| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Psychology, University of Victoria, Canada.
2 School of Aging Studies, University of South Florida, Tampa.
3 Department of Psychology, University of Alberta, Canada.
Address correspondence to Allison Bielak, Department of Psychology, University of Victoria, P.O. Box 3050 STN CSC, Victoria BC V8W 3P5, Canada. E-mail: abielak{at}uvic.ca
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
The extent to which older adults' lifestyles are "engaged" is determined by level of participation in physical, social, and cognitively demanding activities. One informal hypothesis is that a more engaged lifestyle leads to greater use of cognitive resources and practice of cognitive skills, both of which may be associated with relative sparing of cognitive decline. This idea has spawned the "use it or lose it" and the "disuse" notions frequently tested in cognitive aging (see Small, Hughes, Hultsch, & Dixon, 2007
). Although some of the identified risk factors for a faster rate of cognitive decline are fixed or not readily modifiable (e.g., genotype; Anstey & Christensen, 2000), individuals can potentially alter their level of engagement in physical, social, and cognitive activities, all of which can provide stimulation for the cognitive system. According to an active model of cognitive reserve, the environment, through lifestyle activities, can influence a person's neural processing and synaptic organization by allowing neurological processes to become more efficient and adaptive (Stern, 2002). As complemented by research on environmental or occupational complexity, the more complex the demand characteristics of a setting, the greater the potential for maintenance and enhancement of cognitive functioning (e.g., Andel et al., 2005; Schaie, 1996).
Research consistent with the use-it-or-lose-it hypothesis has found that active older adults were more likely to obtain higher scores on cognitive tests at a single point in time (Hultsch, Hammer, & Small, 1993); that participation in intellectually stimulating activities was associated with less decline over a 6-year period (Hultsch, Hertzog, Small, & Dixon, 1999); and that individuals who led more active social or leisure lives show a reduced risk of dementia (e.g., Crowe, Andel, Pedersen, Johansson, & Gatz, 2003; Fratiglioni, Wang, Ericsson, Maytan, & Winblad, 2000). Further, performing household maintenance, doing domestic chores, engaging in volunteer work, and participating in social activities were found to significantly predict 6-year cognitive change in speed, picture naming, and incidental recall, even after controlling for sensory functioning (Newson & Kemps, 2005). Similarly, strenuous physical activity in daily home activities was a direct predictor of 2-year cognitive change on neuropsychological tests (Albert et al., 1995; Plehn, Marcopulos, & McLain, 2004; see also Colcombe & Kramer, 2003, for a meta-analysis).
However, research has not uniformly supported the positive link between an engaged lifestyle and cognitive performance in older adults. Using a 22-item activity inventory that assessed the time spent in a typical week engaged in a given activity, Salthouse, Berish, and Miles (2002) failed to find any evidence that activity participation mediated age-related differences in performance on four composite cognitive measures (i.e., spatial ability, reasoning, episodic memory, and vocabulary). Further, Hultsch and colleagues (1999) found that changes in activity level across a 6-year interval were not associated with corresponding changes in cognitive performance (e.g., verbal fluency, episodic memory, vocabulary, and reading comprehension). Aartsen, Smits, van Tilburg, Knipscheer, and Deeg (2002) reported that initial activity participation had no effect on functioning in immediate recall, learning, fluid intelligence, and information-processing speed at a point 6 years later. Moreover, Mackinnon, Christensen, Hofer, Korten, and Jorm (2003) showed that older adults experienced declines in memory, speed of processing, and crystallized intelligence even when their level of activity was maintained across 7 years.
There are several complications in the literature that may account for the mixed pattern of findings (Salthouse, 2006). First, the causal direction of a relationship between activity engagement and cognitive performance remains unknown. Lifestyle engagement may lead to cognitive buffering, but cognitive decline may also reduce engagement (Hertzog, Hultsch, & Dixon, 1999). Lövdén, Ghisletta, and Lindenberger (2005; see also Ghisletta, Bickel, & Lövdén, 2006) investigated this issue of directionality by using a dual-change model analysis in old and very old adults. Their results supported the intuitively expected direction; that is, social participation influenced subsequent changes in perceptual speed, rather than perceptual speed influencing later changes in social participation. Second, a variety of operational definitions of both lifestyle activities (see Small et al., 2007) and cognitive performance have been used. Regarding cognition, both products and processes of cognition have been tested, but with an array of different tasks. Even at the level of neurocognitive measures (processing speed), there is variation in the indicators of functioning (e.g., digit symbol substitution, Lövdén et al.; comprehension speed, Hultsch et al., 1999). A third complication is that the cognitive benefits of casually engaging in everyday activities may be modest, selective, or threshold related (Hultsch et al.; Small et al.; Mackinnon et al., 2003).
To date, studies investigating the link between an engaged lifestyle and cognitive aging have evaluated only mean-level performance. However, certain cognitive speed tasks permit an evaluation of both mean performance and intraindividual variability, or inconsistency. When measured with a cognitive speed task, inconsistency refers to relatively rapid and short-term changes within an individual (e.g., variation in a person's performance over many trials within a single task). The significance of using inconsistency has intensified as studies have appeared that support the hypothesis that cognitive inconsistency is a plausible behavioral indicator of central nervous system integrity (Hultsch, MacDonald, Hunter, Levy-Bencheton, & Strauss, 2000; Li & Lindenberger, 1999). For example, increased inconsistency in speed is associated with (a) poorer overall level of performance (e.g., Hultsch & MacDonald, 2004; Li, Aggen, Nesselroade, & Baltes, 2001); (b) older compared with younger adults (e.g., Hultsch, MacDonald, & Dixon, 2002; Nesselroade & Salthouse, 2004; Williams, Hultsch, Strauss, Hunter, & Tannock, 2005); (c) neurological disease or injury such as dementia, Parkinson's disease, or traumatic head injury (e.g., Burton, Strauss, Hultsch, Moll, & Hunter, 2006; Stuss, Pogue, Buckle, & Bondar, 1994; Walker et al., 2000); and (d) mild cognitive impairment (Christensen et al., 2005; Dixon et al., 2007; Strauss, Bielak, Bunce, Hunter, & Hultsch, in press). Further, inconsistency in speeded performance may predict some cognitive abilities and diagnostic classifications independent of mean rate of speeded performance (Hultsch et al., 2002; Strauss et al.). This suggests that inconsistency may be more closely representative of the integrity of the neurological system than is mean-level performance. However, this argument must be tempered with research findings that mean-level performance was a more powerful predictor of age and cognitive variables than was within-person variability (e.g., Salthouse & Berish, 2005; Salthouse, Nesselroade, & Berish, 2006). Therefore, it remains to be seen whether inconsistency in speeded performance will be more closely attuned to activity engagement than mean level of performance.
In the present study we investigate the concurrent and longitudinal relationships between older adults' engagement in lifestyle and two indicators of cognitive speed: level and inconsistency. We investigate three main research questions. First, if an engaged lifestyle acts to protect against cognitive decline, we propose that (a) initial activity score will predict the mean-level performance and amount of inconsistency in cognitive speed concurrently, and (b) longitudinal changes in activity scores will be related to changes in mean-level performance and inconsistency in cognitive speed over a 6-year follow-up period. Investigation of this relationship across 6 years will address whether significant relationships between activity participation and cognitive functioning are maintained despite fluctuations in either variable. Second, given findings that cognitively stimulating activities show the strongest relationships to later cognitive functioning (e.g., Hultsch et al., 1999), we expect that participation in such activities will be a better predictor of level and inconsistency in cognitive speed than other activity domains. Third, given that previous literature has shown that the cognitive benefits of activity engagement may be modest and difficult to detect, and that inconsistency may be more sensitive at discriminating among conditions related to cognitive decrements, we explore the possibility that inconsistency in speeded performance may be a more sensitive indicator than is mean level of performance.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Hultsch, Hertzog, Dixon, & Small, 1998).
Participants
VLS participants are community-dwelling older adults (initially between the ages of 55 and 94 years) living in the region of Victoria, British Columbia, Canada, recruited through advertisements in the public media and to community groups. The first wave of Sample 2 began with 530 participants (355 women, 175 men). There was evidence of positive selection in the 76% to 83% retention rate between waves, as participants who returned for testing had higher levels of education, occupation, and previous wave cognitive test scores.
At Wave 1, the sample had an average age of 68.53 years (SD = 7.61) and had completed 14.8 years of education (SD = 3.15). The majority of the sample was retired (82%), had few chronic health conditions (M = 2.6), and took few medications (M = 1.50). Further, 85.4% of the participants rated their health as good or very good relative to a perfect state of health, and 90.5% rated their health as good or very good relative to others their own age.
Measures and Procedure
At each wave of testing, participants completed the test battery over multiple testing sessions scheduled across approximately 4 to 6 weeks. Each testing session was approximately 2 to 3 hours long.
Activity Lifestyle Questionnaire
We used the VLS Activity Lifestyle Questionnaire to measure the typical frequency of engagement in 70 types of everyday cognitive, social, and physical activities over the past 2 years. On the basis of previous validation work (Hultsch et al., 1999), we dropped 3 items that were not associated with any subscale but included 3 items (omitted by Hultsch et al.) that formed a travel activities subscale. We classified the final 67 items into the following seven domains: (a) physical, such as jogging or gardening (n = 4), with a reliability coefficient across the three waves of r =.84; (b) self-maintenance, such as preparing a meal or shopping (n = 6), with r =.83; (c) social, such as attending concerts or visiting friends (n = 7), with r =.88; (d) travel, such as, traveling within Canada (n = 3), with r =.80; (e) passive information processing, such as reading the newspaper or watching a documentary (n = 8), with r =.88; (f) integrative information processing, such as using the computer or playing a musical instrument (n = 12), with r =.91; and (g) novel information processing, such as completing income tax forms or playing bridge (n = 27), with r =.93.
The frequency of participation is rated on a 9-point scale (never, less than once a year, about once a year, two or three times a year, about once a month, two or three times a month, about once a week, two or three times a week, and daily). We scaled the responses such that higher scores were associated with greater frequency of activity. We summed responses on items in each of the seven subscales to create composite activity measures. In cases of missing data (i.e., accidentally skipping a question or page), we prorated the scores from the scales where there were items missing, provided that no more than 6 of the 67 questions were omitted (e.g., if 1 question was missing out of 7, then we multiplied the average for the 6 complete items by 7 to obtain the overall scale score). Twenty participants omitted more than 6 questions; we removed them from the analyses, forming a sample of 510 participants at baseline. Table 1<--CO?3--> describes the correlations between the activity domains, which demonstrate that a higher frequency of engagement in one domain is correlated with greater participation in another domain.
|
Simple reaction time
Simple reaction time (SRT) refers to the latency to press a key when the signal stimulus appeared (50 trials).
Choice reaction time
Choice reaction time (CRT) refers to the latency to press the key corresponding to the location of one of four "plus" (+) signs arranged in a square that changes into a box (60 trials).
Lexical decision
Lexical decision refers to the latency to judge whether a string of five to seven letters forms an English word (e.g., island vs nabion; 30 words, 30 nonwords).
Semantic decision
Semantic decision refers to the latency to judge the plausibility of sentences (e.g., The tree fell to the ground with a loud crash vs The pig gave birth to a litter of kittens this morning; 50 sentences).
Data Preparation
We first examined the latency distributions for outliers at the level of individual trials. Extremely fast or slow responses most likely represent various sources of measurement error (e.g., accidental key press, distraction). Lower bounds for legitimate responses for each task were suggested by prior research (Hultsch et al., 2002; SRT and CRT, 150 ms; lexical decision, 400 ms; semantic decision, 1,000 ms). For upper boundaries, we computed the mean and standard deviation for each task and occasion of measurement for each age group and removed any trials that exceeded the mean by 3 SD or more. On average, 1.8% of responses exceeded these upper and lower boundaries.
We imputed missing value estimates by using a regression substitution procedure that forms individual equations of response times across all trials, which is then used to predict the missing RT entry (Hultsch et al., 2000). The procedures for eliminating outlying trials and imputing estimates for the missing values decrease within-subject variation, so they represent a conservative approach to examining inconsistency in response time.
Computation of level
The computation of level is the mean RT of each individual's latencies, or intraindividual mean (IM) across all trials for each task.
Computation of inconsistency
The computation of inconsistency is the across-trial within-person individual standard deviation (ISD) about each individual's mean RT. Other techniques to calculate inconsistency have been statistically criticized, and our approach follows an alternative logic (see Hultsch, Strauss, Hunter, & MacDonald, in press). We partialed out potential confounding influences (e.g., age differences in mean RT; practice effects) and their higher order interactions by using a split-plot regression:
|
|
To remove effects related to age group, we classified participants into three groups based on age at Wave 1: young-old, 55 to 64 years (n = 176, M = 60.32, SD = 2.95); mid-old, 65 to 74 years (n = 235, M = 69.21, SD = 2.82); and old-old, 75 to 94 years (n = 119, M = 79.30, SD = 3.55), of whom 10 participants were 85 years of age or older. Given the minimal number of errors, we computed ISDs from all trials (correct and incorrect). Burton and colleagues (2006) demonstrated that identical results were obtained whether ISDs were computed from all trials or whether they were computed from only correct trials.
Statistical Analysis
To examine the relationship between individual differences in RT (IM and ISD) and lifestyle activities, we fit a series of random effects models to the RT and lifestyle measures. For each outcome, we fit a linear time model, centered at the midpoint of the follow-up period (Year 3), and we saved each person's intercept and slope parameters. Next, we correlated the intercepts and slopes for RT with the intercepts and slopes for lifestyle activity in order to examine the relationship between concurrent individual differences or longitudinal changes in lifestyle activities and RT, respectively. Finally, we computed multiple regressions whereby concurrent individual differences or longitudinal changes in each of the RT measures were predicted by cross-sectional differences or longitudinal changes in all of the measures of lifestyle activity.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
For lexical decision time, we saw significant longitudinal changes for both IM and ISD. However, the effects for mean-level performance indicated faster responding across the follow-up period, but this was accompanied by increased variability. In both cases, significant random effects for slopes indicated heterogeneity of change over time.
Finally, for semantic processing time, the fixed effect of time was not significant for IM, but it was reliable for ISD. In terms of random effects of change, there were significant effects for mean-level variability, and almost significant effects longitudinally for ISD (p =.062).
There were significant declines in five out of the seven activity domains: physical, self-maintenance, travel, and integrative and novel information processing. For travel, we could not estimate the random effects for time, and therefore we did not include travel activities in our later analyses. Although social and passive information processing activities failed to show significant declines over time, the random effects indicated interindividual variability in rate of change over time.
Correlations Between Cognitive Speed and Lifestyle Activities
Prior to examining the relationship between cognitive speed and lifestyle activities, we examined the relationship between the intercept and longitudinal slope estimates for both IM and ISD. For SRT, the intercepts and slopes were significantly correlated between estimates of IM and ISD (r =.42, p <.001; r =.19, p <.001, respectively), as was the case for lexical decision time (r =.82, p <.001; r =.54, p <.001, respectively) and semantic decision time (r =.75, p <.001; r =.63, p <.001, respectively). In all cases, longer response latency was associated with greater variability for cross-sectional comparisons and greater declines longitudinally were associated with greater variability longitudinally.
Table 3 displays cross-sectional and longitudinal correlations between lifestyle activities and cognitive speed. For the cross-sectional relationships, integrative and novel information processing showed the strongest relationships with both IM and ISD. In addition, physical activity was associated with IM for SRT, CRT, and lexical decision time, and ISD for SRT. Self-maintenance activities were associated with IM and ISD for lexical decision and semantic decision. Finally, passive information processing was associated with lexical decision ISD and social activities were associated with lexical decision IM. In all cases, higher activity participation was associated with decreased IM latencies and lower ISD scores. However, nearly all correlations were larger for IM.
|
Table 4 shows the regression analysis of cognitive speed and lifestyle activities. All activities were entered simultaneously to predict speeded performance. Cross-sectionally, all of the models predicted significant amounts of variance for both IM and ISD, and the magnitude of these correlations was similar across the two speed measures. Novel information processing was a significant predictor in all models. In all cases, greater activity participation was associated with decreased IM latencies and less variability in responding. For the change measures, only IM for CRT and both measures from semantic decision speed had their models account for a significant portion of variance. Despite a significant amount of variance predicted for CRT, none of the individual predictors were significant. For semantic decision, declines in passive and novel information processing activities were associated with greater mean-level slowing and increased variability over time. Finally, although similar, the effects tended to be greater for IM than for ISD.
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The direction of the significant baseline relationships of the individual activity domains was consistent with the use-it-or-lose-it hypothesis; higher frequency of participation was related to better cognitive performance (i.e., less inconsistency and faster mean RT). The strongest correlations involved novel information processing, and this was consistent across all tasks and types of cognitive speed. Integrative information processing was similarly associated with nearly all of the cognitive measures. Physical activity was correlated with IM and ISD on the less challenging tasks (i.e., SRT, CRT, lexical), whereas engagement in self-maintenance activities was negatively associated with the verbal tasks. Change-related correlations, however, were sparse. Half of the significant correlations were found with the semantic decision task, and these involved the passive and novel information processing domains.
The frequency of participation in a variety of activities reported at baseline significantly predicted baseline levels of both inconsistency and mean level of cognitive speed performance in all four tasks. This analysis provides the first known demonstration linking lifestyle activity participation with intraindividual variability. In each case, being more active was associated with faster and more consistent responding. Further, novel information processing provided the largest unique effect in every analysis, which is consistent with expectations from previous research (e.g., Hultsch et al., 1999).
However, analyses addressing the key question of whether activity participation affects changes in cognitive ability did not receive unequivocal support. Changes in activity level only accounted for changes in mean-level CRT and changes in both mean-level performance and inconsistency of semantic decision. Although significant, these effects were modest and in most cases only a fraction of the effect sizes found in the baseline relationship. For example, the benefits of participating in novel information processing activities at baseline varied from β = –.15 to –.28. This is in comparison with β = –.01 to –.10 in predicting longitudinal changes in IM and ISD. These sharp contrasts are similarly illustrated in the drop in overall activity effects (e.g., R2 =.11 at baseline to R2 =.03). Although random effects models are limited in their ability to address lead–lag relations because change and level are defined over the same period, and thus level does not precede change, these findings suggest that activity level at one point in time may be a more important predictor of cognition than an individual's changes in activity level. This proposal is consistent with significant directional findings from short intervals (i.e., 2-year change; Lövdén et al., 2005). Such a relationship would offer a promising picture for older adults as they age, suggesting that it does not matter how active they were across time, but instead how active they are in the present.
Comparisons between the concurrent and longitudinal relationships with activity have varied in the literature. Newson and Kemps (2005) found that the unique contribution of activity in predicting change in speed of processing, picture naming, and verbal fluency was less compared with the concurrent relationship, but prediction of change in incidental recall was greater compared with concurrent prediction. Further, Ghisletta and colleagues (2006) found that increased media and leisure activity was associated with less decline in perceptual speed but was unrelated to yearly changes in verbal fluency. Therefore, although the present results showed that a wider variety of activity domains were significant predictors of cognitive ability when concurrent performance was assessed compared with later cognitive performance, the lag time on cognitive benefits may vary by cognitive domain.
The present results were consistent with previous research demonstrating that the strongest relationships between engagement and cognitive ability involved cognitively challenging activities (e.g., Ghisletta et al., 2006). Furthermore, novel information processing was one of the few domains that significantly predicted longitudinal cognitive change. This may reflect the nature of the stimulation offered by novel information processing activities, as they primarily stimulate cognitive domains, presumably offering the greatest cognitive challenges and thus benefits to the neurological system. Consequently, it appears that although other activity domains show beneficial relationships in the short term, activities that challenge cognitive skills may offer the best predictive power for longer term cognitive ability. Ghisletta and colleagues found this to be true in predicting 1-year changes in perceptual speed, whereas participation in social activities may only allow for greater compensation against the appearance of cognitive difficulties (Fratiglioni et al., 2000).
We expected that cognitive inconsistency might be more sensitive to the benefits of activity stimulation just as it may be more sensitive to the losses associated with cognitive decline and impairment. Generally, however, this was not true for both the baseline and change relationships. Correlations between the activity domains and ISD were fewer and tended to be weaker than those with IM. Across the regressions, in nearly every case, activity participation was a better predictor of mean-level performance than inconsistency, or it demonstrated equivalent effects. It may be that although the effects of an engaged lifestyle are modest, activity operates at a macro-cognitive level (compared with a micro-neurological level), to which standard measurements like the mean are receptive.
Finally, the various tasks showed substantial differences in the activity relationships to cognitive speed. Although the two nonverbal tasks were significantly predicted by activity participation at baseline, only IM in CRT was related to changes in activity. Similarly, baseline activity predicted concurrent lexical decision speed, but activity changes were not related to longitudinal changes in lexical decision speed. In contrast, the semantic decision task demonstrated significant concurrent and longitudinal activity effects. One could argue that the verbal tasks require more cognitive effort than the nonverbal tasks, and that a semantic decision is more demanding than a lexical decision (i.e., plausibility of a sentence vs legitimacy of a word). Therefore, the stronger effects for semantic decision RT suggest that the benefits of activity participation are more likely to appear on cognitively demanding tasks. Such a hypothesis would be consistent with findings showing better discrimination among varying degrees of mild cognitive impairment on more cognitively demanding tasks (Strauss et al., in press), and the suggestion that simple cognitive tests may not tap the same types of cognitive abilities promoted or strengthened by certain activities, and thus are missed in research investigations (Small et al., 2007). Therefore, although the short-term benefits of activities may be apparent in less challenging cognitive tasks, these benefits may be underestimated. For example, SRT was significantly predicted by baseline activity but showed smaller effect sizes than that of semantic decision. Further, if longitudinal effects are minimal in comparison with cross-sectional analyses, as in the present study, then the longitudinal influence of activity may have a greater likelihood of detection in difficult rather than simple cognitive tasks. Finally, Kramer, Bherer, Colcombe, Dong, and Greenough (2004) suggested that lifestyle activities may be more protective for crystallized and general abilities, rather than fluid and more specific abilities, implying that stronger activity effects may be found with higher order cognitive tasks than those used in the present study.
Because of the prior evidence and suggestions that the impact of activity may be greater in old-old adulthood (e.g., Hultsch et al., 1993; Salthouse, 2006), we also investigated whether the present results differ among age groups: young-old (55–64 years), mid-old (65–74 years), and old-old (75–94 years). In general, each age group showed fewer relationships between activity and cognitive speed compared with the overall analyses, but the significant correlations were still in the expected direction (i.e., higher IM or ISD, less activity). Old-old adults demonstrated a higher number of significant correlations than the other groups, and these correlations tended to be higher than the overall results (e.g., self-maintenance and semantic speed: r = –.28 for old-old adults; r = –.20 for overall analysis). Similarly, there were fewer significant correlations between changes in activities and changes in cognitive speed than those in the overall analyses, and the old-old adults tended to show larger correlations than those from the overall analysis and compared with the other groups. This was particularly true for relationships between novel information processing and semantic IM (r = –.23) and ISD (r = –.25). These results corroborate earlier research suggesting that the relationship between activity and cognition may be slightly greater in the later half of older adulthood.
With the exception of lexical decision, participants responded more slowly across the follow-up period, and there was evidence of significant individual differences in change over time. Consistent with previous research that inconsistency increases with age (e.g., MacDonald, Hultsch, & Dixon, 2003), our research showed that ISD increased across the 6 years for lexical and semantic decision. There was also evidence of significant heterogeneity in change over time for SRT ISD. However, failure of model convergence for random effects of rate of change prevented us from evaluating the relationship between ISD in CRT and activity. There were significant declines in the level of participation in physical, self-maintenance, travel, and integrative and novel information processing activities across the 6 years. However, social and passive information processing did show significant individual differences in the rate of change over time.
Overall, the present study shows both promise and limits for the use-it-or-lose-it hypothesis of cognitive aging: there are concurrent and longitudinal cognitive benefits from activity participation, but the relationships are stronger at baseline than they are across time. Despite inconsistency's suggested heightened sensitivity to the neurological system, this measure is not more receptive to the effects of activity than mean level of performance. The present study adds support to the idea that activities that involve novel information processing offer the greatest potential cognitive benefits. Future researchers would benefit from investigating the association between activity participation and cognitive performance on more complex cognitive tasks, and by including more sophisticated analyses to investigate the lead–lag relationship between activity and cognition.
|
Acknowledgments |
|---|
vlslab |
Footnotes |
|---|
Received for publication April 5, 2006. Accepted for publication June 15, 2007.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
