 

H was

1033

550 ,7

mm

Date

0-7 639

l LIBRARY
Michigan State
University

 

 

 

This is to certify that the
thesis entitled

UNITY, RELIABILITY, AND STRUCTURAL STABILITY OF
EXECUTIVE FUNCTION IN A NORMAL AGING POPULATION

presented by

Mark Lawrence Ettenhofer

has been accepted towards fulﬁllment
of the requirements for

M.A. degree in Psychology

 

Major professor

.SI/LI/os

MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:ICIRC/DateDue.p65-p. 15

 

UNITY, RELIABILITY, AND STRUCTURAL STABILITY OF
EXECUTIVE FUNCTION IN A NORMAL AGING POPULATION

By

Mark Lawrence Ettenhofer

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

MASTER OF ARTS
Department of Psychology

2003

ABSTRACT

UNITY, RELIABILITY AND STRUCTURAL STABILITY OF
EXECUTIVE FUNCTION IN A NORMAL AGING POPULATION

By

Mark Lawrence Ettenhofer

Available evidence suggests that the decline of executive functions is a common
part of the normal aging process (Garden, Philips, & MacPherson, 2001; Wecker, et al.,
2000; Schretlen, et al., 2000; Bryan & Luszcz, 2000; Robbins, et al., 1998; West, 1996).
However, many questions have been raised about the reliability, structure, and stability of
executive functions. This study addressed these questions via the collection of
neuropsychological data at two time points ﬁ‘om a sample of 151 individuals aged 55 and
older. This data included ﬁve common measures of executive function (Trail Making
Test, Stroop Color-Word Test, Wisconsin Card Sorting Test, Letter Fluency, and
Category Fluency). Exploratory factor analyses found that a single factor accounted for
greater than 50% of the variance in four measures of executive function at both time
points; Stroop Color-Word Test was dropped from analysis. Structural analyses of the
longitudinal model found that both the regression weights of the executive measures and
the variance of the error terms could be held constant across both time points without a
signiﬁcant reduction in model ﬁt, indicating a large degree of longitudinal stability of
executive function. Structural analyses of this longitudinal model found that the effect of
age on executive function increased signiﬁcantly from Time 1 to Time 2 (12(1, p = 151)

=9Bip<ﬂn.

ACKNOWLEDGEMENTS

This research was supported by a Student Award Program Grant from the Blue
Cross Blue Shield of Michigan Foundation, and a grant from the Pearl Aldrich
Endowment in Gerontology. I also owe many thanks to Norman Abeles for serving as
my mentor and supervising the collection of the data for this study long before my arrival
in East Lansing, and to Zach Hambrick for the countless hours of assistance he provided

in the preparation of this manuscript.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................ v
LIST OF FIGURES .......................................................................................................... vi
INTRODUCTION ............................................................................................................ 1
Reliability of Executive Function ......................................................................... 2
Executive Function as a Uniﬁed Construct .......................................................... 4
Evidence of Age-Related Frontal/Executive Decline ........................................... 6
Overview and Predictions ..................................................................................... 9
METHOD ......................................................................................................................... 1 1
Participants ............................................................................................................ 11
Procedure .............................................................................................................. 1 1
Measures ............................................................................................................... 12
Screening Instruments ............................................................................... 12
Executive Function ................................................................................... 14
Analysis ................................................................................................................. 19
RESULTS ......................................................................................................................... 21
Reliability of Executive Measures ........................................................................ 21
Exploratory Factor Analysis ................................................................................. 21
Structural Analysis of Longitudinal Stability ............................ . ........................... 23
DISCUSSION ................................................................................................................... 25
APPENDIX ....................................................................................................................... 30
LIST OF REFERENCES .................................................................................................. 37
iv

LIST OF TABLES

Table 1. Demographic Characteristics of Sample ............................................................. 32
Table 2. Descriptive Statistics of Executive Variables at Time 1 and Time 2 ................. 32
Table 3. Test-Retest Correlations ..................................................................................... 33

Table 4. Correlations and Factor Loadings of Executive Functioning Measures,

Eigenvalue > 1 Method ..................................................................................................... 33
Table 5. Correlations and Factor Loadings of Executive Functioning Measures,

One Factor Extracted ........................................................................................................ 34
Table 6. Correlations and Factor Loadings of Executive Functioning Measures,

One Factor Extracted and SCWT Excluded ..................................................................... 34
Table 7. Goodness-of-Fit Statistics for Alternative Models ............................................. 34
Table 8. Factor Loadings and Residual Variances for the Longitudinal Factor

Model (03) ........................................................................................................................ 35
Table 9. Variance of Executive Function for the Longitudinal Factor Model (03) .......... 35

LIST OF FIGURES

Figure 1. Accepted model of longitudinal EF factor structure (Model 03) ...................... 36

Figure 2. Accepted model of relationship of age to longitudinal EF factor structure
(Model 05) ........................................................................................................................ 37

INTRODUCTION

Popular beliefs about the effects of aging vary across cultures. Whereas Chinese
individuals generally ascribe wisdom, power, and prestige to advancing age, those living
in the United States often believe that aging inevitably leads to senility and powerlessness
(Silverrnan, Hecht, & McMillin, 2000). Scientiﬁc studies have demonstrated that
although some individuals seem to progress into old age without developing any
signiﬁcant mental deﬁcits (Seeman, Lusignolo, Albert, & Berkrnan, 2001; Silver & Perls,
2001), others may lose aspects of their memory, their visuospatial skills, or their mental
speed (Ylikoski et al., 1999; Mitrushina, Uchiyama, & Satz, 1995). Additionally, the
process of aging often causes subtle but important changes to an individ ’s “executive”
abilities, including volition, inhibition, and attention (W ecker, Kramer, Wisniewski,
Delis, & Kaplan, 2000).

Research examining the causes of cognitive aging has consistently demonstrated
that measures of psychomotor speed can account for a large amount of detectable age-
related variance in test scores, supporting the hypothesis that generalized slowing is at
least partially responsible for decrements in cognitive performance with advanced age
(Schretlen et al., 2000; Salthouse, Fristoe & Rhee, 1996; Salthouse, 1991). However,
other researchers have proposed that the frontal lobes are disproportionately vulnerable to
the effects of aging (Celine, Michel, & Espagnet, 2000; Rympa, Prabhakaren, Desmond,
& Gabrieli, 2001; Goldberg, 2001), and that this deterioration is one causative factor of
the “normal” decline seen in a wide variety of cognitive abilities (West, 1996; Band,
Ridderinkhof, & Segalowitz, 2002). Because executive function as a whole is largely

reliant upon the frontal lobes (Lezak, 1995; Band et al., 2002; Celine, Michel, &

Espagnet, 2000), cognitive functions that are normally coordinated with the aid of proper
executive function may become dysregulated once frontal deterioration has progressed to
a signiﬁcant degree (Goldberg, 2001).

Research has consistently demonstrated the negative relationship between age and
executive function (Garden, Philips, & MacPherson, 2001; Wecker et al., 2000; Schretlen
et al., 2000; Bryan & Luszcz, 2000; Robbins et al., 1998; West, 1996). Although the
precise deﬁnition of “executive function” is a matter of great debate, it is generally
accepted that this category includes the ability to plan and implement goal-directed
actions, voluntarily initiate and terminate behaviors, and purposeﬁrlly direct attention and
behavior (LaRue, 1992). Speaking more generally, Lezak (1995) describes executive
functions as those ﬁmctions that are “necessary for appropriate, socially responsible, and
effectively self-serving adult conduct” (p. 507). Additionally, these functions tend to
cluster together; it is relatively rare to ﬁnd an individual with a deﬁcit in one executive
function without deﬁcits in others as well (Lezak, 1995). Because the category of
functions describedas “executive” is rather abstract and composed of diverse elements,
tasks that assess executive function vary considerably.

Reliability of Executive Function

Although tasks purporting to measure executive function have been used
extensively in aging research, relatively little is known about the psychometric properties
of these tasks, either individually or collectively (Rabbitt, 1997). Additionally, what
evidence is available has led some authors to conclude that executive measures as a
whole cannot be validly administered more than once (Lowe & Rabbitt, 1998; Rabbit,

1997; Phillips, 1997). Phillips (1997) states:

The whole idea of assessing test-retest reliability in executive function is

problematic because a task cannot be novel the second time around. Even

conventional parallel forms of the same test do not overcome the problem: the

content may be new, but the format is not. Also, how could the spontaneous

generation of, say, a particular task strategy be measured twice? (p. 208)
Assertions such as these cast serious doubt on the prospect of examining changes in
executive function across multiple measurement occasions. An evaluation of these
beliefs is a prerequisite to the meaningful examination of age effects on executive
function in a longitudinal context. In order to address these issues in this study, test-
retest correlations have been computed for each of ﬁve measures of executive function.
Additionally, analyses have been conducted of the longitudinal stability of factor models
of executive function, in order to provide a more complete picture of the reliability of the
underlying construct.

As demonstrated by Hertzog and Shaie (1986) and replicated in this study, one
means of testing for longitudinal invariance in a construct such as executive function
involves the analysis of a series of structural models that include data from multiple
measurement occasions. First, a theoretically-based baseline model is created, and model
ﬁt is estimated. Additional constraints, such as restricting factor loadings to be constant
across measurement occasions, are imposed for each of the models that follow.
Individual hypotheses regarding the invariance of factor loadings, error variances, and
factor variances, can then be tested by comparing the overall ﬁt of the successive models

(Hertzog & Shaie, 1986).

Executive Function as a Uniﬁed Construct

Although widespread use of the term “executive function” itself implies a large
degree of convergence between the tasks designed to measure it, it is unclear to what
degree executive function may be more accurately characterized as consisting of multiple
independent functions. For example, Duncan, Emslie, and Williams (1996) cite many
studies employing animal models that seem to indicate anatomical modularity of
executive function.

Support in humans for this type of modularity was found in an exploratory factor
analysis of “dysexecutive” symptoms, which included such indicators as behavioral
deﬁcits in inhibition, intentionality, executive memory, and affect (Burgess, Alderman,
Evans, Emslie, & Wilson, 1998, p. 547). Five orthogonal factors were found, collectively
explaining 67.2% of the variance. The authors concluded that neuropsychological
assessments should include multiple measures of executive function to properly tap
various components of the “dysexecutive syndrome” (Burgess et al., 1998, p. 556).

Additionally, among a sample of head-injured patients, Duncan, Johnson, Swales,
and Freer (1997) found low correlations between executive measures, implying more
divergence than convergence between executive functions. Supporting these results,
Lamar, Zonderman, and Resnick (2002) conducted a factor analysis of a wide range of
neuropsychological tasks (both executive and non-executive) in a sample of nondemented
older adults, and found no evidence that any of the factors that emerged were uniquely
executive in nature. In contrast, in a factor analysis of patients with frontal lesions, Della
Sala, Gray, Spinnler, and Trivelli (1998) found a single factor that accounted for 53% of

the variance among executive measures, as well as an intercorrelation matrix that

averaged r = .48. The conclusion that these authors drew was that a uniﬁed model of
executive function continues to be the best available conceptualization.

Other studies have produced more mixed results. One factor analysis including
multiple indices from four measures of executive function found three separate but
modestly correlated factors (Boone, Ponton, Gorsuch, Gonzalez, & Miller, 1998). These
results, drawn from a sample of patients referred for neuropsychological evaluation, seem
to suggest some degree of both unity and diversity in executive function. Likewise, a
conﬁrmatory factor analysis of nine executive measures in a college sample found that a
model with three distinct executive factors ﬁt the available data better than a single-factor
model (Miyake, Friedman, Emerson, Witzki, & Howerter, 2000). However, much like
the previous study, the authors in this study found moderate correlations between these
three factors, suggesting a signiﬁcant degree of communality between measures as well.

Clearly, the issue of unity versus diversity of executive functions remains
unresolved. Further evidence is necessary in order to conclusively determine if the
implicit one-factor model of executive function that continues to inﬂuence much clinical
judgment is adequate, or if more sophisticated models of executive functioning are
necessary. This study includes a series of factor analyses involving ﬁve common
measures of executive function in order to provide evidence that may help resolve this
discrepancy.

Evidence of Age-Related Frontal/Executive Decline

Although the individual neuropsychological proﬁles of older adults are quite

diverse (Ylikoski et al., 1999), in aggregate form the data from neuropsychological

testing paint a clear picture of executive decline with increased age (Garden, Philips, &

MacPherson, 2001; Wecker et al., 2000; Schretlen et al., 2000; Bryan & Luszcz, 2000;
Robbins et al., 1998; West, 1996). As a whole, neuropsychological tests of executive
function demonstrate a great deal of sensitivity to frontal lobe damage, particularly in the
prefrontal regions (Lezak, 1995). To the degree that these executive tests are sensitive to
frontal dysfunction, the consistent ﬁndings of age-related decline that they yield
demonstrate changes in the functional capacity of the ﬁ'ontal lobes attributable to aging.
Speciﬁc examples of executive tests that have been shown to demonstrate sensitivity to
frontal lobe damage include the Trail Making Test (TMT; Crowe, 1998; Arbuthnott &
Frank, 2000), the Stroop Color-Word Test (SCWT; Stuss, F loden, Alexander, Levine, &
Katz, 2001; West & Bell, 1997; Vendrell et al., 1995), the Wisconsin Card Sorting Test
(W CST; Wang, Kakigi, & Hoshiyama, 2001; Rogers, Andrews, Grasby, Brooks, &
Robbins, 2000; Konishi et al., 1999; Lombardi et al., 1999; Raz, Gunning-Dixon, Head,
Dupuis, & Acker, 1998; Berman et al., 1995; Rezai et al., 1993), and tests of verbal
ﬂuency (Lezak, 1995). Additionally, as will be addressed in more detail below, studies
employing brain imaging techniques have established direct connections between age-
related decline of the frontal lobes and neuropsychological tests of executive function,
including the CF L (Frith, Friston, Liddle, & Frackowiak, 1991), the Wisconsin Card
Sorting Test (W CST; Raz et al., 1998) and the SCWT(Mi1ham et al., 2002).

Structural brain imaging has also proved to be a valuable resource for information
regarding the effects of aging on the brain. Overall brain volume has been reported to
decline steadily ﬁ'om age 16 to 65 (Bilger, 1997). However, an investigation comparing
young adults to individuals in their 7 0s has demonstrated volume reductions in the frontal

cortex of 10-17%, compared to volume reductions of approximately 1% in the temporal,

parietal, and occipital cortices (Haug & Eggers, 1991). A similar study focusing upon the
prefrontal cortex found that this area was more affected by age (volumetrically) than any
other region of interest (Raz et al., 1998). Corroborating the link between age-related
frontal dysfunction and performance on neuropsychologicaltests of executive function,
this study also found that atrophy of the preﬁ'ontal cortex mediates age-related decreases
in performance on the WCST. Together, these data suggest that the frontal cortex is
selectively susceptible to the deleterious effects of aging, and that tests of executive
function may be sensitive to these changes.

Although structural information has proven very useful, age-related changes in
the brain may also occur on a more microscopic level or may be functional in nature, and
may therefore remain invisible to structural brain scans. In these cases, functional brain
imaging is useﬁrl in detecting irregularities or changes in metabolic activity underlying a
functional deﬁcit (Fuster, 1997). Accordingly, this technology also provides a window
into the structures associated with neuropsychological task performance in healthy
individuals, and differences in brain metabolism between young and old individuals. For
example, an age-related reduction of up to 27% in regional cerebral blood ﬂow (rCBF)
has been demonstrated in some cortical regions, as measured by positron emission
tomography (PET) (Shaw et al., 1984).

Generally speaking, reduced blood ﬂow to a brain area is indicative of lower
regional metabolism, which results from lower brain activity (Goldberg, 2001; Fuster,
1997). However, this reduction in overall brain metabolism is not uniform across
different brain regions. Both cross-sectional and longitudinal studies have demonstrated

that older individuals often exhibit a resting pattern of metabolic hypofrontality, in

contrast to the relative hyperﬁontality evident in younger adults (Goldberg, 2001;
Giaquinto, 1988; Gur, Gur, Orbist, Skolnick, & Reivich, 1987; Shaw et al., 1984). This
pattern seems to vary depending upon the task in question. During visual perception
tasks, for example, older individuals have been consistently found to display more frontal
lobe activation than their younger counterparts (Cabeza, 2001). Although it may seem
counterintuitive that older adults would display relatively more regional activation for a
given task, results such as these are often explained in terms of adaptive compensatory
processes. As individuals age, it may become increasingly difﬁcult to coordinate the
myriad component abilities involved in even simple tasks without the recruitment of
supplementary frontal areas. Consistent with this hypothesis, older adults who engage
supplementary areas of the prefrontal cortex often perform better than older adults who
do not, despite the fact that younger adults tend to perform better overall (Cabeza, 2001 ).

Functional imaging has also been used to validate various neuropsychological
tasks by isolating the areas of the brain which are uniquely activated during their
performance. PET examinations of letter ﬂuency (CFL), for example, have demonstrated
that this task is associated with an increase in activity in the left dorsolateral prefrontal
cortex, providing evidence that this measure is a meaningful index of ﬁ'ontal lobe
function. Similar activation of the dorsolateral prefrontal cortex was found in an MRI
study of the SCWT (Milham et al., 2002). In addition, this study found age differences in
the responsiveness of the dorsolateral prefrontal cortex to the demands of the SCWT,
evidence suggestive of age-dependent frontal/ executive decline.

To summarize, the relationships between age, executive function, and frontal lobe

activity are fairly well documented. The presence of these relationships makes greater

conceptual understanding and more accurate assessment of executive function within an
aging population a high priority. In this study, the longitudinal stability of the
relationship between age and executive function has been examined, as it was believed
that an examination of this relationship might provide more information regarding the
general stability of the construct. The hypothesis that age effects on executive function
would be similar across two measurement occasions was tested in the manner
demonstrated by Hertzog and Shaie (1986); that is, the overall ﬁt of a model in which the
effect of age on executive function at two measurement occasions was held constant was
compared to the ﬁt of a baseline model in which the effect of age on executive function
was freely estimated.
Overview and Prediction_s

Available evidence suggests that the decline of executive functions is a common
part of the normal aging process. However, many questions have been raised about the
reliability, structure, and stability of executive functions, both in general as well as within
an aging population. This study was designed to address these questions via the
collection of neuropsychological data from a large sample of aging individuals at two
time points. First, the reliability of various measures of executive ﬁrnction was estimated
by comparing individuals’ scores at two measurement occasions. SeCond, factor analyses
were performed in order to determine the structure of executive function within this
population. Guided by this information, analyses of the longitudinal stability of
executive functions were then conducted. Further analyses were conducted in order to
examine the magnitude of the effect of age on executive function, and the longitudinal

stability of this effect.

We expected that reliability coefﬁcients for the measures of executive function
employed in this study would fall within the moderate-to-high range, and that these
coefﬁcients would be greater for measures that utilized directly observed scores (CFL,
CAT, WCST) than for measures that rely upon composite scores (SCWT, TMT).
Furthermore, we expected that the results of factor analyses would be consistent with a
single-factor model of executive function, and that this factor structure would be stable
across the two time points. Finally, we expected the results of structural analyses to be
consistent with a large and longitudinally stable effect of age upon the latent executive

function variable.

10

METHOD

Participants

Participants consisted of 151 home-dwelling individuals aged 54 to 92 (M =
69.80, SD = 8.47) who had been recruited through local newspaper advertisements and
talks given to local community groups in the greater Lansing area. These individuals
were solicited to participate in educational “Mood and Memory” seminars, and also
received feedback regarding their performance subsequent to all testing. Only
participants who completed both testing sessions (before and after the seminar) were
included. Participants whose MMSE scores suggested dementia (24 or less) and
participants who exhibited moderate to severe depression on the BDI (30 or greater) or
the GDS (20 or greater) were excluded from analysis (F olstein, Folstein, & McHugh,
1975; Beck, Steer, & Garbin, 1988). These cutoffs help ensure that the sample is
representative of the “normal” aging population. Gender and years of education were
also monitored in order to determine sample characteristics. Detailed demographic
characteristics of the sample are presented in Table 1.
Procedure

In addition to self-report demographic information, a standard battery of
neuropsychological test data was collected, including the following measures: the State-
Trait Anxiety Inventory, the MMSE, CF L, CAT, the California Verbal Learning Test,
TMT Part A and B, WAIS-III Digit Span, WAIS-III Visual Memory Span, WAIS-III
Digit-Symbol Coding, the WCST, the SCWT, the American version of the New Adult
Reading Test, the Storandt Mental Control Battery, the Benton Visual Retention Test,

WAIS-III Symbol Search, and the Clock Drawing Test. This battery typically takes 90 to

11

120 minutes to complete. These tests were administered both immediately prior to
participation and subsequent to participation in the “Mood and Memory” seminars,
yielding two sets of data for each participant (Time 1 and Time 2). The amount of time
that passed between Time 1 and Time 2 varied, but was always between four and eight
weeks.
Measures

Many questions have been raised about the interpretation of neuropsychological
test data in regard to assumptions that are often made concerning the localization of brain
damage when neuropsychological test deﬁcits exist. Although any given measure may
demonstrate sensitivity to damage in particular regions of the brain, it may not always be
appropriate to interpret test deﬁcits in terms of damage to that brain region, as similar
deﬁcits could be caused by damage to other brain regions as well (Salthouse, Fristoe, &
Rhee, 1996). In response to this and related issues, great care has been taken to ensure
that the measures selected have demonstrated validity for their intended purpose in this
project.
Screening Instruments

Mini Mental Status Exam. (MMSE; Folstein, Folstein, & McHugh, 1975): The
MMSE is a brief measure of overall orientation widely used in assessments of cognitive
function. The test consists of 30 items, that individually also provide rough measures of
memory, visuospatial function, and the ability to follow instructions. With most
populations, a cutoff score of 24 is recommended in order to screen out individuals with
dementia or other clinically-meaningful cognitive problems (Lezak, 1995).

Test-retest reliability for this test has been estimated at .85 to .99 after 24-hours

12

(Foster, Sclan, Welkowitz, Boksay, & Seeland, 1988). This test has also been shown to
correlate highly with measures intelligence, memory, and executive function (Mitrushina
& Satz, 1991; Axelrod, Goldman, & Henry, 1992). Additionally, studies have show that
the MMSE has adequate speciﬁcity and sensitivity for detecting moderate to severe forms
of dementia (Spreen & Strauss, 1998).

Beck Depression Inventog. (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh,
1961): The BDI is a self-report measure consisting of 21 statements, each addressing a
depressive symptom. The participant is asked to rate the accuracy of each item based
upon their experience on a scale from 0 to 3. Total score for the BDI is the sum of all 21
numbers circled. Scores range from 0 to 63, higher scores indicating higher measured
levels of depression. The following cutoff scores were identiﬁed by the authors of this
test: normal: 0-9; mild depression: 10-15; mild/moderate depression: 16-19;
moderate/severe depression: 20-29; severe depression: 30 or greater (Spreen & Strauss,
1998). Test-retest reliability for this measure has been estimated to be above .90 (Beck,
Steer, & Garbin, 1988).

geriatric Depres_si9n Soak (GDS; Yesavage et al., 1983): The GDS is a self-
report paper-and-pencil measure consisting of 30 items for which scoring directionality
of answers (yes or no) changes pseudo-randomly. Because this test was developed for
use with older adults, it focuses upon items that the authors found to be most relevant to
an aging population. For example, the somatic items in the GDS have been tailored in
such a way in which they are thought to be appropriate for older populations (Spreen &
Strauss, 1998). The participant is asked to read each statement and circle the response

that is most accurate in terms of their own experience. Total score for this test ranges

l3

from 0 to 30, higher numbers indicating higher measured levels of depression. The
following cutoff scores identify graded levels of depressive symptomology: normal: 0-9;
mild depression: 10-19; moderate/severe depression: 20—30 (Spreen & Strauss, 1998).
Correlations between the GDS and other measures of depression are high for the Beck
Depression Inventory (r = .73), the MMPI Depression Scale (r = .72), and the Hamilton
Depression Scale (r = .83; Yesavage et al., 1983).
Executive Function

The Stroop Color-Word Test. (SCWT; Stroop, 193 5): The Stroop Color-Word
test requires participants to read through one list of words and successively name two
lists of colors as rapidly as possible. Each condition has a time limit of 45 seconds. In
the ﬁrst condition (“word” , the list is composed of ﬁve columns of 20 color words
(“blue”, “red”, “green”, or “yellow”) printed in black ink. In the second condition
(“color”), the participant is required to identify the color of ink (blue, red, green, or
yelloW) in which a list of pseudo-words (all “xxxx”) is printed. This list of pseudo-words
is in the same 20 x 5 format as the words in the ﬁrst trial. Finally, the third condition
requires the identiﬁcation of color of ink in which a list of color words (“blue”, “red”,
“green”, or “yellow”) is printed, again in a 20 x 5 format. This is called the
“interference” condition, because there is a discrepancy between the color of ink and the
color word itself, making identifying the color more effortful and difﬁcult. Performance
on the SCWT is assessed by the difference between the number of words achieved in the
interference condition and the color condition (the “interference effec ”; Lezak, 1995;
Spreen & Strauss, 1998).

It is believed that the SCWT measures the degree to which participants are able to

14

shift attention and inhibit prepotent responses, two executive functions associated with
the frontal lobes (Spreen & Strauss, 1998; West, 1996). The validity of this assumption
is supported by Vendrell, et a1. (1995), who found a signiﬁcant difference in number of
interference condition errors committed between frontal lobe patients and matched
controls. Likewise, Stuss, et a1. (2001) found that among 51 patients with frontal and
non-frontal brain lesions and 26 normal controls, only those patients with frontal lesions
exhibited signiﬁcant impairment on the SCWT. Psychophysiological evidence conﬁrms
the importance of frontal functions to performance on the SCWT as well. Pairing
electroencephalography (EEG) with the SCWT, West & Bell (1997) found a relationship
between the magnitude of the interference effect and age-related decline of the anterior
attention system. Finally, support for use of the SCWT as a measure of
prefrontal/executive function within an aging population comes from a study employing
functional magnetic resonance imaging, which found both poorer performance on the
SCWT and decreased responsiveness of the dorsolateral prefrontal cortex among older
individuals (Milharn et al., 2002). Test-retest reliabilities for the word, color, and
interference conditions of the SCWT have been estimated at .90, .83, and .91 ,
respectively (Spreen & Strauss, 1998).

Wisconsin Card Sorting Test. (W CST; Heaton, Chelune, Talley, Kay, & Curtiss,
1993): The WCST is a test that requires the participant to sort up to cards consecutively
under one of four “key cards”. On each card are printed one to four triangles, stars,
crosses, or circles in red, green, yellow, or blue. The participant is not told how to sort
the cards, and instead must deduce the correct principle using only the examiner’s

feedback. The examiner says “correct” or “incorrec ” after each sort depending upon the

15

sorting principle. This sorting principle changes each time the participant sorts 10 cards
in a row correctly, but the participant is not informed of this change, and must deduce
from examiner feedback what rule applies at any given time. The possible dimensions by
which the cards can be sorted are the form (shape) of the symbols, the color of the
symbols, and the number of symbols on the card.

This test was originally designed to measure abstract reasoning and concept
formation abilities in response to a changing context (set-shifting), executive functions
presumed to be reliant upon the frontal lobes (Barcelo, 2001). Since its introduction, the
WCST has endured a great deal of controversy, and remains one of the most widely used
neuropsychological tests in existence. As such, a great deal of research has been
dedicated to validating the WCST for the purpose of measuring executive and ﬁontal
lobe function.

Initial support came ﬁ'om ﬁndings that individuals with prefrontal lesions tend to
exhibit a greater number of perseverative errors on this test (Anderson, Darnasio, Jones,
& Tranel, 1991). Further support for the ability of the WCST to measure ﬁontal lobe
function has come from studies employing neuroirnaging techniques, including: a SPECT
imaging study that found that the WCST task is associated with activation in the left
dorsolateral prefrontal cortex (Rezai et al., 1993), a functional magnetic resonance
imaging (MRI) study that demonstrated that both the working memory and set-shifting
components of the WCST operate within the preﬁontal cortex (Konishi, et al., 1999), a
positron emission tomography (PET) study that demonstrated that metabolic deﬁcits in
the right dorsolateral prefrontal cortex predict greater numbers of perseverative errors

(Lombardi et al., 1999), as well as multiple PET and magneto-encephalography (MEG)

16

studies that demonstrated that performance of the WCST is associated with activation in
frontal areas (Berman et al., 1995; Rogers, etal., 2000; Wang, Kakigi, & Hoshiyama,
2001). Additionally, an MRI study that demonstrated that shrinkage of the prefrontal
cortex due to normal aging predicts greater numbers of perseverative errors on the WCST
(Raz et. a1, 1998) supports the present use of the WCST to measure prefrontal/executive
function within an aging population. Lezak (1995) reports estimates of interscorer and
intrascorer reliabilities on the WSCT to be .88 and .96, respectively (Lezak, 1995).
Performance in this study was assessed by the total number of perseverative errors. To
summarize, Heaton et al. (1993) deﬁne perseverative errors as incorrect responses that
were correct in immediately preceding category, but for which potentially corrective
feedback has already been provided.

Trail Making Test. (TMT; Reitan, 1958): The TMT is a paper-and-pencil test
consisting of Part A and Part B. Part A requires the participant to draw a line, connecting
numbered circles in sequential order. Part B requires the participant to draw a line,
connecting circles with numbers or letters in alternating sequential-alphabetic order. Any
mistakes made during the performance of this test corrected by the examiner as they are
made. Score on each part of the TMT is determined by the time required to complete
each trial. Because errors must be corrected before continuing, they augment this time
score.

By virtue of its relatively simple design, performance on Part A of the Trail
Making Test depends largely upon the participant’s psychomotor speed and visual search
abilities (Crowe, 1998; Arbuthnott & Frank, 2000). Part B, which requires the participant

to maintain two mental sets (both numbers and letters) and alternate between them, places

17

additional demands upon the participant’s working memory and cognitive ﬂexibility, two
important elements of executive function (Crowe, 1998; Arbuthnott & Frank, 2000).
This study employs a (Part B — Part A) derived score as a measure of executive function.
Letter Fluency (CFL) and Category Fluency (CAT ). 1 (Benton, Hamsher, & Sivan,
1983): CFL and CAT, referred to collectively as the “Controlled Oral Word Association
Test” (COWAT), are tests of verbal ﬂuency consisting of three trials. For CFL, the
examiner instructs the participant in each trial to say as many words that begin with a
particular letter of the alphabet as possible, excluding proper nouns as well as different
forms of the same word (i.e., “bookworrn” and “bookshelf”). For this study, the letters C,
F, and L were used, based on the respectively high, moderate, and low frequencies of
words in the English language that begin with these letters. For CAT, the examiner
instructs the participant to say as many words that belong to a certain category as
possible. In this study, the following categories were used: animals, vegetables, fruits.
The COWAT has been found to be a sensitive indicator of brain dysfunction,
especially in the frontal lobes (Lezak, 1995). A study of the CFL using Positron
Emission Tomography (PET) provided further validation for the CFL as a measure of
ﬁontal lobe function by demonstrating that generation of words in this manner was
uniquely associated with activation in the left dorsolateral prefrontal cortex (Frith et al.,
1991). Another study of rCBF has demonstrated that performance of both. CFL and CAT
activate similar, but not identical, brain regions, including left prefrontal regions
(Gourovitch et al., 2000). In addition, previous studies have demonstrated that the CFL is
sensitive to age effects, in that younger adults perform better than older adults by

generating a greater number of words for each letter (W inocur, Moscovitch, & Stuss,

l8

1996). Performance on the CF L and CAT was assessed by the total number of correct
words generated across all three trials.
mm

The SPSS 10.0.5 (1999) statistical package was used. to perform all correlational
and exploratory factor analyses, and the AMOS 4.01 (1999) statistical package was used
for all structural analyses. First, descriptive statistics and test-retest correlations of the
executive function measures at Time 1 and Time 2 were computed.

Exploratory factor analyses were performed for the executive function variables at
Time 1 and Time 2, respectively, in order to examine the factor structure of executive
ﬁrnction in this population. Consistent with the guidelines outlined by Fabrigar,
Wegener, MacCallum, and Strahan (1999), all analyses employed principal axis
factoring, with Promax rotation for multiple-factor solutions. The ﬁrst step in these
exploratory factor analyses included the following variables at Time 1 and Time 2,
respectively: WCST Perseverative Errors, TMT Composite Score, SCWT Interference
score, CAT Total, and CF L Total. Initially, factors with eigenvalues of one or greater
were included in the factor analysis. Further analyses were restricted to one-factor
solutions in the interest of parsimony.

Structural analyses were then performed in which the factor loadings of the
following variables on their respective latent “executive function” variables at Time 1
and Time 2 were freely estimated: WCST Perseverative Errors, TMT Composite Score,
Category Fluency Total, and CFL Total (see Figure 1). As is common practice for
longitudinal factor models of this type (Hertzog & Shaie, 1986), latent executive function

variables and residuals were permitted to autocorrelate across time in this model (denoted

l9

01) in order to achieve maximal model ﬁt. A model similar to O; was then estimated,
with the added restriction that all factor loadings be held constant (model 02). A chi-
square value was computed in order to determine the signiﬁcance level of the difference
in overall model ﬁt between the models 01 and 02. A model similar to 02 was then
estimated, with the added restriction that the residual variance was held constant across
time (model 03). Because only the variance of EF remained freely estimated in this
model, the chi-square comparison of 03 with 02 essentially tested the hypothesis that any
observed differences between Time 1 and Time 2 could be attributed to changes in EF
factor variance. Model 04 was then estimated in which the hypothesis of complete
longitudinal invariance was tested by adding the restriction that EF factor variance was
held constant. In the next step, structural analyses were conducted on a model similar to
model 03 which allowed the observed variable of age to load freely on EF1 and BF;
(model 05; see Figure 2). Model 06 was then estimated by restricting the path

coefficients of age on EF1 and BF; to be held constant.

20

RESULTS

Reliability of Executive MeasLures

First, descriptive statistics and dependent-sample t-tests were computed for all
executive measures (see Table 2). Mean number of WCST perseverative errors was
signiﬁcantly lower (indicating better performance) at Time 2 than at Time 1. Mean CFL
total score was signiﬁcantly higher (indicating better performance) at Time 2 than at
Time 1. Mean TMT time score, Category Fluency total score, and Stoop interference
score were not signiﬁcantly different at Time 2 than at Time 1.

Next, test-retest correlations were computed for all executive variables (see Table
3). Test-retest correlations for CF L total score and Category Fluency total score were
quite high, at .81 and .85, respectively. Test-retest correlations for SCWT interference
score, WCST perseverative errors, and TMT time score were moderate, at .61 , .50, and
.49, respectively. To summarize, performance increased signiﬁcantly from Time 1 to
Time 2 for two of the ﬁve EF measures, and test-retest correlations for all ﬁve measures
were in the moderate-to-high range.
E_Xp.10_rat0ry Fﬁm MIME

Exploratory factor analyses were then performed in order to examine the
similarity of the factor structure for the EF measures at the two measurement occasions.

§_t_ep_1, The executive measures at Time 1 were entered into a principal axis
factor analysis with Promax rotation. This analysis yielded a two-factor solution
accounting for 60.9% of the variance (see Table 4). The ﬁrst factor was weighted more
heavily by WCST Perseverative Errors, TMT Composite Score, Category Fluency Total,

and CFL Total than the second factor, and accounted for 40.6% of the variance. The

21

second factor accounted for 20.3% of the variance, and was weighted by SCWT
interference score more heavily than the ﬁrst factor.

Step; All executive measures at Time 2 were entered into a principal axis factor
analysis. This analysis yielded a one-factor solution accounting for 45.4% of the
variance (see Table 4). Because only one factor was extracted, no rotation was
necessary. SCWT interference score had a weaker loading on this factor than the other
executive measures.

M Consistent with predictions regarding the factor structure of the executive
measures, only one factor contributed meaningfully to interpretation of factors at both
Time 1 and Time 2. Further exploratory analyses were therefore restricted to one-factor
solutions. Principal axis factoring was conducted again with this restriction on all
executive measures at Time 1 (see Table 5). The factor extracted accounted for 40.6% of
the variance. SCWT interference score did not contribute meaningfully to this factor.

_S_t_ep_$ Because SCWT interference score had a substantially lower factor
loading than all other executive measures at both Time 1 and Time 2, this measure was
excluded from further factor analyses. Principal axis factoring, restricted to one factor,
was conducted again with all executive measures except for SCWT interference score, at
Time 1 and Time 2, respectively (see Table 6). Removal of SCWT interference score
from analyses increased the amount of variance accounted for at Time 1 to 50.8%, and at
Time 2 to 54.4%.

To summarize, a one-factor model including four of the ﬁve original measures of

executive function emerged from a series of exploratory factor analyses. The SCWT was

22

excluded from this model because it lacked communality with the other executive

measures.

Structural Analysis of Longitudinal Stabilig

 

Structural analyses were then performed on the four remaining executive
measures in order to examine the longitudinal stability of executive ﬁrnction in this
sample. Model 0., the baseline model for these analyses, provided good overall model
ﬁt (x2 = 30.38, df= 15, Bender-Bonnet normed ﬁt index (NFI) = .987). Model 02, which
constrained the factor loadings to be held constant across time, did not produce a
signiﬁcant change in model ﬁt (x2 = 33.99, df = 18, NF I = .986), and therefore the
hypothesis that the factor loadings are invariant across time could not be rejected. Model
03, which constrained the residual variance to be held constant across time, did not
produce a signiﬁcant change in model ﬁt ﬁom model 02 (x2 = 40.93, df = 22, NFI =
.983), and therefore the hypothesis that the residual variances are invariant across time
could not be rejected. In this model, differences in the structure of EF’ from Time 1 to
Time 2 are necessarily due to changes in the variance of EF itself, not in the degree to
which each variable loads on EF or in the variance of measurement error. Model O4,
which constrained the factor variance of EF to be held constant across time, produced a
signiﬁcant decrease in ﬁt from model 03 (x2 = 45.80, df = 23, NFI = .981), and therefore
the hypothesis that the variance in EF is invariant across time was rejected. Variance of
executive function for the longitudinal model is presented in Table 9. Model 03 therefore
became the accepted model of longitudinal EF factor structure. Goodness-of-ﬁt indices
for models 01 through 04 are presented in Table 6. To summarize, both the loadings of

the executive measures on a single factor (EF) and the error variance associated with

23

these measures were similar across measurement occasions. However, the variance of EF
was greater at Time 2 than at Time 1.

Next, structural analyses were performed in order to examine the similarity of the
effect of age on executive function at the two measurement occasions. Model 05, the
baseline model for structural analysis, allowed age to load ﬁ'eely on EF at both time
points, was otherwise similar to the accepted model of longitudinal EF factor structure
(model 03), and provided good overall model ﬁt (x2 = 53.80, df = 28, NF I = .982).
Standardized regression weights of age on EF 1 and EF2 were estimated to be .50 and .60,
respectively. Model 06, which constrained the loadings of age on EF to be constant
across time, produced a signiﬁcant decrease in ﬁt from model Os (x2 = 63.75, df = 29,
NFI = .979), and therefore the hypothesis that the effect of age on EF is invariant across
time was rejected. Model 05 therefore became the accepted structural model of the
longitudinal effects of age on EF. Goodness-of-ﬁt indices for models 05 and 06 are
presented in Table 6. To summarize, the effect of age on executive function increased

from the ﬁrst to the second measurement occasion.

24

DISCUSSION

The purpose of this study was to investigate the unity and stability of executive
function in an aging population, as measured by ﬁve widely-administered
neuropsychological tests. Overall, the results suggest that executive function is a uniﬁed
and stable construct. First, test-retest correlations for all ﬁve measures were in the
moderate-to high range. Second, exploratory factor analyses and structural analyses
revealed a high degree of unity and stability in the construct across time. Independent
exploratory factor analyses of these measures of executive function at two measurement
occasions yielded similar one-factor models that accounted for more than 50% of the
observed variance. These models included WCST, TMT, CFL, and CAT tasks; SCWT
was dropped from analysis after demonstrating little communality with the other
measures. The degree of convergence demonstrated between the ﬁve executive measures
in this study is remarkable, especially when the relatively heterogeneous nature of these
tasks is taken into account.

Further analyses of the short-term longitudinal stability of this model of executive
ﬁmction supported the notion that executive function is a stable cognitive construct that
can be measured reliably with commonly-used clinical measures. In fact, the only
element of the longitudinal model that was found to be signiﬁcantly invariant across the
two measurement occasions was the variance of the latent executive function variables
themselves.

Several elements separate this study from those that have been conducted
previously, and may account for some of the results. First of all, no known studies have

previously examined the longitudinal stability of the latent executive construct, as was

25

done in this study. Secondly, none of the studies surveyed used the same group of
executive measures that were found in this study (see Burgess et al., 1998; Duncan,
Johnson, Swales & Freer, 1997; Lamar, Zonderrnan, & Resnick, 2002; Della Sala, Gray,
Spinnler, and Trivelli, 1998; Boone et al., 1998; Miyake et al., 2000). It is reasonable to
expect that not all executive measures would exhibit the degree of convergence that has
been demonstrated in this study. Speciﬁcally, those measures that place substantial non-
executive demands upon participants are likely to share less commonality than more
cognitively “pure” executive measures (Miyake et al., 2000).

Although virtually every psychological measure inherently contains both
executive and non-executive task demands, an effort was made in this study to include
indices of executive function that were theoretically and statistically freer from
contamination from non-executive processes (Miyake et al., 2000). For example, this
study used a (Part B — Part A) algorithm in order to remove some of the effects of
proceSsing speed and visual search abilities. In addition, number of perseverative errors
on the WCST was used rather than one or more of the many other calculable indices of
the WCST, based upon the established relationship between executive dysfunction and
the tendency to perseverate (Stern & Prohaska, 1996). This contrasts sharply with the
factor analysis conducted by Lamar, Zonderrnan, and Resnick (2002), which analyzed
TMT Part A and Part B simultaneously, and the factor analysis conducted by Boone, et
a1. (1998), which included four indices from WCST, and resulted in the extraction of a
WCST “factor”.

The relatively high reliability of the executive measures that were used may have

also contributed to the surprising convergence demonstrated in this study. The average

26

test-retest correlation for the ﬁve executive measures used was .65. Contrary to the
notion that executive tasks can be used effectively only once (see Lowe & Rabbitt, 1998;
Rabbitt, 1997; Phillips, 1997), these results suggest that many executive tasks may be
reliable across at least two time points, even after relatively 'short delays (four to eight
weeks). Even more impressive is the correlation in the accepted longitudinal model
between the latent executive function variables at Time 1 and Time 2 of .96. The
strength of this correlation suggests that an individual’s overall picture of executive
functioning, as derived from the four measures discussed here, is remarkably robust, and
unlikely to change signiﬁcantly over a short period of time.

There were, however, several unexpected results. The most important of these is
the lack of communality demonstrated between the SCWT and other executive measures.
It is not possible at the present time to compare this result with results that have been
obtained previously, because none of the factor analyses surveyed included SCWT
interference score as a variable of interest. While no label more speciﬁc than ‘executive
function’ could be meaningfully applied collectively to the WCST, TMT, CF L, and CAT,
which did demonstrate a high degree of communality, it is nevertheless possible that
these measures all fall within some subset of executive function, and that the SCWT lies
within a separate subset. Regardless, this result is informative in that it reduces the
temptation to conclude from this study that all well-validated measures of executive
ftmction are likely to converge in all populations.

Additionally, it is surprising that the variance of the latent executive function
variable at Time 2 was signiﬁcantly greater than the variance at Time 1. It was presumed

that if the variance of executive function were to change within the span of four to eight

27

weeks, it would decrease due to the overall restriction of variance that often accompanies
practice effects. Instead, it is possible that practice increased the performance of some
subset of the sample, whereas the performance of others remained relatively unchanged.
This type of differential practice effect would cause the overall variance in executive
function to increase.

The hypothesis that the effect of age on executive function would be
longitudinally stable was also rejected. This result suggests that age could be the basis
for the potential differential in practice effects discussed above. If this were true, it
would indicate that younger individuals in this study developed tasks skills to a greater
extent than older individuals. Although it is important to note these possibilities, it is also
important to note that the standardized regression weights of age on executive function at
Time 1 (.50) and Time 2 (.60) are not dramatically different, and that the chi-square test
of signiﬁcance was based on a change in one degree of ﬁeedom. Indeed, the increase in
executive ﬁrnction variance discussed above may have been sufﬁcient to cause a
signiﬁcant increase in the relationship between executive function and age. Likewise,
although the difference in the Bender-Bonnet normed ﬁt index changed by only .003, this
difference was statistically signiﬁcant because of the high sensitivity of the chi-square
statistic.

Several aspects of the present study place limitations on the conclusions that can
be drawn from the results. Because this study was conducted on a sample of non-
demented, predominantly Caucasian older individuals with relatively high levels of
education, some caution must be exercised when generalizing these results to the aging

population at large. Similarly, the conclusions drawn in this study may not apply to other

28

populations commonly examined by this research paradigm, such as college students or
neurological patients. Finally, the results of this study are limited by the possibility that
the ﬁve measures of executive ﬁrnction employed in this study may not be representative
of the hundreds of measures used in worldwide psychological research and clinical
practice.

Further investigations of executive function with different measures, and in
different populations, would help clarify the degree to which the unity, reliability, and
stability of executive function found in this study can be generalized. Expanding the
number of measurement occasions to three or more, over longer periods of time, would
also be beneﬁcial in that it would provide a broader picture of the stability in executive
function. Finally, the inclusion of brain imaging techniques within this research
paradigm would also be helpful in clarifying the functional processes underlying both the
unity and diversity in executive function.

’ In conclusion, the results of this study suggest that executive function is a
relatively uniﬁed, stable construct. These ﬁndings challenge views of executive function
as a highly ﬁ'actionated construct (see Burgess et al., 1998; Duncan, Johnson, Swales, &
Freer, 1997; Lamar, Zonderrnan, and Resnick, 2002) that cannot be measured reliably
more than once (see Lowe & Rabbitt, 1998; Rabbitt, 1997; Phillips, 1997). However,
considering the limitations of this study, clariﬁcations of these results might be achieved
through additional research. In particular, designs that include more executive tasks and
more measurement occasions, over longer periods of time, and in a variety of other

samples, are likely to be particularly valuable.

29

APPENDIX

30

Table l

Demoggphic Characteristics of Sample

 

 

 

 

 

 

Age 54-58 59-63 64-68 69-73 74-78 79-83 84-87 88+ M. . Total
rssmg
N 17 23 25 29 33 14 7 1 2 151
% Female 70.6 69.6 68.0 51.7 54.5 71.4 14.3 100.0 0 60.3
Ede” fears 16.06 15.43 16.04 15.69 15.48 14.93 15.00 14.00 1 15.59
“canon
Table 2
Descriptive Statistics of Executive Variables at Time 1 and Time 2
a a Mean
Variable ValidN Mean Mean . b p
(Time 1) (Time 2) Difference
c 17.38 14.39
WCST ‘09 (15.27) (13.50) '2'98 < '05
d 59.87 62.99
M 1 19 (40.80) (48.33) 3'13 "S
e 38.20 39.64
CFL 12‘ (11.42) (12.01) 1'44 < '05
r 44.58 45.37
CAT 120 (10.85) (11.65) '79 "s
.429 -3.58
SCWTlg 116 (7.04) (8.07) .71 ns

 

a Standard deviations in parentheses
b

Calculated as (Time 2 — Time 1); missing data excluded pairwise
C . . . .
Wisconsrn Card Sorting Test perseverative errors
d Trail Making Test time score (Part 13 — Part A)
6
Letter Fluency total score
f
Category Fluency total score
g Stroop Test interference score

31

Table 3

Test-Retest Correlations

 

 

Variable r12
wcsrb 50
me .49
CH} .81
CATe .85
Stroopf .61

 

8 Variables autocorrelated at two time points

b Wisconsin Card Sorting Test perseverative errors
Trail Making Test time score (Part B — Part A)
Letter Fluency total score

6 Category Fluency total score

f Stroop Interference Score

0- 0

Table 4

Correlatipns and Factor Lpﬂings of Execptive Fgctigpm' g Measjrrpsa Eigenvalue > 1 Method

 

 

 

Time 1 1 2 3 4 5 Factor 1 Factor 2
1. WCST -- .27“I -.03 ~29" -.22* -.40 .16
2. TMT -- -.06 -.36** -.38*“ i -.58 .13

3. SCWT -- .10 -.02 .10 -.26
4. CAT -- .50” .74 -.18
5. CFL -- .69 .29
Eigenvalue 2.03 1 .02
% of variance 40.56 20.33
Time 2 1 2 3 4 5 Factor 1

l. WCST -- .53" -.13 -.39** -.18* -.53

2. TMT -- ~29“ -.49** ~30” -.76

3. SCWT -- .22“ .03 .31

4. CAT -- .47” .74

5. CFL -- .44

Eigenvalue 2.27

% of variance 45.44

 

32

Table 5

Correlations and Factgr Loadings of Execptive Fungtioning Measures,

 

 

 

 

One Fi_i§tor Extracted
Time 1 Time 2
1. WCST -.39 -.53
2. TMT -.58 -.76
3. SCWT .07 .31
4. CAT .73 .74
5. CFL .64 .44
Eigenvalue 2.03 2.27
% of variance 40.56 45.44
Table 6

Correlations gd Factor Loadings of Executive Functioning Measures,
One Factor Extracted and SCWT Excluded

 

 

Time 1 Time 2
1. WCST -.40 -.59
2. TMT -.59 -.75
4. CAT .71 .71
5. CFL .65 .46
Eigenvalue 2.03 2.17
% of variance 50.82 54.35

 

Table 7

Goodness-of-Fit Statistics for Alternative Models

 

 

Model X2 df p pa Comparison A12 Adf p Apa
O] 30.38 15 .01 1 .987 -- -- --- --- --

02 33.99 18 .013 .986 01 _ 02 3.61 3 us .001
03 40.93 22 .008 .983 02 _ 03 6.95 4 ns .003
04 45.80 23 .003 .981 O3 _ 04 4.87 l < .05 .002
05 53.80 28 .002 .982 -- --- --- --- ---

0 6 63.75 29 .000 .979 05 _ O 6 9.95 1 < .01 .003

 

Note. p = 151 for all analyses.
a Bender-Bonnet normed ﬁt index.

33

Table 8

Factor Loadings and Residual Variances for the Longitum 1 Factgr Model (03)

 

 

 

 

Standardized Loadings
AMOS Residual - -
Variable , a Time 1 Time 2 , ~ Rcs‘fiua‘ Res‘d‘la‘
estimates Variances Covariances Correlations

wcsr° 1.00%") .46 .51 144.76 48.04 .33
me 3.99 (.741) .59 .64 1193.40 275.76 .23
CFLf -0.98 (.200) -.52 -.58 102.30 77.45 .76
c1413 -122 (.223) -.71 -.76 57.17 41.46 .73

a Standard errors in parentheses

6

Calculated as the proportion of residual variance to total variance; 1 — (residual variance) = communality
c Wisconsin Card Sorting Test perseverative errors
d .
F rxed Parameter
e Trail Making Test time score (Part B — Part A)
f Letter Fluency total score
g Category Fluency total score

Table 9

Variance of Executive Function

for the Longitudinal Factor Model (03)

 

Variance
Time 1 39.79 (13.02)
Time 2 52.48 (16.86)

Covariance 43.74 (14.05)
Correlation .96

 

34

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

//::T~\ \ \R//, y?‘>¢<‘\f’j“\\\\
/// /// \\ \ .\\ \ \
/ [A {x .//_"j\ //A“\ / fl . .L‘Sfﬂ\ :L\\ L \S—
{x e1 /) i e2 ,1 (\ e3 .1 I e4 > (\ e5 ,1 e6 ,1 1 e7 )- 1 e8 1
l . g
WCST TMT CFL CAT WCST TMT CFL CAT
Time1 Time1 Time1 Time1 Time2 Time2 Time2 Time 2
‘\ .59 /‘ 71 / \ 54 9f , 76/
‘-.,_\.‘>46 _.52/ '- / 52 .55;
f/ \\ [l \‘r'
1 EF Time1 ) K EF Time2 ,1
\\‘~._ /,-'// \\\ ~. g ;
\‘WK/ \i-r'“
\Fxra» 5‘9: ,e- f"/ I

Figure 1. Accepted model of longitudinal EF factor structure (Model 03).
Regression weights and residual variances of executive measures have been held constant across time.

Standardized values shown.

35

.31

y,— —\ _ ,,
-_/ ‘:—:

-39

 

 

 

 

 

 

 

 

 

 

 

.50

x ’—

H\\

T. 3x;

\\
‘. I
‘. r,
._ ‘ _.

.76

 

 

 

WCST

 

Time 2

 

TMT
Time

 

 

CFL
Time 2

 

 

CAT
Time

 

‘1. 1
\‘X'55 \\68-.59/ ‘74

 

\
‘. __ .'I
.l‘ \\

/

x
\
\\

 

1
II
I

/
CEO/d EF Time2 J
Age ’ ,

< EF Time1 \j*~~\\\

‘\

 

 

 

\ . .
\\\ ,4”-
\\£/

Ifrr" \\~
/' \\

r

\ I] \

Figure 2. Accepted model of relationship of age to longitudinal EF factor structure (Model 05).
Regression weights and residual variances of executive measures have been held constant across time.
Standardized values shown.

36

References

Anderson, S. W., Damasio, H., Jones, R. D., Tranel, D. (1991). Wisconsin Card
Sorting Test performance as a measure of frontal lobe damage. Joumal of Clinical &
Experimental Neuropsychology, 13(6), 909-922.

Arbuthnott, K., & Frank, J. (2000). Trail Making Test, part B as a measure of
executive control: Validation using a set-switching paradigm. Journal of Clinical and

Exm’mental Neuropsychology, 22(4), 518-528.

Axelrod, B. N., Goldman, R. S., & Henry, R. (1992). Sensitivity of the Mini-
Mental State Examination to frontal lobe dysfunction in normal aging. Journal of
Clinical Psychology, 48(1), 68-71.

Band, G. P., Ridderinkhof, K. R., & Segalowitz, S. (2002). Explaining
neurocognitive aging: is one factor enough? Brain and Cogpition, 49, 259-267.

Barcelo, F. (2001). Does the Wisconsin Card Sorting Test measure prefrontal
function? The Spanish Journal of Psychology. 4(1), 79-100.

Beck, A. T., Steer, R. A., Garbin, M. G. (1988). Psychometric properties of the
Beck Depression Inventory: Twenty-ﬁve years of evaluation. Clinical Psychology
Review 8(1), 77-100.

 

Beck, A.T., Ward, C. H., Mendelson, M., Mock, J ., & Erbaugh, J. (1961). An
inventory for measuring depression. Archives of General Psychiaﬂ, 4, 561-571.

Benton, A. L., Hamsher, K., & Sivan, A. B. (1983). Multilingaal aphasia
examination (3rd ed.). Iowa City, IA: AJA Associates.

Berman, K. F., Ostrem, J. L., Randolph, C., Gold, J ., Goldberg, T. E., Coppola,
R., Carson, R. E., Herscovitch, P., & Weinberger, D. R. (1995). Physiological activation
of a cortical network during performance of the Wisconsin Card Sorting Test: A positron
emission tomography study. Neuropsychologia, 33(8), 1027-1046.

Bilger, E. D. (1997). Neuroirnaging in normal aging and dementia. In
Nussbaum, P. D. (Ed.), Handbook of Neuropsychology and Aging (pp. 409-421). New
York: Plenum Press.

Boone, K. B., Ponton, M. O., Gorsuch, R. L., Gonzalez, J. J., & Miller, B. L.
(1998). Factor analysis of four measures of prefrontal lobe functioning. Archives of
Clinical Neuropsychology, 13(7), 585-595.

37

Bryan, J., & Luszcz, M. A. (2000). Measurement of executive function:
considerations for detecting adult age differences. Journal of Clinical and Experimental

Neuropsychology, 22(1), 40-55.

Burgess, P. W., Alderman, N., Evans, J., Emslie, H., & Wilson, B. A. (1998).
The ecological validity of tests of executive function. Journal of the International

Neuropsycholog'cal Socieg, 4, 547-558.

Cabeza, R. (2001). Cognitive neuroscience of aging: contributions of functional
neuroimaging. Scandinavian Journal of Psychology, 42, 277-286.

Celine, 8., Michel, 1., & Espagnet, L. (2000). Aging, episodic memory F eeling-
of-Knowing, and frontal functioning. Neuropsychology, 14(2), 299-309.

Craik, F., Morris, L. W., Morris, R. G., & Loewen, E. R. (1990). Relations
between source amnesia and frontal lobe functioning in older adults. Psychology and

Aging, 5(1), 148-151.

Crowe, S. F. (1998). The differential contribution of mental tracking, cognitive
ﬂexibility, visual search, and motor speed to performance on parts A and B of the Trail
Making Test. Journal of Clinical Psychology, 54(5), 585-591 .

Della Sala, 8., Gray, C., Spinnler, H., & Trivelli, C. (1998). Frontal lobe
functioning in man: the riddle revisited. Archives of Clirgal Neuropaychology, 13(8),
663-682.

Duncan, J ., Emslie, H., & Williams, P. (1996). Intelligence and the frontal lobe:
the organization of goal directed behavior. Cogr_litive Psychology, 30, 257-303.

Duncan, J., Johnson, R., Swales, M., & Freer, C. (1997). Frontal lobe deﬁcits
after head injury: unity and diversity of function. Cogm'tive Neuropsychology, 14(5),
7 13-741 .

Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. ( 1999).
Evaluating the use of exploratory factor analysis in psychological research.

Psychological Methods, 4(3), 272-299.

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(3), 189-198.

Foster, J. R., Sclan, S., Welkowitz, J., Boksay, 1., & Seeland, I. (1988).
Psychiatric assessment in medical long-term care facilities: reliability of commonly used
rating scales. International Joumal of Geriatric Psychiagy, 3(3), 229-233.

38

Frith, C. D., Friston, K. J ., Liddle, P. F., Frackowiak, R. S. (1991). A PET study
of word ﬁnding. Neuropsychologia, 29(12), 1137-1148.

Fuster, J. M. (1997). The Prefrontal Cortex (3rd ed.). Philadelphia, PA:
Lippincott-Raven Publishers.

Garden, S. E., Phillips, L. H., & MacPherson, S. E. (2001). Midlife aging, open-
ended planning, and laboratory measures of executive function. Neuropsychology, 15(4),
472-482.

Giaquinto, S. (1988). Aging and the Nervous System. New York: John Wiley
and Sons.

Goldberg, E. (2001). The Executive Brag. New York: Oxford University Press.

Gourovitch, M. L., Kirkby, B. S., Goldberg, T. E., Weinberger, D. R., Gold, J. M.,
Esposito, G., Van Horn, J. D., & Berman, K. F. (2000). A comparison of rCBF patterns
during letter and semantic ﬂuency. Neuropsychology, 14(3), 353-360.

Gur, R. C., Gur, R. E., Orbist, W. D., Skolnick, B. E., & Reivich, M. (1987). Age
and regional cerebral blood ﬂow at rest and during cognitive activity. Archives of
General Psychiaga 44, 617-621.

Haug, H., & Eggers, R. (1991). Morphometry of the human cortex cerebri and
corpus striatum during aging. Neurobiology of aging, 12(4), 336-338.

' Heaton, R. K., Chelune, G. J ., Talley, J. L., Kay, G. G., & Curtiss, G. (1993).
Wisconsin Card Sorting Test Manual. USA: Publishing Assessment Resources.

Hertzog, C., & Schaie, K. W. (1986). Stability and change in adult intelligence:
1. Analysis of longitudinal covariance structures. Psychology and Aging, 1(2), 159-171.

Konishi, S., Kawazu, M., Uchida, I., Kikyo, H., Asakura, I., & Miyashita, Y.
(1999). Contribution of working memory to transient activation in human inferior
prefrontal cortex during performance of the Wisconsin Card Sorting Test. Cerebral
Cortex 9(7), 745-753.

 

Lamar, M., Zonderrnan, A. B., & Resnick, S. (2002). Contribution of speciﬁc
cognitive processes to executive functioning in an aging population. Neuropsychology,
16(2), 165-162.

LaRue, A. (1992). Aging and Neuropsychological Assessment. New York:
Plenum Press.

Lezak, M. (1995). Neuropsychological Assessment (3“1 ed.). New York: Oxford
University Press.

39

Lombardi, W. J., Andreason, P. J ., Sirocco, K. Y., Rio, D. E., Gross, R. E.,
Umhau, J. C., & Hemmer, D. W. (1999). Wisconsin Card Sorting Test performance
following head injury: dorsolateral fronto-striatal circuit activity predicts perseveration.
Journal of Clinical and Experimental Neuropaychology. 21(1), 2-16.

Lowe, C., & Rabbitt, P. (1998). Test/retest reliability of the CANTAB and
ISPOCD neuropsychological batteries: theoretical and practical issues.
Neuropsychologia, 36(9), 915-923.

Milham, M. P., Erickson, K. 1., Ballich, M. T., Kramer, A. F ., Webb, A., Wszalek,
T., & Cohen, N. J. (2002). Attentional control in the aging brain: insights from an fMRI
study of the Stroop task. Brain and Cogaition, 49, 277-296.

Mitrushina, M., & Satz, P. (1991). Effect of repeated administration of a
neuropsychological battery in the elderly. Journal of Clinical Psychology,47(6), 790-801.

Mitrushina, M., Uchiyama, C., & Satz, P. (1995). Heterogeneity of cognitive
proﬁles in normal aging: Implications for early manifestations of Alzheimer’s Disease.
Journal of Clinical and Exmp’mental Neuropsychology, 17(3), 374-3 82.

Miyake, A., Friedman, N., Emerson, M. J ., Witzki, A. H., Howerter, A., &
Wager, T. D. (2000). The unity and diversity of executive functions and their
contributions to complex “frontal lobe” tasks: a latent variable analysis. Cogpitive

Psychology, 41, 49-100.

Phillips, L. H. (1997). Do “frontal tests” measure executive function? Issues of
assessment and evidence from ﬂuency tests. In P. Rabbit (Ed.), Methodology of Frontal
and Executive Function (pp. 191-213). Psychology Press: East Sussex, UK.

Rabbitt, P. (1997). Methodology of Frontal and Executive Function. Psychology
Press: East Sussex, UK.

Raz, N., Gunning-Dixon F. M., Head, D., Dupuis, J. H., & Acker, J. D. (1998).
Neuroanatornical correlates of cognitive aging: evidence from structural magnetic
resonance imaging. Neuropsychology, 12(1), 95-114.

Reitan, R. M. (1958). Validity of the Trail Making Test as an indicator of organic
brain damage. Perceptual & Motor SkillsJ. 271-276.

Rezai, K., Andreasen, N. C., Alliger, R., Cohen, G., Swayze, V., & O’Leary, D.
S. (1993). The neuropsychology of the prefrontal cortex. Archives of Neurology, 50(6),
636-642.

Robbins, T. W., James, M., Owen, A. M., Sahakian, B. J ., Lawrence, A. D.,
McInnes, L., & Rabbitt, P. M. (1998). A study of performance on tests from the

40

CANTAB battery sensitive to frontal lobe dysfunction. Journal of the International
Neuropsychological Society, 4, 474-490.

Rogers, R. D., Andrews, T. C., Grasby, P. M., Brooks, D. J., & Robbins, T. W.
(2000). Contrasting cortical and subcortical activations produced by attentional-set
shifting and reversal learning in humans. Journal of Cogm'tive Neuroscience, 12(1), 142-
162.

Rympa, B., Prabhakaren, V., Desmond, J. E., & Gabrieli, J. D. (2001). Age
differences in prefrontal cortical activity in working memory. Psychology and Aging,
EB), 371-384.

Salthouse, T. A. (1991). Theoretical Perspectives on Cognitive Ag'mg, Hillsdale,
New Jersey: Lawrence Erlbaum Associates.

Salthouse, T. A., Fristoe, N., & Rhee, S. H. (1996). How localized are age-related
effects on neuropsychological measures? Neuropsychology, 10(2), 272-285.

Schretlen, D., Pearlson, G. D., Anthony, J. C., Aylward, E. H., Augustine, A. M.,
Davis, A., & Barta, P. (2000). Elucidating the contributions of processing speed,
executive ability, and ﬁontal lobe volume to normal age-related differences in ﬂuid
intelligence. Journal of the International Neuropsychological Society, 6, 52-61.

Seeman, T. E., Lusignolo, T. M., Albert, M., & Berkrnan, L. (2001). Social
relationships, social support, and patterns of cognitive aging in healthy, hi gh-functioning
older adults: MacArthur Studies of Successful Aging. Health Psychology, 20(4), 243-
255.

Shaw, T. G., Mortel, K. F., Meyer, J. S., Rogers, R. L., Hardenburg, J ., & Cutaia,
M. M. (1984). Cerebral blood ﬂow changes in benign aging and cerebrovascular disease.
Neurology, 34, 855-862.

Silver, M. H., & Perls, T. T. (2001). Is dementia the price of a long life? An
optimistic report from centenarians. Journal of Aging and Health, 13(2), 267-286.

Silverrnan, P., Hecht, L., & McMillin, D. (2000). Modeling life satisfaction
among the aged: A comparison of Chinese and Americans. Journal of Cross-Orltural

Gerontology, 15, 289-305.

Spreen, O., & Strauss, E. (1998). A compendium of neuropsychological tests :
administration, norms, and commentary. New York : Oxford University Press.

Stem, R. A., & Prohaska, M. L. (1996). Neuropsychological evaluation of
executive functioning. American Psvahiatlic Press Review of Psychiag, 15, 243 -266.

41

Stroop, J. R. (193 5). Studies of interference in serial verbal reactions. Journal of
Experimental Psychology 18. 643-662.

Stuss, D. T., Floden, D., Alexander, M. R, Levine, B., & Katz, D. (2001). Stroop
performance in focal lesion patients: dissociation of processes and frontal lobe lesion
location. Neuropsychologia, 39, 771-786.

Vendrell, P., Junque, C., Pujol, J ., Jurado, M. A., Molet, J., & Grafrnan, J. (1995).
The role of prefrontal lesions in the Stroop task. Neuropsychologia, 33(3), 341-352.

Wang, L., Kakigi, R., & Hoshiyama, M. (2001). Neural activities during
Wisconsin Card Sorting Test — MEG observation. gagnitive Brain Research. 12. 19-31.

Wecker, S. W., Kramer, J .H., Wisniewski, A., Delis, D. C., & Kaplan, E. (2000).
Age effects on executive ability. Neuropsychology, 14(3), 409-414.

West, R. L. (1996). An application of prefrontal cortex function theory to
cognitive aging. Psychological Bulletin, 120(2), 272-292.

West, R. L., & Bell, M. A. (1997). Stroop color--word interference and
electroencephalogram activation: Evidence for age-related decline of the anterior
attention system. Neuropsychology, 11(3), 421-427.

Winocur, G., Moscovitch, M., Stuss, D. T. (1996). Explicit and implicit memory
in the elderly: Evidence for double dissociation involving medial temporal- and frontal-
lobe functions. NeurOpsychology 10(1), 57-65.

Yesavage, J. A., Brink, T.L., Rose, T.L., Lum 0., Huang, V., Adey, M., & Leirer,
V.O. (1983). Development and validation of a Geriatric Depression Screening Scale: A
preliminary report. Journal of Psychiatric Research 17(1), 37-49.

 

Ylikoski, R., Ylikoski, A., Keskivaara, P, Tilvis, R., Slkava, R., & Erkinjuntti, T.
(1999). Heterogeneity of cognitive proﬁles in aging: successful aging, normal aging, and
individuals at risk for cognitive decline. European Journal of Neurology, 6, 645-652.

42

 

   

rillliiiiiijii111111:1