Journal Article Summary
Z19
Intraindividualvariabilityincognitivespeedandcognitiveperformanceinadulthood.pdf
Covariation of Intraindividual Variability in Cognitive Speed and Cognitive Performance Across Young, Middle, and Older Adulthood
Allison A. M. Bielak Colorado State University
Kaarin J. Anstey University of New South Wales
Intraindividual variability (IIV) in cognitive speed, or moment-to-moment changes in ability, is a developmental phenomenon indicative of neurological integrity that increases gradually across adult- hood. Past research has shown that IIV negatively covaries with cognitive performance, in which higher IIV at one occasion is associated with poorer cognitive ability at the same occasion. However, this association has been demonstrated only in older adulthood. Further, all past examinations of IIV change with cognitive change did not remove the average or between-person effect from within-person change in IIV. Using the PATH Through Life Study, we evaluated whether there were differences across 3 age cohorts (20 –24, 40 – 44, and 60 – 64 years at baseline) in the relationship between 8-year change in IIV and change in cognitive ability (N ! 7,485). Change in IIV was partitioned into between-person and within-person components, and multilevel models covarying for education, sex, diabetes, hypertension, and anxiety and depressive symptoms were conducted. IIV was negatively related to baseline cognitive performance at the between-person level. Notably, this relation was apparent and, in fact, strongest for those in young adulthood. Level of IIV was also negatively associated with cognitive change, but primarily for the youngest cohort. In contrast to previous research, there was minimal evidence of significant covariation in which within-person changes in IIV were associated with changes in cognitive performance, regardless of age group. Overall, IIV is a stable characteristic negatively associated with cognition in adulthood, but this link may primarily exist at the between-person level.
Keywords: intraindividual variability, inconsistency, cognition, adulthood, longitudinal
Supplemental materials: http://dx.doi.org/10.1037/dev0000688.supp
Intraindividual variability (IIV) in cognitive speed refers to fluctuations from trial to trial of a reaction time (RT) task. In adulthood, IIV is not random error but is a stable individual characteristic that is indicative of neurological integrity (Hultsch, MacDonald, & Dixon, 2002). Specifically, adults with various neurological conditions such as dementia (Gorus, De Raedt, Lam- bert, Lemper, & Mets, 2008), Parkinson’s disease (de Frias, Dixon,
& Camicioli, 2012), and traumatic brain injury (Stuss, Murphy, Binns, & Alexander, 2003) have been shown to have higher levels of IIV than healthy adults. Baseline IIV has also been shown to predict the transition to greater neurological disturbance, including mild cognitive impairment (Bielak, Hultsch, Strauss, MacDonald, & Hunter, 2010a; Cherbuin, Sachdev, & Anstey, 2010), dementia (Kochan et al., 2016; Tales et al., 2012), and even death (Mac- Donald, Hultsch, & Dixon, 2008). The hypothesis that IIV is a sensitive indicator of neurological disturbance is further corrobo- rated by neuroscientific evidence showing that greater IIV is correlated with smaller brain structures (Anstey et al., 2007; Fjell, Westlye, Amlien, & Walhovd, 2011), poorer brain functioning (Bellgrove, Hester, & Garavan, 2004), and poorer dopamine mod- ulation (MacDonald, Karlsson, Rieckmann, Nyberg, & Bäckman, 2012).
Interestingly, IIV is also a developmental phenomenon that increases across the adult life span, even in cognitively healthy adults (see Bielak & Anstey, 2015, for a review). Bielak, Cherbuin, Bunce, and Anstey (2014) showed that although adults aged 20 to 24 years showed practice related improvement (i.e., decreases) for IIV over 8 years, small but significant increases in IIV were evident for those aged 40 to 44 years, and those aged 60 to 64 years showed the largest increases over time. Therefore, IIV does not linearly increase throughout adulthood, but rather increases in inconsistency only begin to be detectable in middle age. Similar
This article was published Online First January 28, 2019. Allison A. M. Bielak, Department of Human Development and Family
Studies, Colorado State University; Kaarin J. Anstey, School for Psychol- ogy, Neuroscience Research Australia, University of New South Wales.
We thank the study participants, the PATH interviewers, Trish Jacomb, and Karen Maxwell. We would also like to acknowledge the contribution of Tony Jorm, Helen Christensen, Bryan Rodgers, Peter Butterworth, and Simon Easteal. Kaarin J. Anstey was supported by National Health and Medical Research Council (NHMRC) Fellowships (1002560). The PATH Through Life Study was funded by NHMRC Grants (229936 and 179839). Preliminary analyses were presented at the 69th Scientific Meeting of the Gerontological Society of America in November 2016 in New Orleans, Louisiana.
Correspondence concerning this article should be addressed to Allison A. M. Bielak, Department of Human Development and Family Studies, Colorado State University, 1570 Campus Delivery, Fort Collins, CO 80523-1570. E-mail: [email protected]
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
Developmental Psychology © 2019 American Psychological Association 2019, Vol. 55, No. 5, 994 –1004 0012-1649/19/$12.00 http://dx.doi.org/10.1037/dev0000688
994
patterns have been found in other adult samples (Deary & Der, 2005; Der & Deary, 2006). The values of IIV continue to expand in magnitude throughout older adulthood, as old-old adults have consistently been found to have higher levels of IIV compared with young-old adults both in the cross-sectional (Hultsch et al., 2002) and longitudinal (MacDonald, Hultsch, & Dixon, 2003) literature. However, cross-sectional work has shown that a U-shaped curve exists for IIV across the life span, in which IIV is highest for children, sharply lower for those in adolescence and young adulthood, and gradually higher with age through middle and older adulthood (Williams, Hultsch, Strauss, Hunter, & Tan- nock, 2005). Therefore, IIV appears to be adaptive in childhood (Siegler, 1994) but maladaptive and indicative of neurological aging in adulthood.
The maladaptive nature of IIV in older age is further under- scored given relations to an assortment of important factors, in which higher variability is associated with poorer cognition and everyday functioning (Burton, Strauss, Hultsch, & Hunter, 2009; Rabbitt, Osman, Moore, & Stollery, 2001), poorer health (Bunce, Handley, & Gaines, 2008; Li, Aggen, Nesselroade, & Baltes, 2001), and less activity participation (Bielak, Hughes, Small, & Dixon, 2007). In fact, because IIV is predictive of long-term cognitive changes (Kochan et al., 2016), the sensitivity of the covariation between change in cognition and change in IIV over time has also been evaluated. Three studies have evaluated asso- ciations between longitudinal change in IIV and longitudinal change in cognitive test score across different follow-up periods and retest intervals. The first study by MacDonald et al. (2003) evaluated a sample of adults aged 55 to 89 years who were tested every 3 years for a period of 6 years. They found significant negative covariation between IIV and a range of cognitive mea- sures, in which on occasions in which individuals were more inconsistent, their cognitive score was correspondingly lower. Lövdén, Li, Shing, and Lindenberger (2007) investigated the co- variation relation in a data set of adults 70 to 102 years of age who were followed for 13 years and retested every 2 years. Significant associations were found between longitudinal change in IIV and longitudinal changes in ideational fluency and perceptual speed, in which greater variability at one testing occasion was associated with lower cognitive performance at that occasion. The third study by Bielak, Hultsch, Strauss, MacDonald, and Hunter (2010b) also found significant covariation between IIV and cognitive changes for a sample aged 64 to 92 years who were retested every year for 3 years (i.e., 4 occasions). The relation was strongest for fluid cognitive measures and IIV derived from cognitively challenging RT tasks.
Of note, despite greater increases in IIV with greater older age, all three studies found that the covariation relation between cog- nition and IIV was age-invariant, in which the amount of cognitive change per unit change in IIV was stable across age. Notably, however, each study was limited to participants ranging from early older age to later older adulthood. Consequently, although the coupling between cognition and IIV appears to be well established in older age, it is unknown whether IIV covaries with cognition across the entire adult life span.
Further, all three prior covariation studies maintained IIV as a single variable, confounding between-person averages with within- person change. As described by Hoffman and Stawski (2009), failure to separate a time-varying variable into the between-person and
within-person components can lead to misleading conclusions about the covariation relationship. In such cases, it is possible that the covariation of time-varying IIV with cognitive change was actu- ally driven by between-person differences in IIV (i.e., average IIV score on cognition) and is not actually indicative of within-person covariation with cognition. Rather, it is only when the individual- based mean IIV is removed from each person’s time-varying score that the remaining score represents true within-person variation across time. The differentiation of IIV into its between- and within-person components (Hoffman & Stawski, 2009) clarifies which aspect of IIV is related to cognitive ability: having a higher IIV compared with others or showing greater individualized change in IIV over time. Although Bielak et al. (2010b) and MacDonald et al. (2003) used multilevel modeling, the time- varying IIV predictor was entered as the whole variable in the models rather than being divided into the within- and between- person components. Further, although Lövdén et al. (2007) used a different type of model, the dual change score model (DCSM), the same limitation still applies. In the DCSM, although levels and changes in IIV and cognitive performance are predictive of one another, the estimation of change still includes the average level of performance for that person. As a result, it is unclear whether the significant covariation effects of within-person changes in IIV with within-person changes in cognition will exist when the between- person averages are taken into consideration.
The present study addresses these issues by evaluating change in cognition and change in IIV over 8 years in a sample of adults aged 20 to 24, 40 to 44, and 60 to 64 years at baseline. Our analyses partitioned change in IIV into the between-person and within- person components, which allowed evaluations of the relation between cognition and average IIV compared with others (i.e., between-person differences) and the relation between cognition and individualized change in IIV over time (i.e., within-person change). Given that IIV only begins to increase in middle adult- hood (Bielak et al., 2014), we hypothesized that the between- person association of IIV and cognition would be limited in young adulthood and progressively increase in magnitude in middle and older adulthood. Similarly, despite the fact that past covariation research was based on the whole time-varying IIV index, we expect the covariation relation with cognitive change will still reflect past findings: The relationship will not be significant in young adulthood (20s) and will be smaller in middle adulthood (40s) than that found in older age (60s). This pattern of covariation would show that IIV is present in younger age, but IIV is not yet linked with cognition until middle and older age, when the brain begins to show unfavorable signs of aging. Further, evidence of covariation between IIV and cognition among the older adult cohort of our sample would provide verification of prior research and extension of the finding to 4-year intervals, the longest retest period thus far evaluated. Finally, because past findings have shown that higher IIV is associated with various medical conditions (Bunce et al., 2008; Whitehead, Dixon, Hultsch, & MacDonald, 2011), we controlled for education, sex, diabetes, hypertension, and anxiety and depressive symptoms.
Method
The study sample was drawn from the Personality and Total Health (PATH) Through Life Project, a longitudinal study in
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
995COVARIATION OF VARIABILITY AND COGNITION
which participants from three age cohorts of adulthood (i.e., 20s, 40s, 60s) were tested every 4 years (see Anstey et al., 2012). The first three waves of data were used for the present analyses (i.e., over 8 years). The present article presents secondary data analyses of the completed data set, and as such, Colorado State University’s research ethics committee declared this study exempt from review.
Participants
PATH participants resided in the city of Canberra or the neigh- boring town of Queanbeyan, Australia. Adults who were aged 20 to 24 years on January 1, 1999, 40 to 44 years on January 1, 2000, or 60 to 64 years on January 1, 2001 were invited to participate. Qualifying individuals were identified via the electoral roll, for which registration is compulsory for Australian citizens. The num- ber of participants who returned the survey totaled 7,485 (20s, n ! 2,404; 40s, n ! 2,530; 60s, n ! 2,551), and approximately half of each age cohort was female.
Sample attrition was limited, with 6,680 participants returning for Wave 2 and 5,996 participants also completing Wave 3. Par- ticipants who reported having a history of stroke, epilepsy, Par- kinson’s disease, or brain tumor, and older participants who scored less than 24 on the Mini-Mental State Examination (Folstein, Folstein, & McHugh, 1975) at any time point, were excluded from the present analyses. In addition, participants had to have sufficient RT data (see “Calculation of intraindividual variability” section) and baseline data for all covariates, resulting in a final sample of 7,146 participants. Participants were followed for an average of 7.01 years (SD ! 2.41). Table 1 provides further descriptive information about the sample.
Measures and Procedure
At each testing wave, participants completed a sequence of questionnaires and tests focused on their well-being, mental and physical health, and cognitive function. The majority of the as- sessment was administered on a laptop computer and was com- pleted under the supervision of and with the assistance of an interviewer. All RT and cognitive tasks were administered at each testing wave.
Intraindividual variability. IIV was calculated from the trial latencies on two multitrial RT tasks. Both tasks were completed
using a small box that served as both the response console and the display area. The box was held with both hands, with the left and right top buttons to be depressed by the respective index fingers. The front of the box had three lights: two red stimulus lights under the left and right buttons, respectively, and in the middle beneath these was a green get-ready light. The simple RT (SRT) task was completed first, followed by the choice RT (CRT) task. The time between the get-ready light and stimulus light varied throughout the trials of both tasks. In SRT, participants were presented with a green get-ready light, followed by the right red light. Participants were asked to press a button as soon as the red light appeared. For CRT, participants were presented with the green get-ready light, after which one of the two red lights illuminated. Participants were asked to press the corresponding response button as soon as possible. There were 40 trials presented for CRT and 80 for SRT.
Cognitive ability. Perceptual speed. Perceptual speed was assessed using the
Symbol Digit Modalities Test (Smith, 1982), which presented participants with a coding key pairing numbers 1 through 9 with nine symbols. Participants were given 90 s to transcribe as many randomly ordered numbers to corresponding symbols as possible. The number of correct responses was recorded.
Short-term memory (STM) and episodic memory. Both mem- ory types were measured via the first list of the California Verbal Learning Test (Delis, Kramer, Kaplan, & Ober, 1987). Participants were read a list of 16 words from four taxonomic categories (e.g., fruits, tools) presented in unblocked order, and asked to immedi- ately recall as many words as possible (STM). Following a short interval (i.e., completing a grip strength task), participants were again asked to recall as many words as possible (episodic mem- ory). For both immediate and delayed recall, the number of cor- rectly recalled words was recorded.
Working memory. Working memory was assessed using the digit span backward task from the Wechsler Memory Scale (Wechsler, 1945). Participants were read 10 sets of three to seven numbers, and after each set were asked to repeat the number set in reverse order. The number of correctly recalled sets was recorded.
Vocabulary. Vocabulary was measured by the Spot-the-Word Test (Baddeley, Emslie, & Nimmo-Smith, 1992), which involved presentation of 60 word pairs, each containing a real word and a nonword. Participants were asked to identify the real word. The number of correctly identified real words was recorded.
Fluid Cognitive Ability Composite. The tests for perceptual speed, STM, working memory, and episodic memory all assessed components of fluid intelligence (Horn, 1987) and were addition- ally combined to form a cognitive composite score. As a compos- ite, it was viewed as a potentially more robust metric of cognitive ability. Scores on each of the four tasks were converted to T scores using the baseline sample. The scores were then summed to form a fluid cognitive composite for each individual at each wave.
Covariates. We controlled for the effects of sex, education, diabetes, hypertension, and anxiety and depressive symptoms. Education was assessed by years of formal schooling (M ! 14.35, SD ! 2.28). Diabetes was based on the self-reported presence of the disease at any wave (5.7% of sample). Hypertension was determined from blood pressure readings administered by testers at each wave, and any participant scoring above 140 systolic or 90 diastolic, or reporting taking blood pressure medication at any wave, was coded as having hypertension (49.2% of sample).
Table 1 Descriptive Information About the Sample Covariates
Measure
Age group
20 (n ! 2,351)
M (SD)
40 (n ! 2,447)
M (SD)
60 (n ! 2,348)
M (SD)
Years of education 14.62 (1.56) 14.60 (2.31) 13.83 (2.72) Anxiety at baseline 3.83 (2.71) 3.50 (2.70) 2.17 (2.29) Depressive symptoms
at baseline 2.88 (2.37) 2.41 (2.37) 1.60 (1.82) % Female 51.8 52.8 48.7 % Diabetic .90 3.80 12.5 % Hypertensive 22.0 45.1 80.6
Note. Hypertension was defined as scoring above 140 systolic or 90 diastolic, or reporting taking blood pressure medication at any wave.
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
996 BIELAK AND ANSTEY
Anxiety and depressive symptoms were based on responses to the Goldberg Anxiety and Depression Scale (Goldberg, Bridges, Duncan-Jones, & Grayson, 1988) and were entered into the models separately as time-varying anxiety and depression scores.
Calculation of Intraindividual Variability
The following calculation procedures were completed separately for each task at each wave. First, participants who did not complete more than 50% of the trials1 for the task were removed from the IIV calculation for that wave, and RT values on incorrect CRT trials were removed to remove the confound of slower wrong responses. Next, the remaining RT latencies were examined for outliers. Legitimate lower bounds were set at 50 ms for SRT, and 150 ms for CRT. Upper bounds were set at the mean plus three standard deviations for each individual based on their own distribution of RT latencies.2 Missing values were imputed using a regression substitution procedure that creates individual RT equations, which are then used to predict the missing values (Hultsch, MacDonald, Hunter, Levy-Bencheton, & Strauss, 2000). This is a conservative method of addressing missing data in IIV because it reduces within-subject variation. Approximately 4% of trials were imputed.
There is variation in the literature regarding how to calculate IIV (Stawski et al., 2017). Hultsch, Strauss, Hunter, and MacDonald (2008) described the statistical complications with common alterna- tive quantifications. First, the simple calculation of calculating the standard deviation over trials (often referred to as raw intraindividual standard deviation [ISD]) is not an appropriate method because this fails to control for mean RT. The mean and standard deviation are highly correlated, so that someone with a higher mean RT will likely have a correspondingly higher standard deviation. Failing to remove mean RT means that the raw ISD that is investigated is not purely representative of variability, making it challenging to determine how much of the variability index is true inconsistency in responding and how much of the raw ISD is just a result of differences in response speed. Another method is covarying mean RT via analysis of cova- riance (ANCOVA; i.e., after raw SD has been calculated). Hultsch et al. noted that ANCOVA should only be used in situations in which it is reasonable to assume that the groups are equal on the covariate, which is not the case when comparing young, middle, and older age groups on mean RT. Third, there are also criticisms regarding the coefficient of variation, whereby individual standard deviations are divided by individual performance means. The calculation method for this quantification results in a combination of the main effect of the ISD, the main effect (inverse) of mean performance, and the interac- tion between the two. Therefore, the resulting term could reflect any one of these three effects and not only variability (see Hultsch et al., 2008, for further statistical explanations).
Given these complications, we have chosen to follow the meth- odology developed by Hultsch and colleagues (2000; Hultsch et al., 2008) in which potential confounding influences in the RT data (e.g., age differences in mean RT, practice effects) were accounted for by regressing the trial RT data onto categorical age group, categorical trial, and their interactions:
RT Score ! a " b(age group) " c(trial) " d(Age Group # Trial) " e
The resulting residuals are therefore independent of any system- atic within (i.e., trial) and between-subjects (i.e., age group) sources of variance. Of note, Bielak et al. (2014) demonstrated that
the additional residualization of within-person linear trial effects produced identical IIV values. Each residual was then converted to standardized T scores and each individual’s standard deviation across all trials was calculated. The ISD was used as the indicator of IIV. ISD values were computed for both RT tasks at each wave. IIV was then divided into two components for both SRT and CRT. The between-person effect of IIV (ISD-between) was obtained by calculating each individual’s average ISD value across the waves for that task. The within-person effect of IIV (ISD-within) was created by subtracting each individual’s ISD for that wave from their average level of ISD (i.e., ISD-between).
Statistical Analyses
The data were analyzed using multilevel models in IBM SPSS Statistics Version 24 (IBM, 2016), and cognitive change was modeled using a time in study metric. First, unconditional models confirmed that all cognitive measures and ISD values had significant variation over the 8 years. The intraclass correlations (ICCs), which represent the proportion of between-person variance, were as follows: ISD- SRT ! .392; ISD-CRT ! .411; perceptual speed ! .837; STM ! .535; episodic memory ! .579; working memory ! .637; vocabu- lary ! .838; and fluid ability composite ! .747. Supplementary Table 1 of the online supplemental materials provides further descriptive information on the ISDs and cognitive scores by age group. Condi- tional models were next completed for each cognitive outcome and ISD to further describe change over time (see Supplementary Infor- mation A of the online supplemental materials for multilevel model equations). Age group was the main fixed predictor of interest, and random effects for intercept and slope were estimated.3 Education, sex, diabetes, and hypertension were included as time-invariant co- variate predictors of the intercept,4 and anxiety and depressive symp- toms were included as time-varying covariate predictors. Change in ISD and cognition have each been reported previously using this sample (Bielak, Anstey, Christensen, & Windsor, 2012; Bielak et al., 2014), and conditional models were evaluated to confirm sufficient change was available to evaluate covariation between the variables. Finally, the within and between components of each ISD were in- cluded as time-varying covariates of each cognitive outcome, along with the age group interactions with these effects. Significant effects of ISD-between would indicate the association with average ISD and the intercept and slope of cognitive ability, whereas associations with ISD-within would indicate the existence of time-varying covariation between changes in IIV and changes in cognitive score. Significant interactions of these effects with age group would indicate differences in these effects according to age cohort.
1 Burns et al. (2011) found that the accuracy of imputation is signifi- cantly reduced when there is greater than 50% of item-level missingness.
2 The trimming procedure varied slightly for the Wave 1 60s cohort because a previous trimming of the RT data for the 60s age group accidentally, but permanently, deleted the trials that exceeded the trim cutoffs. We had to apply our trimming method to the existing data after that event.
3 The SRT model with a random slope failed to converge and thus was removed from the model.
4 Covariates were not significant predictors of the slope and thus were not included in the reported models.
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
997COVARIATION OF VARIABILITY AND COGNITION
Results
Changes in Cognitive Score and ISD
Table 2 provides the model estimates for change in each cognitive domain. For all domains except for working memory and vocabulary, the youngest age cohort showed the highest intercept value, followed by 40s with the next highest value, and then the 60s with the lowest score. All age group differences were significant. The same pattern of scoring was maintained for working memory, with the 20s group scoring the highest, but the 40s and 60s groups did not significantly differ from one another. The pattern was reversed for vocabulary, in which the oldest age cohort performed the best, followed by the 40s and then the 20s. All cohorts showed significant changes in their cognitive score over the 8 years for all tasks (except the 60s for working memory), with the 20s age group showing the largest gains, followed by the 40s and then 60s age groups, who often decreased in their performance over time. All age group comparisons for time in the study were significant except that the 40s and 60s groups showed similar gains in vocabulary over time. Significant within- and between-person variation remained to be explained for all cognitive domain models.
Model estimates for ISD change are shown in Table 3. The pattern of baseline ISD was the same for both RT tasks: The oldest cohort showed the highest ISD values, followed by the 40s group and then the 20s group. All age group comparisons were significant. For SRT, although the 60s cohort showed a significant increase in ISD over time, the 20s and 40s cohorts decreased in ISD and did not differ from one another in the magnitude of change. For CRT, both the 40s and 60s cohorts significantly increased in ISD, with the 60s group show- ing a significantly greater increase. However, the 20s cohort remained stable over the 8 years. Significant amounts of within- and between- person variance remained for both ISD types.
Relationship Between Cognition and ISD The addition of the ISD variables significantly improved model
fit for all cognitive outcomes (see Table 4 for SRT and Table 5 for CRT). The between-person effect for SRT-ISD was consistently
Table 2 Parameter Estimates From Multilevel Models Examining 8-Year Change in Cognitive Performance by Age Group
Parameter
Cognitive domain
Perceptual speed Short-term memory Episodic memory Working memory Vocabulary
Fluid ability composite
Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE
Fixed effects Intercept
20 53.18!!! .44 48.87!!! .48 49.19!!! .49 49.94!!! .52 44.80!!! .48 50.26!!! .36 40 49.62!!! .42 47.81!!! .46 47.95!!! .46 49.37!!! .49 49.63!!! .46 48.66!!! .34 60 41.77!!! .38 45.88!!! .42 45.58!!! .42 48.93!!! .45 53.14!!! .42 45.51!!! .31
Time 20 .17!!! .02 .40!!! .03 .41!!! .03 .33!!! .02 .40!!! .02 .33!!! .02 40 ".10!!! .02 .12!!! .03 .13!!! .03 .23!!! .02 .16!!! .01 .09!!! .01 60 ".32!!! .02 ".34!!! .03 ".21!!! .03 .02 .02 .18!!! .02 ".22!!! .02
Random effects Intercept variance 50.51!!! 1.08 45.61!!! 1.56 49.05!!! 1.53 61.59!!! 1.59 63.54!!! 1.27 30.48!!! .72 Slope variance .08!!! .01 .21!!! .39 .14!!! .03 .16!!! .03 .08!!! .01 .07!!! .01 Residual variance 14.18!!! .26 44.13!!! .81 39.99!!! .73 32.89!!! .61 12.75!!! .24 12.42!!! .23
"2LL (df ! 16) 121,790 136,574 135,335 133,524 120,420 115,995 AIC 121,822 136,606 135,367 133,556 120,452 116,027
Note. All estimates are unstandardized. Years of education, sex, diabetes, hypertension, anxiety, and depressive symptoms were included as covariates in the model. SE ! Standard error; df ! degrees of freedom; AIC ! Akaike’s information criterion; "2LL ! "2 log likelihood. !!! p # .001.
Table 3 Parameter Estimates From Multilevel Models Examining 8-Year Change in ISD by Age Group
Parameter
ISD
Simple reaction time
Choice reaction time
Estimate SE Estimate SE
Fixed effects Intercept
20 5.29!!! .19 5.68!!! .14 40 6.46!!! .18 6.73!!! .13 60 6.91!!! .16 7.86!!! .12
Time 20 ".07!!! .01 .02 .01 40 ".09!!! .01 .06!!! .01 60 .20!!! .01 .19!!! .01
Random effects Intercept variance 5.24!!! .16 2.65!!! .15 Slope variance — — .01 .01 Residual variance 9.87!!! .13 5.00!!! .10
"2LL 101,176 (df ! 14) 86,536 (df ! 16) AIC 101,204 86,568
Note. All estimates are unstandardized. Years of education, sex, diabetes, hypertension, anxiety, and depressive symptoms were included as covari- ates in the model. The model for SRT would not converge with a random slope and thus only includes random intercept. ISD ! intraindividual standard deviation; SE ! Standard error; df ! degrees of freedom; AIC ! Akaike’s information criterion; "2LL ! "2 log likelihood. !!! p # .001.
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
998 BIELAK AND ANSTEY
significantly negative across all cognitive domains at baseline (see Figure 1). The negative association meant that each additional unit in average ISD over the 8 years was associated with poorer baseline cognitive performance. Further, there were differences by
age cohort for ISD-between. For SRT, the 20s cohort had the strongest ISD-between effect compared with both the 40s and 60s for working memory, vocabulary, perceptual speed, and the fluid cognitive composite. The stronger effect for the 20s cohort for the
Table 4 Parameter Estimates From Multilevel Models Examining ISD-SRT Predicting Cognitive Performance and Change By Age Cohort
Parameter
Cognitive domain
Perceptual speed
Short-term memory
Episodic memory
Working memory Vocabulary
Fluid ability composite
Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE
Fixed effects ISD-within
20 ".06 .03 ".10 .06 ".10 .06 .00 .05 ".05 .03 ".06! .03 40 ".03 .02 .00 .04 ".04 .04 .02 .03 ".04! .02 ".01 .02 60 ".03 .02 ".02 .03 ".01 .03 ".02 .02 .06 .02 ".01 .02
ISD-between 20 "1.15!!! .09 ".47!!! .10 ".57!!! .10 ".73!!! .10 ".77!!! .09 ".74!!! .07 40 ".73!!! .05 ".41!!! .06 ".44!!! .06 ".44!!! .06 ".55!!! .06 ".51!!! .04 60 ".60!!! .04 ".32!!! .05 ".34!!! .05 ".30!!! .05 ".36!!! .05 ".39!!! .03
ISD-Between $ Time 20 ".04!!! .01 ".04! .02 ".03 .02 ".05!! .01 .02 .01 ".04!!! .01 40 ".01 .01 .00 .01 .00 .01 .00 .01 .01 .01 .00 .00 60 ".01 .00 ".01 .01 .00 .01 .00 .01 .00 .00 .00 .00
Random effects % Intercept variance 11.40% 3.53% 3.00% 3.05% 3.84% 8.02% % Slope variance 8.63% 2.86% 3.07% 0 0 4.29% % Residual variance .31% 0 .27% 0 1.30% .41%
Change in model fit, df% ! 9 4,322!!! 4,383!!! 4,350!!! 4,260!!! 3,577!!! 3,788!!!
Note. All estimates are unstandardized. Years of education, sex, diabetes, hypertension, anxiety, and depressive symptoms were included as covariates. Only reductions in random effect variance are shown. ISD ! intraindividual standard deviation; SRT ! simple reaction time; SE ! Standard error; df ! degrees of freedom. ! p # .05. !! p # .01. !!! p # .001.
Table 5 Parameter Estimates From Multilevel Models Examining ISD-CRT Predicting Cognitive Performance and Change By Age Cohort
Parameter
Cognitive domain
Perceptual speed
Short-term memory
Episodic memory
Working memory Vocabulary
Fluid ability composite
Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE
Fixed effects ISD-within
20 .05 .05 ".21! .10 ".15 .09 .04 .08 ".01 .05 ".06 .05 40 ".05! .04 ".05 .04 ".08 .04 ".01 .04 ".04 .02 ".04 .02 60 ".07!! .03 ".03 .04 .01 .04 ".05 .04 ".02 .02 ".03 .02
ISD-between 20 "1.80!!! .12 ".51!! .15 ".35! .15 ".68!!! .16 ".79!!! .14 ".83!!! .10 40 ".79!!! .06 ".32!!! .08 ".36!!! .08 ".30!!! .08 ".37!!! .07 ".45!!! .05 60 ".84!!! .06 ".34!!! .07 ".27!!! .07 ".40!!! .07 ".27!!! .07 ".46!!! .05
ISD-Between $ Time 20 ".03 .01 ".02 .02 ".04! .02 ".08!!! .02 .01 .01 ".04!! .01 40 ".01 .01 ".02 .01 .00 .01 .01 .01 .01 .01 ".01 .01 60 ".01! .01 .00 .01 ".02 .01 .01 .10 .00 .01 ".01 .01
Random effects % Intercept variance 11.18% 2.45% .28% 1.51% 2.07% 4.84% % Slope variance 9.25% 18.29% 5.0% 0 9.88% 5.86% % Residual variance .69% 0 .22% .52% 3.43% .18%
Change in model fit, df% ! 9 6,962!!! 7,415!!! 8,549!!! 7,229!!! 6,385!!! 6,162!!!
Note. All estimates are unstandardized. Years of education, sex, diabetes, hypertension, anxiety, and depressive symptoms were included as covariates. Only reductions in random effect variance are shown. ISD ! intraindividual standard deviation; CRT ! choice reaction time; SE ! Standard error; df ! degrees of freedom. ! p # .05. !! p # .01. !!! p # .001.
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
999COVARIATION OF VARIABILITY AND COGNITION
majority of the cognitive tasks indicated that each additional unit of higher average ISD value was associated with a higher magni- tude of poorer baseline cognitive performance compared with the other age groups (e.g., perceptual speed: & ! "1.15 for 20s; & ! ".73 for 40s; & ! ".60 for 60s). The 20s cohort was also significantly different from the oldest cohort in the ISD-between effect for episodic memory (p # .05). The 40s group showed a greater between-person ISD effect than the 60s group for vocab- ulary, perceptual speed, and the fluid cognitive composite.
The ISD-between-person effects for cognitive baseline score were similarly consistently negative for CRT for all cognitive domains (see Figure 2). The 20s cohort had the strongest effects across the age groups for vocabulary, perceptual speed, and the cognitive composite, and the 20s group also had a greater effect than the 40s group for working memory (p # .05).
There were significant negative interactions for ISD-between with time (i.e., cognitive slope) for both RT tasks. This can be translated as those with a higher average ISD also showed greater
decline in cognitive performance over time in study. Notably, this effect only existed for those in the 20s cohort for SRT. These effects were found for working memory, perceptual speed, and the fluid cognitive composite, and the 20s age group was significantly different from the two older cohorts, who showed no effect of ISD-Between $ Time. The 20s cohort also had a significant effect for STM, but this effect did not significantly differ from the 40s and 60s cohorts. These effects were mirrored in CRT, as the 20s cohort had significant ISD-Between $ Time effects for working memory and the cognitive composite, and were significantly dif- ferent from the 40s and 60s cohorts, who did not have any significant interaction effects for these tasks. The youngest cohort also showed the negative ISD-Between $ Time effect for episodic memory, and the oldest cohort had this effect for perceptual speed.
Evidence for covariation between the within-person effect of ISD and change in cognitive performance over the 8 years was limited. For SRT, ISD-within was significant for the 20s cohort for the cognitive composite and the 40s cohort for vocabulary, but
Figure 1. Parameter estimates for simple reaction time intraindividual standard deviation - between with baseline cognitive performance.
Figure 2. Parameter estimates for choice reaction time intraindividual standard deviation - between with baseline cognitive performance.
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
1000 BIELAK AND ANSTEY
neither effect significantly differed from the other cohorts, which did not have effects that significantly differed from zero. For CRT, both the 40s and 60s age groups showed significant ISD-within effects in predicting change in perceptual speed, and the 60s age group differed from those in the 20s age group (p # .001), which did not show the effect. For the average individual in the 60s cohort, on occasions when their ISD was one unit higher than usual, their perceptual speed score was correspondingly .07 units lower. There was also a significant ISD-within effect for the youngest age group for STM, but this effect did not differ from the other age groups who showed no effect.
Discussion
The present study used a population-based sample to evaluate the associations of IIV and cognitive performance over 8 years. Our analysis partitioned change in IIV into the between-person and within-person components, which separated average IIV values from fluctuations around an individuals’ own mean from one wave to the next. Although our results supported past findings of higher average IIV being associated with poorer cognitive performance, we found this association was stronger in young compared with middle and older adulthood. Further, there was surprisingly little evidence of significant covariation of within-person changes in IIV being associated with changes in cognitive performance, regard- less of age group.
For each age group and cognitive domain, average IIV score (i.e., ISD-between) was significantly associated with baseline cog- nitive performance. Specifically, the associations were negative, as greater average IIV was linked to a lower cognitive score. The negative association between IIV and cognition has been repeat- edly demonstrated (e.g., Kochan et al., 2016; MacDonald et al., 2003; Rabbitt et al., 2001; Vasquez, Binns, & Anderson, 2016), but the present study is the first to show that the negative associ- ation with cognition is apparent even for those in young adulthood. Further, when age differences existed in the relation between average IIV and baseline cognitive performance, the pattern was always in the direction in which the youngest cohort had the strongest effects, followed by the middle-aged cohort, and, finally, the oldest cohort (but the middle-aged and oldest cohorts were not always significantly different from one another; see Figures 1 and 2). For example, for each additional unit in average IIV, a partic- ipant in their 20s had .74 fewer points on the fluid cognitive composite, whereas an additional unit in average IIV for a partic- ipant in their 40s was associated with only .51 fewer points, and a participant in their 60s had only .39 fewer points.
There was also evidence of negative associations for average IIV and cognitive change, but almost exclusively for the 20s cohort. Young adult participants who had one unit higher of average IIV tended to show greater negative change on various cognitive tasks over the 8 years. However, this did not translate to decline but rather a lesser practice effect or improvement on those tasks. On average, the 20s cohort showed significant gains in cognitive score for all domains, and as such, the individuals with higher IIV did not appear to improve as much over time compared with young adult participants with lower average IIV. Based on previous findings showing that IIV increases throughout adult- hood, and specifically in older age (Bielak et al., 2014; Hultsch et al., 2002), we would have predicted that the oldest cohort would
show the strongest effects between average IIV and cognition rather than the youngest age group. Williams et al. (2005) showed that IIV is at its lowest point in early adulthood, and younger adults (compared with middle-age and older adults) tend to improve or remain stable in variability on RT tasks over time (Bielak et al., 2014). Therefore, the present finding that there is a stronger link between IIV and cognition in younger adulthood may suggest that having higher-than-average IIV in young adulthood is an indicator of potential cognitive or neurological problems in the future. This hypothesis is only speculative, and it remains to be seen how IIV in early adulthood is related to IIV in older age or whether IIV in early life is predictive of later neurological disease (Bielak & Anstey, 2015).
In contrast with multiple studies showing covariation between IIV change and cognitive change (Bielak et al., 2010b; Lövdén et al., 2007; MacDonald et al., 2003), we found few instances of a significant relationship in our sample. Specifically, of the 36 possible covariations between IIV and cognition by age (i.e., 6 Cognitive Types $ 2 RT IIV $ 3 Age Groups), there were only five instances of a significant relationship. These few effects were all negative. That is, on occasions when a person scored one unit higher in ISD than their usual score, their cognitive score was correspondingly lower. Overall, although the direction of the co- variation is consistent with past findings, the scarcity of the cova- riation relations is not. The surprising lack of effects is particularly true for the older adult group, as all three past studies of covari- ation of IIV and cognition focused exclusively on older adult samples.
There are a number of issues to examine regarding the discrep- ancy between our present findings of few covariation relationships and past compelling evidence of the covariation relation between cognition and IIV. First, as noted in the introduction, and perhaps most importantly, there were substantial statistical differences in which variables were considered to constitute the covariation relation- ship. Specifically, all three past examinations of IIV change with cognitive change did not remove the average or between-person effect from within-person change in IIV. Therefore, there is a reasonable possibility that the observation of covariation between IIV change and cognitive change in the past studies was not indicative of true covariation. The paucity of covariation relations in the present results compared with the past studies would cer- tainly fit this conclusion. However, there are a number of other differences between our study and the past studies to also consider. For example, prior studies have found significant within-person covariation over 1- (Bielak et al., 2010b), 2- (Lövdén et al., 2007), and 3-year (MacDonald et al., 2003) retesting intervals. It is possible that IIV and cognition do share a covariation relationship, but it exists only over the short term (i.e., 1 to 3 years) and is not observable over longer time frames, such as the 4-year time frame used in the present study. Our present analyses also relied on only three time points, limiting the potential within-person variability. However, the covariation analyses by MacDonald et al. (2003) also only had three time points.
Next, the three earlier studies focused exclusively on older adulthood: 64 to 92 years in Bielak et al. (2010b), 70 to 102 years in Lövdén et al. (2007), and 55 to 89 years in MacDonald et al. (2003). All three studies also found that the covariation relation was age-invariant throughout older adulthood. Although the pres- ent study also included adults in their 20s and 40s, based on
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
1001COVARIATION OF VARIABILITY AND COGNITION
previous findings we would have expected the covariation relation to be apparent for the oldest group in our study, who were 60 to 64 years at baseline, and 68 to 72 years by the third testing wave. Therefore, our oldest cohort was most similar to the young-old (55– 64 years) and mid-old age groups (65–74 years) in MacDon- ald et al. Despite the similar age range, we did not reproduce the significant negative covariation between IIV and cognition over time, as was found in MacDonald et al. It is possible that some of the previously reported findings were influenced by preclinical dementia cases within samples, as they have mostly been older cohorts. It could be that IIV and cognition become more correlated with neuropathology and not age per se. Separating this distinction is, of course, a common confound in aging studies, however.
Finally, although the exact RT and cognitive measures that were used in the present study differed from the past studies, there were far more similarities than differences. For IIV, Bielak et al. (2010b) found that IIV based on basic and complex RT tasks provided the strongest covariation relationships with cognition, and this included IIV from CRT tasks. Further, Lövdén et al. (2007) relied on a simple matching RT task (identical pictures) and MacDonald et al. (2003) used a IIV composite that included SRT and CRT as two of the four RT tasks. For cognitive tasks, the within-person change relationships in Bielak et al. were found for a range of cognitive tasks, including memory, reasoning, and processing speed. The covariations were greatest for episodic memory tasks compared with perceptual speed in MacDonald et al., and Lövdén et al. found significant covariation with IIV for both perceptual speed and ideational fluency. Overall, the present study’s use of both simple and choice RT measures to calculate IIV and the range of cognitive domains including speed, memory, and vocabulary do indeed have reasonable overlap with the past covariation studies.
The present analysis had many strengths, including a large population-based cohort from young to older adulthood, the lon- gitudinal design, multiple RT tasks for the calculation of IIV, and a range of cognitive outcomes, along with a composite index of fluid cognitive performance. The statistical modeling also incor- porated multilevel modeling, which makes use of all data regard- less of missingness, and the time-varying IIV variable was divided into the between-person average and within-person change com- ponents, permitting evaluation of true within-person covariation with cognition. However, the present analyses were limited by the gaps in the adult age span (i.e., no data for 32–39 and 52–59 years of age), and a fourth wave of data collection would provide greater assurance of the model-implied change of IIV and cognition over time. Further, as described in the methods, we chose the most statistically robust index of ISD for our present analyses. Although we do not compare the results from these various indices, there is evidence that similar findings are found regardless of which index of IIV is used (Lövdén et al., 2007; Stawski et al., 2017), suggest- ing that other quantifications of IIV in our data set would likely not change the findings.
Another issue worth noting is that the ICC for both IIV indices showed a relatively higher proportion of within-person variance, whereas the ICC values for the cognitive tasks indicated a rela- tively higher proportion of between-person variance. One argu- ment could be that it may be challenging to find a significant covariation relation if there is a discrepancy in the proportion of within-person variance between the ICC for the IIV measures and
the ICC values for the cognitive indices. Although this is possible, previous research has shown similarly low between-person corre- lations over time (Lövdén et al., 2007), and the average proportion of within-person variance in the cognitive tasks was in fact even lower in MacDonald et al. (2003; .27) compared with the average of our cognitive tasks (.31). Both of these studies did find evidence of the covariation relationship between IIV and cognition, suggest- ing that the proportions of between- and within-person variance across the IIV measures and cognitive tasks are unlikely to be the root cause of the lack of covariation relations in our sample.
The present study demonstrates that the relations for between- person IIV and cognition are consistently negative across adult- hood, supporting the hypothesis that IIV is an indicator of neuro- logical integrity. However, there was minimal to no evidence for significant covariation of within-person IIV and cognitive perfor- mance, in contrast to previous research. Consequently, there may be a deeper issue at play regarding whether true covariation does in fact exist. At this juncture, in addition to the inclusion of other longitudinal studies of adulthood with varying retest intervals, reanalysis of the previous work using the division of the time- varying IIV into separate within- and between-person components would help to greatly elucidate this issue. Overall, it appears that IIV is a stable characteristic negatively associated with cognition in adulthood, but our present analyses suggest that this link may primarily exist at the between-person level. It remains to be seen whether within-person covariation with cognition is more likely to be evident among those with impending or diagnosed neurological disease.
References
Anstey, K. J., Christensen, H., Butterworth, P., Easteal, S., Mackinnon, A., Jacomb, T., . . . Jorm, A. F. (2012). Cohort profile: The PATH through life project. International Journal of Epidemiology, 41, 951–960. http:// dx.doi.org/10.1093/ije/dyr025
Anstey, K. J., Mack, H. A., Christensen, H., Li, S.-C., Reglade-Meslin, C., Maller, J., . . . Sachdev, P. (2007). Corpus callosum size, reaction time speed, and variability in mild cognitive disorders and in a normative sample. Neuropsychologia, 45, 1911–1920. http://dx.doi.org/10.1016/j .neuropsychologia.2006.11.020
Baddeley, A., Emslie, H., & Nimmo-Smith, I. (1992). The Spot-the-Word Test. Bury St. Edmunds, UK: Thames Valley Test Company.
Bellgrove, M. A., Hester, R., & Garavan, H. (2004). The functional neuroanatomical correlates of response variability: Evidence from a response inhibition task. Neuropsychologia, 42, 1910 –1916. http://dx .doi.org/10.1016/j.neuropsychologia.2004.05.007
Bielak, A. A. M., & Anstey, K. J. (2015). Intraindividual variability in attention across the adult lifespan. In M. Diehl, K. Hooker, & M. J. Sliwinski (Eds.), Handbook of intraindividual variability across the lifespan (pp. 160 –175). New York, NY: Routledge.
Bielak, A. A. M., Anstey, K. J., Christensen, H., & Windsor, T. D. (2012). Activity engagement is related to level, but not change in cognitive ability across adulthood. Psychology and Aging, 27, 219 –228. http://dx .doi.org/10.1037/a0024667
Bielak, A. A. M., Cherbuin, N., Bunce, D., & Anstey, K. J. (2014). Intraindividual variability is a fundamental phenomenon of aging: Evi- dence from an 8-year longitudinal study across young, middle, and older adulthood. Developmental Psychology, 50, 143–151. http://dx.doi.org/ 10.1037/a0032650
Bielak, A. A. M., Hughes, T. F., Small, B. J., & Dixon, R. A. (2007). It’s never too late to engage in lifestyle activities: Significant concurrent but not change relationships between lifestyle activities and cognitive speed.
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
1002 BIELAK AND ANSTEY
The Journals of Gerontology: Series B, Psychological Sciences and Social Sciences, 62, P331–P339. http://dx.doi.org/10.1093/geronb/62.6 .P331
Bielak, A. A. M., Hultsch, D. F., Strauss, E., Macdonald, S. W. S., & Hunter, M. A. (2010a). Intraindividual variability in reaction time pre- dicts cognitive outcomes 5 years later. Neuropsychology, 24, 731–741. http://dx.doi.org/10.1037/a0019802
Bielak, A. A. M., Hultsch, D. F., Strauss, E., MacDonald, S. W. S., & Hunter, M. A. (2010b). Intraindividual variability is related to cognitive change in older adults: Evidence for within-person coupling. Psychology and Aging, 25, 575–586. http://dx.doi.org/10.1037/a0019503
Bunce, D., Handley, R., & Gaines, S. O. J. (2008). Depression, anxiety, and within-person variability in adults aged 18 to 85 years. Psychology and Aging, 23, 848 – 858. http://dx.doi.org/10.1037/a0013678
Burns, R. A., Butterworth, P., Kiely, K. M., Bielak, A. A. M., Luszcz, M. A., Mitchell, P., . . . Anstey, K. J. (2011). Multiple imputation was an efficient method for harmonizing the Mini-Mental State Examination with missing item-level data. Journal of Clinical Epidemiology, 64, 787–793. http://dx.doi.org/10.1016/j.jclinepi.2010.10.011
Burton, C. L., Strauss, E., Hultsch, D. F., & Hunter, M. A. (2009). The relationship between everyday problem solving and inconsistency in reaction time in older adults. Aging, Neuropsychology, and Cognition, 16, 607– 632. http://dx.doi.org/10.1080/13825580903167283
Cherbuin, N., Sachdev, P., & Anstey, K. J. (2010). Neuropsychological predictors of transition from healthy cognitive aging to mild cognitive impairment: The PATH through life study. The American Journal of Geriatric Psychiatry, 18, 723–733. http://dx.doi.org/10.1097/JGP.0b01 3e3181cdecf1
Deary, I. J., & Der, G. (2005). Reaction time, age, and cognitive ability: Longitudinal findings from age 16 to 63 years in representative popu- lation samples. Aging, Neuropsychology, and Cognition, 12, 187–215. http://dx.doi.org/10.1080/13825580590969235
de Frias, C. M., Dixon, R. A., & Camicioli, R. (2012). Neurocognitive speed and inconsistency in Parkinson’s disease with and without incip- ient dementia: An 18-month prospective cohort study. Journal of the International Neuropsychological Society, 18, 764 –772. http://dx.doi .org/10.1017/S1355617712000422
Delis, D., Kramer, J., Kaplan, E., & Ober, B. (1987). California Verbal Learning Test. San Antonio, TX: Psychological Corporation/Harcourt Brace Jovanovich.
Der, G., & Deary, I. J. (2006). Age and sex differences in reaction time in adulthood: Results from the United Kingdom Health and Lifestyle Survey. Psychology and Aging, 21, 62–73. http://dx.doi.org/10.1037/ 0882-7974.21.1.62
Fjell, A. M., Westlye, L. T., Amlien, I. K., & Walhovd, K. B. (2011). Reduced white matter integrity is related to cognitive instability. The Journal of Neuroscience, 31, 18060 –18072. http://dx.doi.org/10.1523/ JNEUROSCI.4735-11.2011
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). “Mini-mental state.” A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, 189 –198. http://dx .doi.org/10.1016/0022-3956(75)90026-6
Goldberg, D., Bridges, K., Duncan-Jones, P., & Grayson, D. (1988). Detecting anxiety and depression in general medical settings. British Medical Journal, 297, 897– 899. http://dx.doi.org/10.1136/bmj.297.6653 .897
Gorus, E., De Raedt, R., Lambert, M., Lemper, J.-C., & Mets, T. (2008). Reaction times and performance variability in normal aging, mild cog- nitive impairment, and Alzheimer’s disease. Journal of Geriatric Psy- chiatry and Neurology, 21, 204 –218. http://dx.doi.org/10.1177/08919 88708320973
Hoffman, L., & Stawski, R. S. (2009). Persons as contexts: Evaluating between-person and within-person effects in longitudinal analysis. Re-
search in Human Development, 6, 97–120. http://dx.doi.org/10.1080/ 15427600902911189
Horn, J. L. (1987). A context for understanding information processing studies of human abilities. In P. A. Vernon (Ed.), Speed of information processing and intelligence (pp. 201–238). Norwood, NJ: Ablex.
Hultsch, D. F., MacDonald, S. W. S., & Dixon, R. A. (2002). Variability in reaction time performance of younger and older adults. The Journals of Gerontology: Series B, Psychological Sciences and Social Sciences, 57, P101–P115. http://dx.doi.org/10.1093/geronb/57.2.P101
Hultsch, D. F., MacDonald, S. W. S., Hunter, M. A., Levy-Bencheton, J., & Strauss, E. (2000). Intraindividual variability in cognitive perfor- mance in older adults: Comparison of adults with mild dementia, adults with arthritis, and healthy adults. Neuropsychology, 14, 588 –598. http:// dx.doi.org/10.1037/0894-4105.14.4.588
Hultsch, D. F., Strauss, E., Hunter, M. A., & MacDonald, S. W. S. (2008). Intraindividual variability, cognition, and aging. In F. I. M. Craik & T. A. Salthouse (Eds.), The handbook of aging and cognition (3rd ed., pp. 491–556). New York, NY: Psychology Press.
IBM. (2016). IBM SPSS statistics for windows (Version 24). Armonk, NY: IBM Corp.
Kochan, N. A., Bunce, D., Pont, S., Crawford, J. D., Brodaty, H., & Sachdev, P. S. (2016). Reaction time measures predict incident dementia in community-living older adults: The Sydney memory and ageing study. The American Journal of Geriatric Psychiatry, 24, 221–231. http://dx.doi.org/10.1016/j.jagp.2015.12.005
Li, S., Aggen, S. H., Nesselroade, J. R., & Baltes, P. B. (2001). Short-term fluctuations in elderly people’s sensorimotor functioning predict text and spatial memory performance: The Macarthur Successful Aging Studies. Gerontology, 47, 100 –116. http://dx.doi.org/10.1159/000052782
Lövdén, M., Li, S.-C., Shing, Y. L., & Lindenberger, U. (2007). Within-person trial-to-trial variability precedes and predicts cognitive decline in old and very old age: Longitudinal data from the Berlin Aging Study. Neuropsy- chologia, 45, 2827–2838. http://dx.doi.org/10.1016/j.neuropsychologia .2007.05.005
MacDonald, S. W. S., Hultsch, D. F., & Dixon, R. A. (2003). Performance variability is related to change in cognition: Evidence from the Victoria Longitudinal Study. Psychology and Aging, 18, 510 –523. http://dx.doi .org/10.1037/0882-7974.18.3.510
MacDonald, S. W. S., Hultsch, D. F., & Dixon, R. A. (2008). Predicting impending death: Inconsistency in speed is a selective and early marker. Psychology and Aging, 23, 595– 607. http://dx.doi.org/10.1037/0882- 7974.23.3.595
MacDonald, S. W. S., Karlsson, S., Rieckmann, A., Nyberg, L., & Bäck- man, L. (2012). Aging-related increases in behavioral variability: Rela- tions to losses of dopamine D1 receptors. The Journal of Neuroscience, 32, 8186 – 8191. http://dx.doi.org/10.1523/JNEUROSCI.5474-11.2012
Rabbitt, P., Osman, P., Moore, B., & Stollery, B. (2001). There are stable individual differences in performance variability, both from moment to moment and from day to day. The Quarterly Journal of Experimental Psychology A: Human Experimental Psychology, 54, 981–1003. http:// dx.doi.org/10.1080/713756013
Siegler, R. S. (1994). Cognitive variability: A key to understanding cog- nitive development. Current Directions in Psychological Science, 3, 1–5. http://dx.doi.org/10.1111/1467-8721.ep10769817
Smith, A. (1982). Symbol Digit Modalities Test. Los Angeles, CA: Western Psychological Services.
Stawski, R. S., MacDonald, S. W. S., Brewster, P. W. H., Munoz, E., Cerino, E. S., & Halliday, D. W. R. (2017). A comprehensive compar- ison of quantifications of intraindividual variability in response times: A measurement burst approach. The Journals of Gerontology: Series B, Psychological Sciences and Social Sciences. Advance online publica- tion. http://dx.doi.org/10.1093/geronb/gbx115
Stuss, D. T., Murphy, K. J., Binns, M. A., & Alexander, M. P. (2003). Staying on the job: The frontal lobes control individual performance
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
1003COVARIATION OF VARIABILITY AND COGNITION
variability. Brain: A Journal of Neurology, 126, 2363–2380. http://dx .doi.org/10.1093/brain/awg237
Tales, A., Leonards, U., Bompas, A., Snowden, R. J., Philips, M., Porter, G., . . . Bayer, A. (2012). Intra-individual reaction time variability in amnestic mild cognitive impairment: A precursor to dementia? Journal of Alzheimer’s Disease, 32, 457– 466. http://dx.doi.org/10.3233/JAD- 2012-120505
Vasquez, B. P., Binns, M. A., & Anderson, N. D. (2016). Staying on task: Age-related changes in the relationship between executive functioning and response time consistency. The Journals of Gerontology: Series B, Psychological Sciences and Social Sciences, 71, 189 –200. http://dx.doi .org/10.1093/geronb/gbu140
Wechsler, D. (1945). A standardized memory scale for clinical use. The Journal of Psychology: Interdisciplinary and Applied, 19, 87–95. http:// dx.doi.org/10.1080/00223980.1945.9917223
Whitehead, B. P., Dixon, R. A., Hultsch, D. F., & MacDonald, S. W. S. (2011). Are neurocognitive speed and inconsistency similarly affected in type 2 diabetes? Journal of Clinical and Experimental Neuropsychology, 33, 647– 657. http://dx.doi.org/10.1080/13803395.2010.547845
Williams, B. R., Hultsch, D. F., Strauss, E. H., Hunter, M. A., & Tannock, R. (2005). Inconsistency in reaction time across the life span. Neuro- psychology, 19, 88 –96. http://dx.doi.org/10.1037/0894-4105.19.1.88
Received March 19, 2018 Revision received July 21, 2018
Accepted December 11, 2018 !
T hi
s do
cu m
en t
is co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
ti on
or on
e of
it s
al li
ed pu
bl is
he rs
. T
hi s
ar ti
cl e
is in
te nd
ed so
le ly
fo r
th e
pe rs
on al
us e
of th
e in
di vi
du al
us er
an d
is no
t to
be di
ss em
in at
ed br
oa dl
y.
1004 BIELAK AND ANSTEY
