| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Department of Sociology and Social Gerontology, Vrije Universiteit, The Netherlands
b Department of Psychiatry and Institute for Research in Extramural Medicine, Vrije Universiteit, The Netherlands
Marja J. Aartsen, Department of Sociology and Social Gerontology, De Boelelaan 1081c, 1081 HV Amsterdam, The Netherlands E-mail: mj.aartsen{at}scw.vu.nl.
Decision Editor: Toni C. Antonucci, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
RESEARCHERS in the field of cognitive aging agree that, on average, cognitive functioning declines with aging. Cognitive decline may begin after midlife, but most often occurs at higher ages (70 or higher). However, major individual differences in rate and onset of decline are observed (Baltes and Baltes 1990
; Schaie 1980, Schaie 1983). Moreover, maintenance or even improvement of cognitive functioning with aging is found among some persons (Baltes, Dittmann-Kohli, and Dixon 1986; Korten et al. 1997). Cognitive decline is associated with increased personal discomfort, loss of autonomy, and increasing societal costs. Research on factors affecting cognitive performance may therefore contribute to the design of intervention strategies that can improve the autonomy and well-being of the aging population.
Over the last decades, increasing evidence has been provided for the beneficial effects of contextual variables suggesting possible points of contact for intervention strategies. Leading an active life is suggested to enhance cognitive functioning (Schaie 1983). This relation is explained by the use of cognitive skills needed to perform activities, especially activities that are cognitively demanding, resulting in maintenance or enhancement of cognitive functioning. Several studies are consistent with this idea (Arbuckle, Gold, and Andres 1986; Fabrigoule et al. 1995; Hultsch, Hammer, and Small 1993; Smits, Van Rijsselt, Jonker, and Deeg 1995). Apart from these cross-sectional studies in which conclusions on causality are limited, a longitudinal study conducted by Gold and colleagues 1995 revealed that being active is related to maintenance of intelligence across 40 years.
Recently, Hultsch, Small, Hertzog, and Dixon 1999 contested the supposed beneficial effect of contextual variables on cognitive functioning. Using longitudinal data collected among 250 middle-aged and older adults, they investigated whether participation in cognitively demanding activities promotes the development and maintenance of cognitive abilities or whether cognitively capable people tend to participate in environments that are cognitively demanding. Both directions of causation were supported by the data. Hultsch and colleagues also demonstrated this reversed causality in the data set used by Gold and colleagues 1995. This finding led to a lively discussion between the two groups of researchers (Hertzog, Hultsch, and Dixon 1999; Pushkar et al. 1999). The discussion focused on the selection of indicators of an active lifestyle, some methodological differences, and differences in study samples. The first and second issues constitute the starting point for the present study, in which we examined the relation between everyday activities and cognitive functioning in later life.
Regarding the measurement of activities, it is noted that Gold and colleagues 1995 used socioeconomic status (SES) as one of the three indicators of an engaged lifestyle, whereas Hultsch and colleagues 1999 selected a broad range of everyday activities, including physical activities, social activities, hobbies, and novel information-processing activities such as learning a language or playing bridge. We share the concern expressed by Hultsch and colleagues 1999 that level of cognitive functioning, which is associated with SES, moderates the positive relation between engaged lifestyle and cognitive functioning observed by Gold and colleagues 1995. However, the relation between novel information-processing activities and cognitive functioning observed by Hultsch and colleagues 1999 may be the result of an unobserved confounding variable such as level of education. It is not unthinkable that these novel information-processing activities, which were operationalized by playing bridge and learning a language, are related to higher levels of education or SES (Ganzeboom 1989). Lack of control for level of education may have resulted in spurious relations between activity and cognitive performance. A procedure to deal with confounding variables, even if these variables are unknown (MacCallum, Wegener, Uchino, and Fabrigar 1993), could have prevented both studies from finding spurious relationships.
Another issue is the specification of all possible directions of causation. In the model used by Gold and colleagues 1995, no reversed causal direction was specified, and hence no reversed causal effect was detected. In contrast, Hultsch and colleagues 1999 specified both directions of causation in two different models. This approach, however, does not reveal whether these effects could still be observed when both directions were implied in one model, or whether one of the two directions was predominant.
In the present study we investigated the extent to which one of the two possible directions of causation was present, or even predominant, in a large population-based sample of older adults, using various types of everyday activities and cognitive functions. We controlled for the effect of age, gender, level of education, and health, as well for the confounding effect of other, unmeasured, variables.
Our first question deals with the direction of causation between different types of everyday activities and cognitive functioning. First the causal effect of activities on cognitive functioning was tested. In line with previous studies (Fabrigoule et al. 1995; Gold et al. 1995) we expected positive effects from everyday activities on cognitive functioning over time (Hypothesis 1). The positive effect of activities on cognitive functioning may be explained by the process of learning generalization (Miller, Slomczynski, and Kohn 1987). Learning generalization means that knowledge and orientations acquired in one situation are generalized or transferred to other situations. For example, people who do intellectually demanding work come to exercise their intellectual abilities not only on the job, but also in their nonoccupational lives.
The idea of learning generalization, however, also pleads for the opposite direction of causation. Individuals who are already at a high level of cognitive functioning may prefer activities that are cognitively demanding. This leads to a competing Hypothesis 1, stating that respondents with good cognitive functioning engage in activities that are cognitively demanding. In line with this is the hypothesis of Hultsch and colleagues 1999, which reads that high-ability individuals lead intellectually active lives until cognitive decline in old age limits their activities.
If everyday activities indeed enhance cognitive functioning, our second question is whether specific activities differ in their impact on cognitive performance. Several studies suggest that specific activities, rather than an activity per se, affect cognitive functioning. Activities suggested to significantly affect cognitive functioning are activities that are cognitively demanding (Hultsch et al. 1993). In line with this are studies of Kohn and Schooler (Kohn and Schooler 1978; Schooler, Mulatu, and Oates 1999), stating that substantively complex work improves intellectual functioning. Finally, Arbuckle, Gold, Andres, Schwartzman, and Chaikelson 1992 revealed that activities in which social support is received results in maintenance of cognitive functioning. However, a classification of everyday activities into cognitively demanding or socially supportive activities is not straightforward. Some activities may be both cognitively demanding and supportive, whereas other activities may well differ across individuals in the extent to which the activity is cognitively demanding or supportive. For example, some people may visit a chess match as a spectator because of their interest in chess, whereas others do so to meet their friends. Lawton 1993 argued that a classification that discriminates best between the universe of possible activities is one based on the meaning of the activity. Adapting Lawton's classification, we distinguish social, experiential, and developmental activities. Social activity includes three subcategories, that is, social interaction, social status, and service, such as volunteering. Experiential activity is characterized by the intrinsic satisfaction of the activity. It includes activities that are engaged in to find relaxation, or relief from social contexts. Developmental activity, including intellectual and creative activities, is meant to help oneself become something, or change in some way. This type of activity thus possesses an instrumental component. Applying the concept of cognitive demand to the categorization of Lawton 1993, we expect developmental activity to best reflect activities that are cognitively demanding, and as such enhance cognitive performance. Furthermore, social support is likely to be generated by social activities, hence we expect social activities to also enhance cognitive functioning. As experiential activity is not necessarily cognitively demanding activity, nor activity in which social support is received, we do not expect any effect on cognitive functioning (Hypothesis 2).
Finally, we focused on the extent to which various cognitive functions are enhanced by everyday activities, or whether everyday activities are enhanced by cognitive functions. On the basis of the speed hypothesis of Salthouse 1996 and arguments of Hultsch and colleagues 1993, we expected that information-processing speed would be less sensitive to everyday activities than memory and various types of nonverbal intelligence. Furthermore, as the slowing of information-processing speed precedes decline of higher order cognitive functions, we expected that whenever the level of activity is affected by cognitive functioning, it is most readily seen for information-processing speed (Hypothesis 3).
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). The LASA sample is stratified by age and sex, and there was an oversampling of older and male participants at baseline. The sample is drawn from the population registers of 11 municipalities in three culturally distinct geographical areas in The Netherlands. The LASA sample was initially recruited for the study, "Living arrangements and social networks of older adults" (LSN; Knipscheer, De Jong Gierveld, Van Tilburg, and Dykstra 1995). Since the beginning of the LASA study, three cycles have been conducted with a time interval of approximately 3 years. The first cycle took place between October 1992 and March 1993. Of the 3,805 participants in the LSN sample, 3,107 persons aged 55 to 85 years took part in the first (T1) LASA cycle (1992/1993). Of the 698 LSN respondents who did not participate in the LASA study, 126 (18%) had died and 134 (19%) were unable to participate in the study because of severe physical and/or mental health problems. Furthermore, 394 (56%) refused to be reinterviewed, and 44 (6%) could not be contacted. In the present study we selected data from the first and third cycles, which yields a 6-year time interval. Of the 3,107 LASA respondents, 2,076 (67% of 3,107) were able to participate in the third (T3) cycle (1998/1999). Attrition between T1 and T3 was caused primarily by mortality (761 persons, 24%). Other reasons for attrition were refusal (160 persons, 5%), frailty (81 persons, 3%), and failure to contact (29 persons, 1%). Attrition, not attributable to mortality, is associated with low cognitive performance, low level of everyday activity, high age, being male, and a lower level of education (all ps < .001) at the first LASA cycle, resulting in a selection of respondents who were relatively younger and more highly educated and active, with relatively good cognitive performance. Although our sample is clearly a survival sample, we still had a relatively large proportion (70%) of respondents with one or more chronic disease suggesting sufficient heterogeneity of general health in the study sample. Furthermore, the stratified sampling frame and the sample size guaranteed the inclusion of sufficient men (T1 49%; T3 45%), respondents in the highest age category (T1 M = 70.8, SD = 8.8; T3 M = 74.7, SD = 8.3), respondents with low level of education (T1 M = 3.4, SD = 2.0; T3 M = 3.5, SD = 2.0), and with low levels of activity (see Table 1 ).
|
Exogenous Variables
Baseline scores of age, gender, health, and level of education are considered exogenous variables as these are known to be associated with cognitive performance (Brayne, Gill, Paykel, Huppert, and O'Connor 1995; Gribbin, Schaie, and Parham 1980; Holland and Rabbitt 1991; Perlmutter and Nyquist 1990). Apart from their potentially confounding effect on cognitive performance, age and gender were included because of their use as sample stratification variables. Functional ability was used as an indicator of health. Scores were based on ability to perform three activities: walking up and down a staircase with 15 steps without having to stop, using own or public transportation, and cutting one's own toenails (Kriegsman, Deeg, Van Eijk, Penninx, and Boeke 1997). The respondents were asked to indicate whether they had difficulty in performing the activity, whether they needed help, or whether they were unable to perform it at all. Functional ability ranged from 0 to 9, with higher test scores indicating better physical functioning. The functional ability index has good reliability (Cronbach's 
= .75 at T1). Level of education was assessed by asking the respondent for the highest educational course completed, resulting in a nine-categories variable ranging from 1 (incomplete elementary education) to 9 (university education).
Endogenous Variables
Everyday activities.
Information was collected on 23 different everyday activities. We first selected those activities in which at least 4% of the respondents were engaged, to ensure sufficient variability in the variables. The selected 16 activities were then assigned to the three main categories of everyday activity—that is, social, experiential, and developmental activity—by eight colleagues of our research team. Agreement on assignment (at least six out of eight raters) was observed for 11 activities, which were used for further analyses.
For social activity three out of five social activities showing the highest intercorrelation were selected: visiting church services (yes/no), visiting neighborhood associations (yes/no), and visiting meetings of an organization for helping older adults, neighbors, or handicapped persons (yes/no). For experiential activity three out of four experiential activities showing the highest intercorrelation were selected: making a trip to the forest, dunes, zoo, or entertainment park; visiting a cultural institution such as a museum, theater, or cinema; and visiting a café or restaurant (all answer categories ranged from 1 [never] to 7 [every day]). For developmental activity two different activities were selected: following an educational course or study during the past 6 months (yes/no), and doing outdoor sports (1 [never] to 7 [every day]). Finally, with the LISREL8 program (Joreskog and Sorbom 1993) we tested whether change in the measurement model of everyday activity could be assumed to be a true change of everyday activity. This change, also referred to as alpha change (Golembiewski, Billingsley, and Yeager 1976), which is the level of change given a constant calibrated instrument and conceptual domain, is assumed to be present when the covariance matrices of the indicators in the measurement model are equal at T1 and T3. We therefore performed a two-group analysis for the latent variables experiential, social, and developmental activity, under the assumption of equal covariances at both time points. The models for experiential activity and social activity revealed good fitting models, 
2 (6, N = 1697) = 13,35, CFI = .99, sRMR = .02, RMSEA = .02, and 2 (6, N = 1693) = 8,81, CFI = 1.00, sRMR = .02, RMSEA = .01, respectively. However, for the two indicators of developmental activity, no such model could be found. It was concluded that, although doing outdoor sports and attending educational courses both were rated as indicators of a developmental activity, their content was too different to represent a homogeneous dimension. As attending educational courses best reflects the idea of a cognitively demanding activity, we selected studying as a single indicator of developmental activity.
Cognitive functioning.
The cognitive functions involved in this study were those commonly found to deteriorate with aging—that is, immediate recall and learning—as indicators of memory performance, fluid intelligence, and information-processing speed. They reflect broadly the cognitive functions currently distinguished in the cognitive aging literature (Baltes 1993; Lindenberger and Baltes 1997). For comparative reasons we further included the MMSE (Folstein, Folstein, and McHugh 1975), because of its widespread use as a screening instrument for cognitive functioning.
The MMSE score involves indications of recall, orientation, registration, attention, language, and construction. Scale scores range from 0 to 30. Higher scores on the MMSE indicate better cognitive performance. At T1, Cronbach's alpha was .69 and at T3, Cronbach's alpha was .61. Although the alpha we observed is low for a 23-item scale, it is comparable to alphas found in other population-based surveys. Moreover, the MMSE is judged to assess the severity of cognitive impairment and cognitive changes satisfactorily (Tombaugh and McIntyre 1992).
The 15 Words Test (Saan and Deelman 1986), derived from the Auditory Verbal Learning Test (Rey 1964), was chosen for the assessment of immediate recall and learning. The procedure started with a verbal presentation by the interviewer of 15 words. Immediately after the presentation, the respondent was asked to repeat as many words as possible. The same procedure took place three times, using the same 15 words, to obtain an indication of immediate recall (score on the first trial) and learning (average score of the three trials). Subsequently, for a duration of approximately 20 min, the respondent performed a different nonverbal task. After this, the respondent was asked to recall as many words as possible of the 15 Words Test, to obtain an indication of the delayed recall function. The respondent was not prepared for this last trial. For the subsequent cycles, parallel versions of the 15 listed words were used. The different words of the parallel versions are comparable with respect to the frequency of daily occurrence, number of syllables, the stage of life at which they are acquired, and mental imagery. As in many tests involving learning (Lezak 1995), possible practice effects were observed in the 15 Words Test. Practice effects may have been due to the fact that respondents remembered the delayed recall test, for which they were unprepared at the first measurement cycle. They may therefore have listened more carefully to the words during the third measurement, which resulted in a better overall score. For this reason we used only those indicators of memory performance in our study that showed no or only small mean improvement. The scores on the 15 Words Test consisted of the number of words correctly remembered per trial, resulting in a score range of 0 to 15 for each attempt. The scores on the immediate recall ranged from 0 to 12. The scores on learning ranged from 0 to 14. The bivariate correlation of the score on the first trial at T1 and the score on the first trial at T3 was .44. The bivariate correlation for the second trial was .53, and, for the third trial, .54.
Raven's Coloured Progressive Matrices (Raven, Raven, and Court 1995) was used to measure fluid intelligence, or the ability to deal with essentially new information. In pilot studies a high correlation was observed between the sum score of the total test and the sum score of Tests A and B (.96). To save time in the interview, Set Ab was not included in this study. Sets A and B each consisted of 12 pages, each page displaying a different pattern, from which one section was missing. At the bottom of each page, six patterns were printed, and the respondent was asked to choose which of these six patterns best fitted into the missing section. The test score was the number of correctly chosen patterns and ranged from 0 to 24. At T1, the internal consistency calculated for ordinal variables (KR 20) was .97, and at T3 KR 20 was .96.
An adaptation of the Coding Task (Savage 1984) was used to assess information-processing speed. The respondents were presented a sheet on which rows of characters were printed. They were asked to name the character that belongs underneath the printed characters, and to work as quickly and accurately as possible. The correct letter combination could be read at the top of the page. This was repeated in three trials of 1 min each. The score for each trial of the Coding Task consisted of the number of completed combinations. Scores on the first trial were used because of its lowest rate of nonresponse.
Procedure
To overcome problems of reversed causation and confounding variables in research on causality, it is recommended to use a linear structural equations approach, which includes reversed effects and confounding variables, as well as a measurement model to account for errors in the measurement (Zapf, Dormann, and Frese 1996). Accordingly, we applied a cross-lagged regression model (Bynner 1994) as in Fig. 1 (full model) to evaluate our hypotheses. Observed variables are enclosed in boxes; latent variables, in ellipses.
|
, we allowed for residual correlations (shown as curved two-way arrows) among the latent variables at T1 (
21) and among the latent variables at T3 (43) to control for the effect of unknown confounding variables. The error terms of the indicators of the latent variables social and experiential activity were allowed to correlate over time. No correlated error terms for the single-indicator latent variables (all cognitive variables and developmental activity) were modeled. The error terms of the single indicators were fixed at a specific value to account for the unreliability of the measures (Bollen 1989; Hayduk 1987). The fixed values were determined by multiplying the proportion of error variance (1 - 
) of the indicator by the variance of the indicator. Reliability was based on Cronbach's alpha (MMSE), KR 20 (fluid intelligence), and test–retest reliability (immediate recall, learning, and information-processing speed). For developmental activity no such reliability measure was available. We therefore made an arbitrary decision about the amount of error variance for the indicator, on the basis of the idea that attending a course does not perfectly measure cognitively demanding activities. We decided to fix the error variance for attending a course at 30%, which seems a reasonable guess compared with other estimated error variances in the models. Apart from these effects, the direct effects (gammas) from the exogenous variables age, gender, level of education, and functional ability on the latent variables (cognition and activity) at T1 and T3 were estimated (arrows not shown in the figure). All effects were standardized. The right time interval for causality to show up, however, is highly dependent on the characteristics of the relationships. For example, when the relation between cognitive functioning and everyday activity is synchronous, that is, the reaction occurs almost immediately after "exposure," then a time interval of 6 years may be too long. To investigate this possibility, we tested another type of model (see Fig. 2) in which the correlation of error terms between the latent variables everyday activity and cognition at T1 was replaced by two direct cross-sectional effects, all other effects being equal.
|
The model shown in Fig. 1 was applied to all combinations of cognitive functions and everyday activity, varying only the two cross-lagged effects between the endogenous latent variables everyday activity and cognitive performance. This procedure was repeated for the model in Fig. 2. Each model was estimated with a step-wise procedure starting with an empty model, which is the full model without the cross-lagged effects ß41 and ß32 (see Fig. 1) and cross-sectional effects ß21 and ß12 (see Fig. 2). Subsequently, when it became necessary to obtain a fitting model, we tested both cross-lagged effects separately, as well as their joint effect within one model. As having too many effects included in one model may result in capitalization on chance (overidentification), we selected the most parsimonious model that still fit the data. The significance of the betas was tested by inspection of the t values and by evaluating several indicators of fit of the total model, on the basis of criteria described by Jaccard and Wan 1996. When both cross-lagged, or cross-sectional, effects appeared to be significant, predominance of one of the two effects was tested by imposing equality constraints on the betas. When equality constraints on the cross-lagged, or cross-sectional effects resulted in an unacceptable fit of the model, we concluded that the effect with the highest beta was predominant.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The estimated gammas reflecting the effects of the exogenous variables on the latent variables revealed that age significantly affects social and developmental activity at T1 (
= -.10 and -.20, respectively). At T3 all types of activity were negatively affected by age ( = -.30 for social activity, -.20 for experiential activity, and -.15 for developmental activity). Cognitive functioning was negatively affected by age at both time points (ranging from -.22 to -.44). Gender was positively related to everyday activities (ranging from .06 to .18), indicating that women have a higher level of activity. Women had higher scores on cognitive functioning (ranging from .07 to .35), fluid intelligence excepted. Level of education positively affected experiential activity and developmental activity (.48 and .21, respectively), but not social activity. A higher level of education was related to higher levels of cognitive functioning (ranging from .26 to .40). Functional ability was positively related to all cognitive functions (ranging from .07 to .19) and experiential activity (.15), indicating that better health was related to higher levels of cognitive functioning and higher levels of everyday activity. Social and developmental activity were not related to functional ability.
The bivariate correlations among all variables by type of activity are presented in Table 2 Table 3 Table 4 . According to the first-order correlations for the single indicators, there is almost no association between the indicators of social activity and cognitive functioning (see Table 2 ). Correlations ranged from -.06 to .09. However, indicators of experiential (see Table 3 ) and developmental activity (see Table 4 ) correlated positively with cognitive functioning (at T1 .07 to .21 and .14 to .28, respectively). The next step was to test whether these correlations are sustained when all the other variables assumed to be associated with these relations are controlled for.
|
|
|
|
Our competing hypothesis stated that positive cross-lagged effects are to be expected of cognitive functioning on everyday activities, at least for activities that can be viewed as being cognitively demanding. With respect to this hypothesis, the cross-lagged effects ß32 were considered. In one of the five models involving developmental activity, which is viewed as a cognitively demanding activity, a positive cross-lagged effect (ß = .14) of cognitive functioning on developmental activity was observed. We may conclude that with respect to information-processing speed the competing hypothesis should not be rejected. However, no evidence was found for other aspects of cognitive functioning.
Hypothesis 2 specified that developmental activity and social activity enhance cognitive performance. However, as we observed already, no cross-lagged effects were present. Thus, the expected positive effects of social and developmental activity on cognitive functioning are clearly not supported by the data. We therefore rejected our second hypothesis.
Hypothesis 3 focused on the extent to which cognitive functions are differently affected by everyday activities and whether information-processing speed was more predictive of everyday activity than other cognitive functions. It was expected that, of the distinguished cognitive functions, information-processing speed would be least affected by everyday activities and would have the largest effect on everyday activities. The first part of this hypothesis was already rejected, and we therefore focused on the last part of the hypothesis, stating that information-processing speed has the largest effect on everyday activities. We observed a positive effect (.14) of information-processing speed on developmental activity, which is in line with our hypothesis.
None of our expectations were supported by the data, except for the effect of speed on developmental activity. One of the reasons for a lack of significant effects may be the length of the time lag. We therefore applied the second model (from Fig. 2), in which the error correlation between the latent variables everyday activity and cognitive function was replaced by two direct effects (ß12 and ß21) to test for a possible synchronous effect between everyday activities and cognitive functions. The results of the additional analyses indicated that the idea of synchronicity does not hold. None of the cross-sectional effects at T1 were significant.
In sum, our study could not provide evidence about a causal effect of everyday activities on cognitive functioning or the other way around. The only effect found (i.e., the effect of information-processing speed on developmental activity) is in favor of the reversed causal hypothesis, stating that respondents with good cognitive functioning prefer cognitively demanding activities. One of the reasons for these small effects compared with other studies in this field may be that we controlled for the effect of unknown confounding variables. Without such a control, the relation between level of activity and cognitive functioning observed in other studies may be the result of spurious relationships. To illustrate that these effects can be observed, we removed the error correlations between the latent variables at T1 and T3 and reanalyzed all 15 models. This revealed additional significant cross-lagged effects from experiential activities on MMSE (.08), immediate recall (.12), and information-processing speed (.06), and one from social activity on learning (.09). These four effects suggest that everyday activities enhance cognitive functioning over time. However, as these effects are the result of fixing the residual correlation at T1 (12) and T3 (34), they reflect spurious effects. A fifth cross-lagged effect (.06) appeared for MMSE on developmental activity, which seems to confirm that the reversed causal direction is also a spurious effect.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
One of the problems we encountered was that attrition has been selective, leaving a relatively younger sample of respondents. However, because of an over-sampling of men and older persons at baseline, the final sample still showed sufficient variability in the study variables, which alleviated the biasing influence of attrition. Furthermore, a study period of 6 years may be too short to demonstrate substantial effects between everyday activities and cognitive functioning. However, increasing the study period undeniably enlarges the biasing impact of selective attrition of respondents. Moreover, the fact that the stability of both cognitive functions and everyday activities was rather high further limited the possibility of finding substantial cross-lagged effects. Finally, only one activity supposed to be cognitively demanding, namely, following a course or study, was included in the analysis. Although this indicator is considered an outstanding example of a cognitively demanding activity, the selection of other cognitively demanding variables may have yielded different results.
Although several studies report positive effects of an active lifestyle on cognitive functioning, our study did not provide evidence for this supposed causality. Regarding the type of cognitive function affected, our findings do not contradict the results reported by Gold and colleagues 1995, as they only found positive effects on crystallized intelligence, and not on indicators of fluid intelligence, which were the only cognitive functions we selected in our study. In the study of Hultsch and colleagues 1999, positive effects of novel information processing and change in information processing were found on working memory, which has a fluid component. However, as working memory was reported to have a high loading on fact recall (.89), which has a crystallized component, we suggest that the positive findings reported by Hultsch and colleagues may be valid for crystallized intelligence, but not for fluid intelligence.
The positive effects of experiential activities on cognitive functions that we found as a result of removing the error correlates between the latent variables cognition and activity point toward the existence of an underlying concept that causes spurious relations between experiential activities and maintenance of cognitive functioning. As this concept is not further specified, we can only speculate on its nature. However, the positive correlation between experiential activities and level of education (see Table 3 ) makes it likely that this concept has much in common with SES. SES is not only related to cognitive functioning, but also to specific activities (Ganzeboom 1989). The partial support for our hypothesis stating that respondents with good cognitive functioning engage in activities that are cognitively demanding is in line with this suggestion. SES is related to a variety of living conditions and lifestyles (Bond and Coleman 1993), such as drinking and smoking behavior and involvement in sports (Tuinstra, Groothoff, Van den Heuvel, and Post 1998). It is found that higher SES decreases the risk of functional decline during later life (Boult, Kane, Louis, Boult, and McCaffrey 1994). Analogous to this finding, a cumulative effect of living conditions and lifestyles over the life course on cognitive decline could exist. In sum, we suggest that not the activity in itself, but rather specific lifestyles and living conditions to which the types of activities engaged in are closely connected, may in fact be responsible for the positive relation between specific activities and cognitive functioning.
The positive effect of information-processing speed on developmental activity indicates that our study is mostly in favor of Hultsch and colleagues 1999, who stated that high-ability people lead intellectually active lives. Whether or not decline in cognitive functions results in the limitation of activities remains an open question.
Further research on the relation between activities and maintenance of cognitive functioning may benefit from a focus on different aspects of living conditions and lifestyles. Therefore, although Lawton's categorization based on the meaning of the activity (Lawton 1993) facilitated a selection of various types of everyday activities, our results indicate that a more fruitful categorization may focus on living conditions and lifestyle. This may shed new light on the underlying mechanisms related to changes of cognitive functioning in old age.
|
Acknowledgments |
|---|
Received for publication November 10, 1998. Accepted for publication October 13, 2000.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. B. Hall, R. B. Lipton, M. Sliwinski, M. J. Katz, C. A. Derby, and J. Verghese Cognitive activities delay onset of memory decline in persons who develop dementia Neurology, August 4, 2009; 73(5): 356 - 361. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Coley, S. Andrieu, V. Gardette, S. Gillette-Guyonnet, C. Sanz, B. Vellas, and A. Grand Dementia Prevention: Methodological Explanations for Inconsistent Results Epidemiol. Rev., November 1, 2008; 30(1): 35 - 66. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. S. Wilson, P. A. Scherr, J. A. Schneider, Y. Tang, and D. A. Bennett Relation of cognitive activity to risk of developing Alzheimer disease Neurology, November 13, 2007; 69(20): 1911 - 1920. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. A. M. Bielak, T. F. Hughes, B. J. Small, and R. A. Dixon It's Never Too Late to Engage in Lifestyle Activities: Significant Concurrent but not Change Relationships Between Lifestyle Activities and Cognitive Speed J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2007; 62(6): P331 - P339. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Fritsch, M. J. McClendon, K. A. Smyth, A. J. Lerner, R. P. Friedland, and J. D. Larsen Cognitive Functioning in Healthy Aging: The Role of Reserve and Lifestyle Factors Early in Life Gerontologist, June 1, 2007; 47(3): 307 - 322. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Ghisletta, J.-F. Bickel, and M. Lovden Does activity engagement protect against cognitive decline in old age? Methodological and analytical considerations. J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2006; 61(5): P253 - P261. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
