 12:00 PM baseline interval and progressively less alert at later test intervals, with minimal alertness usually reached during the final test at about 5:00 AM (Figure 2). This change was reflected in a significant main effect of test interval (F(5,75) = 23.31; P < .001). Posthoc pair-wise comparisons indicated that although there was no significant difference between the 11:00 PM interval and the 12:00 PM baseline, subjects had become sig- nificantly sleepy (P < .01) by the 12:30 AM interval. Eye-movement activity and EEG spectra recorded while the partici- pants rested quietly displayed classic signs of sleepiness, providing con- vergent evidence that alertness decreased approximately linearly over the course of the extended-wakefulness session. For example, past research has indicated that horizontal slow eye movements tend to increase as individuals become increasingly sleepy.49,55,56 To measure Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 58 Figure 2—Average (±SEM) subjective sleepiness ratings assessed with the Karolinska Sleepiness Scale. Sleepiness did not differ significantly between the 12:00 PM alert baseline and the 11:00 PM test interval. It increased in an approximately linear fashion from the 11:00 PM test interval to the 5:00 AM test interval. Figure 3—Average electroencephalographic power recorded from a midline occipital electrode in the resting eyes-closed (top) and eyes-open (bottom) conditions during the 12:00 PM baseline and during the 5 AM test interval. In both conditions, electroencephalographic power below the alpha band was rel- atively higher at the 5 AM test interval. In contrast, alpha-band power was relatively attenuated at 5 AM in the eyes-closed condition.
  4. 4. changes in the incidence of slow eye movements, a bipolar derivation was computed between a pair of electrodes placed at the outer canthi of the eyes, and average power in a 0.5-hertz to 1.0-hertz band was extract- ed from this derivation prior to removal of eye-movement contaminants. By this measure, slow eye movements increased significantly over the test session (F(5,75) = 10.52; P < .001). Past studies have also consistently demonstrated a drowsiness-related pattern of changes in the spectral composition of the resting state EEG.57- 61 In general, particularly during eyes-closed resting conditions over parietooccipital electrode locations, spectral power in the delta and theta bands tends to increase during sleepy states, whereas alpha-band power tends to decrease. These effects were observed in the present data (Fig- ure 3). Specifically, across resting conditions, delta activity measured as the average power between 2 hertz and 4 hertz at electrode Pz increased over test intervals (F(5,75) = 3.87, P = .01), as did posterior theta mea- sured as the average power between 4 hertz and 6 hertz at the same site (F(5,75) = 4.3, P = .01). In contrast, occipital alpha measured as the average power in a 1-hertz band around the peak frequency in the 8-hertz and 12-hertz range at Oz was significantly attenuated by extended wake- fulness in the eyes-closed resting condition but not in the eyes-open rest- ing condition (F(5,75) = 4.61 P < .01). For both slow eye movements and EEG spectral measures during resting conditions, the observed changes paralleled those observed in the subjective ratings, with no sig- nificant differences obtained between the 11:00 PM interval and the 12:00 PM baseline, a significant difference from baseline occurring by the 12:30 AM interval, an approximately monotonic increase (or decrease in the case of the alpha rhythm during the eyes-closed condition) over the course of the extended-wakefulness session, and a maximum differ- ence from baseline during the 5:00 AM test interval. Impact of Sleep Loss on Behavior Measures during Task Performance To examine changes in task performance over the course of the extended-wakefulness session, stimulus classification accuracy was characterized in terms of the ability to detect match trials as character- ized by d’62. Reaction times were measured in milliseconds and then log normalized prior to statistical analyses. Across test intervals, subjects responded more accurately (F(1,15) = 38.22; P < .001) and faster (F(1,15) = 53.58; P < .001) in the low-load task than in the high-load task (Figure 4). These effects of task load did not interact with test inter- val. Classification accuracy was approximately 15% lower late at night than during the 12:00 PM baseline (F(5,75) = 4.90; P < .02), and reaction times were slowed on average by about 100 milliseconds (F(5,75) = 6.53; P < .001). Independently of the overall slowing, reaction times also became relatively more variable over test intervals. That is, there was a significant increase in the “coefficient of variability” (computed for each subject as the ratio of the standard deviation of reaction times over trials within a block of trials to the average of reaction times within the block) over test intervals (F(5,75)=4.28, P < .007). There were no significant interactions obtained between task load and interval for accuracy, reac- tion time, or reaction-time variability. Across tasks, none of these mea- sures significantly differed between the 12:00 PM baseline and the 11:00 PM test interval. Marginal impairment in performance could be observed by the 12:30 AM test interval, and performance reached a nadir by the 1:30 AM test interval. That is, performance in both the low-memory-load and high-memory-load tasks had become significantly impaired by a rel- atively early stage of the extended-wakefulness session. Performance remained at that degraded level through the remaining test intervals, with no further decrement between 1:30 AM and 5:00 AM. To more carefully evaluate the pattern of errors observed during the course of the extended-wakefulness session, the rate of false match responses (1.8% overall) was compared with the rate of false nonmatch responses. Although there was a slightly higher incidence of false non- match response on average (2.2% overall), this difference did not reach significance (F(1,15) =3.0, P = .10), nor was there any interaction between type of false response made and test interval. Furthermore, there was no significant difference between the rate of errors of com- mission (false-positive match and nonmatch responses, 4% of all trials) and errors of omission (failures to respond in the time allowable, 5.2% of all trials) nor was there any interaction between these two types of errors and test interval (both Fs < 1). Finally, there was a high positive Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 59 Figure 4—Average (±SEM) reaction time (top) and accuracy (bottom) in the easy and difficult levels of the working-memory task over test intervals. In both task levels, accuracy decreased significantly and reaction time increased significantly with time; performance decrements reached an asymptotic trough at about 1:30 AM, remaining at that level throughout the rest of the test session. Figure 5—Grand average power spectra for task-related electroencephalography in the low-load and high-load levels of the working-memory task over frontal and parietal midline sites. Over frontal areas, power in the frontal midline theta band (6-7 hertz) was larger in the high-load task than in the low-load task. Over parietal areas, alpha power, in both the slow (8-10 hertz) and fast (10-12 hertz) alpha bands was larger in the low-load task than in the high-load task.
  5. 5. correlation between these two error types over participants and test intervals (Spearman rho = 0.69, P < .001). Impact of Sleep Loss on EEG Spectral Measures during Task Performance Figure 5 compares average power spectra in the low-load and high- load working-memory tasks. As noted in the introduction, past studies indicate that regional power spectral measures in the theta and alpha bands are sensitive to variations in the working-memory load imposed by continuous performance tasks like those employed herein. In partic- ular, the theta rhythm at frontal midline sites tends to be enhanced with higher working-memory loads, whereas the alpha rhythm tends to be attenuated, particularly at parietal sites.17,26,27,30,32-34 Such tonic differ- ences in EEG spectral composition between tasks that differ continu- ously in working-memory requirements might be expected to reflect the integrity of attention-control mechanisms. That is, they would appear to index differences in the effort that subjects willfully make and sustain over the course of several minutes in response to variations in the demands of the task environment. To quantify these phenomena for statistical analyses, the spectral power of the working-memory-load–sensitive frontal theta rhythm was measured at the peak frequency in a 5-hertz to 7-hertz band at frontal midline electrode site AFz. The power of the alpha rhythm was original- ly measured in both an 8-hertz to 10-hertz band and a 10-hertz to 12- hertz band at electrode site Pz. Preliminary analyses indicated no impor- tant differences between these two alpha-band measures in response to sleep loss, so they were collapsed into a single task-related parietal alpha measure. Mean spectral power for frontal midline theta and parietal alpha rhythms at each test interval is presented in Figure 6. As in prior studies, the power of the frontal midline theta rhythm was observed to be greater in the high-load task than in the low-load task (F(1,15) = 7.08, P < .02). In contrast, increasing working-memory load was associated with attenuation of power in the alpha band at parietal sites (F(1,15) =47.25, P < .001). Neither of these signals displayed a test interval by task load interaction (both P values > 0.1), indicating that the magnitude of the task load effect on EEG spectral parameters did not significantly differ over the course of the test session. Despite the lack of any interaction with task load, both the frontal midline theta measure (F(5,75) = 7.59, P < .001) and the parietal alpha measure (F(5,75) = 3.24, P < .04) displayed overall main effects of test interval. As can be seen in Figure 6, the absolute power of both signals was lower at the 11:00 PM test interval of the extended-wakefulness-test session than it was at either the 12:00 PM daytime baseline or at the late- night test intervals. Since this effect does not appear to be related in any consistent fashion to either the subjective or objective measures of sleepiness described above, or to the observed pattern of changes in the task-related behavior measures, it is unlikely that the transient signal attenuation at approximately 11:00 PM is related to the same factors responsible for those other changes. An alternative possibility is that the Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 60 Figure 6—Average (±SEM) EEG power, in decibels, for the anterior frontal midline theta band (6-7 hertz) (top) and the parietal alpha band (8-12 hertz) (bottom). The impact of increasing working-memo- ry task load on these spectral parameters of the EEG remained relatively constant across test intervals. This suggests that although performance degraded over the night, the subjects continued to allocate greater neural resources and make a greater effort to perform the more difficult task throughout the night. Figure 8—Grand average (N,16), stimulus-locked, event-related potentials over central and parietal midline sites. Data are collapsed over low-load and high-load versions of the working-memory task and over matching and nonmatching stimulus categories. The amplitudes of the P215 and late positive com- ponent (LPC) peaks were reduced by the 1:30 AM test period relative to the 12:00 PM baseline period. In contrast, the reduction in the slow wave (SW) peak that followed the LPC did not reach significance. Data were averaged over all intervals, and low-pass filtered at 7 hertz. Figure 7—Grand average (N,16), stimulus-locked, event-related potentials over the right parietooccipi- tal area. Data are collapsed over low-load and high-load versions of the working-memory task and over matching and nonmatching stimulus categories. The amplitude of the N170 peak was reduced by the 1:30 AM test period relative to the 12:00 PM baseline period. Data were bandpass filtered at 2 hertz to 20 hertz.
  6. 6. effect reflects some type of transient beginning-of-session difference in the overall state of arousal. To evaluate this possibility, the measure- ments of the task-related EEG were recomputed as difference scores from corresponding measurements made in the eye-open resting condi- tion (under the presumption that an overall state difference would have similar effects on both the resting and the task-related EEG). In ANOVA performed on these difference scores, the effects of task-load observed above were still obtained for both the frontal midline theta signal and for parietal alpha signal, but no main effects of test interval or test interval by task load interactions were observed for either signal. Impact of Sleep Loss on Stimulus-Locked ERPs Studies of extended wakefulness,63-67 and of medication use that alters alertness,68,69 have found that attention-related and decision-related com- ponents of the ERP tend to be affected by sleepiness. Several compo- nents of the ERP response in the current study appear to be associated with attention and decision making in that they varied systematically between correctly classified matching versus nonmatching stimuli, between task loads, or between stimuli and task loads. These compo- nents included: 1) an N170 recorded over parietooccipital regions (Fig- ure 7); 2) a central P215; 3) a “P300” or late positive component (LPC); and 4) a positive slow wave recorded over central and parietal regions (Figure 8). To quantify these phenomena for statistical analyses, ERP component peak latencies were measured with respect to the time of stimulus onset, and amplitudes were measured with respect to the aver- age amplitude 200-millisecond prestimulus onset. The amplitude of the N170 (as measured at lateral parietooccipital electrode P8) had complex task correlates, displaying a significant task- load-by-stimulus-type interaction (F(1,15) = 11.07, P < .005). This inter- action resulted from a lack of difference in N170 amplitude between the two task types for nonmatch stimuli, whereas for match stimuli the N170 was much larger in amplitude in the low-load task than in the high-load task. The amplitude of the N170 (Figure 9, top) was attenuated over test intervals (F(5,75) = 6.33, P < .001). This attenuation did not interact with stimulus type or task-load factors. Similarly, the peak latency of the N170 was observed to increase over test intervals (F(5,75) = 7.46, P < .001), with a total increase of about 10 milliseconds between the 12:00 PM baseline measure and the 5:00 AM test interval. For both the ampli- tude and the latency measures, posthoc analyses revealed significant (P < .01) effects by the 1:30 AM test interval, with no further significant change over the remaining test intervals. Peak amplitude of the P215 was larger for correctly classified non- match stimuli than for match stimuli (F(1,15)=8.26, P < .02), indepen- dent of task load. A significant main effect of test interval (Figure 9, mid- dle) was also obtained (F(5,75)=4.09, P < .01). There were no interac- tions between test interval and the other factors. As with the N170, rela- tive to the 12:00 PM baseline, there was a significant (P < .01) reduction in P215 amplitude by the 1:30 AM test interval. This attenuation in ampli- tude persisted throughout the remainder of the session, though the P215 displayed a trend toward recovery at the 3:30 AM and 5:00 AM test inter- vals. No significant effects were observed for the peak latency of the P215. Amplitude of the LPC (measured as the average amplitude in a 50- millisecond window around peak latency at parietal midline electrode Pz) did not display significant main effects of task load or stimulus type. There was, however, a significant interaction between these factors (F(1,15)=9.01, P < .01); in the low-load task, the LPC amplitude did not differ between match and nonmatch stimuli, whereas in the high-load task, the amplitude of this potential was much smaller to match stimuli than to nonmatch stimuli. A significant effect of test interval (Figure 9, bottom) was also obtained (F(5,75)=5.58, P < .04). As with the other ERP measures, posthoc analyses revealed that by the 1:30 AM test-inter- val amplitude of the LPC was significantly (P < .01) attenuated, and this attenuation persisted at approximately the same level throughout the remaining test intervals. Peak latency for the LPC did not significantly differ as a function of task load or of test interval. The positive slow-wave component occurring at approximately 500 milliseconds to 600 milliseconds following stimulus onset (measured as the average amplitude in a 100-millisecond window around peak ampli- tude at parietal electrode Pz) was larger to correctly identified nonmatch stimuli than to match stimuli (F(1,15) = 22.68; P < .001), but it did not differ significantly over test intervals. The behavior data reviewed above indicated that accuracy decreased and that reaction times increased and became more variable on a trial- by-trial basis (as measured by the coefficient of variability) in the late- night test sessions relative to the alert baseline. To examine whether there was a systematic relationship between these performance changes and the observed attenuation of ERPs with sleep loss, the change in LPC amplitudes between the daytime baseline measure and the extended- wakefulness session was calculated for each subject and test interval, and correlated with the corresponding change in the accuracy, response Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 61 Figure 9—Average (±SEM) amplitude for the N170 (top), P215 (middle), and LPC (bottom) peaks of the stimulus-locked ERP collapsed across stimulus type and task-load conditions. The data were normal- ized across test intervals within each subject for each peak. Values for the negative N170 peak were inverted to make them graphically comparable to the P215 and the LPC. The average amplitudes for each peak did not differ between the 12:00 PM baseline and the 11:00 PM test interval. Amplitudes were significantly attenuated between the 11:00 PM and 1:30 AM test intervals. For the N170 and LPC, this attenuation reached a nadir at 1:30 AM and displayed little further change over the remaining test inter- vals. For the P215, there was a slight recovery in amplitude at later test periods. These results suggest that the ability to efficiently focus attention on transient stimulus events rapidly decayed in response to a moderate delay in normal sleep-onset time. LPC, late positive component
  7. 7. latency, and response-variability measures. The participants and test intervals that displayed the largest reduction in LPC amplitude also tend- ed to display the largest decrease in response accuracy (Pearson r = 0.48, P < .001), the largest increase in reaction times (r = -0.42, P < .001), and the largest increase in response variability (r = -0.48, P < .001) DISCUSSION The objective of this study was to evaluate whether and how moder- ate sleep deprivation affects behavior and neurophysiologic measures of working-memory task performance. The extended-wakefulness manipu- lation induced significant sleepiness as measured with conventional sub- jective and electrophysiologic criteria. Even moderate levels of sleep loss were found to impair task performance and to disrupt neurophysio- logic signals related to phasic aspects of working-memory task perfor- mance. These findings are discussed below. Effects of Sleep Loss on Overt Task Performance Some recent research has suggested that 1 night of sleep deprivation is not enough to impair performance on tasks designed to assess higher cognitive functions.70 However, most studies have found that modest amounts of sleep loss can significantly degrade performance on tests that explicitly incorporate a working-memory component or that otherwise are thought to require contributions from prefrontal cortical regions.71-74 Such deficits are consistent with the impairments in working memory and other executive functions that have been observed to accompany the sleep fragmentation and hypoxia associated with obstructive sleep apnea.75-77 The effects of sleep loss on behavior measures in the current study appear to be entirely consistent with such findings. The observed effects of the working-memory-load manipulations on behavior data were in good agreement with previous find- ings17,26,27,29,30,32,33; accuracy was lower and reaction times were slower in the high-memory-load task than in the low-memory-load task. Of more interest to the current context, behavior began to degrade by the 12:30 AM test interval relative to the 12:00 PM alert baseline level. The beginning of this decline in overt performance was in good agreement with subjective measures of sleepiness. However, behavior reached a significant, asymptotic level of impairment by the 1:30 AM test interval. This rapid decline to an asymptotic level of functional impairment is in contrast to subjective sleepiness and to objective measures of alertness obtained from resting EEG data, which suggested that drowsiness con- tinued to increase between the 1:30 AM and 5:00 AM test intervals. Successful performance of the working-memory tasks demanded par- ticipants to sustain mental effort over several minutes, to strategically control and quickly focus their attention in order to accurately register transient stimulus events, and to rapidly compare incoming information to that maintained in working memory. The behavior data demonstrate that some critical aspect of this ability was significantly impaired when wakefulness was extended just an hour past the normal time of sleep onset. This result is consistent with the notion that even moderate amounts of sleep loss might increase the likelihood of performance errors in real-world activities—especially those that require sustained attention and working memory.9 Indeed, the magnitude of the behavior impairment observed herein when subjects were performing tasks just a few hours after their normal time of sleep onset exceeded that observed following a legally intoxicating dose of alcohol in these same subjects,34 and in other groups of subjects performing the same tasks.33,35 This observation is consistent with other reports that have compared the effects of sleep deprivation and alcohol on performance.66,78,79 In con- trast, the magnitude of the behavior impairment associated with moder- ate sleep loss was not as great as that observed following a single stan- dard 50-milligram dose of the over-the-counter antihistamine diphenhy- dramine in the same subjects.34 Tonic Modulation of the Ongoing EEG by Sleepiness and Sustained Men- tal Effort Electrophysiologic measures obtained from the passive resting condi- tions varied in a predictable manner with increased sleepiness and men- tal fatigue as evidenced by the subjective ratings and overt task perfor- mance. When comparing brain signals recorded over posterior regions under eyes-closed resting conditions, past studies have demonstrated sleepiness-related increases in the delta and theta EEG bands and decreases in the alpha band.57-61 Slow eye movements have been shown to increase with sleepiness.49,55,56 These findings were replicated in the current study, providing convergent objective evidence of the effective- ness of the extended-wakefulness manipulation for inducing sleepiness in the subjects. These physiologic measures had a timecourse similar to that of the subjective ratings, showing a relatively continuous and grad- ual increase in sleepiness from the 11:00 PM test interval until the 5:00 AM test interval. Although both objective and subjective measures of alertness thus suggested that sleepiness increased in an approximately linear fashion over the intervals of the extended-wakefulness session, as noted above, the corresponding change in behavior was more sigmoidal in nature, showing a rapid asymptotic decline in accuracy and response speed between the 11:00 PM and 1:30 AM test intervals. One possibility that might account for this dissociation is that once subjects started to become sleepy, they rapidly became less motivated to make the sus- tained mental effort required by the continuous performance tasks, resulting in a corresponding rapid decline in performance. However, in order to mitigate such motivational lapses, this study employed a com- petitive, performance-related reimbursement scheme that rewarded con- sistently high levels of effort. Furthermore, other investigators who have studied behavior impairments in simple reaction-time sustained-atten- tion tasks during sleep deprivation have suggested that motivation issues are unlikely to be the major cause of performance decrements. In partic- ular, past results indicate that increases in errors of omission (failures to respond) in such contexts tend to be highly positively correlated with increases in errors of commission (inappropriate key presses), which suggests that participants in such studies often actively try to compen- sate for any sleepiness-related performance failures that might occur.80 Though error rates in the current study were fairly low, the pattern of errors observed was consistent with this view. Measures of ongoing EEG during task performance in the current study were also consistent with this view, suggesting that participants in the extended-wakefulness session continued to exert mental effort toward task performance even at the late-night test intervals. As noted above, past research with n-back working-memory tasks has demon- strated that the theta rhythm at frontal midline sites is enhanced with higher working-memory loads, whereas the alpha rhythm is attenuat- ed.17,26,27,30,32-34 Given that the perceptual and motor requirements of the different versions of the task are identical, such tonic working-memory -load-related differences in the EEG can be interpreted as indexes of the participants’ deliberate differential allocation of mental resources in response to increased demands in the task environment. The data from the current study replicated the previously described changes in spectral EEG characteristics in response to increased working-memory load, indicating that throughout the session participants dedicated a greater proportion of their available mental resources to task performance in the high-load condition than in the low-load condition. Furthermore, there did not appear to be any systematic changes across test intervals in EEG correlates of the subjects’ tonic mental effort. Thus, performance was impaired despite evidence that participants made a relatively constant effort throughout the night to commit neural resources to task perfor- mance. Such results are in sharp contrast to those observed in response to the sedating benzodiazepine alprazolam, the acute effects of which have been found to both dramatically degrade performance and elimi- nate the differences in EEG spectral characteristics otherwise observed to occur in response to variations in working-memory load in tasks sim- ilar to those employed herein38. Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 62
  8. 8. The Impact of Extended Wakefulness on the Phasic ERP Careful analyses of behavior measures during sleep deprivation have led some researchers to posit that sleep loss produces an unstable men- tal state that is characterized by momentary lapses in attention and alert- ness. Such lapses lead to performance errors and an increase in the vari- ability of response times that is disproportional to what might be expect- ed from overall behavior slowing.51,80-82 In the current study, evidence of such “state instability”80 was provided by the observation of decreased accuracy on even the low-load version of the working-memory task, as well as a significant increase in the coefficient of variability measure of reaction times over the course of the extended-wakefulness session. Such drowsiness-related increases in error rate and response variability have been reported to be associated with a decrease in the transient elec- trodermal orienting response, an autonomic measure of the degree to which a stimulus is successful at capturing an individual’s attention.82 Measures of the spectral power of the continuous EEG over a set of task trials—such as those described in the previous section—reflect the relative commitment of attention and working-memory resources to task performance over the course of several minutes. While adequate for characterizing tonic states and for differentiating the overall resource demands of tasks that vary in terms of the effortful attention required for their performance, such measures are likely to be poor indexes of subtle changes in more phasic aspects of attention, such as those required for perceptually analyzing a transient stimulus and for working-memory encoding and updating. In contrast, subsecond neurophysiologic mea- sures such as stimulus-locked ERPs can be highly sensitive to any changes in the manner in which attention is transiently captured and focused in response to a stimulus event. Past research has demonstrated that such measures are affected by sleep deprivation and mental fatigue,63,83,84 and the ERP results from the present study are consistent with the notion that even moderate sleep loss can compromise more pha- sic aspects of attention. For example, the N170 of the visual ERP has frequently been shown to be modulated by the strategic transient focusing of attention.85-87 That is, it is enhanced when stimuli are presented at a target location to which subjects are directing their attention and is attenuated when stimuli are presented away from the target location. Although the task used here dif- fers from the classic visual attention task in which attention is manipu- lated by cues prior to each stimulus, the N170 in the current study also appeared to be enhanced by the strategic phasic focusing of attention. In particular, in the tasks employed here, subjects were required to compare stimuli to a target location indicated on a previous trial; this requirement produced an enhancement of the N170 response to stimuli that then occurred at the anticipated location, a result that replicates similar find- ings from other studies employing these tasks.30,37 With extended wake- fulness, the N170 latency increased and N170 amplitude was signifi- cantly reduced. Thus, moderate sleep loss appeared to rapidly degrade a neural signal related to the selective focusing of attention onto particular stimulus attributes. Similarly, past research has also indicated that ERP components cor- responding to the P215 and the LPC are sensitive to the attention and working-memory requirements of a task.17,29,30 In the current study, these potentials also differed between stimuli that matched or failed to match the target stimulus in working memory, and, in the case of the LPC, this match sensitivity interacted with working-memory load. Extended wakefulness had a similar impact on the amplitude of these potentials as it did on the N170: both rapidly declined in amplitude when task perfor- mance was required after the participants’ normal time of sleep onset. These effects contrast with the lack of a significant attenuation for the parietal slow wave that followed the LPC. The lack of a slow wave effect suggests that extended wakefulness had its primary impact on compo- nents of the stimulus-locked neural response that are primed by selective attention to stimulus processing rather than ERP components related to subsequent operations. Like the change in behavior measures, the overall average timecourse of the ERP changes also showed a rapid decline to an asymptotic floor between the 11:00 PM and 1:30 AM test intervals. Furthermore, correla- tion analyses indicated that there was a tight coupling between the size of the ERP amplitude attenuation observed in individual participants and test intervals over the course of the extended-wakefulness session and the magnitude of the performance impairments observed in those same participants and test intervals. Other investigators have observed similar correlations between the magnitude of ERP attenuation and the behavior impairments that are associated with sleep deprivation.65 The amplitude of an averaged stimulus-locked ERP might decrease either as a result of the latency of the neural response to the stimulus becoming more variable across individual task trials or as a result of a decrease in the average absolute magnitude of the neural response pro- duced across trials. The observed negative correlation between the ERP amplitude reduction and the increase in variability in reaction-time mea- sures provides some indirect evidence that the former mechanism might be at least partly responsible for ERP-amplitude attenuation in response to sleep loss. However, this does not exclude the second possibility from also being a contributing factor, and other considerations are consistent with the notion that sleep loss is associated with a decrease in the mag- nitude of transient cortical responses. For example, in primate models, noradrenergic locus coerulus neurons have been shown to display con- tinuous moderately irregular activity during the alert waking state, punc- tuated by phasic enhancements of firing in the first 200 milliseconds fol- lowing a task-relevant or attention-capturing stimulus. While the magni- tude of locus coerulus background activity might be associated with the animal’s general state or level of wakefulness, the phasic enhancement is thought to provide a brief, stimulus-specific modulating input to cor- tical networks and to produce a transient attention-related increase in the amplitude of the cortically generated ERP that follows task-relevant stimuli.88 Simulation studies suggest that this phasic cortical input from the locus coerulus is also a critical factor in the animal’s ability to make fast and accurate responses.89 Drowsiness and decreased vigilance tend to be accompanied by a reduction in stimulus-related phasic enhance- ment of locus coerulus neuronal activity and a contemporaneous deteri- oration of task related behavior.90 Such reduction of phasic noradrener- gic input to the cortex might have played a central role in both the degraded behavioral performance and the reduced ERP amplitudes that were observed to occur in response to sleep loss in the current study. Detecting Changes in Functional Alertness with Multivariate Methods that Combine Behavior and EEG Measures As noted above, the data described here were collected as part of a larger study in which the same subjects also participated in four separate daytime test sessions in which they performed the same tasks before and after ingesting alcohol, caffeine, an antihistamine, or placebo. The aver- age of the data from the premedication period for each of those sessions comprised the 12:00 PM baseline reference period for the sleep-depriva- tion manipulation. In analyses of the drug-effects data, multivariate methods (in particular, stepwise linear discriminant analysis) were applied to combinations of behavior and EEG variables to detect and quantify the neurocognitive effects of the drugs and to track their phar- macodynamics. Inclusion of EEG variables in discriminant functions led to more sensitive detection of drug effects than that which could be accomplished when only behavior performance variables were used.34 To begin to determine whether such methods could be extended to the problem of detecting changes in cognitive function associated with sleep deprivation, stepwise linear discriminant analyses were performed in which functions were derived (either from behavior measures alone, electrophysiologic measures alone, or from combinations of behavior and electrophysiologic measures) to discriminate each testing interval during the extended-wakefulness session from the 12:00 PM alert base- line data. A leave-out-one resampling procedure91 was used for cross- validation. In this approach, functions are iteratively developed on N-1 of Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 63
  9. 9. the participants and then tested by application to the remaining participant. Figure 10 summarizes the results of the cross-validation analysis, illustrating the percentage of individuals (out of 16 total) detected as dif- ferent from the alert baseline test interval at each night-time test interval for each of the three analysis strategies. Consistent with the univariate analyses, none of the three analysis strategies significantly discriminat- ed data recorded at the 11:00 PM interval from the 12:00 PM baseline. By 12:30 AM, the behavior changes could be significantly detected (binomi- al P < .01), but analyses based on behavior alone did not yield a stable pattern of results when viewed over subsequent test intervals. Electro- physiologic changes from baseline could be detected in 100% of the par- ticipants by the 1:30 AM test session (P < .0001); however, as with behavior, analyses based on electrophysiologic measures alone provided a mixed picture when viewed across test intervals. In contrast, analyses that incorporated both behavior and electrophysiologic measures dis- played a monotonic increase in discriminability from baseline, with the largest increase occurring between the 12:30 AM and 1:30 AM test inter- vals. By the 5:00 AM test interval, 94% of the data samples in the com- bined analysis could be accurately discriminated from baseline (binomi- al P < 0.0005), versus only 69% when behavior measures alone were used (P < .10). Such results indicate that the fairly modest amounts of sleep loss incurred in the current study were sufficient to induce neu- rocognitive changes that could be detected reliably in individual sub- jects. They further indicate that, as with analogous assessments of the psychoactive effects of pharmaceutical interventions, improved sensitiv- ity in detecting such neurocognitive changes can be achieved when neu- rophysiologic measures are added to behavior measures in multivariate classification functions.34 CONCLUSIONS Together, the results are consistent with the notion that even moderate sleep loss can compromise the function of neural circuits that support working memory and the strategic focusing of attention, decreasing the efficiency of rapid information processing and making performance more variable and error prone. The results complement recent imaging studies showing sleep-deprivation-related changes in cortical activation during attention and working-memory tasks,92-95 and extend these find- ings by providing information about the likely relative timing and pha- sic nature of such changes. Given the disruption to executive functions sometimes reported in patients with obstructive sleep apnea,75,76 the high sensitivity of neurophysiologic signals of attention and working-memo- ry task performance to moderate degrees of sleep loss suggests that these types of measures might prove useful in assessing functional alertness in such patients. . ACKNOWLEDGMENTS We thank Jennifer Barber, Gail Chang, Arati Karnik, Caroline Pri- oleau, and Georgia Rush for assistance with data collection and analysis. The National Aeronautics and Space Administration, the National Insti- tute of Mental Health, the National Institute of Neurological Disorders and Stroke, and The Air Force Office of Scientific Research sponsored this research. REFERENCES 1. Owens J, Opipari L, Nobile C, Spirito A. Sleep and daytime behav- ior in children with obstructive sleep apnea and behavioral sleep disorders. Pediatrics 1998;102:1178-84 2. Randazzo AC, Muelbach MJ, Schweitzer PK, Walsh JK. Cognitive function following acute sleep restriction in children ages 10-14. Sleep 1998;21:861-8 3. Wolfson AR, Carskadon MA. Sleep schedules and daytime func- tioning in adolescents. Child Dev 1998;69:875-87 4. Porcu S, Bellatreccia A, Ferrara M, Casagrande M. Sleepiness, alertness, and performance during a laboratory simulation of an acute shift of the wake-sleep cycle. Ergonomics 1998;41:1191-202 5. Scott AJ. Chronobiological considerations in shiftworker sleep and performance and shiftwork scheduling. Hum Performance 1994;7:207-33 6. Tlaviki T, Kajaste S, Partinen M, Koskenovuo M, Salmi T, Kaprio J. Cognitive function in middle-aged snorers and controls: role of excessive daytime somnolence and sleep-related hypoxic events. Sleep 1988;11:454-62 7. Ulfberg J, Carter N, Talback M, Edling C. Excessive daytime sleepiness at work and subjective work performance in the gener- al population and among heavy snorers and patients with obstruc- tive sleep apnea. Chest 1996;110:659-63 8. Bonnet MH, Arand DL. We are chronically sleep deprived. Sleep 1995;18:908-11 9. Dinges DF. An overview of sleepiness and accidents. J Sleep Res 1995;4:4-14 10. Engle RW, Tuholski S, Kane M. Individual differences in working memory capacity and what they tell us about controlled attention, general fluid intelligence and functions of the prefrontal cortex. In: Miyake A, Shah P, eds. Models of working memory. Cambridge: Cambridge University Press; 1999:102-34 11. Oberauer K. Access to information in working memory: exploring the focus of attention. J Exp Psychol Learn Mem Cogn 2002;28:411-21 12. Baddeley A. Working memory. Science 1992;255:556-9 13. Goldman-Rakic P. Circuitry of primate prefrontal cortex and regu- lation of behavior by representational memory. In: Plum F, Mount- castle V, eds. Handbook of physiology, the nervous system - high- er functions of the brain. 1st ed. Vol. 5. Bethesda: American Phys- iological Society; 1987:373-417 14. Goldman-Rakic P. Topography of cognition: parallel distributed networks in primate association cortex. Ann Rev Neurosci 1988;11:137-56 15. Carpenter PA, Just MA, Shell P. What one intelligence test mea- sures: a theoretical account of the processing in the Raven Pro- gressive Matrices Test. Psychol Rev 1990;97:404-31 16. Kyllonen PC, Christal RE. Reasoning ability is little more than Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 64 Figure 10—These results illustrate a multivariate method for detecting changes in an individual’s func- tional alertness relative to a baseline state. Each line indicates the percentage of data samples in each interval of the extended-wakefulness session correctly discriminated from the 12:00 PM alert baseline session, where each data sample represents the results from one subject at that test interval. The three lines reflect the relative cross-validation discrimination accuracies of mathematic functions consisting of weighted linear combinations of either behavior or electroencephalographic measures, or a combination of the two. For N=16, a discrimination accuracy of 75% or better is necessary to achieve a binomial sig- nificance of P < .05 or better. In the combined analyses, the binomial significance of discriminating data samples as coming from the night-time test session from the daytime baseline ranged from a low of P > .10 at the 11:00 PM test interval to a high of P < .0005 at the 5:00 AM interval.
  10. 10. working memory capacity?! Intelligence 1990;14:389-433 17. Gevins A, Smith ME. Neurophysiological measures of working memory and individual differences in cognitive ability and cogni- tive style. Cereb Cortex 2000;10:829-39 18. Petrides M, Milner B. Deficits on subject-ordered tasks after frontal- and temporal-lobe lesions in man. Neuropsychologia 1982;20:249-62 19. Frisk V, Milner B. The relationship of working memory to the immediate recall of stories following unilateral temporal or frontal lobectomy. Neuropsychologia 1990;28:121-35 20. Owen AM, Evans AC, Petrides M. Evidence for a two-stage model of spatial working memory processing within the lateral frontal cortex: a positron emission tomography study. Cereb Cortex 1996;6:31-8 21. Smith EE, Jonides J. Working memory in humans: Neuropsycho- logical evidence. In: Gazzaniga M, ed. The cognitive neuro- sciences. Cambridge: MIT Press; 1995:1009-20 22. Smith EE, Jonides J, Koeppe RA, Awh E, Schumacher EH, Minoshima S. Spatial versus object working memory: PET inves- tigations. J Cogn Neurosci 1995;7:337-56 23. Haier RJ, Siegel BV, Nuechterlein KH, et al. Cortical glucose metabolic rate correlates of abstract reasoning and attention stud- ied with positron emission tomography. Intelligence 1988;12:199- 217. 24. Prabhakaran V, Smith JAL, Desmond JE, Glover GH, Gabrieli JDE. Neural substrates of fluid reasoning: an fMRI study of neo- cortical activation during performance of the Raven's Progressive Matrices test. Cognit Psychol 1997;33:43-63 25. Roland PE. Metabolic measurements of the working frontal cortex in man. Trends Neurosci 1984;7:430-5 26. Gevins A, Smith ME, McEvoy L, Yu D. High-resolution EEG mapping of cortical activation related to working memory: effects of task difficulty, type of processing, and practice. Cereb Cortex 1997;7:374-85 27. Gevins A, Smith ME, Leong H, et al. Monitoring working memo- ry load during computer-based tasks with EEG pattern recognition methods. Hum Factors 1998;40:79-91 28. Smith ME, McEvoy LK, Gevins A. Neurophysiological indices of strategy development and skill acquisition. Brain Res Cogn Brain Res 1999;7:389-404 29. Gevins AS, Smith ME, Le J, et al. High resolution evoked poten- tial imaging of the cortical dynamics of human working memory. Electroencephalogr Clin Neurophysiol 1996;98:327-48 30. McEvoy LK, Smith ME, Gevins A. Dynamic cortical networks of verbal and spatial working memory: effects of memory load and task practice. Cereb Cortex 1998;8:563-74 31. McEvoy LK, Pellouchoud E, Smith ME, Gevins A. Neurophysio- logical signals of working memory in normal aging. Cognit Brain Res 2001;11:363-76 32. McEvoy LK, Smith ME, Gevins A. Test-retest reliability of cogni- tive EEG. Clin Neurophysiol 2000;111:457-63 33. Gevins A, Smith ME. Detecting transient cognitive impairment with EEG pattern recognition methods. Aviat Space Environ Med 1999;70:1018-24 34. Gevins A, Smith ME, McEvoy LK. Tracking the cognitive phar- macodynamics of psychoactive substances with combinations of behavioral & neurophysiological measures. Neuropsychopharma- cology 2002;26:27-39 35. Ilan AB, Gevins A. Prolonged neurophysiological effects of cumu- lative wine drinking. Alcohol 2001;25:137-52 36. Nichols EA, Ilan AB, Page B, Smith ME, Gevins A. Effects of marijuana on neurophysiological correlates of working and inter- mediate-term memory. Paper presented at: Annual Meeting of the Society for Neuroscience; November 2001; San Diego, CA 37. Chung SS, McEvoy LK, Smith ME, Gevins A, Meador K, Laxer KD. Neurophysiological assessment of cognitive dysfunction associated with phenytoin. Clin Neurophysiol 2002;in press. 38. McEvoy LK, Chellaramani R, Smith ME, Gevins A. Effects of a benzodiazepine on behavioral and neurophysiological measures of working memory. Paper presented at: Annual Meeting of the Soci- ety for Neuroscience; November 2001; San Diego, CA 39. McElree B. Working memory and focal attention. J Exp Psychol Learn Mem Cogn. 2001;27:817-35 40. Ross P, Segalowitz SJ. An EEG coherence test of the frontal dor- sal versus ventral hypothesis of n-back working memory. Brain Cogn 2000;43:375-9 41. Wintink AJ, Segalowitz SJ, Cudmore LJ. Task complexity and habituation effects on frontal P300 topography. Brain Cogn 2001;46:307-11 42. Koivisto M, Krause CM, Revonsuo A, Laine M, Hamalainen H. The effects of electromagnetic fields emitted by GSM phones on working memory. Neuroreport. 2000;11:1641-3 43. Oliveri M, Turriziani P, Carlesimo GA, et al. Parieto-frontal inter- actions in visual-object and visual-spatial working memory: evi- dence from transcranial magnetic stimulation. Cereb Cortex 2001;11:606-18 44. Braver TS, Cohen JD, Nystrom LE, Jonides J, Smith EE, Noll DC. A parametric study of prefrontal cortex involvement in human working memory. Neuroimage 1997;5:49-62 45. Cohen JD, Forman SD, Braver TS, Casey BJ, Servan-Schreiber D, Noll DC. Activation of prefrontal cortex in a non-spatial working memory task with functional MRI. Hum Brain Mapping 1994;1:293-304 46. Jansma JM, Ramsey NF, Coppola R, Kahn RS. Specific versus nonspecific brain activity in a parametric n-back task. Neuroimage 2000;12:688-97 47. Jonides J, Smith EE, Koeppe RA, Awh E, Minoshima S, Mintun M. Spatial working memory in humans as revealed by PET. Nature 1993;363:623-5 48. McCarthy G, Blamire AM, Puce A, et al. Functional magnetic res- onance imaging of human prefrontal cortex activation during a spatial working memory task. Proc Natl Acad Sci 1994;91:8690-4 49. Åkerstedt T, Gillberg M. Subjective and objective sleepiness in the active individual. Int J Neurosci 1990;52:29-37 50. Dinges DF, Graeber RC, Connell LJ, Rosekind MR, Powell JW. Fatigue-related reaction time performance in long-haul flight crews. Sleep Res 1990;19:117 51. Dinges DF, Kribbs NB. Performing while sleepy: effects of exper- imentally induced sleepiness. In: Monk TH, ed. Sleep, sleepiness, and performance. New York: Wiley; 1991:97-128 52. Dinges DF, Powell JW. Microcomputer analyses of performance on a portable, simple visual RT task during sustained operations. Behav Res Methods Instruments Comput 1985;17:652-5 53. Comstock JR, Arnegard RJ. (NASA Technical Memorandum). The Multi-Attribute Task Battery for Human Operator Workload and Strategic Behavior Research. 1992. Report #104174 54. Du W, Leong HM, Gevins AS. Ocular artifact minimization by adaptive filtering. Proc Seventh IEEE SP Workshop on Statistical Signal and Array Processing. Quebec City, Canada; 1994:433-6 55. Stern JA. Blink rate: a possible measure of fatigue. Hum Factors 1994;36:285-97 56. Torsvall L, Åkerstedt T. Extreme sleepiness: quantification of EOG and spectral EEG parameters. Int J Neurosci 1988;38:435-41 57. Davis H, Davis PA, Loomis AL, Harvey EN, Hobart G. Human brain potentials during the onset of sleep. J Neurophysiol 1937;1:24-37 58. Gevins AS, Zeitlin GM, Ancoli S, Yeager CL. Computer rejection of EEG artifact. II: contamination by drowsiness. Electroen- cephalogr Clin Neurophysiol 1977;42:31-42 59. Makeig S, Jung TP. Changes in alertness are a principal component of variance in the EEG spectrum. NeuroReport 1995;7:213-6 60. Matousek M, Petersen IA. A method for assessing alertness fluctu- Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 65
  11. 11. ations in vigilance and the EEG spectrum. Electroencephalogr Clin Neurophysiol 1983;55:108-13 61. Oken BS, Salinsky M. Alertness and attention: basic science and electrophysiologic correlates. J Clin Neurophysiol 1992;9:480-94 62. Swets JA. Signal detection and recognition by human observers. New York: Wiley; 1964 63. Gevins AS, Bressler SL, Cutillo BA, et al. Effects of prolonged mental work on functional brain topography. Electroencephalogr Clin Neurophysiol 1990;76:339-50 64. Humphrey DG, Kramer AF, Stanny RR. Influence of extended wakefulness on automatic and nonautomatic processing. Hum Fac- tors 1994;36:652-69 65. Corsi-Cabrera M, Arce C, Del Rio-Portilla IY, Perez-Garci E, Gue- vara MA. Amplitude reduction in visual event-related potentials as a function of sleep deprivation. Sleep 1999;22:181-9 66. Krull KR, Smith LT, Kalbfleisch LD, Parsons OA. The influence of alcohol and sleep deprivation on stimulus evaluation. Alcohol 1992;9:445-50 67. Morris AM, So Y, Lee KA, Lash AA, Becker CE. The P300 event- related potential: the effects of sleep deprivation. J Occup Environ Med 1992;34:1143-52 68. Lorist MM, Snel J, Kok A. Influence of caffeine on information stages in well rested and fatigued subjects. Psychopharmacology (Berl) 1994;113:411-21 69. Lorist MM, Snel J, Kok A, Mulder G. Influence of caffeine on selective attention in well rested and fatigued subjects. Psy- chophysiology 1994;31:525-34 70. Binks PG, Waters WF, Hurry M. Short-term sleep deprivation does not selectively impair higher cortical functioning Sleep 1999;22:328-34 71. Linde L, Bergstrom M. The effect of one night without sleep on problem-solving and immediate recall. Psychol Res 1992;54:127-36 72. Harrison Y, Horne JA. Sleep loss impairs short and novel language tasks having a prefrontal focus. J Sleep Res 1998;7:95-100 73. Harrison Y, Horne JA, Rothwell A. Prefrontal neuropsychological effects of sleep deprivation in young adults—a model for healthy aging? Sleep 2000;23:1067-73 74. Harrison Y, Horne JA. One night of sleep loss impairs innovative thinking and flexible decision making. Organ Behav Hum Decis Process 1999;78:128-45 75. Beebe DW, Gozal D. Obstructive sleep apnea and the prefrontal cortex: towards a comprehensive model linking nocturnal upper airway obstruction to daytime cognitive and behavioral deficits. J Sleep Res 2002;11:1-16 76. Decary A, Rouleau I, Montplaisir J. Cognitive deficits associated with sleep apnea syndrome: a proposed neuropsychological test battery. Sleep 2000;23:369-81 77. Engleman HM, Kingshott RN, Martin SE, Douglas NJ. Cognitive function in the sleep apnea/hypopnea syndrome (SAHS). Sleep 2000;23:S102-8 78. Arendt JT, Wilde GJ, Munt PW, MacLean AW. How do prolonged wakefulness and alcohol compare in the decrements they produce on a simulated driving task? Accid Anal Prev 2001;33:337-44 79. Williamson AM, Feyer AM. Moderate sleep deprivation produces impairments in cognitive and motor performance equivalent to legally prescribed levels of alcohol intoxication. Occup Environ Med 2000;57:649-55 80. Doran SM, Van Dongen HP, Dinges DF. Sustained attention per- formance during sleep deprivation: evidence of state instability. Arch Ital Biol 2001;139:253-67 81. Koslowsky M, Babkoff H. Meta-analysis of the relationship between total sleep deprivation and performance. Chronobiol Int 1992;9:132-6 82. McCarthy ME, Waters WF. Decreased attentional responsivity during sleep deprivation: orienting response latency, amplitude, and habituation. Sleep 1997;20:115-23 83. Lorist MM, Klein M, Niewenhuis S, De Jong R, Mulder G, Meij- man TF. Mental fatigue and task control: planning and preparation. Psychophysiology 2000;37:614-25 84. Raz A, Deouell LY, Bentin S. Is pre-attentive processing compro- mised by prolonged wakefulness? Effects of total sleep deprivation on the mismatch negativity. Psychophysiology 2001;38:787-95 85. Eason R, Harter M, White C. Effects of attention and arousal on visually evoked cortical potentials and reaction time in man. Phys- iol Behav 1969;4:283-9 86. Harter MR, Aine C, Schroeder C. Hemispheric differences in the neural processing of stimulus location and type: effects of selective attention on visual evoked potentials. Neuropsychologia 1982;20:421-38 87. Mangun GR, Buck LA. Sustained visual-spatial attention produces costs and benefits in response time and evoked neural activity. Neuropsychologia 1998;36:189-200 88. Foote SL, Berridge CW, Adams LM, Pineda JA. Electrophysio- logical evidence for the involvement of the locus coerulus in alert- ing, orienting, and attending. Prog Brain Res 1991;88:521-32 89. Usher M, Cohen JD, Servan-Schreiber D, Rajkowski J, Aston- Jones G. The role of locus coerulus in the regulation of cognitive performance. Science 1999;283:549-54 90. Rajkowski J, Kubiak P, Aston-Jones G. Locus coerulus activity in monkey: phasic and tonic changes are associated with altered vig- ilance. Brain Res Bull 1994;35:607-16 91. Efron B. The jackknife, the bootstrap and other resampling plans. Philadelphia: Society for Industrial and Applied Mathematics; 1982 92. Drummond SP, Brown GG, Stricker JL, Buxton RB, Wong EC, Gillin JC. Sleep deprivation-induced reduction in cortical func- tional response to serial subtraction. Neuroreport 1999;16:3745-8 93. Drummond SP, Brown GG, Gillin JC, Stricker JL, Wong EC, Bux- ton RB. Altered brain response to verbal learning following sleep deprivation. Nature 2000;403:605-6 94. Drummond SP, Gillin JC, Brown GG. Increased cerebral response during a divided attention task following sleep deprivation. J Sleep Res 2001;10:85-92 95. Thomas M, Sing H, Belenky G, et al. Neural basis of alertness and cognitive performance impairments during sleepiness. I. Effects of 24h of sleep deprivation on waking human regional brain activity. J Sleep Res 2000;9:335-52 Sleep Loss and Working Memory—Smith et alSLEEP, Vol. 25, No. 7, 2002 66