Learning-dependent changes in sleep spindles and Stage 2 sleep

Transcription

1 J. Sleep Res. (2006) 15, Learning-dependent changes in sleep spindles and Stage 2 sleep STUART M. FOGEL and CARLYLE T. SMITH Department of Psychology, Trent University, Peterborough, Ontario, Canada Accepted in revised form 8 March 2006; received 19 April 2005 SUMMARY It has become increasingly clear that sleep is necessary for efficient memory consolidation. Recently, it has been found that Stage 2 sleep disruption impairs procedural memory performance, and that memory performance is correlated with the duration of Stage 2 sleep; but the mechanisms involved in synaptic plasticity for procedural memory during sleep have not been identified. The present study examined the learning-dependent changes in sleep, including Stage 2 sleep spindles. Following an intense period of simple motor procedural learning, the duration of Stage 2 sleep and spindle density increased. There were no changes observed in the duration of any other stage of sleep or in the density of rapid eye movements. These findings support the hypothesis that sleep spindles are involved in the off-line reprocessing of simple motor procedural memory during Stage 2 sleep. keywords memory, sleep, spindles, stage 2, synaptic plasticity INTRODUCTION The relationship between sleep states and memory processes has recently become of great interest (for review see: Maquet et al., 2003). Past research has focused mainly on memory for a previously learned task following sleep deprivation or the effects of an intense period of learning on subsequent sleep changes (for reviews see: Smith, 1995, 2001). There is evidence to suggest that separate sleep states are differentially involved in the post-learning reprocessing of different categories of memory. Rapid eye movement (REM) sleep has been hypothesized to be involved with the reprocessing of procedural tasks that require a new cognitive strategy, while simple motor tasks involve Stage 2 sleep (see: Smith, 1995, 2001; Smith et al., 2004a for reviews). Improved performance on a complex motor task (mirror-trace) has been found to be associated with the larger amounts of REM sleep found in the last half of the night (Plihal and Born, 1997). Smith et al. (2004b) have reported increases in REM sleep densities in the post-training sleep of subjects that learned the mirror trace task. On the other hand, utilizing selective sleep deprivation, Smith and MacNeill (1994) reported that memory for a simple motor task, the rotary pursuit task, was vulnerable to Stage 2 sleep loss, but not REM sleep loss. By selectively depriving Correspondence: Carlyle Smith, Psychology Department, Trent Sleep Research Laboratory, Trent University, 1600 West Bank Drive, Peterborough, Ontario K9J 7B8, Canada. Tel.: ext 1406; fax: ; csmith@trentu.ca individuals of either REM sleep or by disrupting Stage 2 sleep, Aubrey et al. (1999) reported that REM sleep deprivation impaired performance on the mirror tracing task, but not on the simple tracing task. Conversely, Stage 2 sleep disruption impaired performance on the simple tracing task but not the mirror tracing task. Walker et al. (2002) have reported that acquisition of a simple finger tapping sequence was improved by a night of sleep and performance was positively correlated with amount of Stage 2 in the last quarter of the night. It has been proposed that brain mechanisms active during REM sleep are primarily concerned with off-line reprocessing of more complex procedural tasks that require the development of new cognitive strategies, whereas mechanisms active during Stage 2 sleep are dedicated to tasks for which the individual has had some previous personal experience (Smith et al., 2004a). One of the most prominent features of Stage 2 sleep is the sleep spindle (Carskadon and Dement, 2000). Some researchers have speculated that spindles may be important for synaptic plasticity (Steriade, 1999), and until recently this hypothesis remained to be tested directly. One recent study has demonstrated that stimulation frequencies designed to mimic the sleep spindle induced long-term potentiation (LTP) (Rosanova and Ulrich, 2005). While this study provides provocative evidence to directly link sleep spindles to LTP, the stimulation frequency used in this experiment (10 Hz) was outside of the conventionally defined (12 16 Hz) spindle frequency range. The focus of the present study was to 250 Ó 2006 European Sleep Research Society

2 Learning-dependent changes in sleep spindles 251 investigate the possible electrophysiological changes in sleep architecture including Stage 2 sleep spindles following simple motor procedural learning. It was expected that there would be increases in the duration of Stage 2 sleep, and in the number of Stage 2 sleep spindles following new motor procedural learning. Conversely, no increase in any other stage of sleep was predicted. In particular, we expected that the minutes of REM sleep and the density of rapid eye movements during REM sleep would remain unchanged following task acquisition. METHOD Participants Twelve female participants aged were recruited to participate in this study from Trent University, Peterborough, Ontario, Canada. The participants were selected using the Trent University Sleep Questionnaire to screen for sleep disorders, unusual or irregular sleep habits, confounding medical conditions requiring medication or causing chronic pain, and substance abuse. In addition, participants were instructed to keep a regular sleep wake cycle (sleep between 22:30 and 00:30 hours and wake before 09:00 hours) for a minimum of 3 days before and throughout participating in the experiment. Compliance was verified with the use of a sleep diary. Participants recorded sleep and wake times for a minimum of 3 days prior to the acclimatization night until the end of the experiment. Participants were paid $25.00 for each night in the laboratory, and to maintain motivation, an additional $10.00 was awarded for completion of the study upon re-testing of the learning tasks 1-week later. The research project was approved by the Trent University Human Ethics Committee. Polysomnographic recording parameters The electroencephalogram (EEG) of all participants was recorded for three consecutive nights from 23:00 hours to 07:00 hours. The recording polygraph was a Nihon-Kohden with high frequency filter set to 30 Hz and low frequency filter set to 0.3 Hz. The EEG data were stored and analyzed using paperless polygraph system software developed at the University of Texas Southwestern Medical School. The EEG was recorded from C3 and C4 and referenced to an average of A1 and A2. Eye movement (EOG) was recorded from the outer canthi of both eyes and each eye channel was referenced to A1 + A2. Muscle tension (EMG) was recorded from the left and right chin muscles using a bipolar reference. Procedures Participants came for three consecutive overnight examinations (acclimatization, baseline and test) to the sleep laboratory. They were required to arrive at the sleep laboratory for the first night at 21:00 hours. Until 22:30 hours, they were administered the Multidimensional Aptitude Battery II (MAB-II) to obtain measures of both Verbal and Performance intelligence. Immediately following MAB-II administration, electrode application occurred until 23:00 hours. On the baseline night, only electrode placement (22:30 23:00 hours) took place before overnight recording was done. On the test night, participants were randomly assigned to one of two conditions, either a learning (n ¼ 6) or control task (n ¼ 6). Participants and the experimenter were blind to the experimental condition before the test night. On the evening of the test night, the learning group was required to perform a combination of four learning tasks from 21:00 to 22:30 hours immediately before electrode placement. The participants in the learning condition returned 1 week following the test night from 21:00 to 22:30 hours for re-testing on the learning tasks. Both test and control participants were cautioned not to engage in any vigorous new sports or fine motor activities that were unusual for them during the course of the study. Control subjects were asked to arrive in the sleep laboratory on the evening of training and were allowed to read or watch videos, but did not participate in any motor activity. Learning tasks The learning tasks involved the general refinement of motor skills. They were designed to be conceptually simple and did not require the learning of rules, the use of logic, or development of new cognitive strategies to learn how to perform the tasks. The tasks included the pursuit rotor, the simple tracing task, the ball-and-cup game, and ÔOperationÕ. The pursuit rotor task has been used extensively in psychological testing (Seigel, 1990). Performance on this task is affected by Stage 2 sleep deprivation (Smith and MacNeill, 1994) and has been categorized as a pure motor procedural task (Smith, 2001). The apparatus was a photoelectric pursuit rotor, with a hand-held stylus and photoelectric sensor, revolution counter, and timer. The light target revolved around a 30 cm diameter square track at 45 r.p.m. The duration of time that the light sensor at the end of the handheld stylus was in contact with the revolving light target was recorded during five sets of five 30-s trials. The simple tracing task is a modified version of the mirror tracing task, but is much simpler (Plihal and Born, 1997). Unlike the mirror tracing task, the participant can see his/her own hand. The simple tracing task requires the refinement of fine motor control skills. The goal of the task was to trace around 14 figures with a pen as quickly and accurately as possible while staying within the two concentric lines (5 mm apart) that outlined the figures. The number of errors was used to measure performance. The ball-and-cup game involved the refinement of eye hand coordination and gross motor control. The apparatus consisted of a wooden cup 3.5 cm in diameter, with a handle and 60 cm string attached. A 2.5 cm diameter ball was attached to the other end of the string. The goal of the task was to swing the cup in an upward arc with the intention to catch the ball with the cup. Task performance was scored on a

3 252 S. M. Fogel and C. T. Smith 3-point scale including 0 ¼ miss, 1 ¼ side or rim shot, 2 ¼ caught, but bounced out, and 3 ¼ successfully caught. Participants completed 200 trials in 10 sets with 15-s rests between sets. The Operation skill game is a commonly used children s board game available from the Milton Bradley TM toy company (Hasbro, Inc., Pawtucket, RI, USA). The top surface measured 23 cm by 37 cm with holes of varying shape recessed in the surface of the board. The objective was to remove the small plastic objects (approximately 1.5 cm in length) without contacting the side of the holes that contained the game pieces. Contact with the side of a hole resulted in a buzzer sounding and a red light to register an error. After each piece was successfully removed from its position on the board, the procedure was repeated after a 30-s rest. Participants performed all four tasks twice, once with their preferred hand, and again with their non-preferred hand. This was done to avoid fatigue, to increase the difficulty of the task situation and hopefully maximize any sleep EEG changes. The MAB-II is made up of two separate IQ scales including non-verbal Performance IQ and Verbal IQ, each of which is assessed using five subtests. The Performance scale measures aptitudes for coding, perceptual, analytical, and spatial visualization skills. The Verbal scale measures aptitudes for the use of verbal information, verbal comprehension, arithmetic, abstraction, and vocabulary. Sleep data analysis Sleep stages were scored using standard criteria (Rechtschaffen and Kales, 1968). Sleep spindles were visually identified in all epochs scored as Stage 2 for the entire night. All spindles included in the count were within the Hz frequency band, exceeded 0.5 s, and had typical fusiform (waxing and waning amplitude) spindle morphology. This usually meant that their maximum amplitude also exceeded 10 lv. However, there was no minimum amplitude criterion in accordance with standard sleep scoring procedures (Rechtschaffen and Kales, 1968). Spindle density was calculated as the ratio of the number of sleep spindles counted in Stage 2 sleep to the number of minutes of Stage 2 sleep. Rapid eye movements were also visually identified in all epochs scored as REM sleep. Eye movements were counted if they were conjugal, and their amplitude exceeded 10 lv on either the left or right EOG recording. To verify the reliability of sleep scoring, randomly selected sleep recordings were stage scored by two independent judges. Stage scoring agreement was greater than 95% (>95% of the epochs were stage scored the same by both judges). Points of disagreement were discussed and resolved by mutual agreement between raters. From the same records, randomly selected epochs of Stage 2 and REM sleep were used to assess inter-rater agreement for spindle and REM counts, respectively. The raters continued to randomly select epochs and independently count sleep spindles and REMs until all points of disagreement were resolved until 95% of all spindles or REMs for each epoch were identified correctly by both judges. Sleep scoring agreement was conducted prior to stage scoring, spindle counting and REM counting for this experiment. Both judges were blind to which night or condition the record belonged. RESULTS To verify that improvement on the memory tasks occurred, motor task scores from the evening test session and re-test session (1 week later) were analyzed using paired t-tests for each learning task (Table 1). There was a significant increase in the time the stylus was in contact with the target for the pursuit rotor [t(5) ¼ 9.99, P ¼ 0.001], the accuracy score to catch the ball for the ball and cup task [t(5) ¼ 5.79, P ¼ 0.002], and a decrease in the number of errors made for the Operation task [t(5) ¼ 3.51, P ¼ 0.017] after 1 week. While participants also showed a reduction in the number of errors made from baseline (M ¼ 69.33, SD ¼ 29.54) to re-test (M ¼ 60.83, SD ¼ 18.68) on the direct trace task, the effect was not statistically significant. Separate 2 (baseline, test) 2 (learning, control) anovas were used to analyze the duration (in minutes) of each sleep stage (1, 2, 3/4, REM) and total sleep time (TST). It was found that there was a significant group by night interaction, whereby there was an increase in the duration of Stage 2 sleep from baseline to test night in the learning group [F(1,10) ¼ 5.66, P ¼ 0.04], but not in the control group. There were no group differences, changes from baseline to test night, or significant group by night interactions in the duration of Stage 1, Stages 3/4, or REM sleep. There was an increase in TST from baseline to test night in the learning group, which was similar in magnitude to the increase in Stage 2 sleep duration. However, the increase in TST was not statistically significant (Table 2). There was a 42% increase in the number of sleep spindles following learning [F(1,10) ¼ 17.70, P ¼ 0.002]. The learning Table 1 Mean performance and standard deviation (SD) on the test and re-test for the pursuit rotor (time-on-target), simple tracing task (number of errors), ball and cup (accuracy score) and Operation (number of errors). Results of paired t-tests (t) with respective degrees of freedom (df) and statistical significance (P) Learning task Test Re-test t df P Pursuit rotor (44.14) (43.65) ** Simple tracing (29.54) (18.68) Ball and cup (45.253) (38.74) ** Operation (68.77) (42.38) * Significantly different at *P < 0.05, **P <

4 Learning-dependent changes in sleep spindles 253 Table 2 Mean duration and standard deviation (SD) in minutes of sleep stages 1, 2, 3/4, REM, and total sleep time (TST) as well as REM density (number of REMs per minute of REM sleep) on baseline and test nights in the control and learning groups. Interaction terms from 2 (baseline, test) 2 (learning, control) anovas presented on Stage 1, 2, 3/4, REM sleep, TST and REM density Control Learning Baseline Test Baseline Test F P Stage (16.37) (13.70) 8.92 (4.43) 5.42 (2.63) Stage (72.90) (66.13) (29.84) (28.48) * Stage 3/ (45.90) (35.59) (9.11) (18.78) REM (21.69) (40.15) (41.88) (22.28) TST (43.54) (27.35) (55.90) (28.64) REM density (REMs per minute) 7.25 (5.19) 6.40 (2.95) 8.34 (1.91) 8.42 (3.64) *Significantly different at P < (M ¼ , SD ¼ ) and control (M ¼ , SD ¼ ) groups were not significantly different on the baseline night [t(10) ¼ 1.29, P ¼ 0.23]. However, the learning (M ¼ , SD ¼ ) and control (M ¼ , SD ¼ ) groups differed on the test night [t(10) ¼ 2.67, P ¼ 0.02]. To control for the increase in the duration of Stage 2 sleep, spindle density was used to measure learning-dependent changes in number of sleep spindles. We observed a 24% increase in spindle density (spindles per minute) following acquisition [F(1,10) ¼ 11.19, P ¼ 0.007]. The learning (M ¼ 3.80, SD ¼ 1.53) and control (M ¼ 1.69, SD ¼ 1.94) groups were not significantly different on the baseline night [t(10) ¼ 2.10, P ¼ 0.06]. However, the learning (M ¼ 4.70, SD ¼ 1.57) and control (M ¼ 1.50, SD ¼ 1.90) groups differed on the test night [t(10) ¼ 3.18, P ¼ 0.01]. As there was considerable variation in the number of sleep spindles between subjects, inter-individual differences in sleep spindles were controlled for by using Performance IQ (subscale of total IQ) as a covariate. There was no significant difference between the learning (M ¼ 108, SD ¼ 6.23) and control group (M ¼ 96.17, SD ¼ 12.09) on Performance IQ [t(10) ¼ 2.13, P ¼ 0.06]. The assumption of equal variances was not violated. As has been observed in previous work (Nader and Smith, 2001), Performance IQ was found to be highly correlated with baseline number of sleep spindles [r(10) ¼ 0.71, P ¼ 0.022], and did not violate the assumption of parallel slopes [t(8) ¼ )0.51, P ¼ 0.62]. Thus, to investigate the effects of new motor skill learning on spindle density, a 2 (baseline, test) 2 (learning, control) ancova was used to examine the change in sleep spindles from baseline to test night and between groups. It was found that there was a significant night by group interaction [F(1,9) ¼ 6.44, P ¼ 0.032] for spindle density (Fig. 1). It seemed reasonable to assume that improvement in performance would be positively correlated with spindle density. When the learning scores were standardized and combined into a composite score reflecting the total improvement across the four tasks, it was found that learning on the motor tasks was significantly correlated with the increase in sleep spindle density [r(4) ¼ 0.87, P ¼ 0.025]. In addition, it was found that the decrease in number of errors made on the simple tracing task was correlated with the increase in spindle density [r(4) ¼ 0.93, P ¼ 0.007]. At first Figure 1. Increase in spindle density (number of sleep spindles per minute of Stage 2 sleep) from baseline to test night [indicated by ** at t(9) ¼ 3.63, P < 0.01] following new simple procedural motor learning. Comparison between the learning and control group on test night [indicated by * at t(10) ¼ 2.28, P < 0.05] (learning group ¼ circle, control group ¼ square). The groups did not differ significantly on the baseline night, and there was no change in spindle density from baseline to test night in the control group. Vertical bars indicate standard error of the mean. Adjusted means controlling for interindividual differences in sleep spindles using Performance IQ as a covariate. glance, it would appear that learning on only the simple tracing task was related to spindle density. Because this possibility seemed unlikely, we conducted a multiple regression using the change in performance on each of the four learning tasks to predict the change in spindle density. This analysis revealed that together, the change in performance from test to re-test on four learning tasks accounted for 98% of the variability in the change in spindle density [F(4,1) ¼ 10.58, P ¼ 0.23]. From inspection of the semi-partial correlation coefficients, each task accounted for a large proportion of the variability in the change in spindle density. The pursuit rotor, simple tracing, ball-and-cup and the Operation task uniquely accounted for 43.6%, 93.9%, 81.4% and 81.4% of the variability in the change in spindle density over and above

5 254 S. M. Fogel and C. T. Smith each of the other tasks, respectively. None of the results from the regression analysis were statistically significant (P > 0.05). However, given the lack of degrees of freedom, and the number of predictors used in the analysis, the lack of statistical power is not surprising. Nonetheless, the results do give an indication that overall, the change in performance on the tasks is related to the change in spindle density, and that each task is associated with the change in spindle density. To verify that the change in sleep was limited to sleep spindles and specific to Stage 2 sleep, and not simply represent a general increase in phasic activity (i.e. sleep spindles and rapid eye movements) across Stage 2 and REM sleep, REM density (number of eye movements per minute of REM sleep) was analyzed using a 2 (baseline, test) 2 (learning, control) anova (Table 2). There was no change from baseline to test night in the control [t(5) ¼ 0.40, P ¼ 0.70] or learning group [t(5) ¼ )0.06, P ¼ 0.95] and no difference between groups on the baseline night [t(10) ¼ 0.48, P ¼ 0.64] for REM density. There was no night by group interaction for REM density [F(1, 10) ¼ 0.14, P ¼ 0.72]. DISCUSSION Overall, the results from this study support the hypothesis that sleep spindles are intimately involved with the consolidation of simple motor procedural memory and may be important for the off-line reprocessing of recently acquired simple procedural tasks. It was found that there was an increase in the density of sleep spindles. Furthermore, it was found that the overall improvement on the motor tasks was positively correlated with the increase in sleep spindle density. In addition, there was an increase in the duration of Stage 2 sleep following new learning. The magnitude of this change was very large; overall, there was a 15.8% increase in Stage 2 sleep. The increase in Stage 2 sleep appears to have had some influence on the duration of total sleep time, however, this effect was not statistically significant. The increase in the duration of Stage 2 sleep would be expected to increase the total number of sleep spindles alone. However, in addition to the increased duration of Stage 2 sleep, there was an increase in spindle density. This suggests that motor learning-dependent changes in sleep spindles are independent of the time spent in Stage 2 sleep. Sleep spindles have been proposed as a mechanism for synaptic plasticity (Steriade, 1999) and have been found to increase following new declarative learning (Gais et al., 2002) as well as being correlated with improved performance on a declarative memory task (Schabus et al., 2004). The sleep spindle would appear to be an ideal mechanism for the consolidation of motor procedural memory and synaptic plasticity in the neocortex. Cortical LTP requires spaced and repeated stimulation. Sleep spindles provide stimulation to the cortex in a spaced and repeated pattern. This is consistent with the LTP model presented by Racine s group (Chapman et al., 1998; Trepel and Racine, 1998). It has been recently demonstrated that spindle-like stimulation can produce LTP in neocortical pyramidal cells in vitro (Rosanova and Ulrich, 2005). To determine if the changes to sleep following simple procedural memory were limited to Stage 2 sleep and sleep spindles, REM density was also considered in this experiment. It was found that the duration of REM sleep did not change following new learning, nor did the density of rapid eye movements, which suggests that the changes to sleep following new simple procedural learning affected only Stage 2 sleep, and is specific to sleep spindles. While the current investigation suggests that sleep spindles are intimately involved with the consolidation of simple procedural memory, there are still many unanswered questions regarding the nature and extent of their function in memory consolidation. The current study used a battery of tasks that can be categorized as simple motor procedural tasks. This was done to ensure that the learning manipulation was intense and varied and to avoid physical and mental fatigue or ceiling effects on a given task. The use of multiple tasks was an important manipulation as after the initial improvement in performance on a single task of this nature, with extensive practice, performance deteriorates. However, using a combination of tasks makes it impossible to determine exactly what task characteristics are spindledependent. Furthermore, any one of the tasks performed alone may not have been intense or varied enough to produce an effect of the same magnitude observed in the current experiment. Rather, the unique combination of several motor tasks used here may be responsible for the changes to Stage 2 sleep and sleep spindles. However, this cannot be determined from the present investigation and further, single task studies will have to be done. The results of the present study could be interpreted in two ways as the control group was not engaged in an equivalent amount of motor activity: (1) that the changes in sleep architecture and sleep spindle density were due to non-specific motor activity unrelated to learning on the tasks, or (2) that the observed differences are specific to motor learning. The former explanation seems unlikely as it was found that the overall improvement on the four learning tasks was highly positively correlated with the change in spindle density. The generalizability of these results may have been affected by limitations of using a small sample size. However, despite this, the size of the effect was quite large. Added variability from a larger sample may have minimized the baseline interindividual differences in spindles. In addition, the phenomena were exclusively studied in female subjects. Future studies should include both male and female subjects. In conclusion, an intense period of learning on simple procedural motor tasks increases the duration of Stage 2 sleep and the number of sleep spindles per minute of Stage 2 sleep. This increase in spindle density is positively correlated with the overall improvement in task performance. The results support our hypothesis that sleep spindles during Stage 2 sleep may be involved in the consolidation of simple motor procedural memory, whereas REM sleep mechanisms do not seem to be involved in this type of learning.

6 Learning-dependent changes in sleep spindles 255 ACKNOWLEDGEMENTS This work was supported by the Natural Sciences and Engineering Research Council (NSERC) of Canada and the Canadian Institutes of Health Research (CIHR). REFERENCES Aubrey, J., Smith, C., Tweed, S. and Nader, R. Cognitive and motor procedural tasks are dissociated in REM and stage two sleep. Sleep Res. Online, 1999, 2(Suppl. 1): 220. Carskadon, M. A. and Dement, W. C. Normal human sleep: an overview. In: M. H. Kryger, T. Roth and W. C. Dement (Eds) Principles and Practice of Sleep Medicine. W. B. Saunders Co., Philadelphia, 2000: Chapman, C. A., Trepel, C., Ivanco, T. L., Froc, D. J., Wilson, K. and Racine, R. J. Changes in the field potentials and membrane currents in rat sensorimotor cortex following repeated tetanization of the corpus callosum in vivo. Cereb. Cortex, 1998, 8: Gais, S., Molle, M., Helms, K. and Born, J. Learning-dependent increases in spindle density. J. Neurosci., 2002, 22: Maquet, P., Smith, C. and Stickgold, R. Sleep and Brain Plasticity. Oxford University Press, Oxford, Nader, R. and Smith, C. The relationship between stage 2 sleep and intelligence. Sleep, 2001, 24(Suppl.): A160. Plihal, W. and Born, J. Effects of early and late nocturnal sleep on declarative and procedural memory. J. Cogn. Neurosci., 1997, 9: Rechtschaffen, A. and Kales, A. A Manual of Standardized Terminology, Techniques, and Scoring System for Sleep Stage Scoring of Human Subjects. U. S. Department of Health, Education and Welfare, Bethesda, MD, Rosanova, M. and Ulrich, D. Pattern-specific associative long-term potentiation induced by a sleep spindle-related spike train. J. Neurosci., 2005, 25, Schabus, M., Gruber, G., Parapatics, S., Sauter, C., Klosch, G., Anderer, P., Klimesch, W., Saletu, B. and Zeitlhofer, J. Sleep spindles and their significance for declarative memory consolidation. Sleep, 2004, 27: Seigel, D. A multivariate study of pursuit rotor skill development. Res. Q. Exerc. Sport, 1990, 61: Smith, C. Sleep states and memory. Behav. Brain. Res., 1995, 69: Smith, C. Sleep states and memory processes in humans: procedural versus declarative memory systems. Sleep Med. Rev., 2001, 5: Smith, C. and MacNeill, C. Impaired memory for a pursuit rotor task following Stage 2 sleep loss in college students. J. Sleep Res., 1994, 3: Smith, C., Aubrey, J. and Peters, K. Different roles for REM and stage 2 sleep in motor learning: a proposed model. Psychol. Belg., 2004a, 44: Smith, C. T., Nixon, M. R. and Nader, R. S. Post-training increases in REM sleep intensity implicate REM sleep in memory processing and provide a biological marker of learning potential. Learn. Mem., 2004b, 11: Steriade, M. Coherent oscillations and short-term plasticity in corticothalamic networks. Trends Neurosci., 1999, 22: Trepel, C. and Racine, R. J. Long-term potentiation in the neocortex of the adult, freely moving rat. Cereb. Cortex, 1998, 8: Walker, M. P., Brakefield, T., Morgan, A., Hobson, J. A. and Stickgold, R. Practice with sleep makes perfect: sleep dependent motor skill learning. Neuron, 2002, 35: