Changes in the density of stage 2 sleep spindles following motor learning in young and older adults

Authors


Kevin R. Peters, PhD, Department of Psychology, Trent University, Peterborough, ON K9J 7B8, Canada. Tel.: +1-705-748-1011 x.5395; fax: +1-705-748-1580; e-mail: kevinpeters@trentu.ca

Summary

The purpose of this study was to compare the changes that occur in sleep architecture following the acquisition of a simple motor learning task in young and older adults. Subjects included 14 young (range = 17–24 years) and 14 older (range = 62–79 years) adults, all of whom were in good health. Using in-home recording systems, sleep architecture (sleep stages and the density of Stage 2 sleep spindles) was examined before and after learning the pursuit rotor. To control for possible age differences in baseline motor performance and spindle density, both absolute and relative (percent change) measures were examined. Both groups improved significantly on the pursuit rotor task at Retest (1 week later); however, the magnitude of absolute improvement was larger in the young group than in the older group. There was no group difference when a relative measure of improvement (percent increase across sessions) was used. The density of Stage 2 sleep spindles increased significantly following task Acquisition in the young group but not in the older group. These age differences failed to reach significance when change was measured as a percentage of baseline level of spindle density. The increase in spindle density was correlated with performance level during acquisition in the young group but not the older group. The results of the present study are largely consistent with previous studies on sleep and memory in young adults and suggest that more detailed examination of this relationship in older adults is warranted.

Introduction

The sleep architecture of older adults differs in a number of ways from that of young adults (Bliwise, 2000). In a recent meta-analysis (Ohayon et al., 2004), it was reported that there are significant age-related decreases in total sleep time, sleep efficiency, the percentage of slow-wave sleep, and the percentage of REM sleep. On the other hand, there are significant age-related increases in sleep latency, the percentage of Stage 1 sleep, the percentage of Stage 2 sleep, and the amount of time spent awake after sleep onset. At the micro-level of phasic events, one particularly consistent finding that is relevant to this investigation is an age-related reduction in the number and density of sleep spindles (Crowley et al., 2002; Guazzelli et al., 1986; Nicolas et al., 2001; Wauquier, 1993). Compared to younger adults, sleep spindles in older adults also become shorter in duration (Crowley et al., 2002; Guazzelli et al., 1986; Nicolas et al., 2001), lower in amplitude (Crowley et al., 2002; Guazzelli et al., 1986; Principe and Smith, 1982), and the mean frequency within the sleep spindle increases (Crowley et al., 2002; Guazzelli et al., 1986; Nicolas et al., 2001; Principe and Smith, 1982). Despite these overall age-related differences, it has also been pointed out that there is considerable inter-individual variability in the number of spindles amongst older adults, and that the number of spindles in some older adults fall within the same range as young adults (Guazzelli et al., 1986; Wauquier, 1993). Although the cause of age-related changes in sleep spindle parameters is unknown, it has been suggested that these changes may be due to decreases in thalamo-cortical integrity (Crowley et al., 2002; Guazzelli et al., 1986; Nicolas et al., 2001) or to decreases in plasma melatonin levels (Dijk et al., 1995; Landolt et al., 1996). The functional implications of age-related changes in sleep spindle parameters is also unknown; however, it has been suggested that these changes may help explain age-related difficulties with maintaining sleep (Crowley et al., 2002; Guazzelli et al., 1986; Nicolas et al., 2001).

Guazzelli and colleagues (Guazzelli et al., 1986) tested the interesting hypothesis that age-related changes in spindle density may be related to age-related changes in cognitive functioning and structural brain integrity. In other words, older adults that exhibit fairly high spindle densities may have better cognitive functioning and structural brain integrity than those older adults who have lower spindle densities. Using polysomnography, a wide range of cognitive and intellectual tasks, and computerized-tomography (CT) brain imaging, these authors failed to find any evidence to support this hypothesis. We feel, however, that there is one important limitation to this study that deserves mention. Guazzelli and colleagues examined correlations between the baseline, or native, characteristics of sleep spindles and cognitive functioning. That is, they examined sleep spindles from one night only and then correlated these spindle characteristics with performance on various cognitive tests. Although spindle density is usually quite stable from night to night (Gaillard and Blois, 1981), we and others have shown recently that the density of Stage 2 spindles increases significantly after learning several simple motor tasks in young adults (Fogel and Smith, 2006; Fogel et al., 2007a,b; Milner et al., 2006; Nishida and Walker, 2007; Peters et al., 2007). In addition, sleep spindles have been reported to increase following declarative learning tasks (Clemens et al., 2005, 2006; Gais et al., 2002; Schabus et al., 2004, 2006) and have been associated with general mental ability (Bodizs et al., 2005; Fogel et al., 2007a,b; Schabus et al., 2006) as well. The paradigm in this study involved recording spindle activity on two nights: on a baseline night and then again after learning the motor tasks. The finding of an increase in Stage 2 spindle density after learning is particularly interesting because it has been proposed that sleep spindles may represent an ideal mechanism for promotion of synaptic plasticity (Destexhe and Sejnowski, 2001; Steriade, 1999).

The main objective of the present study was to compare the changes that occur in sleep spindle densities following the acquisition of a simple motor learning task in young and older adults. The present study differs from that of Guazzelli and colleagues (Guazzelli et al., 1986) in two important ways. First, we recorded sleep before and after learning a motor task to determine whether Stage 2 sleep spindle density changes following task acquisition. Second, we used a simple motor task – the pursuit rotor – rather than the high-level cognitive and intelligence tests that were used by Guazzelli and colleagues. As mentioned above, previous findings have suggested that Stage 2 sleep, and the sleep spindles in this stage, are particularly important for consolidating motor procedural tasks (Fogel and Smith, 2006; Fogel et al., 2007a,b; Milner et al., 2006; Nader and Smith, 2003; Nishida and Walker, 2007; Peters et al., 2007; Smith and MacNeill, 1994).

It was predicted that both the young and older groups would show significant improvement on the pursuit rotor task when retested 1-week after acquisition but the amount of improvement would be greater in the young group than in the older group. Similarly, we also predicted that both age groups would show a significant increase in Stage 2 spindle density after learning the pursuit rotor task, but the magnitude of this increase would be greater in the young group than in the older group. Given the large inter-subject variability in spindle density that has been reported for both young and older adults, and the fact that in some older adults spindle density remains as high as in young adults, we performed a correlation analysis to examine the relationships among the Stage 2 sleep variables and the pursuit rotor performance variables.

Methods

Subjects

A total of 28 subjects took part in the present study. There were 14 (7 female) subjects in the young adult group (mean age = 20.14 ± 2.25 years; range = 17–24) and 14 (7 female) subjects in the older adult group (mean age = 69.79 ± 5.13 years; range = 62–79). Young adults were university students at Trent University and received $75, plus research credit for one of their courses if applicable, for participating. The older adults were recruited from the local community and received $100 for participating. All subjects indicated that they were in good health and appeared physically able to take part in the study. Exclusion criteria at the outset of the study included individuals with irregular or unusual sleeping patterns, medical conditions and/or medications as well as conspicuous substance abuse that would disrupt cognitive performance or sleep architecture. In addition, we used the Sleep Disorders Questionnaire (Douglass et al., 1994) and the Beck Depression Inventory (Beck, 1987) to rule out subjects with potential sleep disorders and depression, respectively. Although we did not do a full polysomnographic assessment to rule out sleep disorders, we did review the first night of collection for any obvious signs of increased EEG arousals as a crude way to screen out possible sleep disturbances. An attempt was made to match the young and older subjects as closely as possible on the Multidimensional Aptitude Battery – II (Jackson, 1998), which is an IQ test that is similar to and highly correlated with the Wechsler Adult Intelligence Scale-Revised. We matched participants on the basis of IQ because previous studies have shown that sleep spindle activity is associated with general ability or IQ (Bodizs et al., 2005; Fogel et al., 2007a,b; Schabus et al., 2006), and that changes in sleep architecture that occur following learning are affected by the IQ level of the participants (Smith et al., 2004). T-test results revealed that the young and older groups did not differ on their mean Full-Scale IQ score (108.61 ± 11.92 versus 111.71 ± 20.65), their mean Verbal IQ score (106.29 ± 12.25 versus 115.00 ± 20.03), or their mean Performance IQ score (110.43 ± 13.20 versus 111.79 ± 16.65; P > 0.10 for all). This research project was approved by the Trent University Research Ethics Board. All subjects gave their informed consent to participate.

Measurements

The pursuit rotor task was used as a measure of procedural motor learning. Subjects performed this task by tracking a moving light on a photoelectric device with hand-held stylus. An electronic counter was used to determine how long each subject was able to keep the stylus on target during each 30-s trial (as determined by the experimenter using a stop watch). The light moved around a rectangular pattern in a clockwise direction at a rate of 32 rpm. Subjects performed 30 trials with their non-dominant hand. Each trial lasted for 30 s and trials were grouped into blocks of five with a brief pause in between each block. The main variable of interest for the pursuit rotor was the total time on target (TOT) for the first 30 trials (Acquisition session) and the second set of 30 trials 1 week later (Retest session).

In-home sleep recordings were performed using Suzanne™ (Tyco Healthcare Group LP, Mansfield, MA, USA) portable polysomnographic systems. These systems record physiological data at a sampling rate of 120 Hz onto 32 MB PC flash memory cards. All data were then downloaded off-line onto a PC computer for subsequent analyses. We recorded EEG, EOG (horizontal only), and EMG using silver-plated electrodes. The EEG (C3 and C4) and EOG (left and right eyes) were monopolor recordings and were referenced to electrodes placed onto the contralateral mastoid bones (A1 and A2). The EMG channel was bipolar. Low- and high-pass software filters for the EEG and EOG were set at 0.03 Hz and 30 Hz. The low-pass software filter for the EMG channel was set at 10 Hz and no high-pass was used (i.e. only frequencies above 10 Hz were recorded).

Sleep stages were scored according to standard criteria (Rechtschaffen and Kales, 1968) using Sandman™ software (Version 7.2; Melville Diagnostics, Ottawa, ON, USA). The number of minutes is reported for each of the following measures: time spent awake after sleep onset, time spent in Stage 1 sleep, time spent in Stage 2 sleep, time spent in slow-wave sleep (SWS; Stages 3 and 4 combined), time spent in REM sleep, and the total sleep time (TST). Note that because it is difficult to precisely record a ‘lights out’ time with the Suzanne™ units, we do not report sleep onset latency nor do we report the total recording time (i.e. we do not include the time between lights out and sleep onset). We also report each of the above measures as a percentage of the TST.

Sleep spindles were counted visually in epochs of Stage 2 sleep from the C3 channel. The C4 channel was used on occasion when the signal from the C3 electrode was unusable (e.g. it had fallen off). We used the following criteria: spindles had to be in the 12–16 Hz frequency range, they had to be at least 0.5 s in duration, and they had to resemble the typical fusiform spindle morphology. The visual identification of sleep spindles was aided by displaying an additional copy of the EEG channel with frequencies above and below 12 and 16 Hz filtered. Stage 2 spindle densities were calculated by dividing the total number of spindles by the total number of minutes in Stage 2 sleep throughout the entire night.

One of two research assistants scored the data for sleep stage and sleep spindles. We computed inter-scorer agreement between the two technicians on a random sample of 10 nights. The correlations between the two scorers were above 0.95 for the number of Stage 2 sleep spindles as well as for the time spent in Stage 2, SWS, and REM. The correlation was slightly lower (0.93) for the time spent in Stage 1 sleep.

Procedures

All interested subjects were initially screened for eligibility in the study (see details above). In-home sleep recordings were performed for a total of three consecutive nights. Figure 1 depicts an overview of the study procedure. On the evening of each night, a research assistant would go to the subject’s home and hook-up the sleep recording apparatus. The skin was prepared using alcohol, Nuprep™ abrasive cleanser, and Ten-20™ conductive paste along with tape or gauze to affix the electrodes in place. The research assistant would then leave the home and return in the morning to pick up the recording unit and download the data onto a PC. The first night was an acclimatization night and was used to simply get the subjects used to wearing the recording apparatus. The second night served as the ‘baseline’ night. On the afternoon (approximately 4:00 pm) of the third day (i.e. after the baseline night), each participant performed the pursuit rotor for the first time. This first set of thirty 30-s trials comprised the ‘acquisition’ session and lasted for approximately 20 min. It is important to point out that all subjects were able to perform the task without any physical difficulty. The sleep was recorded later that night and these data served as the ‘post-training’ night which enabled us to determine whether there were any sleep-related changes as a result of the learning task. One week later, subjects completed the second set of thirty 30-s trials on the pursuit rotor (again, at approximately 4:00 pm). This set of trials was referred to as the ‘retest’ session and it allowed us to determine how well the subjects consolidated the motor task.

Figure 1.

 Overview of the study design. In-home polysomnographic recordings were performed on subjects for three consecutive nights. The first night was used as an acclimatization night and the data were discarded. At approximately 4:00 pm on the afternoon of the third night, subjects performed thirty 30-sec trials on the pursuit rotor task. One week later subjects performed an additional thirty 30-s trials on the pursuit rotor task. Using this design, we could determine whether task acquisition had any effect on sleep architecture, and whether such changes were correlated with changes in task performance.

Statistical analysis

Given the size of the samples in the present study and the fact that there is often considerable variability in older adults on many variables, an attempt was made to minimize the influence of outliers on subsequent analysis. Outliers were identified via box plots through the Explore analysis in SPSS (Version 11.5) and defined as cases with values greater than 1.5 box lengths (i.e. the interquartile range) from either the upper or lower edge of the box. Rather than delete these cases, we chose to minimize their influence by replacing their values with those corresponding to one unit above or below the highest or lowest data points falling within 1.5 box lengths from the edge of the box (i.e. the whiskers; Tabachnick and Fidell, 2001). This procedure is particularly beneficial for small samples because the influences of outliers are still present but they are reduced and there is no loss of statistical power. Out of a total of 1232 data values (44 variables × 28 subjects), 46 (3.73%) were modified using the above procedure.

A 2 (group: young, older) × 2 (session: acquisition, retest) mixed anova was carried out to assess whether the two age groups differed in their performance on the pursuit rotor task across sessions. Performance on the pursuit rotor was indexed by the total time on target (TOT) for the 30 trials of each session. Because it was likely that the two age groups would differ in their initial baseline performance on the pursuit rotor, an independent t-test was also carried out to assess whether the percent of change (TOT during retest – TOT during acquisition/TOT during acquisition × 100) differed between the two age groups. In addition, detailed analyses were conducted to determine whether there were age differences in the learning curves during each session. The 30 trials were grouped into six blocks of five trials each for the acquisition and retest sessions. Separate 2 (group: young, older) × 6 (block: trials 1–5, trials 6–10, trials 11–15, trials 16–20, trials 21–25, trials 26–30) mixed anovas were performed for the acquisition and retest sessions. The Greenhouse-Geisser correction for degrees of freedom was applied when the assumption of sphericity was violated (Note, although we adjusted the P-values accordingly when sphericity was violated, we report the uncorrected degrees of freedom).

To test for age differences in sleep architecture, a series of 2 (group: young, older) × 2 (night: baseline, post-training) mixed anovas were conducted for the following variables: total sleep time (TST), time awake after sleep onset (WASO), sleep efficiency, Stage 1, Stage 2, slow-wave sleep (SWS), and REM sleep. A 2 (group: young, older)  × 2 (night: baseline, post-training) mixed anova was performed to determine whether there were age differences in Stage 2 spindle densities before and after learning. An independent t-test was also carried out on the percent of change in Stage 2 spindle density (spindle density during post-training night – spindle density during baseline night/spindle density during baseline night × 100) to control for any age differences in baseline density that were likely to exist. Pearson correlation coefficients were calculated to assess the relationships among the following sleep and performance variables: baseline and post-training Stage 2 spindle density, percent change in Stage 2 spindle density, total time on target (TOT) during acquisition, TOT during retest, and the percent change in TOT. Correlation coefficients were computed separately for each age group.

Results

Pursuit rotor task

The performance of the two age groups on the pursuit rotor task during the acquisition and retest sessions is shown in Figure. 2. There was a main effect of group [F(1,26) = 12.94, P = 0.001], a main effect of session [F(1,26) = 57.27, P < .001], and a significant group × session interaction [F(1,26) = 8.24, P = .008]. The increase in mean TOT in the young group from the acquisition session (mean = 354.98, SD = 71.45) to the retest session (mean = 488.02, SD = 78.43) 1-week later was highly significant [t(13) = 9.19, P < .001]. The increase in mean TOT from the acquisition session (mean = 275.21, SD = 87.42) to the retest session (mean = 335.09, SD = 122.23) in the older group was also significant [t(13) = 2.85, P = 0.014]. In terms of the percent of change across the acquisition and retest sessions, the younger group (mean = 32.84%, SD = 7.88) did not differ significantly from the older group (mean = 27.12%, SD = 24.20) [t(26) = 0.84, P = 0.408].

Figure 2.

 Mean total time on target (TOT) for the pursuit rotor task during the acquisition and retest sessions for the young and older age groups; values are shown as mean ± standard error. Although the performance of both groups improved significantly across the two sessions, the magnitude of the improvement was larger in the young group compared with the older group.

Figure. 3 depicts the performance of the two age groups during the acquisition and retest sessions broken down into six blocks of five trials each. For the acquisition session, the main effects of group [F(1,26) = 7.70, P = 0.010] and block [F(5,130) = 60.08, P < 0.001] were both significant. The group × block interaction was also significant [F(5,130) = 3.35, P = 0.024], indicating that while both groups showed significant learning during acquisition, the magnitude of learning was greater in the young group than in the older group (i.e. the younger subjects demonstrated a steeper learning curve). Regarding performance 1 week later during retest, the main effects of group [F(1,26) = 15.90, P = 0.001] and block [F(5,130) = 9.39, P < .001] were both significant. The group × block interaction during the retest session failed to reach significance [F(5,130) = 1.21, P = 0.312]. These results suggest that although the magnitude of age differences was much larger during the retest session, the two age groups did not differ in their patterns of performance (i.e. there was no difference in the learning/performance curves).

Figure 3.

 Learning curves during the acquisition and retest sessions for young and older groups; values are shown as mean ± standard error. Regarding acquisition, both groups improved significantly over the six blocks of trials; however, the magnitude of the improvement was larger in the young group compared with the older group. For retest, once again, both groups showed significant improvement over the six blocks of trials, but there was no difference in the rates of improvement.

One additional point regarding retest performance is worth noting. Performance of both age groups during retest does not really improve much beyond the second or third block of trials. These data suggest that both age groups have reached a ‘functional ceiling’ in their performance early in the retest session. It should be noted, however, that this ceiling is not a methodological artifact because there is still room for improvement on the task. Even the average total time on target of 84.32 s (SD = 14.23) in the young group at the end of retest does not come close to the maximum possible score of 150 s. Performance during retest seems to represent the maximum level that is attainable by both groups with this training exposure.

Sleep architecture and spindle density

Descriptive data, as well as the anova results, for the sleep stage parameters of the two age groups are reported in Table 1. Compared to young subjects, those in the older group spent more time awake after sleep onset (WASO), and more time in Stage 1 sleep and in Stage 2 sleep (percent only). In contrast, compared to older subjects, those in the young group had higher sleep efficiencies and spent more time in SWS and in REM sleep. The only stage difference that occurred across nights was an overall reduction in Stage 1 sleep. A particularly interesting finding was the group × night interaction for SWS which suggested that older adults spent more time in SWS after learning the pursuit rotor task while there was a no real change in SWS for the young group. Post hoc testing revealed that this interaction was being driven by the changes in the older group: the increase in number of minutes in SWS and the increase in percent of SWS were both significant [t(13) = 2.68, P = 0.019 and t(13) = 2.90, P = 0.012, respectively]. In contrast, neither the change in minutes of SWS nor percent of SWS was significant in the young group [t(13) = 0.98, P = 0.344 and t(13) = 0.42, P = 0.682 respectively].

Table 1.   Sleep parameters for young and older groups on baseline and post-training nights
VariableYoung group (n = 14)Older group (n = 14)anova results*
BaselinePost-trainingBaselinePost-trainingNightGroupInteraction
  1. Values are means ± SD unless otherwise indicated. TST refers to total sleep time; WASO, wakefulness after sleep onset; SWS, slow-wave sleep (Stages 3 and 4 combined); REM, rapid eye-movement.

  2. *F-values are reported with 1 and 26 d.f.

  3. *P < 0.05, **P < 0.01, ***P < 0.001.

TST (min)483.12 ± 35.32471.58 ± 83.57445.43 ± 71.7460.09 ± 92.190.020.921.43
WASO (min)20.90 ± 11.6917.87 ± 16.1645.71 ± 28.9645.61 ± 25.230.1016.71***0.08
WASO (%)4.51 ± 2.553.58 ± 2.9110.40 ± 6.099.94 ± 4.920.4024.71***0.05
Sleep efficiency (%)94.06 ± 3.0595.47 ± 3.5488.54 ± 5.8389.59 ± 4.741.1821.40***0.03
Sleep stage (min)
 16.71 ± 4.962.97 ± 2.2325.68 ± 14.4319.54 ± 9.019.89**35.34***0.59
 2230.07 ± 33.69230.99 ± 37.99255.89 ± 58.81250.54 ± 52.350.092.000.18
 SWS100.00 ± 29.1795.43 ± 23.5122.21 ± 24.4933.68 ± 37.401.2343.52***6.67*
 REM111.25 ± 22.57120.26 ± 47.6985.71 ± 26.1497.09 ± 36.812.604.49*0.04
Sleep stage (%)
 11.44 ± 1.050.57 ± 0.375.78 ± 2.814.13 ± 1.5513.80**52.28***1.37
 246.90 ± 7.3349.24 ± 4.5358.29 ± 9.4155.54 ± 8.740.0211.82**3.38
 SWS21.11 ± 6.0620.84 ± 6.315.33 ± 6.028.00 ± 9.093.0931.40***4.64*
 REM23.86 ± 4.9025.10 ± 7.7619.45 ± 4.6120.85 ± 6.371.145.07*<0.01

Fig. 4 shows the Stage 2 spindle density measures for both age groups. There was a significant main effect of group [F(1,26) = 36.45, P < 0.001], a significant main effect of night [F(1,26) = 7.45, P = 0.011], and a significant group × night interaction [F(1,26) = 5.30, P = 0.030]. These results indicate that, overall, spindle density was greater in the young group than in the older group, and that spindle density was greater on the post-training night than on the baseline night. The presence of a significant group × night interaction, however, indicates that the overall increase in spindle density from the baseline night to the post-training night was driven by the young adults. The increase in spindle density from the baseline night (mean = 6.02, SD = 1.59) to the post-training night (mean = 6.71, SD = 1.58) in the young group was significant [t(13) = 3.30, P = 0.006]. However, the change in mean spindle density from the baseline night (mean = 2.51, SD = 1.72) to the post-training night (mean = 2.57, SD = 1.95) in the older group failed to reach significance (P > 0.70). In terms of the percent change in spindle density, the difference between the young (mean = 13.00%, SD = 15.57) and the older (mean = 2.02%, SD = 32.77) groups did not reach significance [t(26) = 1.13, P = 0.267). The standard deviations in each group underscore the individual variability in the changes in spindle density across the two nights, especially in the older group.

Figure 4.

 Stage 2 sleep spindle density during baseline and post-training nights for the young and older groups; values are shown as mean ± standard error. Spindle density increased significantly in the young group but not in the older group following task acquisition.

Correlation analyses

Table 2 contains the results of the correlation analyses among Stage 2 spindle variables and pursuit rotor variables. Correlation coefficients above the diagonal are for the young group only; whereas coefficients below the diagonal are for the older group only. As one would expect, the correlations between the density of spindles for both nights is significant in the young group [r(12) = 0.878, P < 0.001] and in the older group [r(12) = 0.941, P < 0.001] as are the correlations between the TOT during both sessions [r(12) = 0.742, P = 0.002; r(12) = 0.768, P = 0.001, respectively]. There are several other interesting findings in Table 2. First, the correlation between TOT during acquisition and the percent change in spindle density is significantly positive in the young group [r(12) = 0.619, P = 0.018] but not in the older group, where it is actually in the opposite direction (r = −0.452, P = 0.105). These scatter plots are shown in Fig. 5.

Table 2.   Correlation matrix for young (n = 14) and older (n = 14) adults Thumbnail image of
Figure 5.

 Scatterplot showing the correlation between TOT during acquisition and the percent change in spindle density in the young group (top panel) and in the older group (bottom panel).

Second, the direction of the correlation between the percent change in spindle density and the percent change in TOT is different in the young group [r(12) = −0.521, P = 0.056] than in the older group [r(12) = 0.508, P = 0.064). Because these two correlations came very close to reaching significance, we have included the scatter plots of these correlations in Fig. 6 to get a sense of the relationships and to determine whether it would be useful for future research to follow-up on these findings. These results suggest that larger increases in spindle density after learning are associated with performance improvements in the older group but performance decrements in the young group. The negative correlation found in the young group seems counterintuitive at first, but may be explained by the fact that TOT during acquisition is significantly and negatively correlated with the amount of improvement across sessions in the young group [r(12) = −0.798, P = 0.001]. Thus, young subjects who performed very well during acquisition (who also showed large increases in spindle density) revealed the smallest improvements when tested again 1 week later. Perhaps these subjects were already at, or close to, their ceiling on the task during acquisition so they were not able to demonstrate any further gains during retest.

Figure 6.

 Scatterplot showing the correlation between the percent change in spindle density and the percent change in performance on the pursuit rotor in the young group (top panel) and in the older group (bottom panel).

Third, there is no evidence of a general relationship between the density of Stage 2 spindles and motor performance. For example, the correlation between baseline spindle density and performance during acquisition is not significant in either the young group [r(12) = −0.112, P = 0.703] or in the older group [r(12) = −0.099, P = 0.738].

Discussion

The two main predictions of this investigation were supported in part. Regarding the first prediction, both age groups showed significant improvements on the pursuit rotor task when tested 1 week following acquisition, but the magnitude of the improvement was greater in the young group than in the older group. There was no age difference, however, in the percent of change on the pursuit rotor in each age group – a measure that takes the initial performance level of each group into account. These findings suggest that there are age differences in the absolute amount of improvement on the pursuit rotor but not in the relative amount of improvement. Regarding the second prediction, the density of Stage 2 spindles increased significantly after learning in the young group but not in the older group. However, again, when a relative measure of change was used, the age difference was no longer significant. These results highlight the importance of considering both absolute and relative changes in sleep-related and performance-related variables when comparing young and older adults. The correlation analyses revealed that better performance during acquisition was associated with larger increases in Stage 2 spindle density in the young group but not the older group. However, larger increases in spindle density following acquisition were associated with larger increases in performance across sessions in the older group but not the young group.

The lower initial performance and the smaller gains on the pursuit rotor task exhibited by the old group are consistent with previous studies (Gutman, 1965; Raz et al., 2000; Thumin, 1962; Wright and Payne, 1985). The reduction in Stage 2 sleep spindles in the old group is also consistent with previous studies (Crowley et al., 2002; Guazzelli et al., 1986; Nicolas et al., 2001; Wauquier, 1993). The present study extended these findings by documenting the amount of individual variability in the degree to which Stage 2 spindle density changes following the acquisition of a motor task in older adults. Neither the absolute nor the relative change in spindle density was significant in older adults. This was probably due to the fact that just more than half (n = 8) of subjects in the older group showed proportional decreases in spindle density, compared to only 2 subjects in the young group. However, three of the older adults exhibited proportional increases (48 to 60%) that were greater than the largest increase seen in the young group (41%). These findings complement the positive, albeit non-significant, correlation between the change in spindle density and the change in performance in the older group. Although the change in spindle density was not significant in the older group as a whole, some subjects did show sizable increases in spindle density, which were related to their improvement on the pursuit rotor task.

The increase in Stage 2 spindle density observed in the young group in the present study (13%) was lower than in our previous study (24%; Fogel and Smith, 2006). The larger increase in our previous study may have been due to the fact that those subjects learned three other motor tasks in addition to the pursuit rotor. It is interesting that the increase in spindle density was still fairly robust even after learning just one motor task. Future work should examine the possibility of a dose–response relationship between the amount of learning and the magnitude of changes in spindle density. It should also be pointed out that a number of other recent studies have reported significant increases in spindle density following the acquisition of various declarative memory tasks (Gais et al., 2002; Schabus et al., 2004, 2006) and significant correlations between the number of spindles observed after learning and the amount of improvement on various tasks (Clemens et al., 2005, 2006; Gais et al., 2002; Schabus et al., 2002, Schabus et al., 2006). Given these results, one obvious question is why are there increases in sleep spindles following different types of learning? It has been suggested that sleep spindles may represent an ideal mechanism for synaptic plasticity (Destexhe and Sejnowski, 2001; Steriade, 1999). Briefly, cortical cells are both excited and inhibited by thalamic cells which results in an increase in intracellular calcium, but the cortical cells are prevented from firing action potentials. Since calcium is known to be involved in synaptic plasticity, there is the possibility of a link between sleep spindles and memory consolidation from a cellular point of view. One could also argue, however, that some third, unidentified factor is driving the correlation between increases in sleep spindles and performance improvement. Future research is clearly needed to address these possibilities.

The negative correlation between the changes in spindles and performance in the young group seems counterintuitive and inconsistent with our previous findings (Fogel and Smith, 2006) as well as those from other laboratories (Clemens et al., 2005, 2006; Gais et al., 2002; Nishida and Walker, 2007; Schabus et al., 2004, 2006). One might argue on the basis of this result, combined with the finding of a strong positive correlation between performance during acquisition and changes in spindle density in the young group, that number of spindles is determined by one’s general ability on the task overall, and is not related to consolidation of the task. Indeed there is some evidence that the number of sleep spindles is associated with general ability or IQ (Bodizs et al., 2005; Fogel et al., 2007a,b; Schabus et al., 2006). However, given that there was a positive, albeit non-significant, correlation of 0.51 between the change in spindle density and the change in performance in the older group, we feel that spindles are related to memory consolidation processes that occur during sleep. We feel that the negative correlation between changes in spindles and changes in performance in the young group may be due to the presence of a ceiling effect at retest combined with the robust negative correlation between performance during acquisition and the amount of improvement in performance across sessions. Thus, young subjects who performed very well during acquisition (and who showed large increases in spindle density) revealed smaller increases in performance at retest. Perhaps because they were already at, or close to, ceiling during acquisition, these subjects were not able to show any additional improvements. It is worth pointing out that our previous study that documented a significant positive correlation between changes in spindle density and changes in performance on the pursuit rotor (and other motor tasks) used a faster speed (45 rpm) than in the current study (32 rpm). We chose to use a slower speed in the present study due to the inclusion of an older group. The use of a slower rotor speed may have contributed to the ceiling effect in the present study. This explanation involving a ceiling effect in the young group does not, however, account for the fact that there was also a ceiling effect for the older adults during retest, yet the correlation between change scores was significant. Clearly more research is needed to address these relationships.

There was no evidence of a general relationship between baseline Stage 2 spindle density and baseline motor performance in the present study. These results are consistent with those reported by Guazzelli and colleagues (Guazzelli et al., 1986), who found no significant correlations between spindle characteristics and performance on various high-level cognitive and intellectual tests in older subjects. This finding, however, is not consistent with data from our own laboratory and those of others showing that the baseline density of spindles is highly correlated with measures of intelligence in young adults (Bodizs et al., 2005; Fogel et al., 2007). Baseline spindle density has also been shown to be related to measures of selective attention in healthy individuals and those with schizophrenia (Forest et al., 2007). It will be important for future research to further illuminate the relationships among sleep spindles, attention, memory consolidation, and general ability/intelligence more precisely.

In terms of absolute measures, we have interpreted our findings to suggest that motor learning differentially effects the density of Stage 2 sleep spindles in the younger and older groups, and that since sleep spindles have been shown in previous research to be linked to motor learning, this differential effect on spindle density may explain why the young group performed better than the older group at retest. An alternative explanation could be that the better motor performance at retest of the young group could have been simply due to them getting a better night of sleep after learning than the older group. This explanation focuses on overall sleep quality rather the Stage 2 sleep spindles specifically. We feel that that this alternative explanation is not correct. First, there was no significant difference in sleep efficiency across the two nights in either of the groups. Second, this explanation would not explain why the density of Stage 2 sleep spindles increased significantly in the young group but not the older group.

The finding of an increase in SWS after learning in the older group but not in the younger group was surprising to us as we have found changes in Stage 2 parameters only following acquisition of the pursuit rotor task (Fogel and Smith, 2006; Peters et al., 2007; Smith and MacNeill, 1994). There are at least three possible reasons why we found this result. First, the increase in SWS in the older group may have been a spurious finding. Second, the effect could be real and it may be related to the findings of Huber and colleagues (2004). These investigators found increases in slow-wave activity (SWA) in specific areas of the brain following the acquisition of a rotation adaptation motor task. Huber et al., interpreted these increases as homeostatic responses to the learning episode. We disagree with this possibility because we did not find any increase in SWS in the young group. Also, it is important to note that we did not examine SWA, rather, our analyses focused on the time spent in SWS. A third possibility is that the increase in SWS in the older adults could have been some kind of compensatory response by the brain following task acquisition. Perhaps the Stage 2 mechanisms that are associated with consolidation of simple motor tasks, like the pursuit rotor, are not working effectively in the older adults and the brain was recruiting different mechanisms during SWS, which has been typically linked with consolidation of other types of memory tasks (e.g., declarative learning). This is purely speculation and future research is required to address the finding of increases in SWS in the older adults.

There are several limitations of the present study that deserve mention. First, our study did not include a ‘true’ control group (i.e., a group that did not perform the pursuit rotor task), nor did we employ a counterbalanced cross-over within-subjects design. Consequently, an alternative explanation of our findings may be that the change in spindle density in our young subjects may have been due to some variable other than learning the pursuit rotor task. However, we disagree with this alternative explanation. Previous research has shown that although the number and density of sleep spindles varies across subjects, the number and density of spindles are highly consistent within subjects across nights (Fogel and Smith, 2006; Gaillard and Blois, 1981). These data suggest that, in the absence of any specific intervention (e.g., learning a task), the number and density of sleep spindles does not change very much from night to night. Accordingly, we feel that the increase in spindle density in our young subjects was due to the learning task. However, in the future it would be ideal to examine this issue in more detail.

Second, our sample sizes included only 14 subjects in each age group. Although 14 subjects is a fairly adequate sample size in sleep studies of this nature, our statistical power was limited in certain respects. For example, several of the correlation coefficients in Table 2 came close to reaching significance but did not meet our alpha level. In addition, with the current sample size, we were unable to directly compare the pursuit rotor performance of older adults with high spindle densities to those with low spindle densities. It is highly plausible that old adults with high spindle densities would perform better on the pursuit rotor task than old adults with low spindle densities. This is an important agenda for future research.

A third limitation is that we have examined only one aspect of sleep in detail – the Stage 2 sleep spindle. Although we have focused our attention narrowly in this respect, it is certainly not our intention to suggest that other stages of sleep (e.g., REM, SWS) and/or other phasic events (e.g., REMs) are not as important for consideration in the study of the relationship between sleep and memory. Indeed, evidence is accumulating that other sleep stages, and phasic events within these stages, are related to various aspects of memory consolidation (Smith, 2001; Walker and Stickgold, 2006). For example, sleep spindles are not only seen in Stage 2 sleep, they are also found in Stages 3 and 4 (SWS). Some investigators have reported that SWS is associated with learning and/or synaptic homeostasis (e.g., Huber et al., 2004; Marshall et al., 2004; Plihal and Born, 1997). We did not examine spindles during SWS because many older adults have little or no SWS. We feel that age differences in the relationship between sleep and memory have been a relatively neglected area and that future research needs to address other aspects of sleep as well as replicate/extend our preliminary findings concerning sleep spindles. For example, there is evidence of two different types of sleep spindles: a slower spindle with a frontal predominance and a faster spindle with a central-parietal predominance (Werth et al., 1997; Zeitlhofer et al., 1997). The current study did not examine these two types of spindles separately, rather we examined both types combined. Future research should examine both types of spindles and do so over several nights following acquisition rather than just one night as in the present study. It may be the case that memory-related changes in sleep architecture are slower in older adults and proceed over multiple nights.

A fourth limitation is that the present study was cross-sectional. Consequently, we could only comment on age differences in sleep architecture and performance rather than age changes. Future studies that employ a longitudinal design are needed to address the issue of age changes in sleep and performance.

Finally, we had only one retest session that took place 1 week following acquisition. The advantage of having only one retest session is that there is no confounding effects of repeated practice on consolidation. We decided to have subjects do the retest after 1 week for two reasons. First, we do not currently know how long post-acquisition changes in sleep architecture last for. It is plausible that that these changes may occur over several nights after learning. Indeed there is definitely evidence of such changes in animals (Smith, 1996) and some preliminary evidence in humans (Smith and Smith, 2001). Given this possibility, we thought that a 1-week period would be sufficient time for sleep architecture to return to baseline levels following task acquisition. Second, we were interested in determining whether changes in sleep architecture were associated longer-term changes rather than enhancements that are temporary or short lived. One drawback of this design, however, is that it is conceptually more difficult to link improvements in performance 1 week later to the changes in sleep architecture that occurred right after acquisition.

To conclude, there were several main findings of the present investigation. Both age groups exhibited significant improvements on the pursuit rotor task when retested 1 week following acquisition. Second, the density of Stage 2 sleep spindles increased significantly following task acquisition in young adults but not in older adults. Age differences in the change in performance and spindle density both failed to reach significance when improvement was measured as a percentage of baseline level. The increase in spindle density was correlated with performance level during acquisition in the young group but not the older group. Although spindle density did not change in the older group as a whole, there was some preliminary evidence that older adults who did show large increases in spindle density following acquisition also tended to show larger performance improvements. Overall, the results of the present study are largely consistent with previous studies on sleep and memory in young adults and suggest that more detailed examination of this relationship in older adults is warranted.

Disclosure statement

This was not an industry supported study. Drs. Peters and Smith, Mrs Ray, and Miss Smith have indicated no financial conflicts of interest.

Ancillary