Neuropsychology
1996, Vol. I0, No. 2, 272-285
Copyright 1996 by the American Psychological Association, Inc.
0894-4105/96/$3.00
How Localized Are Age-Related Effects
on Neuropsychological Measures?
Timothy A. Salthouse and Nathanael Fristoe
Soo Hyun Rhee
Georgia Institute of Technology
Emory University
A battery of neuropsychologicaltests (e.g., Wisconsin Card Sorting Test, Trail Making Test, Rey
Auditory-Verbal Learning Test, Wechsler Adult Intelligence Scale--Revised Block Design,
Object Assembly, and Digit Symbol tests) was administered to 259 adults ages 18 to 94 who
reported themselves to be in good to excellent health. Moderate age-related declines were
apparent in performance measures that could be postulated to be sensitive to damage in frontal,
parietal, and temporal lobes. However, correlation-based analyses revealed that the age-related
influences on the different measures were not independent. Across all variables examined, an
average of about 58% of the age-related variance in a given variable was shared with that in other
variables. These results indicate that only a portion of the age-related influences on many
commonlyused neuropsychologicalmeasures is specificand potentially localized.
The primary goal of the current study was to examine the
degree of independence of the age-related influences on
measures sometimes postulated to be sensitive to functioning
in different regions of the brain. It is often assumed that levels
of performance on different cognitive measures and the
age-related effects on those measures are determined by
separate and distinct mechanisms. In the case of measures
from neuropsychological tests, the discovery of selective impairments by individuals with damage in particular regions of the
brain has led to the inference that various brain structures are
specialized for different types of processing. When combined
with results indicating age-related deficits on those measures,
these findings have led to speculation that certain regions of
the brain are more sensitive to age-related decline than others.
For example, at different times it has been suggested that the
right hemisphere ages more rapidly than the left (e.g., Goldstein & Shelly, 1981; Klisz, 1978; Schaie & Schaie, 1977) or that
frontal or prefrontal regions are the first to deteriorate with
advancing age (e.g., Daigneault, Braun, & Whitaker, 1992;
Mittenberg, Seidenberg, O'Leary, & Di Giulio, 1989; Whelihan & Lesher, 1985).
Although it is clearly informative to consider different types
of influences on the same variables, questions can be raised
about the practice, which can be termed first-order functional
localization, of relying on performance on a test previously
found to be impaired with lesions in specific regions as a
marker for the level of functioning in that region in neurologi-
cally intact individuals (e.g., for discussion of this issue see
Kertesz, 1994b; Reitan & Wolfson, 1994; Sergent, 1988). One
problem with this reasoning is that although the measures may
be sensitive to damage in particular regions, the measures are
not necessarily specific because damage to other regions could
lead to similar patterns of impairment. Another major problem with this approach is that there are many determinants of
test performance, of which effectiveness of functioning in a
particular brain region is only one. An even more dubious
practice is what can be termed second-order functional localization, in which researchers have inferred that a new measure
reflects functioning in a specific structure on the basis of
correlations between the new measure and one or more of the
first-order markers (e.g., Craik, Morris, Morris, & Loewen,
1990; Parkin & Walter, 1991). Second-order relations seem
especially questionable when the first-order relations are not
yet convincingly established and when other sources of the
correlations are not considered.
The latter problem is particularly relevant in age-comparative studies because recent research with psychometric and
experimental cognitive tests has revealed that there is considerable commonality, or shared variance, in the age-related
effects on different cognitive measures. In other words, because many measures have been found to share large proportions of their age-related variance (e.g., Salthouse, 1992a,
1993b, 1994a, 1994b), the age-related influences on those
measures are not independent. We propose that a more
productive approach than assuming that all age-related influences on a given measure are specific is to assume that both
common (or shared) and specific (or unique) age-related
influences often operate simultaneously (Salthouse, 1992a). If
this assumption is valid, then both types of influences need to
be considered because the contribution of any specific influences that might exist cannot be accurately determined unless
the general influences are first controlled.
Two analytical methods are used in the current study. One
method relies o(! structural equation modeling (i.e., LISREL)
to investigate the assumption that both common and specific
age-related influences contribute to performance on the ob-
Timothy A. Salthouse and Nathanael Fristoe, School of Psychology,
Georgia Institute of Technology; Soo Hyun Rhee, Department of
Psychology, Emory University.
This research was supported by National Institute on Aging Grant
R37 6826. We thank Alice Hardee, Robin Harrison, Doug Robbins,
and Troy Surdick for assistance in testing research participants, and
Ulman Lindenberger and Naftali Raz for helpful comments on an
earlier version of this article.
Correspondence concerning this article should be addressed to
Timothy A. Salthouse, School of Psychology, Georgia Institute of
Technology,Atlanta, Georgia 30332-0170. Electronic mail may be sent
to tim.salthouse@psych.gatech.edu.
272
LOCALIZATION OF AGE EFFECTS
served variables. Of particular interest with this analytical
procedure are the relative strengths of the linkages from age to
the specific factors and from age to the hypothesized common
factor. In order to obtain estimates of the specific factors that
were distinct from the observed variables, we defined each
specific factor in terms of two or more observed variables.
It is also possible to derive estimates of the contribution of
the hypothesized common influence by determining the proportion of age-related variance in a variable that is shared with
other variables. Figure 1 portrays a Venn diagram of the total
variance (circles) and shared variance (overlapping regions) in
two variables. The areas of greatest interest in the present
context are regions b and d because together they represent
the age-related variance in the criterion variable. However, the
figure illustrates that the age-related variance can be partitioned into that unique to the criterion variable, represented
by region d, and that common to the controlled variable
represented by region b. Because this latter region corresponds to shared variance, it can serve as an estimate of the
potentially common or general influences associated with age.
It is only a potential estimate because it represents variance
shared between a single pair of variables, but different combinations of variables can be examined to evaluate generalizability of the inferred common influence.
The relevant proportions of variance can be estimated with
hierarchical regression analyses in which the initial age-related
variance in the criterion variable (b + d) is compared with the
increment in variance associated with age after control of the
controlled variable (d). Subtracting the latter from the former
and dividing by the former yields a ratio of b/(b + d). If the
age-related effects on different variables are largely independent of one another, then the ratiO will be small because most
of the age-related variance in the variable is unique (i.e., d is
large). However, if b is large relative to d, then the ratio will
also be large, and one could infer that a considerable proportion of the age-related variance in the two variables is shared.
It is important to note that this analytical procedure will
yield substantial estimates of shared age-related variance only
if the controlled variables are correlated both with age and
with the criterion variable. If a variable (e.g., grayness of hair)
Controlled
Criterion
Age
Figure 1. Regions of overlapping variance among age, a controlled
variable, and a criterion variable.
273
is primarily correlated only with age then it will have a
near-zero value of b, and there will be little or no age-related
variance shared with the criterion. Conversely, if the variable
(e.g., quantity of domain-relevant knowledge) is primarily
correlated only with the criterion measure and not with age,
then it will have a large a region and a near-zero b region.
The measures examined in the current study were selected
from frequently used neuropsychological tests (cf. Butler,
Retzlaff, & Vanderploeg, 1991). The primary selection criteria
were that there was evidence of age-related declines on the
measures and that there was some reason to believe that the
measures might be sensitive (although not necessarily specific)
to functioning in different neuroanatomical regions. The first
criterion is important to ensure the investigation of age-related
influences, and the second criterion is included to maximize
the possibility that different anatomical regions might be
contributing to the age-related influences on the various
measures.
Several rather gross localization assumptions can be identified in the contemporary literature. For example, planning,
integration, abstraction, and executive functions are often
thought to involve the frontal lobe, visuo-spatial construction
and manipulation processes are frequently postulated to involve the right parietal lobe, and memory for verbal information has been hypothesized to involve the left medial temporal
lobe or hippocampus complex (e.g., Kertesz, 1994a; Kolb &
Whishaw, 1990; McCarthy & Warrington, 1990; Squire, 1987;
Stuss & Benson, 1984; Walsh, 1994). These basic assumptions,
supplemented by information about individual tests when it
was available, guided the selection of specific tests.
Four tests were selected to assess functions postulated to
involve the frontal lobe. One was the Wisconsin Card Sorting
Test (WCST; Heaton, Chelune, Talley, Kay, & Curtiss, 1993),
which is a concept identification test in which the identity of
the concept changes throughout the course of the test. Several
researchers have reported that patients with damage to the
frontal lobe perform more poorly on this test than neurologically intact controls or than patients with brain damage in
other regions (e.g., Drewe, 1974; Janowsky, Shimamura,
Kritchevsky, & Squire, 1989; Milner, 1963; Nelson, 1976;
Robinson, Heaton, Lehman, & Stilson, 1980; see Anderson,
Damasio, Jones, & Tranel, 1991; Reitan & Wolfson, 1994, for
arguments against frontal lobe specificity of this test). There
are also numerous reports of age-related differences favoring
young adults on measures of performance on this test (e.g.,
Anderson et al., 1991; Axelrod & Henry, 1992; Crockett,
Bilsker, Hurwitz, & Kozak, 1986; Daigneault, Braun, &
Whitaker, 1992; Heaton et al., 1993; Libon et al., 1994; Nelson,
1976; Parkin & Walter, 1991).
A second test hypothesized to be sensitive to frontal lobe
functioning because of its sequence planning requirements is
the Trail Making Test (Reitan, 1992). This test requires
research participants to draw lines connecting a series of
designated targets as rapidly as possible. No studies could be
located in which trail-making performance was impaired by
damage restricted to the frontal lobes, but a large number of
studies have reported negative relations between age and
performance on this test (e.g., desRosiers & Kavanaugh, 1987;
Elias, Robbins, Walter, & Schultz, 1993; Heaton, Grant, &
274
SALTHOUSE, FRISTOE, AND RHEE
Matthews, 1986; Lyness, Eaton, & Schneider, 1994; Moehle &
Long, 1989; Nielsen, Knudsen, & Daugbjerg, 1989; Salthouse
& Fristoe, 1995; Seines et al., 1991; Stuss, Stethem, & Poirier,
1987; Van Gorp, Satz, & Mitrushima, 1990; Wiederholt et al.,
1993).
The Shipley Abstraction Test (Zachary, 1986), consisting of
series completion problems with different types of stimulus
material, was also expected to be sensitive to frontal lobe
functioning because of the abstraction requirements of the
test. No reports of selective impairments among patients with
frontal lobe damage could be located, but there are many
reports of poorer performance with increased age (e.g., Hooper,
Hooper, & Colbert, 1984; Johnson, 1993; Salthouse, 1991;
Salthouse & Mitchell, 1990; Shelton, Parsons, & Lever, 1982).
Frontal lobe damage has also been implicated as a factor in
performance on assorted verbal fluency or controlled word
association tasks (e.g., Benton, 1968; Bolter, Long, & Wagner,
1983; Butler, Rorsman, Hill, & Tuma, 1993; Crockett et al.,
1986; Janowsky et al., 1989; Kolb & Whishaw, 1990; Miller,
1984; Milner, 1964; Pendleton, Heaton, Lehman, & Hulihan,
1982; Perret, 1974; see Reitan & Wolfson, 1994, for a critique
of this research). Age relations on verbal fluency tests have
been mixed, with some reports of negative relations (e.g.,
Ardila & Rosselli, 1989; Pendleton et al., 1982; Whelihan &
Lesher, 1985) and other reports of no significant age relations
(e.g., Axelrod & Henry, 1992; Crockett et al., 1986; Daigenault
et al., 1992; Mittenberg, Seidenberg, O'Leary, & DiGiulio,
1989; Seines et al., 1991). Because there is evidence that
vocabulary knowledge moderates the age relations on fluency
tests (Salthouse, 1993a), the Shipley Vocabulary test (Zachary,
1986) was also administered to all participants in this study.
The two visual-spatial tests that might be expected to be
sensitive to functioning of the right parietal lobe are the Object
Assembly and Block Design tests from the Wechsler Adult
Intelligence Scale--Revised (WAIS-R; Wechsler, 1981). Only
two studies could be located reporting impairment of Block
Design performance with damage to the right parietal lobe
(McFie, 1960; Warrington, James, & Maciejewski, 1986), and
no studies were found implicating this structure in Object
Assembly performance. However, robust age relations have
been reported with both tests in large representative samples
(e.g., Birren & Morrison, 1961; Heaton, Grant, & Matthews,
1986; Kaufman, Reynolds, & McLean, 1989).
The two verbal memory tests hypothesized to be sensitive to
damage in the medial temporal lobe were the Rey AuditoryVerbal Learning Test (RAVLT; see Spreen & Strauss, 1991),
and a locally developed paired associate test. Moderate-tolarge negative age relations have been reported on the RAVLT
(e.g., Bolla-Wilson & Bleecker, 1986; Geffen, Moar, O'Hanlon,
Clark, & Geffen, 1990; Hartman & Hasher, 1991; Mitrushina,
Satz, Chervinsky, & D'Elia, 1991; Nielson et al., 1989; Query &
Megran, 1983; Seines et al., 1991) and on paired associate tests
similar to that used here (e.g., Salthouse, 1993b; Salthouse,
Kausler, & Saults, 1988). No studies were found in which
performance on these particular tests was impaired with
damage to specific brain regions, but performance on similar
tests has been found to be impaired with damage to the left
temporal lobe (e.g., Blakemore & Falconer, 1967; Weingartner, 1968).
Finally, because prior research has revealed that measures
of perceptual comparison speed share considerable proportions of age-related variance with a wide range of cognitive
measures (e.g., Salthouse, 1991; 1992a; 1993b; 1994a; 1994b),
several tests of processing speed were also administered. One
test was the Digit Symbol Substitution Test from the W A I S - R
(Wechsler, 1981). The other tests were locally developed, with
two hypothesized to assess sensory-motor speed (Digit Copying and Symbol Copying), and two hypothesized to involve
sensory-motor and perceptual comparison speed (Letter Comparison and Pattern Comparison). Although these tests are
similar in several respects, they have been found to form
distinct factors and to have different patterns of relations with
other variables (e.g., Earles & Salthouse, 1995; Salthouse,
1993b, 1994b).
In summary, a battery of tests that might be expected to be
sensitive to functioning in different regions of the brain was
administered to a moderately large sample of healthy adults
from a wide age range. The principal goal was to investigate
the degree to which the age-related influences on the different
measures were independent. If the measures reflect functioning in anatomically distinct regions or modules of the brain,
then the age-related effects on the measures would be predicted to be largely independent. In contrast, if substantial
proportions of the age-related variance in the measures were
found to be shared with other measures, then the measures
would lack specificity with respect to age-related influences,
and common or general age-related mechanisms would presumably be implicated.
Method
Participants
Participants in the project consisted of 259 adults between 18 and 94
years of age, with a mean age of 51.4 years, and a standard deviation of
18.4. Demographic and (self-reported) health variables by decade are
summarized in Table 1. The average education rating for the entire
sample was 3.82 (SD = 1.0), which, because a rating of 3 represented
some college and a rating of 4 represented graduation from college,
corresponds to approximately 3 years of college. Only 3 participants,
ages 60, 61, and 90, had less than a high school education. The average
health rating for the entire sample was in the very good range (i.e.,
2.2), and 93% of the participants rated their health as good, very good,
or excellent.
Tests
The tests were administered in the same order to all research
participants. This order was Letter Comparison, Pattern Comparison,
Digit Copying, Symbol Copying, Paired Associates Memory, Digit
Symbol, Block Design, Object Assembly, WCST, Trail Making Test,
RAVLT, Shipley Vocabulary and Abstraction, Delayed Portion of the
RAVLT, and Controlled Word Association Test (verbal fluency).
The four speed tests (Letter Comparison, Pattern Comparison,
Digit Copying, and Symbol Copying), as well as two additional tests not
described here, were each administered twice with 30-s time limits.
The materials for the two comparison tests consisted of a page
containing pairs of three, six, or nine letters (Letter Comparison), or
line segments arranged in a pattern (Pattern Comparison), with a line
between the two members of the pair. The research participant was
275
LOCALIZATION OF AGE EFFECTS
Table 1
Demographic Characteristics o f Research Participants According to Age Group
Age group
Characteristic
Number
W o m e n (%)
20s
30s
40s
50s
60s
70-94
41
59
40
65
30
60
58
66
43
58
47
70
24.6
2.6
34.5
2.9
43.7
2.7
54.3
2.7
64.4
3.0
78.4
5.7
3.9
0.9
3.8
1.0
4.1
0.9
3.9
0.9
3.7
1.2
3.5
1.1
2.0
0.7
2.1
0.7
2.2
0.8
2.1
0.7
2.1
0.8
2.4
0.9
2.0
0.9
2.0
0.8
2.2
0.9
2.1
0.9
2.2
0.9
2.5
1.0
1.3
0.7
1.5
0.8
1.8
1.1
1.4
0.7
1.6
0.7
2.2
1.I
0.00
0.00
0.00
0.00
0.03
0.18
0.07
0.26
0.14
0.35
0.21
0.41
0.00
0.00
0.08
0.27
0.13
0.35
0.14
0.35
0.33
0.47
0.40
0.50
0.00
0.00
0.00
0.00
0.13
0.35
0.03
0.18
0.02
0.15
0.04
0.20
0.02
0.16
0.05
0.22
0.13
0.35
0.03
0.18
0.02
0.15
0.02
0.15
Age
M
SD
Education a
M
SD
Health satisfaction b
M
SD
Health rating b
M
SD
Health-related activity limitations b
M
SD
Cardiovascular s u r g e r f
M
SD
Hypertension medications c
M
SD
Head injuryc
M
SD
Neurological treatment c
M
SD
Note.
n = 259.
aCategorized on a 5-point scale where 1 = less than 12years, 2 = high school graduate, 3 = 13-15years, 4 =
college graduate, and 5 = more than 16 years, bRatings on a 5-point scale where lower n u m b e r s indicate
better health. CResponses were yes/no, and thus m e a n s correspond to the proportion of individuals
reporting a positive response.
instructed to decide whether the two m e m b e r s of the pair were
identical and to write an S on the line between them if they were the
same or to write a D on the line between t h e m if they were not. To
adjust for guessing, the m e a s u r e of performance in these tests was the
n u m b e r of items answered correctly minus the n u m b e r of items
answered incorrectly.
The materials for the two copying tests consisted of a page
containing pairs of boxes with a digit (Digit Copying), or a simple line
pattern (Symbol Copying) in the top box and nothing in the bottom
box. The task for the research participant was to copy the item from
the top box into the bottom box and to complete as many items as
possible within the allotted time. Performance was assessed in terms of
the n u m b e r of items completed within 30 s.
Two trials with six word pairs each were presented in the paired
associates memory test. The pairs for List 1 were sign-baby, hill-ring,
race-girl, home-suit, snow-case, and frog-neck. Pairs for List 2 were
foot--tree, bank-milk, room-face, mile-spot, yard-body, and coal-year.
Each word pair was read by the examiner at a rate of approximately
one word per second, with a longer pause between words from
different pairs. Response forms contained the first words in the pair,
with participants instructed to write the response word associated with
each stimulus word. O n e minute was allowed for recall with each list.
Performance was assessed in terms of the n u m b e r of response words
correctly recalled in each list.
T h e three subtests from the W A I S - R (Digit Symbol, Block Design,
and Object Assembly) were administered according to the instructions
in the manual (Wechsler, 1981). Very briefly, in the Digit Symbol test
90 s was allowed for the research participant to write the symbols
associated with digits according to a code table, in the Block Design
test the speed and accuracy of reproducing designs with colored blocks
was determined, and in the Object Assembly test participants were
evaluated with respect to the speed and accuracy with which they could
assemble simple jigsaw puzzles. Except where noted, performance in
each test was represented by the raw scores and not the scaled
(age-adjusted) scores.
The W C S T was administered according to the instructions in the
manual (Heaton et al., 1993). This test requires participants to classify
cards varying in the color, form, and n u m b e r of geometric patterns into
categories according to the feedback provided by the examiner. After
every 10 correct classifications for a given dimension, the relevant
dimension (i.e., color, form, and number) was changed. A total of 10
different m e a s u r e s of performance were examined in this test. Because
the scoring of perseveration errors is complex, the scoring reliability
was checked by having a second evaluator independently score 40
protocols. There was complete agreement on 85% of the protocols,
and the correlation between the two sets of perseveration error scores
was .995. These values are similar to reports in the Heaton et al. (1993)
manual and indicate that the measures were assessed with a high
degree of consistency.
T h e Trail Making Test was also administered according to the
published instructions (Reitan, 1992). In this test, the examinee is
instructed to draw a line connecting the circled targets in sequence as
rapidly as possible. Both Trails A consisting of a series of 25 numbered
targets and Trails B consisting of an alternating sequence of 25 n u m b e r
276
SALTHOUSE, FRISTOE, AND RHEE
and letter targets were administered. Because errors were infrequent,
the measures of performance were the total time required to draw
lines connecting all 25 targets in each version.
The RAVLT was administered with the materials and procedures
described in Spreen and Strauss (1991). This test involves multipletrial free recall with an interference list, a delay interval, and a
recognition test. Five successive listen-recall trials were administered
with the same 15 words in the same order, followed by a new list of 15
words, and then an attempt to recall the words from the original list
without another presentation of the words. After a delay interval of
approximately 20 min, occupied by the Shipley Vocabulary and
Abstraction tests, the participant was again asked to recall the words
from the original list without another presentation of the items.
Finally, a recognition test consisting of the 15 target words from List 1,
the 15 target words from List 2, and 20 new words was administered.
Measures of performance in this test consisted of the number of
correctly recalled items in each trial, and the number of target (List 1)
and distractor (List 2) recognitions in the recognition test.
The Shipley Vocabulary Test (Zachary, 1986) consists of 40 multiplechoice vocabulary items of progressively increasing difficulty. The
Shipley Abstraction test consists of 20 series completion items with
content ranging from digits to letters to words. We allowed 10 min for
the participant to complete as many items as possible in each test.
Scores in these tests were computed according to procedures in the
manual (Zachary, 1986), with an adjustment for the number of
unattempted items in the Vocabulary test and multiplication of the
number of correct items by two in the Abstraction test.
The Controlled Word Association (Fluency) Test required research
participants to write as many words as possible beginning with the
letters F, A, and S. We allowed 60 s for each letter but the participants
were instructed to draw a line on the paper after 30 s. Therefore, two
scores were obtained in each version of the test; the number of
different words produced in 30 s and the number produced in 60 s.
Most of the tests are from published sources, and reliability
estimates for some of the measures can be obtained from the manuals.
For example, internal consistency estimates of reliability for the
Shipley (Zachary, 1986) measures were .87 for the Vocabulary
measure and .89 for the Abstraction measure. Test-retest reliabilities
for the WAIS-R (Wechsler, 1981) tests were .82 for Digit Symbol, .87
for Block Design, and .68 for Object Assembly. Test-retest or
alternate-form reliabilities from age-heterogeneous samples for other
tests were .67 to .88 for the Trail Making Tests (Matarazzo, Matarazzo, Wiens, Gallo, & Klonoff, 1976), and .60 to .77 for the RAVLT
(Ryan, Geisser, Randall, & Georgemiller, 1986). Each of the locally
developed speed tests was administered twice, and the estimated
reliabilities, obtained by boosting the correlation between scores on
each administration by the Spearman-Brown formula, were .97 for
Digit Copying, .95 for Symbol Copying, .81 for Letter Comparison, and
.86 for Pattern Comparison.
Results
Because age norms were available for several of the measures, it was possible to use those data to examine the
representativeness of the present sample. The age-scaled
scores revealed that the sample was generally above average,
but somewhat less select in the decade of the 30s than in other
decades. Means for the entire sample were (for tests with
scaled means of 10 and SDs of 3) Digit Symbol = 12.3, Block
Design = 11.9; Object Assembly = 10.8; (and for tests with
scaled means of 50 and SDs of 10) Shipley Vocabulary = 58.0;
Shipley Abstraction = 60.3, and W C S T number of errors =
49.3. It is important to note that when level of education is
considered in addition to age, as in the W C S T measure, the
current sample is very similar to the normative sample for that
test.
A factor analysis (exploratory with promax rotation) conducted on the health variables (see Table 1) revealed two
factors. O n e factor had high (i.e., greater than .73) Ioadings on
the three subjective rating measures and was correlated .28
with age. The second health factor was correlated only .09 with
the first factor and only .06 with age. The variables loading on
this factor were reports of past treatment for cardiovascular
disease, current medications for high blood pressure, reports
of head injury resulting in loss of consciousness, and past
treatment for neurological disorder. Because the health rating
factor correlated with age and is more similar to the measures
used by other researchers, a composite health index formed by
averaging the z scores for the three self-rating measures was
used in the subsequent analyses.
Table 2 contains means, standard deviations, and the proportions of variance associated with age before and after control
of health and education. It is apparent that there was relatively
little alteration of the age effects when health or education
were controlled, suggesting that the influences were largely
independent of one another. Nonlinear age relations were
significant on a few variables, and this was invariably because
of larger age-related effects with increased age.
Small but significant interactions of age and health were
present on the Trails A, W C S T percentage of conceptual
responses, and W C S T number of categories measures. In all
cases, the interactions reflected larger age relations among the
individuals reporting themselves to be in poorer health.
Interactions of age and education were significant on the Trails
A and B measures, reflecting smaller age relations among
participants with more education. G e n d e r differences were
significant on several measures after control of age: women
performed at higher levels than men on the Digit Symbol and
on R A V L T Trials 3, 4, 5, and 7 measures, but at lower levels on
the Block Design measure. No interactions of age and gender
were significant on any measures.
Selection of Measures for Later Analyses
To select the most meaningful measures from each test for
later analyses, several preliminary analyses were conducted.
No formal analyses were conducted with the speed measures;
instead, the two perceptual comparison measures (Letter
Comparison and Pattern Comparison) were retained for
subsequent analyses because they have been found to exhibit
the strongest relations with other cognitive measures (e.g.,
Salthouse, 1993b, 1994b).
An hierarchical regression analysis was conducted with the
Trail Making measures in which the presumably simpler Trails
A measure was controlled before examining the age-related
variance in the Trails B measure. This analysis revealed that
there was a significant increment in variance (i.e., .056)
associated with age in the Trails B measure after control of'the
Trails A measure, thus suggesting that unique age-sensitive
processes were involved in the Trails B measure. Therefore,
both the Trails A and Trails B variables were retained for
subsequent analyses.
LOCALIZATION OF AGE EFFECTS
277
Table 2
Means, Standard Deviations, and Age Relations for Primary Dependent Variables (n = 259)
Proportion of variance
Test variable
M
SD
Age
Age after
health a
Age after
education
Age after health
and educationa
Digit Copying
Symbol Copying
Letter Comparison
Pattern Comparison
WAIS-R
Digit Symbol
Block Design
Object Assembly
Trail Making
A
B
Word Association (Fluency)b
F30
F60
A30
A60
$30
$60
Shipley
Vocabulary
Abstraction
Paired Associates
Trial 1
Trial 2
WCST
Number of trials
Errors (%)
Perseverative responses (%)
Perseverative errors (%)
Nonperseverative errors (%)
Conceptual responses (%)
Number of categories
Trials to Category 1 (n = 256)
Failure to Maintain Set
Learning to Learn (n = 246)
RAVL'Ix
A1
A2
A3
A4
A5
B1
A6
A7
Target Recognition
Distractor Recognition
49.8
13.6
8.7
15.6
10.3
4.0
2.9
3.8
.310"
.315"
.243*
.436*
.267*
.281"
.207*
.394*
.270*
.281"
.213"
.403*
.241"
.258*
.187"
.373*
57.1
31.1
29.4
14.5
10.0
6.2
.429*
.219"
.170"
.375*
.196"
.166"
.384*
.188"
.152"
.346*
.175"
.153"
33.4
76.4
16.5
44.2
.256*
.348"
.210"
.322*
.229*
.306*
.192"
.219"
8.1
12.7
7.3
11.2
8.1
13.4
2.4
4.0
2.4
3.9
2.4
4.1
.021
,018
,009
.002
.059*
.052*
.018
.019
.001
.001
.049*
.042*
.011
.010
.003
.000
.043*
.038*
.011
.012
.004
.000
.038*
.033*
33.9
28.8
4.7
8.5
.063"
,199 *
.056*
.187 *
.092*
.076*
.155 *
.155 *
2.6
2.4
1.8
1.5
.261"
.123"
.246*
.113"
.224*
.106"
.220*
.100"
100.7
27.6
17.7
15.6
11.9
65.7
4.96
15.1
0.69
-3.64
23.4
15.2
14.8
11.4
6.8
20.5
1.59
11.5
1.04
6.90
.165"
.170"
.184"
.184"
.042*
.175"
.167*
.018
.069*
.063*
.148"
.161"
.172"
.171"
.043*
.163"
.140*
.014
.046*
.060*
.139"
.140"
.153"
.153"
.034*
.144"
.141 *
.010
.067*
.058*
.131"
.139"
.150"
.148"
.037*
.141"
.122*
.009
.046*
.056*
6.5
9.2
10.7
11.5
12.3
6.4
10.2
10.1
12.9
9.4
1.8
2.3
2.4
2.4
2.1
2.2
3.3
3.5
2.4
3.2
.142"
.220*
.187"
.200*
.164"
.184"
.199"
.177"
.081 *
.088*
.127"
.228*
.200*
.197"
.164"
.172"
.206*
.181"
.089*
.100"
.126'
.205*
.170"
.181"
.146"
.172"
.180"
.164"
.080*
.087*
.116"
.216"
.187"
.183"
.150"
.163"
.191"
.171"
.088*
.099*
Note. WAIS-R = Wechsler Adult Intelligence Scale--Revised; WCST = Wisconsin Card Sorting Test; RAVLT = Rey Auditory-Verbal
Learning Test.
aHealth is a composite that averages the ratings from Health Satisfaction, Health Rating, and Health-Related Limitations items, bSubentry
variables refer to the first letter (F, A, or S) of the words written and the time in seconds taken for the responses, cVariable subentries refer to
successive recall trials with an initial list (A) of words, another trial (B) with the interference word list, and a final recognition trial.
*p < .01.
Correlations b e t w e e n the scores after 30 s and after 60 s in
the word association fluency tests w e r e quite high (i.e.,
F = .90, A = .88, and S = .90), and t h e r e was no significant
i n c r e m e n t in R z associated with age in any of the 60-s m e a s u r e s
after control o f the c o r r e s p o n d i n g 30-s measure. This latter
result implies that all of the unique or i n d e p e n d e n t age-related
effects are a p p a r e n t in the first 30 s o f the test. Nevertheless,
the 60 s m e a s u r e s were r e t a i n e d for s u b s e q u e n t analyses
instead of the 30-s m e a s u r e s because of their p r e s u m e d greater
reliability. However, only the scores from the F and S versions
were used because the m e a s u r e from the A version had a
near-zero correlation with age.
Significant age effects w e r e evident on the n u m b e r of trials
m e a s u r e in the W C S T (cf. Table 2), and t h e r e f o r e p e r c e n t a g e s
were used instead of absolute frequencies for o t h e r m e a s u r e s
from this test. However, it is i m p o r t a n t to n o t e that the
correlations b e t w e e n the absolute and the p e r c e n t a g e measures were very high (e.g., .99 for both total errors and
278
SALTHOUSE, FRISTOE, AND RHEE
perseverative errors), and thus the pattern of results would
have been very similar had the absolute measures been used.
An exploratory factor analysis (see Table 3) suggested the
existence of two factors, with the first and largest factor
including most measures. This pattern implies that even
though the measures varied in terms of their relations to age
(see Table 2), the measures do not appear to be measuring
fundamentally different aspects of performance. The number
of categories and percentage of perseverative error measures
are frequently reported in analyses of WCST performance, but
the number of categories measure was highly skewed and was
not suitable for the planned covariance structural modeling.
The percentage of conceptual level responses, corresponding
to the number of consecutive correct trials occurring in
sequences of three or more divided by the total number of
trials, was therefore used as a second WCST measure in later
analyses. The number of categories measure was highly correlated with both the percentage of perseveration errors
(r = -.82) and the percentage of conceptual level responses
(r = .88).
An exploratory factor analysis on the 10 measures available
from the RAVLT revealed that only a single factor could be
identified with an eigenvalue greater than 1.0. Therefore, at
least in this sample, all of the RAVLT memory measures can
be conceptualized as reflecting a single factor of memory.
Performance on the second trial and on the sixth trial, after the
interference list, were used in subsequent analyses because
they had two of the highest loadings on the factor. Additional
analyses of trial-by-trial performance in this test has been
reported in a separate article by Dunlosky and Salthouse (in
press).
Interrelations of Measures
The measures retained for later analyses were transformed
where necessary such that higher values represent better
performance, and then correlations were computed. These
correlations are summarized in Table 4, where it can be seen
Table 3
Results of Exploratory Factor Analysis (Promax Rotation) on
Measures From the Wisconsin Card Sorting Test
Measure
Factor I
Number of trials
Errors (%)
Perseverative responses (%)
Perseverative errors (%)
Nonperseverative errors (%)
Conceptual responses (%)
Number of categories
Trials to category 1
Failure to maintain set
Learning to learn
Proportion of variance
Eigenvalue
.881
.978
.926
.945
.693
-.979
-.891
.575
.212
-.593
.661
6.610
1. Age
2. Factor 1
Factor 2
.633
.464
.308
.336
.497
-.476
-.565
-.227
.779
-.694
.114
1.137
Correlation
h2
.876
.964
.860
.895
.541
.969
.850
.570
.616
.604
Factor 1
Factor 2
.36
--
.29
.39
that most were moderate in magnitude, with a range from .25
to .89.
Proportions of shared age-related variance for the variables
in Table 4 are displayed in Table 5. Entries in this table
correspond to ratios of b/b + d from Figure 1, where the
column is the criterion variable and the row is the controlled
variable. The entries are not symmetric across the criterion
and controlled variables because regions c and d, representing
the unique or unshared age-related variance in each variable,
are not necessarily equal for a given pair of variables.
It can be seen that an average of 58% of the age-related
variance in these variables is shared. The means for the
columns, which range from 44% to 80%, indicate the average
amount of age-related variance in the criterion variable that is
shared with other variables. The row means indicate the
average amount of variance that the controlled variable shared
with other variables. Therefore, variables with the highest row
averages account for the largest proportions of age-related
variance in the other variables. Inspection of the table reveals
that the speed variables have the highest means.
Speed has been hypothesized to be a central construct in
cognitive aging (e.g., Salthouse, 1991, 1993b, 1994b), and the
results just described indicate that the speed measures shared
considerable age-related variance with the other measures.
Therefore, the proportion of age-related variance and the
regression coefficient for age was determined for each of the
measures after control of composite measures of sensorymotor speed (average ofz scores for Symbol Copying and Digit
Copying) and perceptual speed (average ofz scores for Letter
Comparison and Pattern Comparison). (The correlation between the Symbol Copying and Digit Copying measures was
.69, and that between the Letter Comparison and Pattern
Comparison measures was .66.) The R 2 and B values from
these analyses are reported in Table 6.
Table 6 shows that there is considerable reduction of the
age-related variance in the measures after control of the speed
composites, and particularly after control of the perceptual
speed composite. For example, the age-related variance was
reduced 67% (from .261 to .085) for the first paired associates
measure, 74% (from.184 to .047) for the WCST percentage of
Perseveration Errors measure, and 88% (from .170 to .021) for
the Object Assembly measure. There was no significant residual age-related variance for the Shipley Abstraction measure, and with some variables (e.g., Letter Comparison, F60,
and $60) the direction of the age relation was actually reversed
after control of the composite perceptual speed measure.
A reduction was also evident in the magnitude of the
regression coefficient (B), although it is not necessarily the
same magnitude as that evident in the proportion of agerelated variance because the two measures reflect different
aspects of the age relation. That is, the variance measure
indicates how strong the relation is in terms of the proportion
of variance in the scores associated with age, whereas the
regression coefficient indicates the amount by which the
dependent variable changes with each additional year of age.
Stated somewhat differently, the regression coefficient corresponds to the slope of the line relating performance to age, and
the proportion of variance indicates the degree to which that
279
LOCALIZATION OF AGE EFFECTS
Table 4
Correlation Matrix
1. Letter Comparison
2. Pattern Comparison
3. Digit Symbol
4. Trails A
5. Trails B
6. Paired Associates 1
7. Paired Associates 2
8. RAVLT, Trial 2
9. RAVLT, Trial 6
10. Shipley Abstraction
11. WCST-PE
12. WCST-CL
13. Object Assembly
t4. Block Design
15. F60
16. $60
Age
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1.0
.66
.74
.53
.55
.36
.30
.30
.36
.57
.37
.41
.37
.44
.34
.42
-.49
1.0
.75
.58
.60
.46
.38
.40
.47
.51
.38
.39
.49
.51
.35
.43
-.66
1.0
.63
.65
.46
.43
.36
.44
.59
.43
.45
.48
.56
.37
.43
-.66
1.0
.76
.29
.29
.28
.35
.48
.39
.35
.46
.48
.26
.38
-.51
1.0
.38
.36
.40
.43
.59
.49
.46
.40
.50
.35
.39
-.59
1.0
.59
.48
.52
.45
.35
.36
.36
.43
.27
.33
-.51
1.0
.42
,44
.38
.28
.27
.30
.33
.26
.30
-.35
1.0
,69
.36
.27
.29
.27
.30
.31
,35
-.47
1.0
.41
.31
.32
.33
.38
.33
.36
-.45
1.0
.57
.57
.51
.65
.49
.53
-.45
1.0
.89
.34
.47
.25
.29
-.43
1.0
.33
.50
.26
.28
-.42
1.0
.64
.26
.34
-.41
1.0
.26
.34
-.47
1.0
.72
-.13
1.0
-.23
Note. RAVLT = Rey Auditory-Verbal Learning Test; WCST = Wisconsin Card Sorting Test; PE = perseverative errors; CL = conceptual
learning; F60 = refers to a 60-s response with words beginning with letter F; $60 = refers to a 60-s response with words beginning with letter S.
regression line characterizes the data. With many of the
variables, it a p p e a r s that control o f the s p e e d m e a s u r e s has a
g r e a t e r a t t e n u a t i n g effect o n the d e g r e e to which the regression line accurately describes t h e data than o n the m a g n i t u d e
o f the change in score associated with each additional year
o f age.
(Paired Associates 1 and 2; R A V L T Trials 2 and 6), and right
parietal (Object Assembly and Block Design). This model did
not fit the data well; ×z(62, N = 259) = 493.35, s t a n d a r d i z e d
root m e a n residual ( R M R ) = .14, g o o d n e s s o f fit ( G F I ) = .77,
adjusted g o o d n e s s o f fit ( A G F I ) = .67, comparative fit index
(CFI) = .77, and examination o f the modification indices
suggested that n u m e r o u s major alterations would be n e e d e d to
improve the fit.
Instead of attempting to introduce a series o f data-based
modifications of the model derived from anatomical assumptions, two exploratory analyses were c o n d u c t e d on the data
after including the t h r e e variables hypothesized to reflect
perceptual s p e e d (i.e., Digit Symbol, L e t t e r Comparison, a n d
P a t t e r n Comparison). The initial analysis was a principal
c o m p o n e n t s analysis because the first c o m p o n e n t in this type
Structural Analyses
A confirmatory factor analysis (J6reskog & S6rbom, 1993)
was c o n d u c t e d to examine the plausibility of a m o d e l organized in t e r m s o f t h r e e anatomical regions, that is, frontal
(Trails A, Trails B, Shipley Abstraction, W C S T p e r c e n t a g e o f
Perseveration Errors, W C S T p e r c e n t a g e o f Conceptual Level
R e s p o n s e s , Fluency F60, and Fluency $60), medial temporal
Table 5
Proportions of Shared Age-Related Variance
Controlled
variable
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
ll.
12.
13.
14.
15.
16.
M
Letter Comparison
Pattern Comparison
Digit Symbol
Trails A
Trails B
Paired Associates 1
Paired Associates 2
RAVLT, Trial 2
RAVLT, Trial 6
Shipley Abstraction
WCST-PE
WCST-CL
Object Assembly
Block Design
F60
$60
1
2
3
4
5
6
7
-.98
.99
.73
.83
.47
.30
.35
.44
.72
.44
.49
.43
.57
.16
.32
.55
.66
-.89
.59
.67
.44
.27
.34
.42
.47
.31
.31
.42
.48
.12
.24
.44
.77
.89
-.64
.73
.44
.32
.29
.39
.55
.36
.38
.41
.54
.13
.24
.47
.68
.89
.94
-.84
.32
.27
.29
.40
.58
.45
.38
.52
.60
.11
.28
.50
.61
.81
.87
.98
-.39
.30
.40
.43
.62
.49
.45
.37
.54
.13
.24
.51
.44
.71
.71
.32
.52
-.60
.60
.63
.54
.39
.40
.39
.53
.12
.24
.48
.56
.86
.94
.55
.76
.97
-.76
.76
.67
.47
.45
.49
.60
.17
.32
.62
Criterion variable
8
9
10
.38
.66
.57
.34
.62
.69
.46
-.86
.46
.30
.34
.30
.37
.15
.27
.45
.51
.83
.78
.51
.71
.77
,51
.90
-.56
.39
.40
.41
.53
.17
.30
.55
.81
.89
.97
.72
.92
.68
.43
.50
.56
-.75
.73
.66
.87
.39
.44
.69
11
12
13
14
15
16
M
.56
.69
.80
.61
.84
.54
.32
.36
.43
.80
-.98
.49
.60
.96
.74
.65
.64
.73
.85
.55
.81
.57
.32
.41
.46
.81
.99
-.47
.63
.97
.71
.66
.59
.92
.91
.75
.72
.59
.37
.39
.49
.76
.45
.45
-.77
.96
.81
.66
.61
.86
.92
.69
.79
.62
.35
.37
.49
.82
.70
.75
.91
-.99
.88
.72
.90
.06
.0l
.99
.47
.99
.90
.98
.97
.25
.13
.14
.14
.12
-.64
.51
.99
.90
.90
.97
.99
.90
.66
.90
.89
.99
.25
.24
.31
.49
.92
-.80
.65
.78
.80
.66
.75
.63
.43
.52
.57
.64
.46
.46
.45
.55
.43
.45
.58
Note. RAVLT = Rey Auditory-Verbal Learning Test; WCST = Wisconsin Card Sorting Test; PE = perseverative errors; CL = conceptual
learning; F60 refers to a 60-s response with words beginning with the letter F; $60 refers to a 60-s response with words beginning with letter S.
280
SALTHOUSE, FRISTOE, AND RHEE
Table 6
Proportions of Age-Related Variance and Regression Coefficients Before and After Control
of Speed Composites
After motor speed
Age alone
Measure
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
M
Letter Comparison
Pattern Comparison
Digit Symbol
Trails A
Trails B
Paired Associates I
Paired Associates 2
RAVLT-2
RAVLT-6
ShipleyAbstraction
WCST-PE
WCST-CL
Object Assembly
Block Design
F60
$60
Incremental
After perceptual
speed
Incremental
Rz
B
R2
B
R2
B
.243*
.436*
.429*
.256*
.348*
.261'
.123"
.220*
.199"
.199"
.184"
,175"
.170"
.219"
.018
.052*
.221
-.077
-.138
-.517
.453
1.416
-.050
-.029
-.059
-.079
-.207
.264
-.464
-.139
-.253
-.029
-.050
.029*
.088*
.053*
.024*
.087*
.123"
.044*
.120"
.068*
.037*
.060*
.053*
.028*
.033*
.014
.002
.054
-.033
-.078
-.227
.176
.892
-.043
-.022
-.055
-.058
-.113
.189
-.320
-.071
-.124
.032
.012
.012"
.0t1"
.032*
.024*
.059*
.085*
.022
.085*
.042*
.009
.047*
.034*
.021"
.031"
.019
.008
.034
.022
-.028
-.183
.179
.755
-.037
-.016
-.047
-.047
-.054
.172
-.263
-.063
-.122
.039
.026
Note. RAVLT = Rey Auditory-Verbal Learning Test; WCST = Wisconsin Card Sorting Test; PE =
perseverative errors; CL = conceptual learning; F60 refers 60-s responses with words beginning with F;
$60 refers to words beginning with the letter S.
*p < .01.
of analysis has been postulated to provide an estimate of the
general or common factor in a set of variables (Jensen, 1980;
Ree & Earles, 1991). The primary result from this analysis was
that the first component accounted for nearly 47% of the total
variance in the variables. Furthermore, all measures had
moderately high loadings on the first principal component,
with a range of .543 to .819 and a mean of .679. This finding
indicates that a substantial proportion of the total variance in
the variables was shared with the hypothesized general factor,
which had a correlation of - . 6 7 with age.
A factor analysis was then conducted to identify factors that
might more closely correspond to theoretical constructs than
the mathematical composites derived from the principal components analysis. An exploratory factor analysis with promax
(oblique) rotation resulted in five interpretable factors. The
factor solution is shown in Table 7, and correlations involving
the factors are presented in Table 8.
Table 8 reveals that moderate negative age correlations
were evident with all factors, with the smallest age relation
apparent in the fluency factor. Positive correlations with
education of around .3 were evident for all factors, but only
one correlation, to the first (speed) factor, was significant with
the health composite.
Figure 2 portrays the age trends on the factors. The
functions shown are all very similar, with the exception of
somewhat shallower relations with the fluency factor. The
slight dip in the 30s probably reflects the fact that the research
participants in this decade were likely of somewhat lower
ability than those in other decades, as suggested by the scaled
scores described earlier.
Finally, a structural model was examined in which percep-
tual comparison speed functioned as the hypothesized common construct contributing to the mediation of age-related
effects on the other constructs. The Trail Making measures
were eliminated from the analyses because the factor analysis
in Table 7 suggests that they were largely redundant with the
other speed variables, and the Shipley Abstraction measure
was eliminated because it was factorially complex, as reflected
by moderate loadings distributed across several factors (cf.
Table 7).
The measurement model (from a confirmatory factor analysis) provided an acceptable fit to the data, and therefore the
age variable was introduced and the fit examined when age was
directly related only to the speed construct. This initial model
provided an adequate fit, ×2 (71,N = 259) = 153.26, and thus it
was compared with successive models containing paths from
age to the memory, reasoning, space, and fluency constructs.
The models with paths from age to memory and from age
to fluency provided significantly better fits; such as A×2(1,
N = 259) = 9.62, for memory, and A×z(1, N = 259) = 22.54,
for fluency. Therefore, the final model included paths to both
of these constructs, and it provided a good fit to the data; X2
(69, N = 259) = 123.88, Standardized RMR = .048, GFI =
.94, AGFI = .90, CFI = .97). Standardized coetficients for this
model are in Figure 3.
The existence of the direct path from age to memory (i.e.,
-.24) indicate that there are significant specific, or direct,
age-related influences on memory independent of speed. This
is consistent with the moderate residual age-related variance
in the memory measures after control of the perceptual speed
composite apparent in Table 6. However, the absence of direct
paths from age to either the reasoning or the space constructs
281
LOCALIZATION OF AGE EFFECTS
Table 7
Results of Factor Analysis (Promax Rotation) on the Measures in Table 4
Factor
1
2
3
4
5
Label
Speed
Memory
Reasoning
Fluency
Space
h2
1. Letter Comparison
2. Pattern Comparison
3. Digit Symbol
4. Trails A
5. Trails B
6. Paired Associates 1
7. Paired Associates 2
8. RAVLT, Trial 2
9. RAVLT, Trial 6
10. Shipley Abstraction
11. WCST-PE
12. WCST-CL
13. Object Assembly
14. Block Design
15. F60
16. $60
Proportion of variance
.821
.831
.875
.843
.847
.414
.371
.408
.480
.640
.469
.465
.501
.573
.380
.480
.468
.385
.528
.509
.336
.466
.787
.744
.831
.836
.462
.353
.366
.358
.414
.352
.401
.091
.409
.373
.449
.383
.530
.374
.277
.299
.334
.660
.963
.964
.327
.532
.284
.305
.078
.444
.419
.435
.327
.402
.305
.284
.375
.383
.615
.303
.308
.328
.331
.929
.913
.072
.434
.544
.574
.471
.442
.517
.429
.210
.313
.685
.420
.428
.888
.867
.278
.380
.054
.681
.722
.782
.724
.741
.674
.578
.720
.711
.736
.929
.929
.798
.783
.869
.840
Note. Entries in bold have the highest loadings on the factor and were used to assign labels for the
factors. RAVLT = Rey Auditory-Verbal Learning Test; WCST = Wisconsin Card Sorting Test; PE =
perseverative errors; CL = conceptual learning; F60 refers to 60-s responses with words beginning with F;
$60 refers to 60-S responses with words beginning with the letter S.
indicates that all of the age-related variance in these constructs
is apparently mediated through the speed construct.
The path from age to fluency was positive rather than
negative, and thus this relation was examined with another
structural model in which vocabulary was included as an
additional exogenous variable, and speed and fluency were the
only latent constructs. This model, in which there was no direct
path from age to fluency and instead all of the age-relations
were mediated through slower speed and higher vocabulary,
provided a good fit to the data; ×2 (10, N = 259) = 26.62;
standardized R M R = .035, G F I = .97; A G F I = .92, and CFI =
.98. Therefore, it can be inferred that the age effects on the
fluency measure are smaller than would be expected from the
mediation of speed because of the offsetting influence of a
higher vocabulary with increased age.
Discussion
Before discussing implications of the current results, it is
useful to mention their similarity in several respects to results
of other studies. First, as expected because one of the criteria
for selection of the tests was documented age sensitivity, most
of the measures were found to have m o d e r a t e correlations
with age (cf. Table 2). Second, as reported by Salthouse
(1992b), the values presented in Table 6 indicate that a large
proportion (i.e., 92.5%) of the age-related variance in the Digit
Symbol measure is shared with the Letter Comparison and
Pattern Comparison perceptual speed measures. Third, the
greater attenuation of the age-related variance after control of
perceptual speed measures than after control of sensory-motor
speed measures (see Table 6) replicates similar findings by
Salthouse (1993b, 1994b). Fourth, consistent with the finding
by Salthouse and Fristoe (1995), some of the age-related
variance in the Trails B measure was unique, and independent
of the age-related variance in the Trails A measure. This
suggests that the requirement to alternate between letter and
number targets imposes additional unique demands with
increased age. Fifth, the existence of a significant relation
between age and measures of memory after measures of
Table 8
Correlations With Age and Factors
Factor
1
1. Age
2. Education
3. Health
Factor I Speed
Factor 2 Memory
Factor 3 Reasoning
Factor 4 Fluency
Factor 5 Space
*p < .01.
--
2
-.13
--
3
1
2
3
4
5
.25*
-.14
-.67*
.33*
-.26*
-.55*
.26*
-.08
-.43*
.37*
-.11
-.20*
.34*
-.06
-.46*
.31"
-.13
--
.50*
--
.49*
.38*
--
.47*
.40*
.34*
.55*
.41"
.45*
--
.36*
282
SALTHOUSE, FRISTOE, AND RHEE
1.5
F1 (Sepeed)
F2(Memory)
1.2
F3 (Reasoning)
0.9
F4(Ruency)
0.6
"~',
O
O
¢/)
II.. \'~
0.3
.,tl...
FS(space)
..";';" %'"...
~"' 4 , , . , , , - ~ _ ~
~--,~.':.
0
-0.3
-0.6
-0.9
-1.2
-1.5
:~0
I
I
ChronologicalAge
Figure 2. Mean factor (F) scores as a function of age.
perceptual speed have been controlled replicates the finding
by. Salthouse (1993b) and suggests that not all of the agerelated influences on memory are mediated by slower speed.
Sixth, the discovery that the age-related effects on fluency
measures were small because increased age was associated
with higher levels of vocabulary knowledge replicates results of
Salthouse (1993a).
Finally and perhaps most importantly, the results indicating
that much of the age-related variance in a variety of different
variables is shared and is not all independent and specific, are
consistent with many earlier reports (Salthouse, 1991, 1993b,
1994a, 1994b). What is new about the current results is that
this pattern has now been found to hold for measures that are
sometimes assumed to be sensitive to functioning in different
neuroanatomical regions. As an example, the values in Table 5
indicate that the measure of percentage of perseverative errors
in the WCST, which is often considered a prototypical measure
of frontal lobe functioning, shares 60% of its age-related
variance with a visual-spatial measure (Block Design) often
hypothesized to be sensitive to right parietal lobe functioning,
and 54% of its age-related variance with a verbal memory
measure (Paired Associates 1) that might be suspected to
.23
.77
Memory
.13
.56
-.24
/
\1/
/
~
Reasoning
.87
-.74
Speed
.75
.42
.88
Space
.75
Ruency
Figure 3. Standardized coefficients for the structural model found to fit the data. DIGSYM = Digit
Symbol; PATCOM = Pattern Comparison; LETCOM = Letter Comparison; RVLT = Rey Verbal
Learning Test; PA = Paired Associate; WCSTCL = Wisconsin Card Sorting Test conceptual learning;
WCSTPE = Wisconsin Card Sorting Test perseverative errors; BLKDES = Block Design; OBJASSM =
Object Assembly; S-60 = 60-S responses with words beginning with letter S; F-60 = 60-S responses with
words beginning with letter F.
283
LOCALIZATION OF AGE EFFECTS
reflect functioning in the medial temporal lobe. These results
resemble those by Salthouse (1995) in which a substantial
proportion of the age-related variance in verbal and spatial
memory measures sometimes hypothesized to reflect functioning in different cerebral hemispheres was found to be shared.
An important implication of the current results is that the
neuropsychological measures used in this study do not merely
reflect functioning in discrete and independent structures or
modules, nor are the age-related effects on them exclusively
determined by distinct and unique sets of influences. This does
not mean that a certain amount of specificity and localization
could not exist, but it does suggest that inferences about
specific or discrete age-related influences should be interpreted cautiously until there is confidence that those effects
are truly distinct and independent of the age-related influences on other types of measures (also see Reitan & Wolfson,
1994, for a similar argument concerning the importance of
general factors affecting neuropsychological performance).
It is possible that somewhat different results might have
been obtained with measures in which there was more evidence of both sensitivity and specificity. For example, two
measures of equal sensitivity but one specific to damage in the
frontal lobe and one specific to damage in the right parietal
lobe may share much lower proportions of age-related variance than the measures examined in this study. Unfortunately,
at present there appears to be little consensus with respect to
which neuropsychological measures possess the desired properties. Thus this speculation cannot be investigated at the
present time. However, it is important to note that the
methodological procedures described in this study should be
applicable to the examination of the degree of independence
of age-related influences with virtually any combination of
measures.
At least two quite different interpretations could be proposed to account for the existence of large common or shared
age-related influences on a wide range of neuropsychological
and cognitive variables. One interpretation is that despite the
speculations about different functional localization, all of the
measures may be dependent on functioning of a single neuroanatomical structure. For example, it is possible that many
cognitive measures require the involvement of the frontal lobe
regardless of any other structures that might contribute to
performance on those measures. Indeed, there are reports that
patients with frontal lobe damage are impaired in both Block
Design and RAVLT performance (Janowsky et al., 1989), as
well as in paired associate learning (Benton, 1968). The key
aspect of this interpretation is that the high degree of shared
age-related influences might be explained by effects on a single
neuroanatomical structure required by all measures. Research
with neuroimaging techniques may be useful in identifying
structures that are active during the performance of many
different cognitive tasks and might be the source of the
common age-related influences.
A second interpretation might maintain that the lack of
independence of the age-related variance on different neuropsychological and cognitive measures is attributable to the
existence of broad systemic factors that have effects on many
different measures. For example, commonality of age-related
influences could be a consequence of diffuse cell loss induced
by cerebrovascular problems, loss of myelination, or reductions in the quantity of various neurotransmitters. The critical
feature of this interpretation is that the source of the common
or shared age-related variance in different cognitive measures
is not discrete and localized, but instead corresponds to
anatomical or physiological characteristics distributed across
many regions of the cortex.
Although it does not yet appear possible to distinguish
unequivocally between these alternative interpretations for
the relative lack of specificity of age-related effects on the
neuropsychological measures used in this study, some insight
might be gained by detailed consideration of the nature of the
hypothesized common or general factor. There are several
reasons for believing that the speed with which simple cognitive operations can be executed plays an important role in the
hypothesized common factor underlying age-related influences on different cognitive measures. First, measures of speed
were found to share the highest proportions of age-related
variance with the other variables (see Table 5). Second, the
values in Table 6 indicate that an average of almost 85% of the
age-related variance in the other variables is eliminated after
control of the perceptual speed composite. And third, the
structural model with speed as the hypothesized common
factor (see Figure 3) provided a good fit to the observed
covariances.
It is important to emphasize that we are not claiming that
slower processing speed is the exclusive source or proximal
cause of the age-related effects in these measures. Not only
was there a significant path from age to the memory construct
in the structural model (see Figure 5), but the residual
age-related variance after control of the perceptual speed
composite was significantly greater than zero for many of the
measures in Table 6. Instead, our proposal is that something
analogous to processing speed appears to play a key role in the
age-related influences that have been found to be shared
across different cognitive measures, including those obtained
without external time limits (cf. Salthouse, 1994b).
Of course, we realize that the neuroanatomical and neurophysiological factors responsible for the age-related reduction
in processing speed still need to be identified. For example,
research is needed to establish linkages between behavioral
measures of speed and neurophysiological properties, such as
number of functional neurons, amount of dendritic branching,
number of synapses, quantity of neurotransmitters, variability
in temporal patterns of excitation, or degree of myelination.
Nevertheless, focusing on the processing speed construct may
have substantial heuristic value because results from the types
of correlational analyses reported here suggest that a substantial amount of the age-related effects on a wide range of
cognitive neuropsychological and cognitive measures could be
accounted for if the negative relation between age and
measures of processing speed could be explained.
References
Anderson, S. W., Damasio, H., Jones, R. D., & Tranel, D. (1991).
Wisconsin Card Sorting Test performance as a measure of frontal
lobe damage. Journal of Clinical and Experimental Neuropsychology,
13, 909-922.
284
SALTHOUSE, FRISTOE, AND RHEE
Ardila, A., & Rosselli, W. (1989). Neuropsychological characteristics
of normal aging. Developmental Neuropsychology, 5, 307-320.
Axelrod, B. N., & Henry, R. R. (1992). Age-related performance on
the Wisconsin Card Sorting, Similarities, and Controlled Oral Word
Association tests. Clinical Neuropsychologist, 6, 16--26.
Benton, A. (1968). Differential behavioral effects in frontal lobe
disease. Neuropsychologia, 6, 53-60.
Birren, J. E., & Morrison, D. F. (1961). Analysis of the WAIS subtests
in relation to age and education. Journal of Gerontology, 16, 363-369.
Blakemore, C. B., & Falconer, M. A. (1967). Long-term effects of
anterior temporal lobectomy on certain cognitive functions. Journal
of Neurology, Neurosurgery, and Psychiatry, 30, 364-367.
Bolla-Wilson, K., & Bleecker, M. L. (1986). Influence of verbal
intelligence, sex, age, and education on the Rey Auditory Verbal
Learning Test. Developmental Neuropsychology, 2, 203--211.
Bolter, J. F., Long, C. J., & Wagner, M. (1983). The utility of the
Thurstone Word Fluency Test in identifying cortical damage.
Clinical Neuropsychology, 5, 77-82.
Butler, M., Retzlaff, P., & Vanderploeg, R. (1991). Neuropsychological test usage. Professional Psychology: Research and Practice, 22,
510-512.
Butler, R. W., Rorsman, I., Hill, J. M., & Tuma, R. (1993). The effects
of frontal brain impairment on fluency: Simple and complex
paradigms. Neuropsychology, 7, 519-529.
Craik, F. I. M., Morris, L. W., Morris, R. W., & Loewen, E. R. (1990).
Relations between source amnesia and frontal lobe functions in
older adults. Psychology and Aging, 5, 148-151.
Crockett, D., Bilsker, D., Hurwitz, T., & Kozak, J. (1986). Clinical
utility of three measures of frontal lobe dysfunction in neuropsychiatric samples. International Journal of Neuroscience, 30, 241-248.
Daigneault, S., Braun, C. M. J., & Whitaker, H. A. (1992). Early effects
of normal aging on perseverative and non-perseverative prefrontal
measures. Developmental Neuropsychology, 8, 99-114.
desRosiers, G. & Kavanagh, D. (1987). Cognitive assessment in closed
head injury: Stability, validity, and parallel forms for two neuropsychological measures of recovery. International Journal of Clinical
Neuropsychology, 9, 162-173.
Drewe, E. A. (1974). The effect of type and area of brain lesion on
Wisconsin Card Sorting Test performance. Cortex, 10, 159-170.
Dunlosky, J. & Salthouse, T. A. (in press). A decomposition of
age-related differences in multi-trial free recall.Aging and Cognition.
Earles, J. L., & Salthouse, T. A. (1995). Interrelations of age, health, and
speed.Jotonal of Gerontology: PsychologicalSciences, 50B, P33-P41.
Elias, M. F., Robbins, M. A., Walter, L. J., & Schultz, N. R. (1993).
The influence of gender and age on Halstead-Reitan neuropsychological test performance. Journal of Gerontology, 48, P278-P281.
Geffen, G., Moar, K. J., O'Hanlon, A. P., Clark, C. R., & Geffen, L. B.
(1990). Performance measures of 16- to 86-year-old males and
females on the auditory-verbal learning test. The Clinical Neuropsychologist, 4, 45-63.
Goldstein, S., & Shelly, C. (1981). Does the right hemisphere age more
rapidly than the left? Journal of Clinical Neuropsychology, 3, 67-78.
Hartman, M., & Hasher, L. (1991). Aging and suppression: Memory
for previously relevant information. Psychology andAging, 6, 587-594.
Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G., & Curtiss, G.
(1993). Wisconsin Card Sorting Test manual: Revised and expanded.
Odessa, FL: Psychological Assessment Resources.
Heaton, R. K., Grant, I., & Matthews, C. (1986). Differences in
neuropsychological test performance associated with age, education, and sex. In I. Grant & K. M. Adams (Eds.), Neuropsychological
assessment in neuro-psychiatric disorders (pp. 100-120). New York:
Oxford University Press.
Hooper, F. H., Hooper, J. O., & Colbert, K. C. (1984). Personality and
memory correlates of intellectual functioning: Young adulthood to old
age. Basel, Switzerland: Karger.
Janowsky, J. S., Shimamura, A. P., Kritcheysky, M., & Squire, L. R.
(1989). Cognitive impairment following frontal lobe damage and its
relevance to human amnesia. Behavioral Neuroscience, 103, 548-560.
Jensen, A. R. (1980). Bias in mental testing. New York: Free Press.
Johnson, M. M. S. (1993). Thinking about strategies during, before,
and after making a decision. Psychology andAging, 8, 231-241.
J6reskog, K., & S6rbom, D. (1993). LISREL 8." Structural equation
modeling with the SIMPLIS command language. Chicago, IL: Scientific Software International.
Kaufman, A. S., Reynolds, C. R., & McLean, J. E. (1989). Age and
WAIS-R intelligence in a national sample of adults in the 20-to-74year age range: A cross-sectional analysis with educational level
controlled. Intelligence, 13, 235-253.
Kertesz, A. (1994a). Frontal lesions and function. In A. Kertesz (Ed.),
Localization and neuroimaging in neuropsychology (pp. 567-598). San
Diego, CA: Academic Press.
Kertesz, A. (1994b). Localization and function: Old issues revisited
and new developments. In A. Kertesz (Ed.), Localization and
neuroimaging in neuropsychology (pp. 1-33). San Diego, CA: Academic Press.
Klisz, D. (1978). Neuropsychological evaluation of older persons. In
M. Storandt, E. Siegler, & M. Elias (Eds.), The clinicalpsychology of
aging (pp. 71-95). New York: Plenum.
Kolb, B., & Whishaw, I. Q. (1990). Fundamentals of human neuropsychology. New York: Freeman.
Libon, D. J., Glosser, G., Malamut, B. L., Kaplan, E., Goldberg, E.,
Swanson, R., & Sands, L. P. (1994). Age, executive functions and
visuospatial functioning in healthy older adults. Neuropsychology, 8,
38-43.
Lyness, S. A., Eaton, E. M., & Schneider, L. S. (1994). Cognitive
performance in older and middle-aged depressed outpatients and
controls.Journal of Gerontology:PsychologicalSciences, 49, P129-P136.
Matarazzo, J. D., Matarazzo, R. G., Wiens, A. N., Gallo, A. E., &
Klonoff, H. (1976). Retest reliability of the Halstead Impairment
Index in a normal, a schizophrenic, and two samples o f organic
patients. Journal of Clinical Psychology, 32, 338-349.
McCarthy, R. A., & Warrington, E. K. (1990). Cognitive neuropsychology: A clinical introduction. San Diego: CA: Academic Press.
McFie, J. (1960). Psychological testing in clinical neurology. Journal of
Nervous and Mental Disorders, 131, 383-393.
Miller, E. (1984). Verbal fluency as a function of a measure of verbal
intelligence and in relation to different types of cerebral pathology.
British Journal of Clinical Psychology, 23, 53-57.
Milner, B. (1963). Effects of different brain lesions on card sorting.
Archives of Neurology, 9, 90-100.
Milner, B. (1964). Some effects of frontal lobectomy in man. In J. M.
Warren & K. Akert (Eds.), The frontal granular cortex and behavior
(pp. 313-331). New York: McGraw-Hill.
Mitrushina, M., Satz, P., Chervinsky, A., & D'Elia, L. (1991). Performance of four age groups of normal elderly on the Rey AuditoryVerbal Learning Test. Journal of Clinical Psychology, 47, 351-357.
Mittenberg, W., Seidenberg, M., O'Leary, D. S., & DiGiulio, D. V.
(1989). Changes in cerebral functioning associated with normal
aging. Journal of Clinical and Experimental Neuropsychology, 11,
918-922.
Moehle, K. A., & Long, C. J. (1989). Models of aging and neuropsychological test performance decline with aging. Journal of Gerontology:
Psychological Science, 44, P176-P177.
Nelson, H. E. (1976). A modified card sorting test sensitive to frontal
lobe deficits. Cortex, 12, 313-324.
Nielsen, H., Knudsen, L., & Daugbjerg, O. (1989). Normative data for
eight neuropsychological tests based on a Danish sample. Scandinavian Journal of Psychology, 30, 37-45.
Parkin, A. J., & Walter, B. M. (1991). Aging, short-term memory, and
frontal dysfunction. Psychobiology, 19, 175-179.
LOCALIZATION OF AGE EFFECTS
Pendleton, M. G., Heaton, R. K., Lehman, R. A. W., & Hulihan, D.
(1982). Diagnostic utility of the Thurstone World Fluency Test in
neuropsychological evaluations. Journal of Clinical Neuropsychology,
4, 307-317.
Perret, E. (1974). The left frontal lobe of man and the suppression of
habitual responses in verbal categorical behavior. Neuropsychologia,
12, 323-330.
Query, W. T., & Megran, J. (1983). Age-related norms for AVLT in a
male patient population. Journal of Clinical Psychology, 39, 136-138.
Ree, M. J. A., & Earles, J. A. (1991). The stability ofg across different
methods of estimation. Intelligence, 15, 271-278.
Reitan, R. M. (1992). TrailMaking Test. Tucson, AZ: Reitan Neuropsychology Laboratory.
Reitan, R. M., & Wolfson, D. (1994). A selective and critical review of
neuropsychological deficits and the frontal lobe. Neuropsychology
Review, 4, 161-198.
Robinson, A. L., Heaton, R. K., Lehman, R. A. W., & Stilson, D. W.
(1980). The utility of the Wisconsin Card Sorting Test in detecting
and localizing frontal lobe lesions. Journal of Consulting and Clinical
Psychology, 48, 605-614.
Ryan, J. J., Geisser, M. E., Randall, D. M., & Georgemiller, R. J.
(1986). Alternate form reliability and equivalency of the Rey
Auditory-Verbal Learning Test. Journal of Clinical and Experimental
Neuropsychology, 8, 611~516.
Salthouse, T. A. (1991). Mediation of adult age differences in
cognition by reductions in working memory and speed of processing.
Psychological Science, 2, 179-183.
Salthouse, T. A. (1992a). Shifting levels of analysis in the investigation
of cognitive aging. Human Development, 35, 321-342.
Salthouse, T. A. (1992b). What do adult age differences in the Digit
Symbol Substitution Test reflect? Journal of Gerontology: Psychological Sciences, 47, P121-P128.
Salthouse, T. A. (1993a). Speed and knowledge as determinants of
adult age differences in verbal tasks. Journal of Gerontology: Psychological Sciences, 48, P29-P36.
Salthouse, T. A. (1993b). Speed mediation of adult age differences in
cognition. Developmental Psychology, 29, 722-738:
Salthouse, T. A. (1994a). How many causes are there of aging-related
decrements in cognitive functioning? DevelopmentalReview, 14, 413-437.
Salthouse, T. A. (1994b). The nature of the influence of speed on adult
age differences in cognition, Developmental Psychology, 30, 240-259.
Salthouse, T. A. (1995). Differential age-related influences on memory
for verbal-symbolic information and visual-spatial information. Journal of Gerontology: Psychological Sciences, 50B, P193-P201.
Salthouse, T. A., & Fristoe, N. (1995). A process analysis of adult age
effects on a computer-administered trail-making test. Neuropsychology, 9, 518-528.
Salthouse, T. A., Kausler, D. H., & Saults, J. S. (1988). Investigation of
student status, background variables, and the feasibility of standard
tasks in cognitive aging research. Psychology andAging, 3, 29-37.
Salthouse, T. A., & Mitchell, D. R. D. (1990). Effects of age
andnaturally occurring experience on spatial visualization performance. Developmental Psychology, 26, 845-854.
285
Schaie, K. W., & Schaie, J. P. (1977). Clinical assessment and aging. In
J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of
aging (pp. 692-723). New York: Van Nostrand Reinhold.
Seines, O. A., Jacobson, L., Machado, A. M., Becker, J. T., Wesch, J.,
Miller, E. N., Visscher, B., McArthur, J. C. (1991). Normative data
for a brief neuropsychological screening battery. Perceptual and
Motor Skills, 73, 539--550.
Sergent, J. (1988). Some theoretical and methodological issues in
neuropsychological research. In F. Boiler & J. Grafman (Eds.),
Handbook of neuropsychology (Vol. 1, pp. 69-81). Amsterdam:
Elsevier.
Shelton, M. D., Parsons, O. A., & Leber, W. R. (1982). Verbal and
visuospatial performance with aging: A neuropsychological approach. Journal of Gerontology, 37, 336-341.
Spreen, O., & Strauss, E. (1991).A compendium ofneuropsychological
tests. New York: Oxford University Press.
Squire, L. R. (1987). Memory and brain. New York: Oxford University
Press.
Stuss, D. T., & Benson, D. F. (1984). Neuropsychological studies of the
frontal lobes. Psychological Bulletin, 95, 3-28.
Stuss, D. T., Stethem, L. L., & Poirier, C. A. (1987). Comparison of
three tests of attention and rapid information processing across six
age groups. The Clinical Neuropsyehologist, 1, 139-152.
Van Gorp, W. G., Satz, P., & Mitrushina, M. (1990). Neuropsychological processes associated with normal aging. Developmental Neuropsychology, 6, 279-290.
Walsh, K. (1994). Neuropsychology: A clinical approach. London:
Churchill Livingstone.
Warrington, E. K., James, M., & Maciejewski, C. (1986). The WAIS as
a lateralizing and localizing diagnostic instrument: A study of 656
patients with unilateral cerebral lesions. Neuropsychologica, 24,
223-239.
Wechsler, D. (1981). Manual for the WechslerAdult Intelligence Scale-Revised. New York: Psychological Corporation.
Weingartner, H. (1968). Verbal learning in patients with temporal
lobe lesions. Journal of Verbal Learning and Verbal Behavior, 7,
520-526.
Whelihan, W. M., & Lesher, E. L. (1985). Neuropsychological changes
in frontal functions with aging. Developmental Neuropsychology, 1,
371-380.
Wiederholt, W. C., Cahn, D., Butters, N. M., Salmon, D. P., KritzSilverstein, D., & Barrett-Connor, E. (1993). Effects of age, gender,
and education on selected neuropsychological tests in an elderly
community cohort. Journal of the American Geriatrics Society, 41,
639-647.
Zachary, R. A. (1986). Shipley Institute of Living Scale--Revised. Los
Angeles, CA: Western Psychological Services.
Received D e c e m b e r 5, 1994
Revision received May 3, 1995
Accepted September 8, 1995 •