• volume 3 no 8 • august 2000
ed brain areas that were more active in trained than in non-trained subjects, in REM sleep as compared to resting wakeful-ness. Finally, to formally test that these brain regions, possiblyreactivated during REM sleep, would be among the structuresthat had been engaged by executing and learning the task, a con- junction analysis was done
. This analysis identified those regionsthat were both more active during REM sleep in trained subjects(group 2) compared to non-trained subjects (group 3) and acti-vated during the execution of the task during waking (group 1).These cerebral regions, located in occipital and premotor cor-tices, were actually re-activated during REM sleep after training.
Task performance and sleep
Reaction times in group 2 significantly decreased across sessions(A, 568 ms; B, 533 ms; C, 444 ms,
<0.0001) andwithin sessions (24 blocks, ANOVA,
). The interaction between block and session factors wassignificant (
<0.0001), showing that reaction-timeimprovements differed across sessions. Separate analyses showedthat reaction times significantly improved within blocks duringsessions A and C (
<0.0001), but not during session B(
Tukey HSD tests revealed significant reac-tion-time differences between sessions A and B (
<0.05) andhighly significant RT differences between post-sleep (C) and eachpre-sleep (A, B) session (
<0.0005), suggesting that subjectsrefined their task-dependent visuomotor capabilities overnight.Sleep in groups 2 and 3 was comparable in all respects; no sig-nificant differences could be found in terms of any sleep parame-ter (
). Thus, any differences in brain activity between thetwo groups must be due to the prior training session in group 2.
In group 1 (waking SRT group), as compared to rest, execution of the SRT task was related to significant rCBF increases, in bilat-eral striate and extrastriate, premotor and superior parietal cor-tices and in the left primary sensorimotor cortex, supplementarymotor area, precuneus, anterior cingulate cortex and cerebellum(
). In group 2 (trained sleep group), during REMsleep as compared to wakefulness, significant rCBF increases wereobserved in bilateral striate, extrastriate, motor, premotor andsuperior parietal cortices as well as in the supplementary motorarea, anterior cingulate cortex, pons, mesencephalon and cere-bellum (
). In group 3 (non-trained sleep group), during REM sleep ascompared to wakefulness, significant bilateralrCBF increases were observed in striate,extrastriate, motor and superior parietal cor-tices as well as in supplementary motor areaand anterior cingulate cortex (
).Although both groups were scanned incomparable conditions, a significant group(trained versus non trained) by condition(REM sleep versus wakefulness) interactionwas found bilaterally in the cuneus and adja-cent striate cortex, the left premotor cortex,the inferior part of the left thalamus and themesencephalon (
). These brainareas were more active in trained than in non-trained subjects during REM sleep. The con- junction between the group-by-conditioninteraction and the waking results (group 1)was significant in the cuneus and adjacent stri-ate cortex bilaterally, left premotor cortex, and mesencephalon(
). These regions are both more active during REMsleep in the trained subjects (group 2) compared to the non-trained subjects (group 3) and activated by task performanceduring waking (group 1).
In group 1, task performance was accompanied by significantactivation of occipital, parietal, anterior cingulate, motor andpremotor cortices and the cerebellum. This pattern is consistentwith engagement of several mental processes necessary for thetask, such as visual and spatial perception
, spatial attention
,mental motor representation
, upper limb movements
, tim-ing processes
.In group 2, intensive pre-sleep training on the SRT task pre-sumably engaged this set of cerebral areas. On the behaviorallevel, performance improved within session A and C, as well asacross night. During the learning of perceptual-motor skills, theexperience of sequences of some external stimuli is proposed tolead to the formation of novel sequences of muscle activity
.Consequently, speed of motor execution is improved without anydeterioration in accuracy. During this process, the system acquiresa new capability, which is more than the simple functional adap-tation of any pre-existing property of the network
. Our datasuggest that similar processes may occur during REM sleep, asthey show a temporal association between the subjects’ perfor-
Table 1. Sleep parameters on the third night.
Sleep parameterGroup 2Group 3
Mean ±s.d.Mean ±s.d.
Total recording period (TRP, min)417 ±52403 ±100.46*Total sleep time (TST, min)352 ±46339 ±290.58*Stage II duration (min)185 ±43184 ±290.86*SWSduration (min)106 ±5991 ±491.00*REM sleep duration (min)61 ±2164 ±151.00*Stage II latency (min)17 ±1316 ±181.00*REM sleep latency (minus awake, min)112 ±51107 ±530.58*Sleep efficiency (TST/TRP)0.85 ±0.130.84 ±0.070.36*Sleep quality (SWS+REM sleep/TST)0.48 ±0.060.46 ±0.090.85*REM density in REM sleep17 ±1712 ±130.30**REM density in wakefulness24 ±2116 ±110.26**Theta power in REM sleep (
)4.88 ±2.776.22 ±4.190.22**
*Mann–Whitney U test; **unpaired two-tailed Student’s
Average reaction times (and standard deviations) for pre- andpost-sleep sessions (group 2). Global reaction time improves within ses-sions A and C, as well as overnight, between sessions.
Session ASession CSession B
R e a c t i o n t i m e ( m s )
2000 Nature America Inc. • http://neurosci.nature.com
© 2 0 0 0 N a t u r e A m e r i c a I n c . • h t t p : / / n e u r o s c i . n a t u r e . c o m