Effects of Early and Late Nocturnal Sleep on Declarative and Procedural Memory

Home , Jan Born, Procedural memory, Sleep and memory

Werner Plihal and Jan Born University of Bamberg, FRG Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021

Abstract 1 Recall of paired-associate lists (declarative memory) and early sleep, and recall of mirror-tracing skills improved more mirror-tracing skills (procedural memory) was assessed after during late sleep. The effects may reflect different influences retention intervals defined over early and late nocturnal sleep. of slow wave sleep (SWS) and rapid eye movement @EM) In addition, effects of sleep on recall were compared with sleep since time in SWS was 5 times longer during the early those of early and late retention intervals filled with wakeful- than late sleep retention interval, and time in REM sleep was ness. Twenty healthy men served as subjects. Saliva cortisol twice as long during late than early sleep @ < 0.005).Changes concentrations were determined before and after the retention in cortisol concentrations, which independently of sleep and intervals to determine pituitary-adrenal secretory activity.Sleep wakefulness were lower during early retention intervals than was determined somnopolygraphically. Sleep generally en- late ones, cannot account for the effects of sleep on memory. hanced recall when compared with the effects of correspond- The experiments for the first time dissociate specific effects of ing retention intervals of wakefulness. The benefit from sleep early and late sleep on two principal types of memory, decla- on recall depended on the phase of sleep and on the type of rative and procedural, in humans. H memory: Recall of paired-associate lists improved more during

INTRODUCTION practice sessions subjects learned to make the discrimi- nation with shorter exposure times of the stimulus Numerous studies in animals and humans have sug- configuration. While subjects still improved on this task gested that an essential function of sleep is the strength- following a night of SWS deprivation, deprivation of REM ening of memories (e.g., Cipolli, 1995; Dujardien, sleep completely abolished any beneficial effect of sleep. Guerrien, & Leconte, 1990). The sleep-dependent mecha- These results led the authors to suggest that REM sleep nisms underlying memory facilitation, however, are still facilitates the consolidation of procedural memory (i.e., poorly understood. Most studies approaching this issue memories considered to stem from repeated practicing have focused on the role of different sleep stages for of tasks of perceptual or motor skill learning; Squire, memory consolidation, and findings from these studies 1992). have merged into a still ongoing discussion contrasting Despite the large body of positive findings, quite a few the functions of slow wave sleep (SWS) and rapid eye studies failed to demonstrate impairing effects of REM movement @EM) sleep in memory formation and con- sleep deprivation on human memory mainly in tasks of solidation (Dujardien et al., 1990). declarative memory (e.g., Ekstrand, Sullivan, Parker, Sr Mnemonic effects related to REM sleep have been West, 1971; Levin & Glaubman, 1975). In light of these mainly investigated in studies employing REM sleep dep- inconsistencies,it has been questioned whether the deprivation (for review, see Levin & Glaubman, 1975). In rivation of REM sleep is the most appropriate way to humans, impairment of memory following REM sleep identify the role of this sleep stage for memory pro- deprivation has been observed in a great variety of tasks cesses. The disruption of REM sleep may per se affect involving declarative memory, such as the recall of sen- cognitive and emotive aspects of stimulus processing tences and stories (e.g., Empson & Clarke, 1970;Tilley & more than disruption of any other sleep stage and may Empson, 1978). However, a strong deleterious effect of thus obscure the functions of this sleep stage essential REM sleep deprivation on memory has been recently for retention processes (Cipolli, 1995). This objection reported also for recall of procedural memories (Karni, appears to be supported by the fact that studies that did Tanne, Rubenstein, Askenasy, & Sagi, 1994). Those authors not employ selective sleep deprivation techniques over- trained subjects to discriminate the orientation of a all suggested a more positive effect on retention of SWS configuration of visually presented lines. With repeated than of REM sleep (Barrett & Ekstrand, 1972; Ekstrand,

0 I997 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 94,pp. 534-547

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Barrett, West, & Maier, 1977;Fowler, Sullivan, & Ekstrand, RESULTS 1973;Yaroush, Sullivan, & Ekstrand, 1971). In those stud- Sleep ies, retention rates following undisturbed early and late nocturnal sleep were compared. If learning was followed Results from sleep recordings are summarized in Table by a 4-h retention interval placed in the early period of 1. Sleep time between the learning and recall period was nocturnal sleep characterized by extended epochs of nearly identical during the “early sleep” (189.65 f 7.68 SWS, recall of word pairs was superior to recall after a min) and “late sleep’’ condition (188.55 f 6.45 min). 4-h retention interval placed in the REM sleep predomi- However, sleep architecture substantially differed be- nated late period of sleep. However, evidence for an ad- tween both conditions. Time spent in SWS was about 5 vantage of SWS over REM sleep on memory consolidation

times longer during early sleep than during late sleep Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 provided by examinations of periods of undisturbed (refer to Table 1 for statistical results). In contrast,during early and late nocturnal sleep so far is based solely on early sleep, time spent in REM sleep on the average an assessment of declarative memories. Tasks of proce- reached only 35% of the time spent in this sleep stage dural memory have not been employed with this strategy. during late sleep. Sleep onset latency and latencies of the However, it is important to point out in this context different sleep stages were not different for the early and that early and late nocturnal sleep apart from the diver- late sleep conditions. gent distribution of SWS and REM sleep differ also with Both the condition of late sleep and late wake were respect to other factors that might well influence mem- preceded by 3-h intervals of sleep. Sleep in these inter- ory consolidation. Thus, secretory activity within the vals (prior to the learning period) were comparable, hypothalamo-pituitary-adrenal (HPA) system is at a mini- except that time in stage 1 sleep was slightly enhanced mum during the early part of nocturnal sleep and in the late wake condition. However, this difference did reaches a diurnal maximum in the late part of the night. not reach the 5% level of significance. This secretory pattern very likely reflects a tonic regula- tory influence of the septo-hippocampal system on the Cortisol HPA axis (Born, Steinbach,Dodt, & Fehm, 1997;De Kloet, 1991 ; De Kloet & Reul, 1987). Interestingly, several stud- Cortisol concentrations were substantially lower during ies in awake healthy humans indicated that high plasma the early sleep retention interval (0.34 f 0.13 pg/dl) than corticosteroid concentrations impair declarative mem- during late sleep retention interval (0.64 f 0.09 pg/dl; ory (eg,Kirschbaum, Wolf, May, Wippich,& Hellhammer, F(1, 9) = 7.29,p c 0.05). A parallel difference was ob- 1996; Newcomer, Craft, Hershey, Askins, & Bardgett, served in the wake control group with markedly lower 1994).This influence is probably mediated via binding to cortisol concentrations during the early wake (0.34 f hippocampal corticosteroid receptors (e.g., Bliss & 0.09 pg/dl) than during the late wake retention interval Collingridge, 1993; Oitzel & De Kloet, 1992), and it may (0.91 k 0.17; F(1, 9) = 13.47, p < 0.01). Differences express itself also in a differential impairment of recall between the experimental sleep group and the wake following early and late nocturnal sleep. control group for the corresponding early and late reten- The present study was designed to compare the role tion intervals were not significant (F(1, 18) = 0.00 and of early nocturnal sleep, dominated by SWS, and of late F(1, 18) = 1.82 for the early and the late night, respec- sleep, dominated by REM sleep, for the consolidation of tively). declarative and procedural memory. Since the selective dependency of procedural memory from REM sleep, so Paired-Associate far, has been shown only in the context of REM sleep Lists deprivation procedures, it was of particular interest to Table 2 summarizes results from performance on paired- examine whether procedural memory would improve associate lists for the learning and recall period of the also from undisturbed late sleep to a greater extent than four retention conditions (early sleep, late sleep, early from undisturbed early sleep. For this purpose, healthy wake, late wake). In the experimental sleep group, com- men were first trained (to a criterion) in recall of a parison of performance during the learning period (as paired-associate word list (a task of declarative memory) indicated by the number of trails until the criterion of and in a mirror-tracing task (a procedural memory task). 60% correct responses was reached and by the number Recall was again tested after 3-h retention intervals po- of correct responses during the final learning trial) did sitioned during early and during late sleep and also after not indicate any systematic difference between the con- corresponding intervals of wakefulness. In addition, ditions of early and late sleep. cortisol concentrations (in saliva) were assessed imme- Following early and late sleep, recall of words was diately before and after the retention intervals to indi- markedly improved compared with recall on the final cate the tonic state of HPA activity and to control for learning trial immediately prior to sleep. However, this possible feedback influences of cortisol on the retention improvement was distinctly greater following early sleep process. than late sleep (F(1, 9) = 5.21,p c 0.05; Table 2). Early

Plihal and Born 535

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Table 1. Sleep parameters. Values are means f SEM. Sleep onset (with reference to 2300 h; lights off), sleep time (i.e.,time from sleep onset until the end of the respective retention interval, when subjects were wakened), percentage of awake time (Wake), stage 1 sleep (Sl), stage 2 sleep (S2), slow wave sleep (SWS), and rapid eye movement sleep (REM). The right column indicates results from pairwise Wicoxon's test.

Sleep During Early versus Late Retention Interval Sleep During Early Sleep During Late Retention Interval Retention Interval P Sleep onset (min) 12.85 f 4.32 11.85 f 2.89

Sleep time (min) 189.65 f 7.68 188.55 f 6.45 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 Wake % 5.78 f 3.11 2.91 f 0.51 s1 % 8.85 f 2.87 10.95 f 2.31 s2 o/o 53.97 f 4.08 57.60 f 3.01 sws o/o 23.02 f 5.46 4.61 k 1.94 <0.005 REM % 8.38 f 2.33 23.93 f 1.10 <0.005

Sleep Prior to Late Wake versus Late Sleep Retention Interval

~~ Sleep Prior to Late Wake Sleep Prior to Late Sleep Retention Interval Retention Interval P Sleep onset (min) 17.85 f 5.23 35.42 f 12.02 Sleep time (min) 191.85 f 8.13 204.71 f 3.01 Wake % 0.19 f 0.11 3.73 k 2.41 s1 % 8.61 f 2.41 9.98 * 4.70

sleep increased recall from 17.6 k 0.84 to 23.3 f 0.33 improvement following early sleep was superior to that words (F(1, 9) = 33.19,p c 0.001), and late sleep from obtained during the three other retention conditions. 18.1 k 0.69 to 20.1 f 0.70 words (F(1, 9) = 5.04,p c Learning prior to the retention interval in both groups 0.05;Table 2, Figure 1). A greater improvement following was comparable:For both, in the early and late retention early sleep than late sleep was likewise confirmed for intervals no 'differences were observed between the two percentage of improvement (F(1, 9) = 5.01, p c 0.05). groups on the number of trials to criterion and the Also, the absolute number of words correctly recalled number of correct responses on the final criterion trial. was markedly higher following early sleep than late However, as is shown in Figure 1, the absolute number sleep (F(1,9) = 13.72,p c 0.01). of recalled words improved to a significantly greater In the wake control group, performance during the extent following the early sleep retention interval than learning period also did not differ for the early and late following the corresponding early wake retention inter- wake conditions (Table 2). Moreover, both the early and val (F(1, 18) = 8.63,~< O.Ol), while no such difference late wake retention intervals improved recall of words was observed between the late sleep and late wake compared to recall on the final learning trial immediately retention intervals (F(1, 18) = 2.68, ns; Group x Reten- prior to the retention interval. However, unlike in the tion interval - F(1, 18) = 12.62,p < 0.01).Absolute and experimental sleep group, in the wake control group, the percentage of improvement scores likewise showed a improvements in recall did not differ between the early particular pronounced enhancement of recall following (2.92 f 0.79 words) and late wake retention interval (2.1 the early sleep but not following the late sleep retention f 0.96 words; F(1, 9) = 0.34, ns). Also, percentage of interval when compared with the respective retention improvement scores were comparable for the early (16.5 interval of wakefulness (refer to Table 2 for results from f 5.1%)and late wake retention intervals (12.2 f 5.8%; pairwise statistical comparisons between the effects of F(1,9) = 0.28,ns; Figure l), and the absolute number of the corresponding retention intervals of sleep and wake- recalled words following the retention interval was, on fulness). the average, even slightly smaller following the early retention interval (20.6 0.86 words) than the late wake f Mirror Tracing retention interval (21.6 rt 0.60 words).Yet, this difference did not approach any significance (F(1,9) = 1.56, ns). Table 3 summarizes results on mirror-tracing perfor- A direct comparison between the experimental sleep mance for the learning and recall periods of the four group and the wake control group confirmed that recall retention wnditions (early sleep, late sleep, early wake,

536 Journal of Cognitive Neuroscience Volume 9, Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Table 2. Performance on paired-associate lists. Values are means f SEM. In the learning condition, presentation of a list of 24 paired associates was immediately followed by a cued recall. This procedure was repeated until the subject reached the criterion of at least 15 correctly recalled associates (criterion trial). The recall condition began with a cued recall of associates learned in the learning condition (relearn trial). Improvement refers to the difference in the number of words recalled in the learning and the recall condition. Percentage of improvement refers to improvement with recall on the criterion trial of the learning condition set to 100%.The right column indicates F values from pairwise comparisons.

Early versus Late SleeD Response Measure Early Sleep Late Sleep F(1, 9)

Learning Trials to criterion 1.4 f 0.16 1.5 f 0.17 3.10 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 Recall on criterion trial 17.6 f 0.84 18.1 f 0.69 2.00 Recall Recall on relearn trial 23.3 f 0.33 20.1 f 0.70 13.72R Improvement 5.7 f 0.99 2.0 f 0.89 5.21b Percentage of improvement 32.4 f 6.85 11.0 f 5.16 5.01b

Early versus Late Wakefulness

Response Measure Early Wake Late Wake F(1, 9)

~~~ Learning Trials to criterion 1.4 f 0.22 1.4 f 0.16 0.00 Recall on criterion trial 17.7 f 0.64 19.5 f 0.89 3.22 Recall Recall on relearn trial 20.6 f 0.86 21.6 f 0.60 1.56 Improvement 2.9 f 0.79 2.1 f 0.96 0.34 Percentage of improvement 16.5 f 5.10 12.2 f 5.79 0.28

Sleep versus Wakefulness Retention interval Early Late Response Measure F(1, 18) F(1, 18) Learning Trials to criterion 0.00 1.80 Recall on criterion trial 0.00 1.56

~ Recall Recall on relearn trial 2.68 Improvement 1.oo Percentage of improvement 2.80

< 0.01. bp < 0.05.

late wake). In the experimental sleep group, perfor- made fewer errors (4.4 _+ 1.1) than during the learning mance during the learning period was very comparable period before this interval (draw time: 67.2 f 7.0 sec, prior to the early and late sleep retention intervals for error time: 6.5 f 1.9 sec, error count: 9.9 k 2.1;F(l, 7) = mirror tracing a star as well as for average performance 25.97,p < 0.01; F(1, 7) = 7.83,~< 0.05, and F(1, 7) = on the six linedrawn figures. 10.89,p < 0.05, respectively). While the improvements Sleep improved subsequent mirror-tracing perfor- with regard to error time and error count were compa- mance. Thus, after the early sleep retention interval, sub- rable for the early and late sleep retention intervals, the jects needed less time to draw a human figure (50.5 f decrease in draw time following the late sleep retention 2.7 sec) and needed less time to correct errors (2.5 f interval was more than twice as strong as following the 0.3 sec) than before this sleep interval (draw time: 62.8 early sleep retention interval (F(1, 7) = 6.60, p < 0.05; f 5.4 sec, error time: 5.9 f 1.4 sec; F(1, 7) = 9.35,p < Table 3). A strikingly more pronounced reduction in 0.05, and F(1,7) = 8.18,p < 0.05, respectively). A parallel draw time following the late sleep retention interval decrease in the number of errors failed to reach sig- than early sleep retention interval was likewise indicated nificance. Also, after the late sleep retention interval, by percentage scores of improvement (41.8 f 5.1 versus draw time (39.1 f 2.5 sec) and error time (1.7 f 0.3 sec) 19.6 f 4.7%,F(1, 7) = 13.15,p < 0.01; Figure 1). Also, on the six figures were markedly shorter, and subjects absolute draw time was shorter after the late sleep

Plibal and Born 537

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 % Improvement

% 50 % 50 ** ** * I I I * I ii- 40 n 40 T Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 30 30

20 20

I0 10

0 0 early late early late retention interval retention interval Paired-associate list Mirror tracing

Figure 1. Comparisons of percentage of improvement in the experimental sleep (shaded bars) and the wake control (white bars) group of re call of paired-associate lists (left panel) and in mirror-tracing speed (right panel) after the early and the late retention interval (*p < 0.05;**p < 0.01).

retention interval (39.1 f 2.5 sec) than after the early A direct comparison between the experimental sleep sleep retention interval (50.5 k 2.7 sec, F(1,7) = 6.43,p group and the wake control group confirmed that mir- < 0.05). ror-tracing improvement following late sleep was supe- In the wake control group, performance during the rior to that obtained during the three other retention learning period also did not differ for the early and late conditions. Mirror-tracing performance during the learn- wake conditions (Table 3). Both retention intervals of ing period prior to the retention intervals did not differ wakefulness improved mirror-tracing performance:Thus, between the experimental sleep and the wake control following the early wake retention interval, subjects group. Also, mirror-tracing performance did not differ drew the figures faster (57.9 f 5.5 sec) and corrected following the early sleep and early wake retention inter- errors faster (2.2 f 0.9 sec) than before the early reten- vals. However, absolute draw time was distinctly shorter tion interval (draw time: 69.5 f 7.6 sec, error time: 3.3 f following the late sleep retention interval than the late 1.2 sec; F(1,7)= 7.03,~< 0.05 and F(1,7)= 6.80,p < wake retention interval (39.1k 2.5 sec versus 59.5 k 5.6 0.05, respectively). Again, a parallel decrease in the sec;F(l, 16) = 9.35,p< 0.01;Group x Retention interval number of errors failed to reach significance. Also, after - F(2, 30) = 3.48,p < 0.08).Likewise, improvement in the late wake retention interval, subjects were faster in draw time was distinctly more pronounced following mirror tracing the figures (59.5 k 5.6 sec) and made the late sleep retention interval (28.0f 5.5 sec) than the fewer errors (6.5f 2.1) than during the learning period late wake retention interval (10.9 f 4.8 sec, F(1, 16) = before this interval (draw time: 70.5 f 7.9 sec, error 5.50,p< 0.05;Group x Retention interval - F(1, 16) = count: 8.2 f 2.9;F(1, 7) = 5.18,p< 0.05 and F(1, 7) = 2.28,p< 0.1).A stronger reduction in draw time follow- 6.48,p< 0.05,respectively). However, improvement,per- ing the late sleep retention interval than the late wake centage of improvement and absolute measures of draw retention interval was similarly indicated by the per- time, error time, and error count were highly compara- centage of improvement (41.8& 5.1%versus 15.6 f 5.6%; ble following the early and late wake retention intervals. F(1, 16) = 11.21,p< 0.01;Group x Retention interval - Differences in these measures between both the early F(1, 16) = 4.17,p < 0.05).In contrast to draw time, and late wake conditions were generally small and far changes in error count and error time were similar after from approaching any significance (Table 3). the late sleep and late wake retention intervals.

538 Journal of Cognitive Neuroscience Volume 9, Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 DISCUSSION these conditions during the learning epoch. While learning prior to the early sleep retention interval took place The present study compared effects of early and late after the usual period of daytime wakefulness, learning nocturnal sleep on indicators of declarative and proce- prior to the late sleep retention interval was preceded dural memory. Effects of sleep, in addition, were com- by a 3-h interval of sleep. In fact, Worchel and Marks pared with those of retention intervals filled with (1951) reported on an improved list learning when the wakefulness. presentation of the lists was preceded by a period of Sleep, in general, improved recall of paired-associate sleep rather than wakefulness, suggesting a proactive word lists and performance on the mirror-tracing task, if supporting effect of sleep on learning.In contrast, Stones

compared with effects of corresponding retention inter- (1973,1977) reported on a negative effect of sleep prior Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 vals of wakefulness. These findings confirm a large num- to learning which, in addition, depended on the stage of ber of foregoing observations in animals and humans sleep from which the subject was aroused. The present (for reviews, e.g., Fishbein & Gutwein, 1977; Perlman, data did not indicate any effect of early nocturnal sleep 1979; Dujardin et al., 1990; Rotenberg, 1992; Cipolli, on subsequent learning (prior to the late sleep retention 1995). Yet, as a major result the present experiments interval). Learning of paired associate word lists as well demonstrate that the beneficial effects of sleep on mem- as mirror-tracing skills was comparable for both early ory are strongly influenced by the period of nocturnal and late sleep retention conditions. These findings argue sleep (early versus late) and by the type of memory task against an essential proactive influence on memory pro- employed (declarative versus procedural). Declarative cesses of the periods of, respectively, sleep and wake- memory as indicated by the recall of paired-associate fulness prior to the learning period, although a weak word lists improved to a distinctly greater extent over influence of this type cannot be entirely ruled out. The periods of early nocturnal sleep than of late sleep. Con- lack of any effect of prior sleep on learning,in this study, versely, recall of mirror-tracing skills, indicating proce- may be ascribed to the fact that on all occasions subjects dural memory, was facilitated by periods of late were aroused from stage 1 or 2 NonREM sleep, which nocturnal sleep more than by early sleep. in foregoing studies had been found to prevent proactive The differential effects of early and late nocturnal interfering effects of sleep (Stones, 1973, 1977). sleep cannot be reduced to influences of a circadian The finding of a stronger increase in recall of paired- oscillator mediated, for example, by general changes in associate word lists after an epoch of early nocturnal body temperature and basal metabolism (Aschoff, Fatran- sleep than of late confirms earlier findings by Ekstrand ska, Gerecke, & Gliedke, 1974). This type of influence and coworkers (Ekstrand et al., 1977;Fowler et al., 1972; would have been expected to result in a comparable Yaroush et al., 1971). Those authors supposed that the pattern of changes while subjects were asleep and effects of early sleep reflect a protective influence of awake. However, unlike retention intervals of sleep, re- SWS on the decay process of memory. As expected, the tention intervals filled with wakefulness did not differ in present recordings confirmed a prevalence of SWS dur- their effects on recall between the early and late part of ing early sleep.Time spent in SWS during the early sleep the night. Nevertheless, it should be noted that with the retention interval was about 5 times longer than during wake control group of the present study, circadian ef- the late sleep retention interval. In fact, the amount of fects and effects of sleep could not be completely dis- SWS during the late sleep retention interval, being less sected. While the early retention interval was preceded than 5%, appeared to be neghgible. by a (daytime) 1Gh period of wakefulness, the late re- Alternatively, the superior retention of paired-associate tention interval was preceded by a 3-h period of early word lists over intervals of early sleep may reflect a nocturnal sleep. Resulting differences in sleep propen- disruptive influence of REM sleep on memory. Consis- sity may have obscured circadian effects in the wake tent with the distribution of REM sleep observed under condition similar to those observed during early versus normal undisturbed sleeping conditions, time spent in late retention intervals of sleep. Nevertheless, this view REM sleep was more than twice as long during the late seems very unlikely considering that respective antece- sleep retention interval than during the early one. Hence, dent conditions were identical for the wake control a disruptive effect of this sleep stage would be expected group and the experimental sleep group. Also, the late to be less pronounced during early sleep. retention interval in both groups was preceded by a In this way, superior effects of early nocturnal sleep comparable period of early nocturnal sleep, which ex- on retention of paired-associate word lists have been cludes an explanation of differences in recall between consistently reduced to differential effects of SWS and the retention periods of wakefulness and sleep solely in REM sleep on the decay of declarative memory, the rate terms of the accumulated time of wakefulness (i.e., of of which is slow during SWS and fast during REM sleep. sleep deprivation). However, this concept so far has not been tested with It could be likewise argued that differential effects of other types of memory and obviously cannot account early and late sleep on the recall of paired associates and for the present findings that indicate that recall of mir- mirror-tracing skills were due to differences between ror-tracing skills (i.e., procedural memory), conversely, is

Plibal and Born 539

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Table 3. Performance on mirror tracing. Values are means f SEM. Times are given in seconds. Error count refers to the number of instances when the stylus left the trace. In the learning condition, first a star was repeatedly presented until it was traced with less than six errors. Thereafter, subjects were required to mirror trace six linedrawn figures. In the recall condition, the six line-drawn figures were presented again. Improvement values refer to the differences in the respective values of the learning and the recall condition. Percentage of improvement refers to improvement with the respective values of the learning condition set to 100%.The right column indicates F values from pairwise comparisons.

Earlv versus Late SleeD Response Measure Early Sleep Late Sleep FU, 7)

Learning STAR Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 Trials to criterion 1.9 f 0.3 2.4 f 0.5 0.70 Draw time 143.8 f 33.6 152.6 f 46.1 0.08 Error time 16.1 k 4.1 29.8 f 12.9 1.17 Error count 17.1 f 8.4 23.1 f 9.9 0.22 SIX FIGURES Draw time 62.8 f 5.4 67.2 f 7.0 0.22 Error time 5.9 f 1.4 6.5 f 1.9 0.06 Error count 9.0 f 1.5 9.9 f 2.1 0.12 Recall SIX FIGURES Draw time 50.5 f 2.7 39.1 f 2.5 6.43b Error time 2.5 f 0.3 1.7 f 0.3 2.33 Error count 5.9 f 0.4 4.4 f 1.1 1.49 IMPROVEMENT Draw time 12.3 f 4.0 28.0 f 5.5 6.60b Error time 3.4 f 1.2 4.8 f 1.7 0.48 Error count 3.2 f 1.4 5.6 f 1.7 1.05 PERCENTAGE OF IMPROVEMENT Draw time 19.6 f 4.7 41.8f 5.1 13.15a Error time 57.6 f 9.3 73.8 f 10.7 1.12 Error count 34.4 f 14.9 55.0 f 10.8 1.83

Early versus Late Wakefulness Response Measure Early Wake Late Wake FU,9) Learning STAR Trials to criterion 2.0 f 0.4 2.3 k 0.5 0.31 Draw time 140.2 f 35.8 166.9 + 50.0 0.13 Error time 11.3 f 4.1 26.4 f 15.9 1.01 Error count 22.4 k 8.3 26.0 f 10.3 0.09 SIX FIGURES Draw time 69.5 k 7.6 70.5 f 7.9 0.01 Error time 3.3 f 1.2 4.4 f 2.0 0.27 Error count 7.4 f 1.8 8.2 f 2.9 0.38 Recall SIX FIGURES Draw time 57.9 f 5.5 59.5f 5.6 0.10 Error time 2.2 f 0.9 2.4 f 0.8 0.12 Error count 5.2 f 1.8 6.5 f 2.1 0.91 IMPROVEMENT Draw time 11.6 f 4.4 10.9 f 4.8 0.01 Error time 1.1 f 0.4 2.0 f 1.5 0.28 Error count 2.3 f 1.1 2.4 f 0.9 0.00 PERCENTAGE OF IMPROVEMENT Draw time 16.7 f 4.7 15.6 f 5.6 0.06 Error time 33.3 f 20.7 45.5 f 22.7 0.14 Error count 30.7 f 15.4 26.9 f 12.3 0.25

up < 0.01. bp < 0.05.

540 Journal of Cognitive Neuroscience Volume 3, Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Table 3. Continued.

Sleep versus Wakefulness Retention Interval

Response Measure Learning STAR Trials to criterion 0.06 0.01 Draw time 0.01 0.04 Error time 0.64 0.03 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 Error count 0.19 0.04 SM FIGURES Draw time 0.47 0.09 Error time 2.15 0.56 Error count 0.41 0.09

~ ~~~ Recall SIX FIGURES Draw time 1.26 9.35a Error time 0.07 0.64 Error count 0.11 0.68 IMPROVEMENT Draw time 0.02 5.50b Error time 4.06 1.51 Error count 0.25 3.04 PERCENTAGE OF IMPROVEMENT Draw time 0.15 11.21U Error time 0.00 5.74O Error count 0.05 0.59

< 0.01. bp < 0.05.

facilitated more by late sleep than by early sleep. These during previous learning that helps to strengthen long- results support earlier work by Karni et al. (1994) that term consolidation. The spatial behavioral task used in suggest a beneficial effect of REM sleep on recall of that study, in which the animals learned environmental perceptual skills that likewise reflects procedural mem- locations, is a paradigm as typical for the assessment of ory. However, that study employed a REM sleep depriva- declarative memories as is the recall of paired-associate tion procedure that may induce emotional and cognitive lists. Moreover, the findings by Wilson and McNaughton disturbances that per se can affect the retention process (1994) and others (for a review, see Squire, 1992) suggest (Cipolli, 1995). that declarative memory, more than other types of mem- The present findings show a selective facilitation of ory, relies on hippocampal mechanisms supporting its indicators of declarative memory by early sleep and a consolidation. An essential part of this consolidation selective facilitation of indicators of procedural memory process appears to be a hippocampal replay of pre- by late nocturnal sleep. Although the mechanisms re- viously encoded events by which information is trans- sponsible for the specifity of sleep effects on different ferred to the neocortex. This hippocampal replay may be types of memory are as yet completely unknown, the most efficient during SWS supposed to be an “off-line” data seem to be consistent with the view that discrete time with little interfering background electricity (Wil- sleep stages, such as SWS and REM sleep, actively en- son & McNaughton, 1994; Barinaga, 1994). hance certain aspects of memory formation and consoli- While animal studies have provided a basis to explain dation. A significant role of SWS for the consolidation of the improvement of declarative memories during SWS, spatial memories has been suggested by recent animal it is at present difficult to account for the selective studies (Wilson & McNaughton, 1994) that show that facilitation of procedural memory by REM sleep. Since spatio-temporal patterns of neuronal firing in the hippo- neither the present nor previous studies (Karni et al., campus during SWS were similar to firing patterns ob- 1994) found any evidence for a particular improving served in the same neuronal cell assemblies during a influence of SWS on procedural memory, a hippocampal learning epoch preceding sleep.Those authors suggested replay of previously encoded procedural skills seems not that the hippocampal mode of neuronal firing during to be essential for the long-term storage of these memo- SWS represented a replay of neural patterns encoded ries. In fact several lines of evidence support the view

Plibal and Born 541

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 that the formation of procedural memories is less de- the experiment, including the placement of electrodes. pendent on hippocampal activity than is consolidation The experiment was approved by the local ethics com- of declarative memories (Squire, 1992; Ungerleider, mittee. All subjects gave written consent and were told 1995). Hence, it may be that consolidation of procedural that they could leave the experiment at any time. memory primarily involves a strengthening of stimulus or motor representations within cortical areas that are Material nonspecifically supported by the generally enhanced cortical activity during REM sleep (McGinty & Siegel, A paired-associate list learning task with category in- 1977; Ravagnati, Halgren, Babb, & Candrall, 1979). stance pairs was used to assess declarative learning, and

Saliva cortisol concentrations were determined, here, a mirror-tracing task was designed to reflect procedural Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 to indicate the tonic state of secretory activity within the learning. hypothalamus-pituitaryadrenal system and to control for hormonal feedback effects on memory processes. Corre- Paired-Associate Word Lists sponding with its well-known circadian rhythm (Weitz- man, Zimmerman, Czeisler, & Ronda, 1983), cortisol Stimuli. Two different paired associate word lists were concentration in both the experimental sleep group and used, enabling measurement on two experimental nights. the wake control group were higher in the late part of Each list consisted of 24 pairs of German nouns pre- the night than in the early part. Since these changes were viously evaluated in a normative study (Naumann & Bar- comparable in both groups, a feedback influence of cor- tussek, unpublished). Both stimulus words and response tisol would not explain the different effects on memory words of the two lists were matched in length, emotion- observed selectively between the early and late noctur- ality, meaningfulness, and concreteness. In addition to the nal retention periods of sleep. 24 word pairs, 4 dummy pairs of words at the beginning Nevertheless, differences in memory consolidation of each list served as primacy-effect buffers, and 4 during nocturnal sleep may not be completely inde- dummy pairs at the end of the list were used to buffer pendent from regulation of pituitary-adrenal activity. Pre- recency effects. Response words represented instances vious human studies have demonstrated that stimulation for the categories of the respective stimulus words (e.g., of pituitary-adrenal secretion is inhibited during early Animal-Dog;Bird-Eagle). The sequence of word-pair pres- sleep, with this inhibition presumably imposed by the entations within a list was randomized over repeated hippocampus (Spath-Schwalbe,Uthgenannt, Voget, Kern, trials in order to prevent serial learning. Born, & Fehm, 1993;Jacobson & Sapolsky, 1991). Consid- ering the key roles of the hippocampus for the tonic Apparatus and Procedure. In the learning condition regulation of pituitary-adrenal secretory activity and for the list was presented by a personal computer on a color the formation of memory (De Kloet, 1991; De Kloet, monitor (NEC MultiSync 5FGe). The presentation rate Rosenfeld, van Ekelen, Sutanto, & Levine, 1988), a more was 1/5 sec and the interstimulus interval was 100 msec. direct assessment of the sleep-related inhibition of pitui- Presentation of a list was followed by a cued recall in tary-adrenal secretion in future studies could provide an which the 24 stimulus words appeared on the screen in indicator more closely related to the facilitating influ- a different succession than in the foregoing presentation ences of early sleep on declarative memory than the of the list of word pairs. The subject was required to discrete EEG signs of SWS. recall the appropriate response word and had unlimited time to respond by typing the word on a keyboard. Responses were processed immediately and the presen- METHODS tation of lists for learning and cued recall was repeated until the subject achieved a minimum of 60%(i.e., 15 Subjects words) correct responses. Subjects were 20 healthy men who were nonsmokers In the recall condition after the retention interval the and had no history of sleep disturbances. Ten of them 24 stimulus words appeared in a randomized succession (mean age 28.2 years, range 24-38 years) were randomly on the screen. The subject was required to recall the assigned to the experimental sleep group. The remaining appropriate response words and had unlimited time to 10 men (mean age 26.2 years, range 21-37 years) con- respond. stituted a wake control group. Subjects were psychology students at the University Mirror tracing of Bamberg and were paid for participation. They were obliged to get up before 0700 h on the days prior to Stimuli. Two different sets with seven different stimuli experimental nights and to not take any naps during the were used in order to enable measurement in two ex- day. They were acclimated to the experimental sleep perimental nights. Set 1 consisted of a linedrawn five- conditions by spending one night under conditions of pointed star and six linedrawn human figures made up

542 Journal of Cognitive Neuroscience Volume 9, Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 of 26 to 28 straight segments joined by 25 to 27 angles criteria outlined by Rechtschaffen and Kales (1968). Fig- (i.e., changes in direction). In set 2 the same figures were ure 3 outlines the experimental design. used, but the straight segments were smoothed (Figure The subjects of the experimental sleep group on one 2).The overall width of a figure was 15 cm and the height of these occasions (early sleep retention interval) were was 20 cm. Lines were 0.8 cm thick. presented with the two learning tasks just before retir- ing, i.e., between 2215 and 2300 h (learning). After the Apparatus. The apparatus consisted of a 33- by 33-cm learning criteria had been reached, lights were turned platform made of mdky glass that was illuminated from off at 2300 h to enable sleep. When 3 h of sleep had the underside. Adjoining one end of the platform was a been observed in the on-line recordings, the subject was 16- by 20-cm steel angle on which was mounted a 12- awakened and cued recall of the list and performance Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 by 18-cm mirror. Direct visual access to the platform was on the mirror-tracing task were tested 15 min later prevented by a 31- by 33-cm plastic panel that was fixed (recall). On the other night (late sleep retention interval), at a height of 17 cm above the platform. A stylus with an following lights off at 2300 h, subjects first slept for 3 h, electronic light sensor counted errors and assessed error and 15 min after awakening were presented with the time whenever it was moved off the thick black line of two learning tasks within 45 min (learning). Thereafter, a figure.An electronic device synchronized the activation they returned to bed for 3 more hours of sleep, after of the stylus with an electronic stopwatch that measured which they were awakened, again. Fifteen min later cued drawing time. recall on the paired-associate list and performance on the mirror-tracing task was tested (recall). Since arousal from certain sleep stages may influence Procedure. In the learning condition, the star was pre- subsequent recall (Stones, 1977), subjects were always sented first. The subject traced the star, starting and end- awakened from stage 1 or stage 2 sleep. The sequence ing at the same marked point, with an electronic stylus of tasks was balanced across subjects. Moreover, the that measured drawing time, number of errors, and total order of presentation of the two paired-associate word error time. An error consisted of moving the stylus off lists and the two sets of mirror-tracing stimuli during the the line of a figure. Each time the stylus was moved off two testing occasions of a subject was balanced across the figure was considered as a separate error. As long as subjects. Five subjects were run first on the early sleep the stylus was beyond the thick line of the figure the condition and then on the late sleep condition. For the light sensor counted error time. Error time was not meas- other five subjects this order was reversed. The interval ured separately for single errors but was accumulated to between the two experimental nights of a subject was total error time. Mirror tracing the star was repeated until at least 1 week. the subject reached a criterion of a maximum of six The subjects of the wake control group were also errors. Then, the six line-drawn figures were presented tested on two nights (early wake and late wake retention one after the other. interval). During the early wake condition (as during the In the recall condition, the star was presented first in early sleep condition), they were presented with the order to warm up subjects and to keep conditions com- learning tasks between 2215 and 2300 h (learning). But parable. Thereafter, the six line-drawn figures were pre- then they stayed awake, and cued recall of the paired-as- sented in a randomized sequence. The subject traced sociate word lists and performance on the mirror-tracing each figure starting at the scalp and ending at the scalp. task was tested 3 h and 15 min later (recall). During the Draw time, error count, and error time were recorded late wake condition (as during the late sleep condition) for each figure. Performance measures were (1) the following lights off at 2300 h, subjects first slept for 3 h, mean draw time, (2) the mean error count, and (3) the and 15 min after awakening were presented with the mean error time collapsed across the six linedrawn two learning tasks within 45 min (learning). Thereafter, figures. they stayed awake and recall of the paired-associate word list and performance on the mirror-tracing task was tested 3 h and 15 min later (recall). Design and General Procedure During the intervals of wakefulness subjects of the The experiments took place in the sleep laboratory of wake control group were offered a slide show of a the Department of Physiological Psychology at the Uni- variety of parks and wilderness areas around the world. versity of Bamberg. After having spent an adaptation Balancing procedures with regard to the sequence of night of sleep in the laboratory, each man was tested on task presentation, the order of presenting the two paral- two experimental nights. Each experimental night lel versions of the tasks on the two testing occasions and started at 2130 h by preparing the subject for som- the order of experimental conditions were the same as nopolygraphic sleep recordings. Sleep was monitored by for the experimental sleep group. standard recordings of EEG, EMG, and vertical and hori- To determine cortisol concentrations during the reten- zontal EOG. Sleep stages were classified according to the tion intervals, saliva was sampled immediately before

Plibal and Born 543

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021

Set 1

Set 2

Figure 2. Mirror-tracing stimuli. Two different sets with seven different stimuli were used in order to enable measurement in two experimental nights. Each set consisted of a linedrawn five-pointed star and six linedrawn figures. Figure segments were straight in set 1 and smoothed in set 2. With visual access enabled by a mirror, subjects traced the figures with an electronic stylus that measured draw time, number of errors, and error time. An error consisted of moving the stylus off the trace.

544 Journal of Cognitive Neuroscience Volume 9,Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 STUDY DESIGN 2215-2300h 2300-0200h 0215-0300h 0300-0600h 0615-0700h

Sleep group

~ ~~ ~~~ Early Learning retention Reca 11 Night A interval MT SLEEP PAL

PAL MT Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 Late Learning retention Reca 11 Night B interval Sleep MT SLEEP PAL PAL MT

Wake qroup Early Learning retention Recall Night A interval MT WAKEFULNESS PAL PAL MT Late Learning retention Recall Night B interval Sleep MT WAKEFULNESS PAL PAL MT

Figure 3. Study design. Both sleep and wake groups were tested on two experimental nights (Night A and Night B), with the order of nights balanced across subjects. When the retention interval was placed into the early night, a paired-associate list (PAL) and a mirror-tracing (MT) task were presented between 2215 and 2300 h (learning). After learning, a 3-h retention interval (filled with sleep or wakefulness) followed. At 0215 h (i.e., 15 min after the retention interval), cued recall of the list and performance on the mirror-tracing task were tested. When the retention interval was placed into the late night, all subjects (sleep as well as wake group) slept from 2300 to 0200 h. Fifteen min after awakening, the learning tasks were presented. The retention interval (filled with sleep versus wakefulness) lasted from 0300 to 0600 h. Fifteen min later, cued recall of the word list and mirror-tracing performance were tested between 0615 and 0700 h.

and after the retention interval with the Salivette sam- Cortisol pling device (Sarstedt Inc., Rommelsdorf, Germany). Sam- ples were kept at -20°C until analyzed. Saliva cortisol was measured by radioimmunoassay (Herrman Biermann, Bad Nauheim, FRG; sensitivity 0.01 pg/dl, intraassay coefficient of variation <3% between Dependent Variables and Data Analysis 0.1 to 5 pg/dl, interassay coefficient of variation

Plibal and Born 545

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 and (2) the difference in the number of recalled words sleep.Journal of Clinical Endocrinology and Metabo- between the first recall trial and the final learning trial lism, 82, 1106-1 110. Cipolli, C. (1995). Sleep, dreams and memory: An overview. (improvement). The improvement was also expressed as Journal of Sleep Research, 4, 2-9. a percentage value with the number of words recalled De Kloet, E. R. (1991). Brain corticosteroid receptor balance on the final learning trial set to 100%. and homeostatic control. Frontiers Neuroendocrinology, 12, 95-164. De Kloet, E. R., & Reul, J. M. H. M. (1987). Feedback action Mirror Tracing and tonic influence of corticosteroids on brain function: A concept arising from the heterogeneity of brain recep- Differences in the original learning performance among tor systems. Psychoneuroendocrinology, 12, 83- 105.

the four experimental conditions were assessed with De Kloet, E. R., Rosenfeld, F?, van Ekelen, J. A. M., Sutanto, W., Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 respect to (1) the number of trials to criterion for mirror & Levine, S. (1988). Stress, glucocorticoids and develop- ment. Progress in Brain Research, 73, 101-120. tracing a star, (2) the mean draw time, (3) the mean error Dujardin, K., Guerrien, A., & Leconte, F! (1990). Sleep, brain time, and (4) the mean error count across the trials to activation and cognition. Physiology and Behaviol; 47, criterion. The three latter measures were evaluated sepa- 1271-1278. rately for mirror tracing the star and for performance on Ekstrand, B. R., Barrett, T. R., West, J. N., & Maier, W. G.(1977). the six line-drawn figures. Regarding the performance The effect of sleep on human long-term memory. In R. R. Drucker-Colin & J. L. McGaugh (Eds.), Neurobiology of during the later recall period, again, (1) mean draw time, sleep and memory (pp. 419-438). New York: Academic mean error count, and mean error time were calculated Press. (across the six trials of mirror-tracing line-drawn figures). Ekstrand, B. R., Sullivan, M. G., Parker, D. E, & West, J. N. (2) Improvement was determined by calculating individ- (1971). Spontaneous recovery and sleep.Journa1 of Ex- ual differences for the measures of mean draw time, perimental Psychology, 88(1), 142- 144. Empson, J. A. C., & Clarke, F? R. E (1970). Rapid eye move- mean error count, and mean error time derived from the ments and remembering. Nature, 227, 287-288. learning and recall periods. Again this improvement was Fishbein, W., & Gutwein, B. M. (1977). Paradoxical sleep and also expressed as a percentage with the respective val- memory storage processes. Behavioral Biolou, 19, 425- ues obtained during learning set to 100%. 464. Statistical analysis in general based on multivariate Fowler, M. J., Sullivan, M. G., & Ekstrand, B. R. (1973). Sleep and memory. Science, 179, 302-304. analyses of variance (MANOVA) including a Group factor Jacobsen, L., & Sapolsky, R. (1991). The role of the hippocam- (sleep versus wake) and a repeated measures factor pus in feedback regulation of the hypothalamic-pituitary- representing the Retention interval (early versus late adrenocortical axis. Endocrine Reviews, 12, 118- 134. nighttime). In addition, contrasts were specified to deter- Karni, A,, Tanne, D., Rubenstein, B. S. Askenasy,J. J. M., & mine effects of the early and late retention intervals Sagi, D. (1994). Dependence on REM sleep of overnight improvement of a perceptual skill. Science, 265, 679- separately in the experimental sleep group and wake 681. control group. Since two subjects completely failed to Kirschbaum, C., Wolf, 0. T., May, M., Wippich, W., Sr Hellham- reach the criterion after six attempts, mirror-tracing mer, D. H. (1996). Stress- and treatment-induced elevations analyses were restricted to the remaining eight subjects. of cortisol levels associated with impaired declarative A Greenhouse-Geisser corrected p value less than 0.05 memory in healthy adults. Life Sciences, 58(17), 1475- 1483. was considered significant. Lewin, I., & Glaubman, H. (1975). The effect of REM deprivation: Is it detrimental, beneficial, or neutral? Psychophysiol- Reprint requests should be sent to Werner Plihal, Physiolo- o@, 12(3), 349-353. gische Psychologie, Universitat Bamberg, Markusplatz 3,96045 McGinty, D. J., & Siegel,J. M. (1977). Neuronal activity pat- Bamberg, FRG. E-mail: [email protected]. terns during REM sleep: Relation to waking patterns. In R. R. Drucker-Colins & J. L. McGaugh (Eds.), Neurobiology of sleep and memory (pp. 135-158). New York: Academic REFERENCES Press. Naumann, E., & Bartussek, D. (1992). [Concreteness, emotion- Aschoff, J., Fatranska, M., Gerecke, U., & Giedke, H. (1974). ality, imagery, meaningfulness, potency, and valence of 800 Twenty-four-hour rhythms of rectal temperature in hu- German nouns]. Unpublished raw data. mans: Effects of sleep-interruptions and of test sessions. Newcomer, J. W., Craft, S., Hershey, T., Askins, K., & Bardgett, Pflugers Archives, 346, 215-222. M. E. (1994). Glucocorticoid-induced impairment in deckara- Barinaga, M. (1994). To sleep, perchance to . . . learn? New tive memory performance in adult humans. Journal of studies say yes. Neuroscience, 263, 603-604. Neuroscience, 14, 2047-2053. Barrett, T. R., & Ekstrand, B. R. (1972). Effect of sleep on mem- Oitzel, M. S., & De Kloet, E. R. (1992). Selective corticosteroid ory: 111. Controlling for time-of-day effects. Journal of Ex- antagonists modulate specific aspects of spatial orienta- perimental Psychology, 96, 321-327. tion learning. Behavioral Neuroscience, 106, 62-7 l. Bliss, T. V. F?, & Collingridge, G. L. (1993). A synaptic model of Pearlman, C. (1979). REM sleep and information processing: memory: Long-term potentiation in the hippocampus. Na- Evidence from animal studies. Neuroscience and Biobe- ture, 361, 31-40. havioral Reviews, 3, 57-68. Born, J., Steinbach, D., Dodt, C., & Fehm, H. J. (1997). Block- Ravagnati, L., Halgren, E., Babb, T. L., & Candrall, l? H. (1 979). ing of central nervous mineralocorticoid receptors (MR) Activity of human hippocampal formation and amygdala conteracts inhibition of pituitary-adrenal activity in human neurons during sleep. Sleep, 2(2), 161-173.

546 Journal of Cognitive Neuroscience Volume 9, Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021 Rechtschaffen, A,, & Kales, A. (1967).A manual of stand- different sleep stages. British Journal of Pycbology, 68, ardized terrninoloa, techniques and scoring system for 177- 181. sleep stages of human subjects. Bethesda, MD: N.I.H. Publi- Tilley, A. J., & Empson, J. A. C. (1978). REM sleep and memory cation No. 204. consolidation. Biological Psycbology, 6, 293-300. Rotenberg, V S. (1992). Sleep and memory 11: Investigations Ungerleider, L. G. (1995). Functional brain imaging studies of on humans. Neuroscience and Behavioral Reviews, 16, cortical mechanisms for memory. Science, 270, 769-775. 503-505. Weitzman, E. D., Zimmerman,J. C., Czeisler, C. A., & Ronda,J. Spath-Schwalbe,E., Uthgenannt, D., Voget, G. Kern, W., Born, (1983). Cortisol secretion is inhibited during sleep in nor- J., & Fehm, H. L. (1993). Corticotropin-releasing-hormone mal men.Journal of Clinical Endocrinology and Metabc- induced ACTH and cortisol secretion depends on sleep lism, 56 352-360. and wakefulness.Journal of Clinical Endocrinology and Wilson, M. A., & McNaughton, B. L. (1994). Reactivation of

Metabolism, 77, 1170-1 173. hippocampal ensemble memories during sleep. Science, Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/9/4/534/1754819/jocn.1997.9.4.534.pdf by guest on 18 May 2021 Squire, L. R. (1992). Memory and the hippocampus: A synthe- 265, 676-679. sis from findings with rats, monkeys, and humans. Psycho- Worchel, €?,& Marks, M. H. (1951). The effects of prior sleep logical Review, 99(2), 195-231. on learning.Journal of Experimental Pycbology, 42, 313- Stones, M. J. (1973). The effect of prior sleep on rehearsal, re- 316. coding and memory. British Journal of Psycbology, 64, Yaroush, R., Sullivan, M. J., & Ekstrand, B. R. (1971). Effect of 537-543. sleep on memory: 11. Differential effect of the first and sec- Stones, M. J. (1977). Memory performance after arousal from ond half of the night.Journal of Experimental Psycbol- O~Y,88, 361-366.

Plibal and Born 547

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1997.9.4.534 by guest on 30 September 2021