Content uploaded by Ken Nakayama
Author content
All content in this area was uploaded by Ken Nakayama
Content may be subject to copyright.
As life in our culture becomes more demanding, average nightly
sleep is decreasing in all segments of the population (Stein, M.,
National Sleep Foundation Poll, 2001, www.sleepfoundation.org/
PressArchives/lessfun_lesssleep.html). ‘Power naps’—brief periods
of daytime sleep lasting an hour or less—improve alertness, pro-
ductivity and mood
1,2
, especially under sleep-deprived conditions
3,4
,
during nightshift work
5
and during prolonged periods of driving
6
.
Although naps have been shown to enhance psychomotor speed as
well as short-term memory acquisition
7,8
, the effect of daytime naps
on previously learned information is not known. The finding that
power naps are common among people reporting daily informa-
tion overload indicates that napping supports a previously unknown
mechanism of off-line information processing, perhaps related to
that which normally occurs during nocturnal sleep
9–16
.
We investigated the phenomenon of information overload at
the perceptual level. Typically in visual perception tasks, fast learn-
ing happens in the first minutes to hours of training
17,18
. Previous
studies using a visual texture discrimination task (TDT)
19
show
that a slower phase of perceptual learning also exists, which
depends on nocturnal sleep after training
9,14,19–22
. The slow phase
of improvement becomes evident only after at least six hours of
nocturnal sleep
14
, and sleep deprivation the night after training
eliminates the normal post-sleep improvement, even when mea-
sured after two full nights of recovery sleep
15
. The improvement
seen in subjects who sleep for eight hours during the night after
training correlates with the proportion of deep, slow wave sleep
(SWS) in the first quarter of the night and with the proportion of
rapid eye movement sleep (REM) in the last quarter
14
. These results
indicate that a full night of sleep is important for maintenance and
consolidation of experience-dependent learning, and that without
at least six hours of sleep, this potential consolidation is lost. These
studies do not, however, address the question of how power naps
of an hour or less could aid in such information processing.
Here we show that perceptual performance declined on the
TDT with repeated, within-day training. In the context of this
The restorative effect of naps on
perceptual deterioration
Sara C. Mednick
1
, Ken Nakayama
1
, Jose L. Cantero
2
, Mercedes Atienza
2
, Alicia A. Levin
2
,
Neha Pathak
2
and Robert Stickgold
2
1
Department of Psychology, Harvard University, 33 Kirkland Street, Cambridge, Massachusetts 02138, USA
2
Laboratory of Neurophysiology and Department of Psychiatry, Massachusetts Mental Health Center, Harvard Medical School,
74 Fenwood Road, Boston, Massachusetts 02115, USA
Correspondence should be addressed to S.C.M. (smednick@wjh.harvard.edu)
Published online: 28 May 2002, doi:10.1038/nn864
Human performance on visual texture discrimination tasks improves slowly (over days) in the
absence of additional training. This ‘slow learning’ requires nocturnal sleep after training and is lim-
ited to the region of visual space in which training occurred. Here, we tested human subjects four
times in one day and found that with repeated, within-day testing, perceptual thresholds actually
increased progressively across the four test sessions. This performance deterioration was prevented
either by shifting the target stimuli to an untrained region of visual space or by having the subjects
take a mid-day nap between the second and third sessions.
deterioration, we found that (i) a daytime nap, but not an equiv-
alent period of rest without visual input, reversed the deteriora-
tion, (ii) the deterioration was retinotopically specific and (iii)
neither an increase in subject motivation nor a decrease in task
difficulty improved performance.
R
ESULTS
Can too much practice be detrimental?
To investigate whether repeated within-day testing on a percep-
tual learning task can impair performance, subjects were tested on
the TDT four times in a single day (at 9 a.m., 12 p.m., 4 p.m. and
7 p.m.). Each session lasted approximately 60 minutes. For each
session, the speed of perceptual processing was calculated as the
threshold target-to-mask interstimulus interval (ISI) needed to
achieve 80% accuracy. Thirty subjects were randomly assigned
to one of three groups: control, long nap or short nap. Control
subjects (n = 10) showed a 52% slowing in perceptual process-
ing across the four test sessions (Fig. 1, filled circles; P = 0.0003,
repeated measures analysis of variance (ANOVA) and post hoc
tests). Thus, with each successive session, subjects needed increas-
ingly longer exposures to the stimuli to reliably identify targets.
Performance deteriorated despite all testing being done within
12 hours of morning awakening, a time when one would not nor-
mally expect to see cognitive impairment, and without prior sleep
deprivation. Subjects averaged 6.92 ± 0.77 (s.d.) hours of sleep
on the night before testing.
Can daytime sleep reverse perceptual deterioration?
As nocturnal sleep is known to enhance alertness and to consol-
idate TDT learning
9,14,19,22
, we asked whether a daytime nap
might stop or even reverse the process of deterioration seen with
repeated within-day testing. The remaining 20 subjects were ran-
domly assigned to a long (60-minute) or short (30-minute) nap
condition. All subjects, including no-nap controls, performed
the task four times during the day; experimental subjects took a
articles
nature neuroscience • volume 5 no 7 • july 2002 677
© 2002 Nature Publishing Group http://neurosci.nature.com
678
nature neuroscience • volume 5 no 7 • july 2002
nap at 2 p.m.—midway between the second and third test ses-
sions. As predicted, napping significantly affected subsequent
performance (P = 0.001, group × session interaction, mixed-
model ANOVA): short naps prevented the normal deterioration
that was seen in test sessions 3 and 4 (Fig. 1, open circles) and
long naps reversed the deterioration seen in the second test ses-
sion (Fig. 1, open triangles). Thus, whereas controls showed a
14.1-ms increase in ISI threshold between the second and third
sessions, the short nap group showed no change (<1-ms increase)
and the long nap group showed a 20.9-ms decrease (P =0.03,
paired t-test). The short nap group showed no significant change
in ISI thresholds across the last three sessions (P = 0.94), but did
show significantly lower thresholds in the fourth session than did
controls (P = 0.01). The long nap group showed significantly bet-
ter performance than controls on both the third and fourth ses-
sions (P = 0.03). We compared the distributions of sleep stages
in long naps versus short naps (Ta ble 1) and found no signifi-
cant differences in the amount of time spent in stage 1 or stage 2
sleep. There was, however, a difference for SWS and REM sleep:
the long naps contained 4.1–4.4 times more SWS and REM sleep
than the short naps did. This matches the SWS and REM depen-
dency reported for overnight improvement on the task
14
, sug-
gesting that processes occurring during SWS and/or REM
underlie sleep-induced perceptual recovery.
We compared naps taken on the day of TDT testing with base-
line naps taken on a different day (see Methods) and found that
the long nap group spent significantly more time in SWS on the
test day (27.4 versus 20.9 minutes, P < 0.05, one-tailed paired t-
test) at the expense of time in stage 2 sleep (17.9 versus
25.0 minutes, P = 0.06) (Fig. 2). This large increase in SWS in test-
day naps (31% over baseline) suggests that SWS is crucial for post-
nap performance, perhaps through stabilizing and consolidating
plastic neuronal changes from earlier in the day. Such a function
has previously been proposed for nocturnal SWS (refs. 23–25).
The role of REM, however, is less clear.
The increase in time spent in REM during
test-day naps compared with baseline naps
(60% increase) was larger than that seen for
SWS, but was not statistically significant
(Fig. 2). Similar post-training REM sleep
increases have been reported in animal stud-
ies of sleep-dependent learning
26
. In the
short nap condition, there was no signifi-
cant change in SWS (9.2 minutes on test
day versus 6.7 minutes baseline) or REM
(0.8 minutes on test day versus 2.0 minutes
baseline) duration. Nap duration was controlled in both groups
and neither group showed a significant difference in nap duration
between baseline and test-day naps (P > 0.27).
Sleep versus rest
To t est whether the benefits of naps resulted from the absence of
visual input rather than from sleep itself, subjects (n = 9) repeat-
ed the long nap protocol; but instead of sleeping, they rested qui-
etly while blindfolded. Wake–sleep state was continuously
monitored physiologically to ensure maintained wakefulness.
The hour of rest without visual input did not have a restorative
effect: subjects still showed performance decrements at 4 p.m.
and 7 p.m. ISI thresholds increased by an average of 29.0 ms
between the second and fourth session (P < 0.05), a performance
decrement nearly identical (P = 0.31) to the 32-ms increase seen
in controls (Fig. 1).
Motivation and task difficulty
To t est whether a decrease in motivation contributed to the per-
formance deterioration, we informed subjects (n = 10) after their
second session that their performance had worsened, and that they
would receive a cash bonus if they subsequently returned to their
baseline performance. Even with this incentive, none of the sub-
jects regained baseline performance during the third or fourth ses-
sions, and mean ISI thresholds were 32.2 ms longer on the fourth
session compared to the first session (P = 0.001).
To t est whether the performance decrement resulted specif-
ically from exposure to more difficult trials, we tested ten sub-
jects throughout the day in four sessions, with all blocks in the
first three sessions run at the longest ISI (400 ms). Subjects were
then tested in the fourth session with the standard 25 blocks of
decreasing ISIs, and their performance in the fourth session did
not differ from that of controls (P = 0.50). Thus, despite hav-
ing an easier task in the first three sessions, subsequent perfor-
mance still declined.
What aspect of processing is impaired?
Several mechanisms might underlie this deterioration. A gener-
alized fatigue effect, mediated by a decrease in alertness or atten-
tional resources, is one possibility that is consistent with our data.
Alternatively, we propose that specific neural networks in pri-
mary visual cortex gradually become saturated with information
through repeated testing, preventing further perceptual process-
ing. This would cause a training-specific deterioration in per-
articles
Fig. 1. Beneficial effect of napping. Performance of control and both
short and long nap groups across same-day sessions. ISI thresholds from
the first session defined baseline performance, and were subtracted
from thresholds for subsequent sessions to determine relative change in
performance.
Ta b le 1. Characteristics of short and long naps.
S1 (min) S2 (min) SWS (min) REM (min) ∆Threshold (ms)
Short naps 6.2 ± 1.1 14.3 ± 1.5 6.6 ± 2.1 2.0 ± 1.4 4.5 ± 10.7
Long naps 5.5 ± 0.7 17.9 ± 3.0 27.4 ± 3.8 8.9 ± 3.4 20.8 ± 7.6
P value 0.85 0.22 0.0001 0.07
The mean times spent in each sleep stage during short (30-min) and long (60-min) naps are presented as
min ± s.e.m. S1, stage 1; S2, stage 2; SWS, slow wave sleep (stages 3 and 4); REM, rapid eye movement
sleep. P values from unpaired t-test comparing times in each stage for long versus short naps.
∆Threshold, the difference between ISI thresholds on the second (pre-nap) and third (post-nap) tests,
expressed as ms ± s.e.m.
© 2002 Nature Publishing Group http://neurosci.nature.com
ceptual processing. Whereas the generalized fatigue hypothesis
predicts that decrements in performance would be widespread
and largely task-independent, our hypothesis of a training-specific
deterioration predicts that the performance decrements would
be restricted to behaviors mediated by the specific neural net-
works previously involved in processing the target stimuli. We
reasoned that because TDT learning does not transfer to
untrained portions of the visual field
20
, there should be no
training-specific deterioration if stimuli are presented to an
untrained region of the visual cortex.
To test this hypothesis, we trained 24 subjects as before, but
for half of them we switched the target stimuli to the con-
tralateral visual field for the fourth and final same-day test ses-
sion. Performance of the switch group did not differ
significantly from that of the control group across the first three
sessions, but, unlike the control group, the switch group
showed significant recovery in the fourth session (Fig. 3, P =
0.002, ANOVA group × session interaction and post hoc test).
Performance during the switch condition was not significant-
ly worse than it was during the first session, indicating that the
behavioral deterioration observed in the trained visual quad-
rant did not transfer to the untrained contralateral quadrant.
These results strongly support the training-specific deteriora-
tion hypothesis, and are contrary to the predictions of the gen-
eralized fatigue hypothesis.
Further evidence against the generalized fatigue hypothesis
comes from the dissociation between improved performance and
subjective levels of sleepiness. If the steady decrease in perfor-
mance throughout the day in the control group resulted from a
general fatigue effect, then one should see a parallel increase in
reported sleepiness. But no such increase was seen, and mean
levels of subjective sleepiness on the first and last tests were iden-
tical (P = 0.45, repeated measures ANOVA). Similarly, the switch
group showed no significant change in sleepiness across sessions
(P = 0.49). In contrast, sleepiness decreased from the first to the
last session in the nap groups (P < 0.03, ANOVA and post hoc
tests). Thus, the switch group showed the same improvement in
performance as the nap groups did, but without a similar decrease
in subjective sleepiness; compared to controls, the switch group
showed the same maintenance in degree of sleepiness, but no
deterioration in performance.
Data from the switch group also eliminated another possible
explanation, that of a strictly circadian effect. Although the con-
trol data could have been explained as a circadian rather than
repetition effect, the fact that shifting the stimulus to the con-
tralateral visual field for the last session reversed this decrease
eliminates this possibility. Thus, when subjects were tested at
7 p.m. with stimuli in an untrained region of visual space, they
performed as well as they had at 9 a.m. the same morning.
D
ISCUSSION
Tw o learning components occur with TDT testing: a fast, within-
session component and a slow, sleep-dependent component
16
.
Our study identified a third consequence of TDT training: with
repeated, same-day training, people require progressively longer
ISIs for texture discrimination. Such perceptual deterioration
has not previously been reported for repeated same-day testing
on other visual tasks. Task procedure may contribute to this dif-
ference, such as whether the task measures vernier acuity
27
, res-
olution acuity
27
or texture discrimination
19
, whether stimuli
are presented foveally
17,18
, parafoveally
27
or at more peripher-
al eccentricities
19
, and whether stimulus presentations are long
(100–150 ms)
17,18,27
or short (17 ms)
19
. Furthermore, a number
of perceptual learning protocols train subjects over several
days
17,18,27–29
rather than within-day, as in the current study,
making it unclear whether fast or slow learning is occurring.
With these caveats in mind, the present study shows that some
forms of neural plasticity that require sleep for subsequent con-
solidation and improvement of perception may actually hinder
performance before sleep.
Our three main findings—that there was a normal decline in
TDT performance across the day with repeated exposure to the
task, that this decline was specific to previously trained regions
of visual space and that performance was restored by daytime
napping—have important implications. First, as circadian influ-
ences have been ruled out, the performance decline must result
from specific neuronal changes induced by the initial testing peri-
od. Second, as brain regions involved in higher levels of visual
articles
nature neuroscience • volume 5 no 7 • july 2002 679
Fig. 2. Comparison of test-day and baseline long naps. The number of
minutes spent in stage 1 (S1), stage 2 (S2), rapid eye movement (REM)
and slow wave sleep (SWS; stage 3 and 4) during baseline (gray bars) and
test-day (solid bars) naps. * P < 0.05; (*) P = 0.06.
Fig. 3. Beneficial effect of shifting stimulus location. Solid bars, deterio-
ration in performance of control subjects during the third (T3) and
fourth (T4) sessions of the day. Gray bars, deterioration in performance
of experimental ‘switch group’ subjects during T3 followed by recovery
during T4 when stimulus location was shifted.
© 2002 Nature Publishing Group http://neurosci.nature.com
680
nature neuroscience • volume 5 no 7 • july 2002
processing lack retinotopic specificity, the critically affected neu-
rons are most likely located in early visual processing areas. Final-
ly, these initially affected neural networks must be further altered
during napping to reverse the performance decrement.
We propose that the performance decrement seen here was a
direct consequence of a mechanism for preserving information
that has been processed but has not yet been consolidated into
memory by sleep. As this hypothetical limited-capacity mecha-
nism becomes saturated with task-specific information, the local
neural network’s ability to process on-line information during task
performance worsens, resulting in the performance decrement.
Several findings support this relationship between information
processing and the performance decrement. First, the retinotopic
specificity of the performance decrement is consistent with previ-
ous studies showing that TDT learning is similarly retinotopic
19
and dependent on early stages of visual processing
19,30
. Second,
post-training sleep is known to be critical for stabilization and con-
solidation of TDT learning
9,14,15
and we now have shown that a
60-minute nap reverses the performance decrement.
Although we cannot exclude a function for REM in this
process, it seems most likely that SWS has the central role.
Although both SWS and REM have previously been implicated
in the nocturnal, sleep-dependent consolidation of this task
9,14
,
the REM-dependent period appears four hours after the SWS-
dependent period has ended
14
—well beyond the timeframe of
these naps. We posit that during SWS, mechanisms of cortical
plasticity lead to secondary changes in the TDT-trained neural
networks. Thus SWS serves as the initial processing stage of expe-
rience-dependent, long-term learning and as the critical stage for
restoring perceptual performance. Roles for SWS in memory con-
solidation have been proposed by others
9,20,23–25
.
This example of a training-induced deterioration in perfor-
mance has several additional implications. First, it indicates that
the cognitive benefits of sleep can be studied over a very short
time period and do not require sleep deprivation or overnight
sessions of sleep. This provides a more favorable set of conditions
to study the involvement of sleep in information processing and
performance. Second, it suggests that the psychological sensa-
tion of ‘burnout,’ described anecdotally as increased irritation
and frustration along with decreased effectiveness after prolonged
cognitive effort, may not reflect a general mental fatigue, but
rather the specific need of an overused local neural network to
enjoy the restorative benefits of sleep.
M
ETHODS
A total of 129 undergraduates gave informed consent to participate in the
study, which was approved by the Harvard University Department of Psy-
chology internal review board. All su
bjects had normal or corrected-to-
normal vision and no history of neurological, mental or physical illness.
Each subject was tested on the TDT four times in one day: at 9 a.m.,
12 p.m., 4 p.m. and 7 p.m. Tests normally lasted 60–75 min and includ-
ed 1,250 trials. Participants were asked to discriminate the shape of a
target in one of the lower quadrants of the display at 2.5–5.0° eccen-
tricity from the center. Either a ‘T’ or an ‘L’ appeared in the center,
where fixation was maintained. (Except where noted, target arrays
always appeared in the lower left quadrant.) The target consisted of a
horizontally or vertically oriented array of three diagonal bars against
a background of horizontal bars. For each trial, the following sequence
was shown: target screen for 16 ms, blank screen for a variable period
of time, and then a mask for 16 ms. After each trial, subjects report-
ed both the letter (T or L) at the central fixation point and the orien-
tation of the diagonal bar array (horizontal or vertical). For each
session, the speed of perceptual processing was calculated as the
threshold target-to-mask ISI needed to achieve 80% accuracy. The ISI
threshold provides a measure of the minimal effective stimulation
needed for perception of the target. This task was carried out as pre-
viously described
14,19
. To examine the amount of change from base-
line, difference scores were calculated by subtracting the threshold for
the first session from those of the second, third and fourth. Subjects
in all groups kept sleep logs for the week before testing, and no sig-
nificant between-group differences were found in sleep patterns.
Thirty subjects were randomly assigned to one of the nap conditions
or to the previously described control condition, permitting statistical
comparisons between all three groups. Naps began at 2 p.m. and were
recorded polysomnographically, with standard electroencephalographic
(EEG), electro-oculographic (EOG) and electromyographic (EMG) mea-
sures. Subjects were allowed to sleep until they had completed either a
full half-hour or a full hour of polysomnographically identified
31
sleep
and then were woken by the experimenter. Sleep stages were subsequently
rescored off-line. On re-scoring, one subject in the short nap group was
found to have slept for only 9.5 min, and was excluded from all analy-
ses. Subjects were recorded during naps on two separate days: the TDT
test day and a control day (‘baseline nap’) either one week before or one
week after the test day, with the order balanced across subjects. Subjects
in the short nap group averaged 29.1 ±2.9 (mean ± s.d.) min of sleep on
the test day; subjects in the long nap group averaged 59.6 ± 5.5 min.
For the quadrant switch study (24 switches; 22 controls), peripheral
targets were presented in the lower left or lower right quadrant (balanced
across subjects) for the first three test sessions. Then, for the fourth session,
the targets were switched to the opposite lower quadrant for the switch
group. For the quiet rest condition, nine subjects followed the long nap
protocol but were instructed not to sleep during the nap hour. Subjects
were blindfolded to prevent visual stimulation and listened to an audio
tape of short stories. Each subject’s wake state was monitored with the
Nightcap
32
(HealthDyne Technologies, Marietta, Georgia, USA) and sub-
jects were alerted at the first indication of impending sleep. At the start
of the third session, subjects (n = 10) in the motivation protocol were
offered a bonus of $25 if they could regain their initial performance level
during the next two sessions. Subjects (n = 12) in the fixed ISI protocol
performed the standard protocol but with the ISI set to 400 ms for all 25
blocks during test sessions 1–3. For the fourth session, subjects followed
the standard protocol with decreasing ISIs across the 25 blocks. All subjects
rated their sleepiness using the Stanford sleepiness scale
33
, which rates
subjective sleepiness on a seven-point scale.
Acknowledgments
This research was supported by grants from the National Institutes of Health
(MH 48,832 and NS 26,985) and AFOSR (83-0320) and by fellowships to J.L.C.
and M.A. from the Spanish Ministry of Education and the NATO Scientific
Program, respectively.
Competing interests statement
The authors declare that they have no competing financial interests.
R
ECEIVED
10 DECEMBER
2001; ACCEPTED
2 M
AY
2002
1. Takahashi, M. & Arito, H. Maintenance of alertness and performance by a
brief nap after lunch under prior sleep deficit. Sleep 23, 813–819 (2000).
2. Dinges, D. F. & Broughton, R. J. (eds.) Sleep and Alertness: Chronobiological,
Behavioral and Medical Aspects of Napping (Raven, New York, 1989).
3. Bonnet, M. H. & Arand, D. L. Impact of naps and caffeine on extended
nocturnal performance. Physiol. Behav. 56, 103–109 (1994).
4. Bonnet, M. H. The effect of varying prophylactic naps on performance,
alertness and mood throughout a 52-hour continuous operation. Sleep 14,
307–315 (1991).
5. Rosa, R. R. Napping at home and alertness on the job in rotating shift
workers. Sleep 16, 727–735 (1993).
6. Horne, J. A. & Reyner, L. A. Counteracting driver sleepiness: effects of
napping, caffeine, and placebo. Psychophysiology 33, 306–309 (1996).
7. Harma, M., Knauth, P. & Ilmarinen, J. Daytime napping and its effects on
alertness and short-term memory performance in shiftworkers. Int. Arch.
Occupational Environ. Health 61, 341–345 (1989).
8. Taub, J. M. Effects of habitual variations in napping on psychomotor
performance, memory and subjective states. Int. J. Neurosci. 9, 97–112
(1979).
9. Gais, S., Plihal, W., Wagner, U. & Born, J. Early sleep triggers memory for
articles
© 2002 Nature Publishing Group http://neurosci.nature.com
early visual discrimination skills. Nat. Neurosci. 3, 1335–1339 (2000).
10. Plihal, W. & Born, J. Effects of early and late nocturnal sleep on priming and
spatial memory. Psychophysiology 36, 571–582 (1999).
11. Smith, C. Sleep states and memory processes. Behav. Brain Res. 69, 137–145
(1995).
12. Smith, C. T., Conway, J. M. & Rose, G. M. Brief paradoxical sleep deprivation
impairs reference, but not working, memory in the radial arm maze task.
Neurobiol. Learn. Mem. 69, 211–217 (1998).
13. Stickgold, R. Sleep: off-line memory reprocessing. Trends Cogn. Sci. 2,
484–492 (1998).
14. Stickgold, R., Whidbee, D., Schirmer, B., Patel, V. & Hobson, J. A. Visual
discrimination task improvement: a multi-step process occurring during
sleep. J. Cogn. Neurosci. 12, 246–254 (2000).
15. Stickgold, R., James, L. & Hobson, J. A. Visual discrimination learning
requires post-training sleep. Nat. Neurosci. 2, 1237–1238 (2000).
16. Karni, A. & Sagi, D. The time course of learning a visual skill. Nature 365,
250–252 (1993).
17. Fahle, M. Human pattern recognition: parallel processing and perceptual
learning. Perception 23, 411–427 (1994).
18. Fahle, M. & Edelman, S. Long-term learning in vernier acuity: effects of
stimulus orientation, range and of feedback. Vision Res. 33, 397–412 (1993).
19. Karni, A. & Sagi, D. Where practice makes perfect in texture discrimination:
evidence for primary visual cortex plasticity. Proc. Natl. Acad. Sci. USA 88,
4966–4970 (1991).
20. Plihal, W. & Born, J. Effects of early and late nocturnal sleep on declarative
and procedural memory. J. Cogn. Neurosci. 9, 534–547 (1997).
21. Stickgold, R., Schirmer, B., Patel, V., Whidbee, D. & Hobson, J. A. Visual
discrimination learning: both NREM and REM are required. Sleep 21
(Suppl.), 256 (1998).
22. Karni, A., Tanne, D., Rubenstein, B. S., Askenasy, J. J. M. & Sagi, D.
Dependence on REM sleep of overnight improvement of a perceptual skill.
Science 265, 679–682 (1994).
23. Contreras, D., Destexhe, A. & Steriade, M. Intracellular and computational
characterization of the intracortical inhibitory control of synchronized
thalamic inputs in vivo. J. Neurophysiol. 78, 335–350 (1997).
24. Destexhe, A., Contreras, D. & Steriade, M. Spatiotemporal analysis of local
field potentials and unit discharges in cat cerebral cortex during natural wake
and sleep states. J. Neurosci. 19, 4595–4608 (1999).
25. Sejnowski, T. J. & Destexhe, A. Why do we sleep? Brain Res. 886 (Suppl.)
208–223 (2000).
26. Smith, C. & Rose, G. M. Evidence for a paradoxical sleep window for place
learning in the Morris water maze. Physiol. Behav. 59, 93–97 (1996).
27. Bettina, B. L., Levi, D. M. & Reich, L. N. Vision Res. 34, 1679–1690 (1995).
28. Ball, K. & Sekuler, R. Direction-specific improvement in motion
discrimination. Vision Res. 27, 953–963 (1987).
29. Watanabe, T., Nanez, J. E. & Saksaki, Y. Perceptual learning without
perception. Nature 413, 844–848 (2001).
30. Karni, A., Weisberg, J., Lalonde, F. & Ungerleider, L. G. Slow changes in
primary and secondary visual cortex associated with perceptual skill learning:
an fMRI study. Neuroimage 3 (Suppl.), S543 (1995).
31. Rechtschaffen, A. & Kales, A. A Manual of Standardized Terminology,
Techniques and Scoring System for Sleep Stages of Human Subjects (Brain
Information Service, University of California, Los Angeles, 1968).
32. Cantero, J. L., Atienza, M., Hobson, J. A. & Stickgold, R. Nightcap: a reliable
home monitor for detecting normal sleep onset. Sleep 25, 238–245 (2002).
33. Hoddes, E., Zarcone, V., Smythe, H., Phillips, R. & Dement, W. C.
Quantification of sleepiness: a new approach. Psychophysiology 10, 431–436
(1973).
articles
nature neuroscience • volume 5 no 7 • july 2002 681
© 2002 Nature Publishing Group http://neurosci.nature.com
