| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Prince of Wales Medical Research Institute, University of New South Wales, Randwick, Australia
b School of Psychology, Flinders University of South Australia, Adelaide
c Department of Speech Pathology, Flinders University of South Australia, Adelaide
Kaarin J. Anstey, Prince of Wales Medical Research Institute, Barker St., Randwick, 2031, Australia E-mail: k.anstey{at}unsw.edu.au.
Toni C. Antonucci, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
SEVERAL empirical studies have demonstrated a robust association between sensory and cognitive function in old age (Anstey 1999
; Anstey, Lord, and Williams 1997; Anstey and Smith 1999; Baltes and Lindenberger 1997; Lindenberger and Baltes 1994; Salthouse, Hambrick, and McGuthry 1998; Salthouse, Hancock, Meinz, and Hambrick 1996; Stankov 1986). A common cause explanation suggests that this association occurs because "both sets of measures are an expression of the physiological architecture of the aging brain" (Baltes and Lindenberger 1997, p. 13). Measures of sensory acuity have been reliably shown to explain large amounts of age-related variance in cognition. For example, Lindenberger and Baltes 1994 found that sensory variables explained more than 90% of the age-related variance in a general cognitive factor. Likewise, Anstey and Smith 1999 found that a latent variable of indicators of biological age (including vision, hearing, forced expiratory volume, vibration sense, and grip strength) fully mediated the relationship between age and a general cognitive factor. Studies such as these indicate that age differences in sensory functioning may provide a window through which to view age differences in cognitive function (Baltes and Lindenberger 1997).
Interpreted in the broadest sense, a common factor theory of cognitive aging predicts that all factors mediating age-cognition relations are measures of the same common factor. For example, both processing speed and sensory function should explain the same portion of age-related variance in cognitive function. The role of processing speed has been central to theories of cognitive aging for more than 30 years (Birren 1965; Salthouse 1991). Many studies have shown that speed reduces or eliminates the age effect on a range of cognitive tasks (e.g., Bors and Forin 1995; Bryan and Luszcz 1996; Lindenberger, Mayr and Kliegl 1993; Nettelbeck and Rabbitt 1992; Salthouse 1992a, Salthouse 1992b, Salthouse 1996). It is therefore pertinent for researchers to evaluate the importance of sensory function as a mediator of the relationship between age and cognition concurrently with processing speed.
Lindenberger and Baltes 1994 evaluated the relative importance of speed and sensory function in cross-sectional models of age differences in cognition. In one model, they found that speed did not fully mediate the effect of age on sensory function but did fully mediate the effect of age on cognition. In a second model, they found that the effect of age on speed was fully mediated by sensory function, and the effect of sensory function on cognition was fully mediated by speed. They concluded that, although speed and sensory function were equivalent in their capacity to mediate the association between age and cognition, vision and hearing were more important predictors of age differences because they explained all age-related variance in cognition (including speed) whereas speed did not explain all the age-related variance in sensory function. These authors did not find that sensory function explained any more age-related variance in cognitive function than processing speed. In terms of models of cognitive aging, sensory function and speed were equally powerful mediators. However, in terms of explaining overall age differences, sensory function was a more powerful predictor than speed. That is, compared with speed, sensory function explained a larger proportion of the variance in age, consistent with the view that sensory function is a reliable biomarker (Anstey and Smith 1999).
Another approach to evaluating mediators of the relationship between age and cognition is in terms of the mediators' theoretical independence from cognitive function (Lindenberger and Potter 1998). Measures of processing speed are essentially cognitive variables may overlap theoretically and empirically with measures of cognitive function. For example, many measures of cognitive function are conducted under time limits, so the explanatory power of processing speed may be partly due to the fact that it measures this speeded aspect of cognitive performance. Furthermore, measures of processing speed often entail a small memory load, as do many measures of cognitive function (Piccinin and Rabbitt 1999). For example, participants must often hold elements of the cognitive task in their mind while performing cognitive operations. Measures of sensory function are from a qualitatively different domain than measures of cognitive function. Although they involve some cognitive processing in that participants must understand test instructions, sensory acuity tests are not speeded or timed.
On the other hand, the fact that sensory function alone (Lindenberger and Baltes 1994) and in conjunction with other biomarkers (Anstey and Smith 1999) has been shown to explain all age differences in measures of cognitive abilities may indicate that sensory acuity is just a proxy for age (Salthouse et al. 1998). It is possible that sensory function is no more than an index of age or time. If this is the case, sensory acuity may not be a meaningful mediational construct in this context and does not provide a substantive explanation of age differences in cognition. Consequently, the recent emphasis placed on sensory function as a mediator of age-cognition relations may be misguided, and the empirical findings relating to the importance of sensory function may be spurious, a proposition raised by Salthouse and colleagues 1998. At the conceptual level, this criticism may be refuted: one would argue that even if the association between sensory variables and cognition is due to the fact that sensory variables are good measures of age, then it is still more informative to relate age differences in cognition to a physiological variable than to a measure of time. This is particularly true in late adulthood, when individual differences in biological aging increase.
Empirical evidence against the "spuriousness" interpretation of the association between sensory function and cognition in old age would be found if sensory acuity explained individual differences in cognitive function that were independent of age differences, or if experimental manipulations of sensory acuity resulted in changes in cognitive performance. Anstey and Smith 1999 reported that biomarkers including measures of sensory acuity explained individual differences in measures of crystallized intelligence that were independent of age. Studies of young adults have also shown associations among sensory variables and individual differences in intelligence (Li, Jordanova, and Lindenberger 1998; Roberts, Stankov, Pallier, and Dolph 1997). Experimental evidence for the effect of sensory deficit on cognitive performance was also reported by Dickinson and Rabbitt 1991, although Lindenberger, Scherer, and Baltes 1999 did not find a significant effect of simulated sensory deficit on cognitive performance.
Explanations such as the common cause model (Lindenberger and Baltes 1994) relate the domains of sensory and cognitive performance at the level of brain functioning. The common cause model does not necessarily exclude the possibility of small specific associations existing between sensory and cognitive variables that are not shared with the more significant common factor. It is possible that, in addition to generalized brain aging, peripheral changes in sense organs affect perceptual and cognitive processing. Age-related loss of sensory receptors and neurons may result in slowing of perceptual processing and less effective and slower encoding of new information. This may directly lead to slowing of cognitive processing and increased errors on cognitive tasks. In the present study we reexamined the role of sensory function as a mediator of the association between age and cognitive abilities in late adulthood in a large population-based sample.
Our specific goal in this study was to see whether the empirical findings Lindenberger and Baltes 1994—of a large proportion of shared age-related variance among sensory function, age, and cognition—would be replicated in the Australian Longitudinal Study of Ageing (ALSA). The sensory measures used in the ALSA and the age range of the sample are similar to those used in the Berlin Aging Study (BASE), but the cognitive battery is much smaller and includes measures of episodic memory that were not included in BASE. Nevertheless, if a common factor is responsible for the observed association among cognitive and sensory function generally, it is reasonable to expect that this factor would operate for a wide range of cognitive measures. Similar to Lindenberger and Baltes 1994, in this study we used speed as a part of the cognitive factor for evaluating the general model, and then we removed speed and used it as a mediating factor in subsequent models.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
, for more details). The ALSA used the South Australian Electoral Roll as a sampling frame to identify households with residents 70 years and older. The sample was stratified by age and sex into three 5-year cohorts: 70–74, 75–79, 80–84, and a fourth cohort of individuals older than 85. Individuals randomly sampled within these cohorts were invited to participate in the ALSA on a voluntary basis. The participation rate for the baseline data collection (Wave 1) was 55%. Data were collected in two phases with two different formats. A comprehensive 2-h home interview was followed by a further individual clinical assessment conducted approximately 2 weeks later. The home interview yielded demographic data and information on self-rated health, depression, medical conditions, cognitive status, memory, and subjective measures of vision, audition, and physical performance. Individual clinical assessments provided objective cognitive and sensory data. For the first wave of the study, 1,947 participants (1,039 men) were interviewed, and 1,511 underwent portions of the clinical assessment. The sample for the present study comprised participants who completed the clinical assessment and interview and had complete data on the variables used in structural equation modeling. This included 894 participants aged 70–98 (M = 78.16, SD = 6.69) of whom approximately 51% were male.
Measures of Cognitive Function
Most of the cognitive measures have been described more fully elsewhere (Luszcz et al. 1997). Some were based on measures developed as part of the Canberra Inventory for the Elderly (CIE; see Christensen et al. 1994); these included Similarities, Definitions, and Address Memory. Measures are grouped according to the latent variables used in the structural equation models.
Verbal
Verbal skills were assessed with four measures.
) assessed verbal reasoning. They were scored 0 (incorrect) or 1 (correct) to give a possible total of 3. ), participants defined three items (repairman, terminate, compassion), which were scored 0 (incorrect) or 1 (correct) for a possible total of 3. Test-retest reliability was .69. ) was used for assessment of confrontation naming. The four different versions of this test were counterbalanced across participants. They were shown drawings, one at a time, and asked to name the object illustrated. According to standard instructions, if participants were unable to name the object or seemed confused about what it was, they were given a stimulus cue (semantic cue). If they were still unable to name the item, a phonemic cue was given. Up to 20 s were allowed for each response. If participants were still unable to name the item after both cues had been presented, the label was provided. This procedure ensured that participants knew the name of each item. A score of 1 was given for each item correctly named without cues, for a maximum possible total of 15. Observed scores on confrontation naming ranged from 6 to 15. ) measures verbal knowledge as an element of crystallized ability (Crawford 1992). It comprises 50 infrequent words of irregular pronunciation that respondents are asked to read aloud. The number of items named correctly was used. Test-retest reliability was .83 (Luszcz et al. 1997).
Speed
The Digit Symbol Substitution (DSS) subscale of the WAIS-R (Wechsler 1981) was used for assessment of processing speed (Bryan and Luszcz 1996; Salthouse 1991). The participant was required to substitute symbols corresponding to the numbers 1 through 9 into a randomly ordered array of 93 digits. Symbols to be used were available throughout the task on a code sheet illustrating the 9 digit-symbol pairs. The participant was required to make substitutions as rapidly as possible. The number of substitutions completed correctly in 90 s made up the measure of processing speed. Observed scores ranged from 0 to 67. Test-retest reliability was .79 (Luszcz et al. 1997).
Memory
Symbol, picture, and address memory were assessed.
). ). Participants were asked (without prior warning) to recall the 15 pictures immediately after the test. ). The name and address were repeated up to five times until the respondent remembered them. Immediate recall after the first trial for the five components of the item and delayed recall a few minutes later were recorded. The maximum score was 10.
Vision and hearing measures.
Vision and hearing were measured as follows.
Statistical Analysis
We conducted structural equation modeling to evaluate alternative multivariate models of the interrelationships among the sensory and cognitive variables and age. For all analyses we analyzed the raw data matrix using Maximum Likelihood. If necessary, the direction of scoring of sensory and cognitive variables was reversed so that higher scores indicated better functioning. Isolated missing values of pure tone thresholds were imputed on the basis of the participant's entire audiogram by the third author, a practising audiologist. No other missing data were imputed.
In the measurement model, all loadings were free, factor variances were fixed at 1, and covariances among all factors were estimated. We then specified structural models to depict alternative interpretations of the data and tested them for significance. There were three cognitive factors (Speed, Memory, and Verbal) and two sensory factors (Vision and Hearing). Three groups of models were tested. The first group was based on the common cause model reported by Lindenberger and Baltes 1994. The second and third models were also based on models reported by Lindenberger and Baltes 1994 and involved removing Speed as a cognitive factor and using it as a mediator.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
2 = 36.6, p < .01).
|
|
; Horn 1987).
|
. Memory, Speed, and Verbal loaded onto a single factor called Cognition. In this model the effect of Age on Cognition was mediated by Vision and Hearing, and there was no direct effect of Age on Cognition. The larger than desirable ratio of 2 to degrees of freedom in these analyses is partly due to the large sample size in the present study. All other goodness of fit indices indicated that this model was highly acceptable (Table 2 ). A second version of this model was tested that included the path from Age to Cognition and resulted in a significant improvement in fit (2 = 17.44, df = 1, p < .01). This model is depicted in Fig. 1.
|
Estimation of the Age-Related Variance in Cognition Shared With Sensory Function
To allow for a direct comparison of results between the present study and that of Lindenberger and Baltes 1994, we calculated the proportion of age-related variance in Cognition shared with sensory function. This analysis was conducted in latent space (Bentler 1995) where the square of the residual coefficient subtracted from 1 gives the multiple correlation coefficient for the equation. The variance in Cognition explained by Age was 22.56%, the variance in Cognition explained by Vision and Hearing was 31.94%, and the variance in Cognition explained by Age, Vision, and Hearing was 36.80%. Therefore, 78.46% of the Age-related variance in Cognition was shared with sensory function.
Speed as a Mediator of Sensory and Age Effects on Cognition
In the next model (SENSPEED1) the Speed factor was removed from the Cognition factor and used as a mediator of the effect of Age and sensory function on Cognition. In this model, Age had direct paths to Vision and Hearing, Vision and Hearing had direct paths to Speed, and Speed had a direct path to Cognition. This model also provided acceptable fit of the data but did not provide information about direct paths from Age, Vision, and Hearing to Cognition and Age to speed. In SENSPEED2, a direct path from Age to Cognition was included, resulting in a significant improvement in fit (2 = 15, p < .01; Table 2 ). In SENSPEED3, an additional direct path from Age to Speed was also included, and this also resulted in a significant improvement in fit (2 = 14.96, p < .01). In SENSPEED4, a path was added from Vision to Cognition. Although this path was significant, it did not improve the fit of the model. In SENSPEED5, the path from Vision to Cognition was removed and a path from Hearing to Cognition was added (Fig. 2). This resulted in a significant improvement in fit (2 = 13.23, p < .01) and was accepted as the final model of the relationships among Speed, Vision, Hearing, Age, and Cognition. When an additional model with additional direct paths from Age to Memory was tested, linear dependencies occurred for the path from Age to Cognition. A model testing the additional direct path from Age to Verbal did not converge.
|
that examined Speed as a mediator of the relationships between Age and sensory function and Age and Cognition. In this model, Speed was used as a mediator between Age and Vision, Age and Hearing, and Age and Cognition. SPEED1 included a direct path from Age to Speed and direct paths from Speed to Vision, Speed to Hearing, and Speed to Cognition. The fit of this model was barely acceptable (Table 2 ). The addition of a path from Age to Cognition in SPEED2 resulted in a significant improvement in fit (2 = 13.56, p < .01). In SPEED3 a path was added from Age to Vision, also resulting in a better fitting model (2 = 83.24, p < .01). Finally, in SPEED4, a path was added from Age to Hearing, resulting in a further improvement in fit (2 = 102.43, p < .01). Fig. 3 shows the standardized path coefficients for SPEED4. This model showed that Speed did not fully mediate the effect of Age on sensory function or Cognition as measured by Verbal and Memory factors.
|
Estimation of the Amount of Variance in Cognition That Speed Shared With Sensory Function and Age
We calculated multiple correlations in latent space to enable an estimate of the amount of variance shared between Speed and Age and Speed and sensory function. For these analyses Cognition was a second-order factor onto which Verbal and Memory loaded. For these analyses, the variance in Cognition explained by Age was only 19.9%, the variance in Cognition explained by Vision and Hearing was 21.50%, and the variance in Cognition explained by Speed was 51.30%. Of the Age-related variance in this cognitive factor, 54.27% was shared with Speed and 68.34% was shared with Vision and Hearing. Of the Vision and Hearing–related variance in cognitive factor, 53.02% was shared with Speed.
Summary of Results
We tested three groups of models that aimed to evaluate whether a single mediational factor could explain the relationship between Age and Cognition. The first group of analyses showed that the common factor model fitted the data well but that a model including a unique effect of Age in addition to sensory function provided the best fit. Nearly 80% of the Age-related variance in Cognition was shared with sensory function. The second group of analyses evaluated the role of Speed as a mediator of the effect of sensory function and Age on Cognition. Additional paths from Age to Cognition and Hearing to Cognition improved the fit of this model significantly, indicating that Speed did not mediate all the Age-related and Hearing-related variance in Cognition. A third group of models was tested in which Speed mediated the effect of Age on both sensory and cognitive function. These models showed that additional direct paths from Age to Vision, Age to Hearing, and Age to Cognition were significant. This indicates that a significant proportion of the Age-related variance in Vision, Hearing, and Cognition was not mediated by Speed. Compared with Age and sensory function, Speed shared a larger total amount of total variance with Cognition (i.e., the factor that comprised Verbal and Memory). However, of the total age-related variance in Cognition, a larger proportion was shared with sensory function than with Speed.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
and support the view that a common factor explains most of the shared variance among cognition, age, speed, and sensory function. However, significant unique effects of age, speed, and sensory function were not shared, indicating that small, unique effects of these factors must be accounted for in theoretical accounts of the relationships among these variables. The finding of unique effects of age and sensory function in the present study, which were not present in the Lindenberger and Baltes study, is likely due to the much larger sample size and smaller cognitive test battery we used.
The cognitive battery in the present study was limited by the small number of test items on some of the verbal tasks, the single indicator of speed, and the limited number of cognitive factors represented. This was a result of time limitations placed on the cognitive component of this large epidemiological study. Although the DSS is a well validated measure and widely used in cognitive aging research (Salthouse 1992b), additional of indicators of speed would have improved the measurement of this construct for use in structural equation modeling. Nevertheless, speed was retained as a separate factor in this study because of the important role that it has had historically as a mediator in cognitive aging research. Because of the limited number of cognitive factors represented in this study, analyses involving speed were conducted on a cognitive factor comprised only of verbal and memory factors. Verbal did not have a strong association with age, which reduced the proportion of age-related variance in the cognitive factor. Previous analyses of these data have also shown that age remains a significant predictor of memory after control of cognitive and noncognitive variables (Luszcz et al. 1997). It may be that memory does not have as strong a connection with sensory function as do measures of general cognitive ability reported in other studies (Anstey and Smith 1999; Baltes and Lindenberger 1997). This may be due to age-related differences in strategy use being present on memory tasks but not present on tasks of spatial and verbal abilities (e.g., Moshe and Craik 1996).
In the present study, sensory function explained more age-related variance in cognition than that explained by speed, even though speed explained more of the total variance in cognition than sensory function. These results are consistent with those of Lindenberger and Baltes 1994 in showing that at the statistical level, sensory function is equally, if not more, important than speed as a mediator of age-cognition relations.
The measures of hearing did not include the highest frequencies, 6 and 8 kHz. These highest frequencies are known to be unreliable (Hickling 1966), and the measurable rate of change in the very old is greater in the speech range (0.5–2 kHz) frequencies because of loss of hearing in the highest frequencies (Brant and Fozard 1990). If age differences in cross-sectional patterns of hearing can be inferred from longitudinal patterns of hearing changes in old age, it is unlikely that the lack of these highest frequencies resulted in a significant reduction in the size of the association between hearing and age. Consequently, not including these frequencies would not reduce the likelihood of finding support for the common cause hypothesis.
The limitations of the sensory and cognitive test battery are somewhat compensated by the large sample used in this study in that the study had a larger statistical power. This increased statistical power also revealed unique effects of age and sensory function that may not emerge in a smaller sample. Readers should therefore consider the size of the effects found in the present study when considering the importance of the findings.
Compared with results reported by Lindenberger and Baltes 1994, we did not find that as large a proportion of the age-related variance in cognition was explained by speed, although the general results were replicated. It is likely that the discrepancy in the effect sizes between studies is due to the fact that in the present study the cognitive factor examined in relation to speed comprised only verbal and memory tasks, whereas in the BASE study the cognitive factor included measures of reasoning, memory, knowledge, and fluency.
The results of the models involving speed as a mediator also provide some evidence that the relationship between sensory and cognitive function in old age is not fully explained by age and is therefore not spurious. An independent effect of hearing on cognition was observed that was not mediated by speed.
Altogether these results present a complex picture of the relationships among sensory and cognitive function. It is possible that a common factor representing general age-related changes in neurophysiological integrity, along with specific age-related and sensory-related factors, contributes to individual differences in cognitive performance in very old adults. The specific factors may relate to both test taking and cognitive processing. Possible specific causes of sensory effects on cognition include age-related changes in sense organs and the effects of disease on specific parts of the brain.
Another possibility raised by our results is that different relationships pertain between memory and sensory function, compared with other general cognitive abilities and sensory function. Further research is required to determine whether specific disease processes are responsible for these specific relationships and whether a decline in both cognitive and sensory aging is indicative of pathological aging. At this stage we still do not know if changes in specific sensory abilities indicate changes in specific cognitive abilities.
Most of the research conducted in this field has used only threshold measures of sensory function. It is possible that other measures and methods, such as sensory discrimination tasks and signal detection analysis, may provide useful approaches to understanding the specific relationships among these factors. Sensory and cognitive performance in old age is also influenced by a number of contextual factors not included in the models presented here (Anstey and Smith 1999; Luszcz 1998). Researchers must conduct longitudinal and experimental analyses of the relationships among sensory and cognitive function to test and develop further hypotheses about the general versus specific nature of the effects of sensory function on cognition in old age.
|
Acknowledgments |
|---|
Received for publication October 7, 1998. Accepted for publication June 5, 2000.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  I. J Deary and G D. Batty Cognitive epidemiology J Epidemiol Community Health, May 1, 2007; 61(5): 378 - 384. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Evert, E. Lawler, H. Bogan, and T. Perls Morbidity Profiles of Centenarians: Survivors, Delayers, and Escapers J. Gerontol. A Biol. Sci. Med. Sci., March 1, 2003; 58(3): M232 - 237. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
