SEARCH

SEARCH BY CITATION

Keywords:

  • declarative memory;
  • intelligence;
  • memory;
  • motor memory;
  • sleep;
  • sleep duration

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Recent findings clearly demonstrate that daytime naps impart substantial memory benefits compared with equivalent periods of wakefulness. Using a declarative paired associates task and a procedural motor sequence task, this study examined the effect of two lengthier durations of nocturnal sleep [either a half night (3.5 h) or a full night (7.5 h) of sleep] on over-sleep changes in memory performance. We also assessed whether subject intelligence is associated with heightened task acquisition and, more importantly, whether greater intelligence translates to greater over-sleep declarative and procedural memory enhancement. Across both tasks, we demonstrate that postsleep performance gains are nearly equivalent, regardless of whether subjects obtain a half night or a full night of sleep. Remarkably, the over-sleep memory changes observed on both tasks are very similar to findings from studies examining performance following a daytime nap. Consistent with previous research, we also observed a strong positive correlation between amount of Stage 2 sleep and motor skill performance in the full-night sleep group. This finding contrasts with a highly significant correlation between spectral power in the spindle frequency band (12–15 Hz) and motor skill enhancement only in the half-night group, suggesting that sigma power and amount of Stage 2 sleep are both important for optimal motor memory processing. While subject intelligence correlated positively with acquisition and retest performance on both tasks, it did not correlate with over-sleep changes in performance on either task, suggesting that intelligence may not be a powerful modulator of sleep’s effect on memory performance.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

It is becoming clear that specific stages of sleep contribute to the processing of different classes of memory. Several studies clearly suggest a role for rapid eye movement (REM) sleep in the processing of procedural/perceptual memories (Karni et al., 1994; Smith and Lapp, 1991; Stickgold et al., 2000a,b), while more recent findings shed light on the mnemonic significance of non-REM (NREM) sleep. Specifically, periods of sleep rich in slow wave sleep (SWS) are associated with enhanced recall of declarative information (Gais and Born, 2004; Peigneux et al., 2004; Plihal and Born, 1997, 1999; Takashima et al., 2006), while Stage 2 sleep is important for the processing of motor skill tasks, such as the motor sequence task (Walker et al., 2002, 2003) and the pursuit rotor task (Smith and Fazekas, 1997; Smith and MacNeill, 1994). These findings have led to a broad effort from multiple disciplines to describe more clearly the physiological mechanisms underlying sleep’s beneficial effect on memory.

Another experimental approach that has received less attention is the examination of the impact of sleep duration on memory performance. If one of the functions of sleep is to consolidate information acquired during wakefulness, then longer durations of sleep might be expected to impart greater performance gains. Thus far, separate studies have described the declarative and motor memory enhancements that follow a short daytime nap, a half night of sleep or a full night of sleep. However, the findings of these individual studies preclude a determination about the amount of sleep required to obtain optimal performance improvements. Interestingly, daytime nap studies, in which subjects obtain 45–90 min of sleep, have revealed sleep-dependent improvements on visual discrimination (Mednick et al., 2003), declarative memory (Tucker et al., 2006), procedural memory (Backhaus and Junghanns, 2006) and motor memory tasks (Nishida and Walker, 2007). Even daytime naps as short as 6 min have been shown to benefit memory more than an equivalent period of wakefulness (Lahl et al., 2008). Do these findings imply that, while sleep is important for memory formation, only small amounts are necessary to observe optimal benefits? If so, what is the value of larger quantities of sleep for memory processing? To assess the impact of longer durations of sleep on memory, the present study examined performance on a standard declarative memory task (semantically related paired associates) and a motor learning task (motor sequence learning) following a half night of sleep (∼3.5 h) and a full night of sleep (∼7.5 h). This manipulation reveals not only the relative benefit of a full night, compared to a half night, of sleep on memory performance, but also it also clarifies the effect of sleep duration across different memory domains. A number of studies have already found that sleep early in the night (the first 3.5 h), which contains most of the night’s SWS, promotes the recall of declarative information (Gais and Born, 2004; Plihal and Born, 1997), while sleep obtained late in the sleep period, which is imbued with REM sleep and Stage 2 sleep, does not appear to benefit recall of this type of information. Conversely, motor skill learning benefits most from Stage 2 sleep (Smith and MacNeill, 1994; Walker et al., 2002), approximately half of which is obtained in the second half of the sleep period. Based on these findings, we put forth the tentative hypotheses that declarative memory performance will receive a significant boost after the first 3.5 h sleep, but that additional sleep will not further enhance performance. Motor skill performance, on the other hand, should show more robust enhancement after a full night of sleep, compared to 3.5 h of sleep, by virtue of the additional Stage 2 sleep obtained during the second half of the night.

These hypotheses center on the idea that specific sleep stages are important for memory processing (i.e. SWS early in the night for declarative memory processing, and Stage 2 sleep throughout the night for motor skill enhancement). However, the mere passage of time also leads to changes in memory. Declarative memories tend to deteriorate over time, either through interference from incoming sensory information or newly learned information, or through natural decay processes. If the memory trace becomes weaker over increasing durations of sleep, one would expect performance to be affected detrimentally. However, if sleep serves to consolidate memories actively (Ellenbogen et al., 2006), then one might expect declarative memory performance to remain stable or even improve over longer periods of sleep. On the other hand, performance on non-declarative visual discrimination (Gais et al., 2000; Stickgold et al., 2000a,b) and motor memory tasks (Walker et al., 2003) has been shown to improve gradually over time, suggesting that a full night of sleep may impart greater motor skill enhancement than a half night of sleep. This study represents a first look at how the memory trace changes as a function of time spent asleep.

In addition to the effect of sleep duration on performance across declarative and procedural memory domains, the present study also set out to explore the relationship between subjects’ intelligence, ability to acquire the tasks prior to sleep, and over-sleep memory enhancements. A number of findings validate such an inquiry by revealing a link between strength of information acquisition and sleep-related memory processing. One study found that only subjects who acquired a simple motor task strongly (pursuit rotor) showed increased postacquisition sleep spindle density, which correlated with improved pursuit rotor performance 1 week later (Peters et al., 2007). In a related vein, it was revealed that subjects with higher intelligence quotient (IQ) scores performed better than subjects with lower scores on two procedural memory tasks during acquisition and exhibited higher postacquisition REM densities (Smith et al., 2004). Finally, a recent study examining verbal and non-verbal declarative memory task performance showed that the effect of a daytime nap on memory is modulated by the relative strength of information acquisition (Tucker and Fishbein, 2008). Specifically, sleep enhanced memory for each of three declarative memory tasks only if subjects performed in the top half of the sample during the training session.

While these findings suggest that general intelligence may be an important modulator of sleep-related memory processing, few studies have examined this relationship directly. One study found that intelligence, as assessed by the Raven’s Progressive Matrices Test, was not related to sleep-related memory enhancements (Schabus et al., 2006). However, when subjects were placed into high and low memory groups based on their scores on the Wechsler Memory Scale, they found that the high memory group benefited more from sleep, suggesting that trait memory skills are predictive of greater sleep-related performance gains.

This study adds to the findings of these studies by examining the relationship between subject intelligence, the subject’s ability to acquire information prior to sleep, and the extent to which intelligence is predictive of over-sleep performance enhancements.

Method

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Participants

Twenty-four undergraduate students (12 female, 12 male; mean age = 20.9 years) of diverse ethnic composition participated in the study. Eleven subjects were assigned randomly to the half-night sleep group and 13 to the full-night sleep group. Originally, there were 13 subjects in the half-night group. However, two of these failed to complete the study. All subjects were in good health, and medication-free. Subjects were instructed to abstain from alcohol and caffeine on the day of the intelligence testing session and the day of the overnight portion of the study. Subjects completed a sleep log to document their average bedtime, wake time and number of hours of sleep for the three nights prior to the overnight study. All subjects reported that habitual bedtimes were earlier than 24:00 hours. Subjects were paid and received course credit for their participation.

Procedures

Subjects were scheduled for a 1.5-h test session prior to the overnight study to acclimate them to the laboratory environment and to take the Multidimensional Aptitude Battery-II (MAB-II) (Jackson, 1998), a 10-subtest multiple choice test of general intelligence. Several subtests of the MAB-II are similar in content to those of the WAIS III-R, and the performance, verbal and full scale IQ values correlate strongly with the WAIS III-R scales (Jackson, 2003). The MAB-II is a timed test in which subjects are allowed 7 min to complete each subtest.

Upon arrival at the sleep laboratory subjects signed consent and were then seated in a well-lit, sound-attenuated room (one of the sleep chambers in the sleep laboratory). A researcher gave instructions for each subtest over the intercom, and all subjects were video monitored during the test session to ensure compliance with test instructions. The researcher signaled when to start and stop each 7-min test. After subjects completed the test, they were scheduled for the overnight part of the study.

On the day of the overnight study subjects arrived at the sleep laboratory at 21:30 hours. Subjects completed a demographic questionnaire and the Epworth Sleepiness Scale (ESS) (Johns, 1991). After completing the forms, nine electrodes were applied to monitor sleep [central electro-encephalogram (EEG) (C3-A2, C4-A1), electro-oculogram (EOG) and chin electromyogram (EMG)]. Sleep data were acquired using Grass-Telefactor Gamma software (Grass Technologies, West Warwick, RI, USA), and were stage-scored according to standard criteria. Spectral analyses were performed using Hanning windowing on all artifact-free 30-s epochs. The EEG was sampled at 128 Hz and was filtered between 0.5 and 35 Hz. Averaging over EEG channels C3 and C4, absolute power in the theta (4–8 Hz), delta (1–4 Hz) and sigma (12–15 Hz) bands was analyzed.

After electrode application, subjects relaxed and watched television before the learning session. At 22:45 hours subjects sat at a computer in a sound-attenuated room. The Stanford Sleepiness Scale (SSS) was administered to assess level of sleepiness at that moment. Subjects were then administered the digit span task to assess general level of concentration and to prepare them mentally for the subsequent learning tasks. Following the digit span task, the paired associates task and motor sequence learning were administered in a balanced fashion across subjects. After performing each task subjects informed the researcher, who instructed them on the next task. After the training session, which lasted approximately 40 min, subjects remained awake for another 15–20 min, during which time they were informed about when they would be awakened, either halfway into the sleep period or after a full night of sleep. Subjects in the half-night condition were told that they would be able to sleep several hours before being awakened.

Prior to retest, subjects were awakened either directly following an arousal from sleep, or from Stages 1 or 2 sleep. Subjects were never awakened from SWS or REM sleep. After awakening, electrodes were removed and subjects were instructed to wash their faces to remove any residual electrode paste. The retest session took place at least 30 min after awakening to eliminate the effects of sleep inertia.

At retest, each subject completed the SSS and then completed each memory task in the same order as during baseline training. After the retest session subjects completed the Student Opinion Scale (SOS; Wolf and Smith, 1995), which surveys subjects’ interest in their performance and general motivation regarding the study. After completion of the SOS subjects were paid for their participation. Subjects in the half-night sleep group were then allowed to sleep out the rest of the night in the laboratory, while full-night sleep subjects were allowed to leave the laboratory.

Memory tasks

Paired associates task

Forty-eight semantically related word pairs (e.g. clock-hands) were selected from a larger pool of word pairs used by Gais and Born (2004), and have been used in a previous study from our laboratory (Tucker et al., 2006). Eight word pairs (four at the beginning and four at the end) were excluded from the response phase to eliminate primacy/recency effects. Word pairs in this task are presented on a 15′′ VGA monitor for 5 s each with a 100-ms interstimulus interval (ISI). Immediately following presentation of the word pair list subjects are shown, in random order, the first word of each of the 40 word pairs (minus the four at the beginning and the four at the end) and are asked to type the word that completes the pair. After each response is entered, the correct answer is displayed for 2 s. At retest subjects are shown the same 40 target words in a different random order, and are asked to type the word that completes the word pair. Performance is measured as the number of correctly completed word pairs, while improvement is measured as number of word pairs recalled at retest minus number recalled at baseline training. This improvement score is also calculated as a percentage improvement over the original baseline score (no. correct at retest–no. correct at training/no. correct at training).

Motor sequence task

The motor sequence learning task used in the present study was modeled on the task used by Walker (Walker et al., 2002, 2003). A five-digit series of numbers (4-1-3-2-4) is displayed 20 times in the center of a 15′′ VGA monitor. Subjects can advance to the next number only by entering the correct number. After each 20-sequence trial is completed, time (s) and accuracy (no. incorrect keystrokes per 20-sequence trial) are calculated. To ensure optimal performance subjects are instructed to type as fast as they can without compromising accuracy. Twelve trials are performed during training, and three are performed at retest. Performance is measured as the difference between the average of the last three trials of the training session and the three trials of the retest session.

Digit span

The digit span test is based on the WAIS-III subtest, using only the forward learning part of the test. It was used in this study as a measure of attention prior to training on the paired associates task and motor sequence task, and to acclimate subjects to the testing environment. Numbers in this task are presented serially on a 15′′ VGA monitor for 1 s each followed by a 1-s ISI. After each series is presented, subjects write down on a response sheet the numbers in the order they were presented. The first test series consists of three numbers with each series thereafter increasing by one number until the last series, which contains 10 numbers. After responding to the last series of 10 numbers, the task is performed a second time using different number sequences. The digit span task is always administered first. The same task is administered at retest using different number series. Performance is measured as the average number of correctly recalled number series during training and retest.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Presleep variables

The half-night and full-night sleep groups scored similarly on the verbal, performance and full IQ scales MAB-II (see Table 1). Sleep log data for the three nights prior to the overnight session revealed no differences between the half-night and full-night sleep groups for time awake prior to the training session [half-night 14.7 ± 0.44 h, full-night 14.0 ± 0.44 h (mean ± SEM, t(22) = 1.06, = 0.30)] or average total sleep time for the three nights prior to the day of the study (half-night 7.5 ± 0.34 h, full-night 7.4 ± 0.32 h, t(22) = 0.25, = 0.80). Sleepiness prior to sleep was similar between groups [SSS: half-night 3.2 ± 0.31, full-night 2.8 ± 0.23, t(22) = 1.20, = 0.24]. Sleepiness did not change significantly from presleep to pre-retest for either group (SSS: paired t-tests, half-night +0.2, > 0.7, full-night −0.5, = 0.13; time (training versus retest) × sleep condition (half-night versus full-night) interaction, (F(1,22) = 1.67, = 0.21). General sleepiness, assessed by the Epworth Sleepiness Scale, did not differ between groups (half-night 7.8 ± 0.91, full-night 6.8 ± 0.65, t(22) = 0.96, = 0.35), and sleep onset latencies were also similar [(SL): half-night 16.0 ± 6.3, full-night 18.4 ± 3.9, t(22) = 0.33, = 0.74). There was no difference in general motivation level, assessed by the SOS (half-night 40.9 ± 1.00, full-night 39.5 ± 1.10, t(22) = 0.98, = 0.34).

Table 1.   IQ scores for the half-night and full-night sleep groups
 Sleep condition
Half-night sleep groupFull-night sleep groupt(22)P
  1. IQ, intelligence quotient.

  2. IQ values are presented as mean ± SEM.

Verbal IQ103.7 ± 3.997.0 ± 3.41.30.20
Performance IQ99.8 ± 4.3102. ± 3.10.440.67
Full scale IQ102.1 ± 4.199.5 ± 3.20.500.62

Sleep data

Sleep EEG data for one subject were not included in the sleep analyses due to technical problems. A comparison of relevant sleep variables for the half-night and full-night sleep groups is presented in Table 2. The average total sleep time (TST) for the half-night group was approximately 3.5 h, while the full-night sleep group obtained nearly 7.5 h of sleep. In keeping with previous studies, subjects in the half-night sleep group obtained a large proportion of SWS (31.2%) relative to REM sleep (9.7%). The full-night sleep group obtained approximately equal proportions of Stage 2 sleep in the first half (50%) and second half (53%) of the night.

Table 2.   Sleep parameters
 Sleep condition
Half-night sleep groupFull-night sleep group
Minutes ± SD% of TST ± SDMinutes ± SD% of TST ± SD
  1. TST, total sleep time; SE, sleep efficiency; SL, latency to sleep onset (first epoch of sleep); WASO, wake after sleep onset; S1–S4, Stages 1–4; SWS, slow wave sleep; REM, rapid eye movement sleep; SD, standard deviation.

TST211.5 ± 19.4 449.4 ± 27.1 
SE89.0 ± 9.0 92.2 ± 4.0 
SL16.0 ± 21.1 18.2 ± 14.8 
WASO7.8 ± 8.9 19.2 ± 10.4 
S112.0 ± 4.15.8 ± 2.429.3 ± 9.66.5 ± 2.0
S2111.9 ± 17.953.3 ± 9.3231.3 ± 15.751.5 ± 3.4
S333.2 ± 19.015.5 ± 8.449.4 ± 15.211.0 ± 3.4
S433.4 ± 24.115.7 ± 11.024.4 ± 17.85.4 ± 3.9
SWS (S3 + S4)66.6 ± 21.231.2 ± 8.873.8 ± 16.416.4 ± 3.6
REM20.5 ± 11.59.7 ± 5.1 115.1 ± 21.825.5 ± 4.1

Task performance

Paired associates task

Performance on the semantically related paired associates prior to sleep was similar between groups (half-night 21.3 ± 1.9, full-night 20.6 ± 2.2, t(22) = 0.31, = 0.22). Recall at retest increased by 8.3 word pairs in the half-night group (paired samples t-test, t(10) = 5.8, < 0.001), and by 8.6 word pairs in the full-night group (paired samples t-test, t(12) = 6.7, < 0.001). This increase in paired associates recall was very similar between groups for raw score improvement (half-night 8.3 ± 1.4, full-night 8.6 ± 1.3; time × sleep condition interaction, F(1,22) = 0.032, = 0.86, ηp2 = 0.001; Fig. 1) and percentage improvement over pre-sleep training (half-night 45.3 ± 9.0%, full-night 53.2 ± 11.2%, t(22) = 0.54, = 0.60).

image

Figure 1.  Paired associates performance. Bars represent means ± SEMs.

Download figure to PowerPoint

Motor sequence task

At training, average time per trial (speed in seconds) (half-night 29.6 ± 2.2, full-night 29.5 ± 1.6, t(22) = 0.04, = 0.97) and accuracy (number of errors per trial; half-night 3.1 ± 0.9, full-night 2.5 ± 0.5, t(22) = 0.64, = 0.53) were similar between groups. Both groups demonstrated a significant over-sleep improvement in speed (paired samples t-tests, < 0.001), but showed non-significant improvement in accuracy (paired samples t-test: half-night, = 0.14; full-night, = 0.08). Over-sleep improvement was nearly identical between groups for improvement in speed (half-night 5.4 ± 1.0 s, full-night 5.2 ± 1.2 s; time × sleep condition interaction, F(1,22) = 0.01, = 0.92, ηp2 = 0.0004; Fig. 2) and percentage improvement in speed (half-night 17.3 ± 3.2%, full-night 16.7 ± 3.7%, t(22) = 0.13, = 0.90). Improvement in accuracy was also similar between groups (half-night 1.7 ± 1.0, full-night 0.8 ± 0.4, F(1,22) = 0.61, = 0.44, ηp2 = 0.03). In the full-night sleep group amount of Stage 2 sleep correlated with improvement in speed (= 0.57, = 0.04; Fig. 3). However, amount of Stage 2 sleep was correlated only weakly with improvement in speed in the half-night sleep group (= 0.20, = 0.55).

image

Figure 2.  Performance on the motor sequence task. Bars represent means ± SEMs.

Download figure to PowerPoint

image

Figure 3.  Correlation between amount of stage 2 sleep and over-sleep improvement in speed on the motor sequence task in the full night sleep group.

Download figure to PowerPoint

Digit span

Half-night and full-night sleep subjects recalled a similar number of correct sequences during training (half-night 4.8 ± 0.43, full-night 4.3 ± 0.43, t(22) = 0.90, = 0.38). Change in over-sleep performance was similar for the half-night and full-night groups (half-night −0.05 sequences, full-night +0.5 sequences, sleep group × time interaction, F(1,22) = 2.20, = 0.15, ηp2 = 0.09) (Fig. 3).

Relationship between intelligence and task performance

Paired associates task

Across all subjects acquisition of paired associates correlated with verbal (= 0.41, = 0.05), performance (= 0.45, = 0.03) and full scale IQ (= 0.47, = 0.02; Fig. 4a). Paired associates recall at retest also correlated with verbal (= 0.49, = 0.02) and full scale IQ (= 0.44, = 0.03). However, over-sleep recall enhancement did not correlate significantly with IQ measures (all P-values >0.17; Fig. 4b).

image

Figure 4.  Correlations between full scale IQ and task performance. (a) Paired associates recall at training. (b) Over-sleep improvement in paired associates recall. (c) Motor sequence speed per trial at training (in seconds). (d) Over-sleep motor skill improvement. Paired associates improvement is defined as correct recall at retest minus correct recall at training, while motor sequence improvement is defined as the average speed across the last three training trials minus the average speed per trial across the three retest trials.

Download figure to PowerPoint

Motor sequence task

Across all subjects speed on the motor sequence task during training correlated marginally with full scale IQ (= −0.35, = 0.09), and at retest speed correlated with full scale IQ (= −0.42, = 0.04; Fig. 4c) and performance IQ (= −0.42, = 0.04). Over-sleep improvement in speed did not correlate with any IQ measures (all P-values > 0.6; Fig. 4d).

Relationship between EEG power spectra, task performance and intelligence

Power spectral analysis revealed a highly significant correlation between absolute power in the sleep spindle range (sigma: 12–15 Hz) and over-sleep improvement in speed on the motor sequence task for the half-night group (= 0.77, = 0.005; Fig. 5), but not for the full-night group (= 0.28, = 0.38). The strength of the half-night correlation was equally robust across all NREM sleep (Stage 1: = 0.62, = 0.04; Stage 2: = 0.78, = 0.005; SWS: = 0.73, = 0.01). Spectral power in the theta band was also highly correlated with over-sleep improvement in motor sequence speed in the half-night group (r = 0.90, < 0.001), but not for the full-night group (= 0.16, = 0.62).

image

Figure 5.  Correlation between sigma power (12–15 hz) and over-sleep improvement on the motor sequence task in the half night sleep group.

Download figure to PowerPoint

Correlations between all spectral bands and measures of intelligence (i.e. full scale, verbal and performance IQ) were non-significant (all P-values >0.45), as were correlations between spectral power and paired associates performance (all P-values >0.1).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study set out to address two central issues by assessing the effect of sleep duration on processing of declarative and motor memory tasks, and by examining the relationship between subject intelligence, task acquisition facility and memory enhancement following sleep.

For the semantically related paired associates task the half-night and full-night subjects showed a strikingly similar degree of over-sleep improvement, confirming our hypothesis that the first 3.5 h of sleep, which are dominated by SWS, confer the greatest performance enhancement, while additional hours of sleep do not add to this initial level of performance. Interestingly, the paired associates enhancement demonstrated by both groups in this study (8.3 word pairs for the half-night sleep group and 8.6 for the full-night sleep group) was comparable to a previous study that found an 8.75 word pair increase in recall across a 6-h time interval in subjects who took a short daytime nap (∼45 min), compared to a 6.5 word increase in subjects who did not nap (Tucker et al., 2006).

As with the paired associates task, the results for the motor sequence task again demonstrated very similar reductions in time per trial at retest (17.3% for the half-night sleep group and 16.7% for the full-night sleep group), which is surprising in light of previous findings that show a relationship between amount of Stage 2 sleep and over-sleep motor skill enhancement (Smith and MacNeill, 1994; Walker et al., 2002).

As observed with the performance improvement on the paired associates task, the percentage gains in motor skill speed observed in the half-night and full-night sleep groups are very similar to findings from a recent nap study demonstrating a 16% improvement in speed across an 8-h time interval that contained a 60-min nap, compared to a non-significant 3.8% gain in the no-nap condition (Nishida and Walker, 2007). These observations suggest that while sleep is important for processing of memories from multiple domains, greater amounts of sleep may not be required for optimal performance. Importantly, this pattern of findings also generalizes to perceptual learning tasks (visual texture discrimination), in which performance following a 90-min daytime nap containing REM sleep is equivalent to the performance gains observed after a full night of sleep (Mednick et al., 2003).

In keeping with findings establishing a relationship between the amount of Stage 2 sleep and motor memory performance (Walker et al., 2002), we found that greater amounts of Stage 2 sleep in the full-night sleep group correlated with over-sleep performance improvement on the motor sequence task, while we found only a weak positive correlation between these variables in the half-night group. However, when we looked at spectral power in the spindle frequency range (sigma; 12–15 Hz), we found the inverse was true: there was a strong correlation between sigma power and motor memory performance in the half-night sleep group, but not the full-night sleep group, suggesting that sigma power in the half-night group may represent a mechanism by which performance was boosted to the level observed in the full-night group. However, an alternative interpretation would suggest that performance on this task is facilitated by a unique combination of spindle activity and amount of Stage 2 sleep during the night. Indeed, Nishida and Walker (2007) found that the amount of Stage 2 sleep, as well as the numerical difference between spindle activity generated in task-related brain regions (regions contralateral to the hand used for the task) and ipsilateral (non-learning) regions were associated with greater over-sleep motor skill performance. Another study showed that both amount of Stage 2 sleep and spindle density increased during sleep following intense training on a simple motor task (Fogel and Smith, 2006). Taken together, these findings suggest that Stage 2 sleep duration, in conjunction with spindle activity, may represent a mechanism by which experience-dependent plastic changes in neocortical circuits might be achieved, as has been theorized previously (Sejnowski and Destexhe, 2000). Interestingly, we also found that spectral power in the theta frequency band was also correlated highly with motor memory performance in the half-night sleep group, but not the full-night group. A clear interpretation of the importance of theta frequency EEG activity for motor memory performance is difficult to formulate. However, the theta rhythm (4–8 Hz), which is another hallmark feature of the Stage 2 sleep EEG, suggests the possibility that this rhythm, in addition to spindle activity and Stage 2 duration, may each contribute uniquely, or in combination, to motor memory processing.

While changes in declarative and motor memory performance following varying amounts of sleep are very similar following a half night and full night of sleep, the influence of the mere passage of time on memory processing must be considered. In the case of declarative memory, the passage of time has a detrimental effect on memory, either due to interference from information acquired after task training (Wixted, 2005), or natural decay processes. If these time-modulated processes act to erode memories, then it is impressive that the sleeping brain appears to sustain the memory trace across a full night of sleep. This suggests that sleep may play an active role in the maintenance or consolidation of the memory trace in a way that counteracts factors that would have a much more negative influence over equivalent durations of wakefulness (Ellenbogen et al., 2006). Conversely, with the motor sequence task sleep-dependent performance enhancements emerge after a brief daytime nap relative to an equal amount of wakefulness (Nishida and Walker, 2007), and these sleep–wake differences appear to remain stable across a 12-h time interval. It may be that sleep immediately following learning is most important for motor skill consolidation, but that slower sleep-dependent processes continue to exert a beneficial influence over subsequent nights of sleep (Walker et al., 2003). However, the precise contribution of sleep and wake neurophysiology across lengthier intervals (>12 h) has yet to be adequately characterized for declarative and motor memories.

The decision to examine of subject intelligence as a modulator of the effect of sleep on memory was derived from a number of intriguing studies showing that subjects who acquire information more strongly experience greater sleep-related performance gains. Here, we aimed to tap into the relationship between intelligence, task acquisition and over-sleep memory enhancement. Interestingly, while intelligence was predictive of heightened training and retest performance on the declarative and motor learning tasks (as well as the digit span task), intelligence did not predict a greater degree of over-sleep memory enhancement. This suggests that, while some subjects may perform better during task training, general intelligence, per se, may not be a trait that permits the sleeping brain to better process these memories. It is conceivable that factors such as the subject’s level of motivation to perform well, or the emotional salience of the task, may facilitate task acquisition, which would then translate to an augmentation of sleep-related memory enhancement.

Overall, this study draws into question the idea that extended periods of sleep are necessary for optimal sleep-related memory gains. The findings also suggest that, while intelligence is associated clearly with ability to acquire declarative and procedural information, it does not appear to be a subject variable that modulates over-sleep memory enhancement appreciably. Future research should attempt to identify subject variables and task characteristics that not only enhance acquisition, but also lead to more pronounced performance improvements following sleep.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study was supported by PSC-CUNY Grant no. 67648-00 36 to W. F. Special thanks to Erin Wamsley for insightful comments and suggestions on the manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References