how localized are age-related effects on neuropsychological … · 2002-03-18 · neuropsychology...

Neuropsychology Copyright 1996 by the American Psychological Association, Inc. 1996, Vol. I0, No. 2, 272-285 0894-4105/96/$3.00

How Localized Are Age-Related Effects on Neuropsychological Measures?

Timothy A. Salthouse and Nathanael Fristoe Georgia Institute of Technology

Soo Hyun Rhee Emory University

A battery of neuropsychological tests (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 commonly used neuropsychological measures is specific and 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., Gold- stein & 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; Wheli- han & 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-

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 [email protected].

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-compara- tive 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-

272

LOCALIZATION OF AGE EFFECTS 273

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 parti- tioned 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 combi- nations of variables can be examined to evaluate generalizabil- ity 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.

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 neurologi- cally 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 Auditory- Verbal Learning Test (RAVLT; see Spreen & Strauss, 1991), and a locally developed paired associate test. Moderate-to- large 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; Weingart- ner, 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 WAIS-R (Wechsler, 1981). The other tests were locally developed, with two hypothesized to assess sensory-motor speed (Digit Copy- ing and Symbol Copying), and two hypothesized to involve sensory-motor and perceptual comparison speed (Letter Com- parison 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 pre- dicted 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.

M e t h o d

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

L O C A L I Z A T I O N OF A G E EFFECTS 275

T a b l e 1

Demographic Characteristics o f Research Participants According to Age Group

Age group

Characteristic 20s 30s 40s 50s 60s 70-94

Number 41 40 30 58 Women (%) 59 65 60 66

Age M 24.6 34.5 43.7 54.3 SD 2.6 2.9 2.7 2.7

Education a M 3.9 3.8 4.1 3.9 SD 0.9 1.0 0.9 0.9

Heal th satisfaction b M 2.0 2.1 2.2 2.1 SD 0.7 0.7 0.8 0.7

Heal th rating b M 2.0 2.0 2.2 2.1 SD 0.9 0.8 0.9 0.9

Health-related activity limitations b M 1.3 1.5 1.8 1.4 SD 0.7 0.8 1.1 0.7

Cardiovascular s u r g e r f M 0.00 0.00 0.03 0.07 SD 0.00 0.00 0.18 0.26

Hypertension medications c M 0.00 0.08 0.13 0.14 SD 0.00 0.27 0.35 0.35

Head injury c M 0.00 0.00 0.13 0.03 SD 0.00 0.00 0.35 0.18

Neurological t reatment c M 0.02 0.05 0.13 0.03 SD 0.16 0.22 0.35 0.18

43 58

47 70

64.4 78.4 3.0 5.7

3.7 3.5 1.2 1.1

2.1 2.4 0.8 0.9

2.2 2.5 0.9 1.0

1.6 2.2 0.7 1.I

0.14 0.21 0.35 0.41

0.33 0.40 0.47 0.50

0.02 0.04 0.15 0.20

0.02 0.02 0.15 0.15

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 numbers indicate better health. CResponses were yes/no, and thus means correspond to the proportion of individuals reporting a positive response.

instructed to decide whether the two members 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 them if they were not. To adjust for guessing, the measure of performance in these tests was the number of i tems answered correctly minus the number 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 pat tern (Symbol Copying) in the top box and nothing in the bot tom box. The task for the research participant was to copy the item from the top box into the bot tom box and to complete as many items as possible within the allotted time. Performance was assessed in terms of the number of i tems 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. One minute was allowed for recall with each list. Performance was assessed in terms of the number of response words correctly recalled in each list.

The 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 WCST 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 number 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 measures 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.

The 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 number


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 multiple- trial 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 multiple- choice 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, Mata- razzo, 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.

R e s u l t s

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 o ther 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 WCST number of errors = 49.3. It is important to note that when level of education is

considered in addition to age, as in the WCST 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. One 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 t reatment for cardiovascular disease, current medications for high blood pressure, reports of head injury resulting in loss of consciousness, and past t reatment 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, WCST percentage of conceptual responses, and WCST 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. Gender 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.


Table 2 Means, Standard Deviations, and Age Relations for Primary Dependent Variables (n = 259)

Proportion of variance

Age after Age after Age after health Test variable M SD Age health a education and education a

Digit Copying 49.8 10.3 .310" .267* .270* .241" Symbol Copying 13.6 4.0 .315" .281" .281" .258* Letter Comparison 8.7 2.9 .243* .207* .213" .187" Pattern Comparison 15.6 3.8 .436* .394* .403* .373* WAIS-R

Digit Symbol 57.1 14.5 .429* .375* .384* .346* Block Design 31.1 10.0 .219" .196" .188" .175" Object Assembly 29.4 6.2 .170" .166" .152" .153"

Trail Making A 33.4 16.5 .256* .210" .229* .192" B 76.4 44.2 .348" .322* .306* .219"

Word Association (Fluency) b F30 8.1 2.4 .021 .018 .011 .011 F60 12.7 4.0 ,018 .019 .010 .012 A30 7.3 2.4 ,009 .001 .003 .004 A60 11.2 3.9 .002 .001 .000 .000 $30 8.1 2.4 .059* .049* .043* .038* $60 13.4 4.1 .052* .042* .038* .033*

Shipley Vocabulary 33.9 4.7 .063" .056* .092* .076* Abstraction 28.8 8.5 ,199 * .187 * . 1 5 5 * . 1 5 5 *

Paired Associates Trial 1 2.6 1.8 .261" .246* .224* .220* Trial 2 2.4 1.5 .123" .113" .106" .100"

WCST Number of trials 100.7 23.4 .165" .148" .139" .131" Errors (%) 27.6 15.2 .170" .161" .140" .139" Perseverative responses (%) 17.7 14.8 .184" .172" .153" .150" Perseverative errors (%) 15.6 11.4 .184" .171" .153" .148" Nonperseverative errors (%) 11.9 6.8 .042* .043* .034* .037* Conceptual responses (%) 65.7 20.5 .175" .163" .144" .141" Number of categories 4.96 1.59 .167* .140* .141 * .122* Trials to Category 1 (n = 256) 15.1 11.5 .018 .014 .010 .009 Failure to Maintain Set 0.69 1.04 .069* .046* .067* .046* Learning to Learn (n = 246) -3.64 6.90 .063* .060* .058* .056*

RAVL'I x A1 6.5 1.8 .142" .127" .126' .116" A2 9.2 2.3 .220* .228* .205* .216" A3 10.7 2.4 .187" .200* .170" .187" A4 11.5 2.4 .200* .197" .181" .183" A5 12.3 2.1 .164" .164" .146" .150" B1 6.4 2.2 .184" .172" .172" .163" A6 10.2 3.3 .199" .206* .180" .191" A7 10.1 3.5 .177" .181" .164" .171" Target Recognition 12.9 2.4 .081 * .089* .080* .088* Distractor Recognition 9.4 3.2 .088* .100" .087* .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.

Correlat ions be tween the scores after 30 s and after 60 s in the word association fluency tests were quite high (i.e., F = .90, A = .88, and S = .90), and there was no significant increment in R z associated with age in any of the 60-s measures after control of the corresponding 30-s measure. This lat ter result implies that all of the unique or independen t age-related effects are apparent in the first 30 s of the test. Nevertheless, the 60 s measures were re ta ined for subsequent analyses instead of the 30-s measures because of their p resumed greater

reliability. However, only the scores from the F and S versions were used because the measure from the A version had a near-zero correlation with age.

Significant age effects were evident on the number of trials measure in the WCST (cf. Table 2), and therefore percentages were used instead of absolute frequencies for o ther measures from this test. However, it is important to note that the correlat ions be tween the absolute and the percentage measures were very high (e.g., .99 for both total errors and


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 Factor 2 h 2

Number of trials .881 .633 .876 Errors (%) .978 .464 .964 Perseverative responses (%) .926 .308 .860 Perseverative errors (%) .945 .336 .895 Nonperseverative errors (%) .693 .497 .541 Conceptual responses (%) -.979 -.476 .969 Number of categories -.891 -.565 .850 Trials to category 1 .575 -.227 .570 Failure to maintain set .212 .779 .616 Learning to learn -.593 -.694 .604 Proportion of variance .661 .114 Eigenvalue 6.610 1.137

Correlation

Factor 1 Factor 2

1. Age .36 .29 2. Factor 1 - - .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 sensory- motor 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 age- related 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


Table 4 Correlation Matrix

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1. Letter Comparison 1.0 2. Pattern Comparison .66 1.0 3. Digit Symbol .74 .75 1.0 4. Trails A .53 .58 .63 1.0 5. Trails B .55 .60 .65 .76 1.0 6. Paired Associates 1 .36 .46 .46 .29 .38 1.0 7. Paired Associates 2 .30 .38 .43 .29 .36 .59 1.0 8. RAVLT, Trial 2 .30 .40 .36 .28 .40 .48 .42 1.0 9. RAVLT, Trial 6 .36 .47 .44 .35 .43 .52 ,44 ,69 1.0

10. Shipley Abstraction .57 .51 .59 .48 .59 .45 .38 .36 .41 11. WCST-PE .37 .38 .43 .39 .49 .35 .28 .27 .31 12. WCST-CL .41 .39 .45 .35 .46 .36 .27 .29 .32 13. Object Assembly .37 .49 .48 .46 .40 .36 .30 .27 .33 t4. Block Design .44 .51 .56 .48 .50 .43 .33 .30 .38 15. F60 .34 .35 .37 .26 .35 .27 .26 .31 .33 16. $60 .42 .43 .43 .38 .39 .33 .30 ,35 .36 Age - .49 -.66 - .66 -.51 - .59 -.51 -.35 - .47 -.45

1.0 .57 1.0 .57 .89 1.0 .51 .34 .33 1.0 .65 .47 .50 .64 1.0 .49 .25 .26 .26 .26 1.0 .53 .29 .28 .34 .34 .72 1.0

-.45 - .43 - .42 -.41 - .47 -.13 -.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, i t appears that control of the speed measures has a greater a t tenuat ing effect on the degree to which the regression line accurately describes the data than on the magni tude of the change in score associated with each additional year of age.

Structural Analyses

A confirmatory factor analysis (J6reskog & S6rbom, 1993) was conducted to examine the plausibility of a model orga- nized in terms of three anatomical regions, that is, frontal (Trails A, Trails B, Shipley Abstract ion, WCST percentage of Perseverat ion Errors, WCST percentage of Conceptual Level Responses , Fluency F60, and Fluency $60), medial temporal

(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 tandardized

root mean residual ( R M R ) = .14, goodness of fit (GFI) = .77, adjusted goodness of fit (AGFI) = .67, comparat ive fit index

(CFI) = .77, and examination of the modification indices

suggested that numerous major alterations would be needed to

improve the fit.

Instead of a t tempting to introduce a series of data-based modifications of the model derived from anatomical assump-

tions, two exploratory analyses were conducted on the data

after including the three variables hypothesized to reflect perceptual speed (i.e., Digit Symbol, Let ter Comparison, and

Pat tern Comparison) . The initial analysis was a principal components analysis because the first componen t in this type

Table 5 Proportions of Shared Age-Related Variance

Controlled Criterion variable

variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 M

1. Letter Comparison - - .66 .77 .68 .61 .44 .56 .38 .51 .81 .56 .64 .59 .61 .90 .99 .65 2. Pattern Comparison .98 - - .89 .89 .81 .71 .86 .66 .83 .89 .69 .73 .92 .86 .06 .90 .78 3. Digit Symbol .99 .89 - - .94 .87 .71 .94 .57 .78 .97 .80 .85 .91 .92 .0l .90 .80 4. Trails A .73 .59 .64 - - .98 .32 .55 .34 .51 .72 .61 .55 .75 .69 .99 .97 .66 5. Trails B .83 .67 .73 .84 - - .52 .76 .62 .71 .92 .84 .81 .72 .79 .47 .99 .75 6. Paired Associates 1 .47 .44 .44 .32 .39 - - .97 .69 .77 .68 .54 .57 .59 .62 .99 .90 .63 7. Paired Associates 2 .30 .27 .32 .27 .30 .60 - - .46 ,51 .43 .32 .32 .37 .35 .90 .66 .43 8. RAVLT, Trial 2 .35 .34 .29 .29 .40 .60 .76 - - .90 .50 .36 .41 .39 .37 .98 .90 .52 9. RAVLT, Trial 6 .44 .42 .39 .40 .43 .63 .76 .86 - - .56 .43 .46 .49 .49 .97 .89 .57

10. Shipley Abstraction .72 .47 .55 .58 .62 .54 .67 .46 .56 - - .80 .81 .76 .82 .25 .99 .64 l l . WCST-PE .44 .31 .36 .45 .49 .39 .47 .30 .39 .75 - - .99 .45 .70 .13 .25 .46 12. WCST-CL .49 .31 .38 .38 .45 .40 .45 .34 .40 .73 .98 - - .45 .75 .14 .24 .46 13. Object Assembly .43 .42 .41 .52 .37 .39 .49 .30 .41 .66 .49 .47 - - .91 .14 .31 .45 14. Block Design .57 .48 .54 .60 .54 .53 .60 .37 .53 .87 .60 .63 .77 - - .12 .49 .55 15. F60 .16 .12 .13 .11 .13 .12 .17 .15 .17 .39 .96 .97 .96 .99 - - .92 .43 16. $60 .32 .24 .24 .28 .24 .24 .32 .27 .30 .44 .74 .71 .81 .88 .64 - - .45 M .55 .44 .47 .50 .51 .48 .62 .45 .55 .69 .65 .66 .66 .72 .51 .80 .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.


Table 6 Proportions o f Age-Related Variance and Regression Coefficients Before and After Control of Speed Composites

Measure

After perceptual After motor speed speed

Age alone Incremental Incremental R z B R 2 B R 2 B

1. Letter Comparison .243* -.077 .029* -.033 .012" .022 2. Pattern Comparison .436* -.138 .088* -.078 .0t1" -.028 3. Digit Symbol .429* -.517 .053* -.227 .032* -.183 4. Trails A .256* .453 .024* .176 .024* .179 5. Trails B .348* 1.416 .087* .892 .059* .755 6. Paired Associates I .261' -.050 .123" -.043 .085* -.037 7. Paired Associates 2 .123" -.029 .044* -.022 .022 -.016 8. RAVLT-2 .220* -.059 .120" -.055 .085* -.047 9. RAVLT-6 .199" -.079 .068* -.058 .042* -.047

10. Shipley Abstraction .199" -.207 .037* -.113 .009 -.054 11. WCST-PE .184" .264 .060* .189 .047* .172 12. WCST-CL ,175" -.464 .053* -.320 .034* -.263 13. Object Assembly .170" -.139 .028* -.071 .021" -.063 14. Block Design .219" -.253 .033* -.124 .031" -.122 15. F60 .018 -.029 .014 .032 .019 .039 16. $60 .052* -.050 .002 .012 .008 .026 M .221 .054 .034

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 - .67 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; X 2 (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


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 h 2

1. Letter Comparison .821 .385 .409 .444 .434 .681 2. Pattern Comparison .831 .528 .373 .419 .544 .722 3. Digit Symbol .875 .509 .449 .435 .574 .782 4. Trails A .843 .336 .383 .327 .471 .724 5. Trails B .847 .466 .530 .402 .442 .741 6. Paired Associates 1 .414 .787 .374 .305 .517 .674 7. Paired Associates 2 .371 .744 .277 .284 .429 .578 8. RAVLT, Trial 2 .408 .831 .299 .375 .210 .720 9. RAVLT, Trial 6 .480 .836 .334 .383 .313 .711

10. Shipley Abstraction .640 .462 .660 .615 .685 .736 11. WCST-PE .469 .353 .963 .303 .420 .929 12. WCST-CL .465 .366 .964 .308 .428 .929 13. Object Assembly .501 .358 .327 .328 .888 .798 14. Block Design .573 .414 .532 .331 .867 .783 15. F60 .380 .352 .284 .929 .278 .869 16. $60 .480 .401 .305 .913 .380 .840 Proportion of variance .468 .091 .078 .072 .054

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, GFI = .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 mediat ion of speed because of the offsetting influence of a higher vocabulary with increased age.

D i s c u s s i o n

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 modera te correlations with age (cf. Table 2). Second, as reported by Salthouse (1992b), the values presented in Table 6 indicate that a large proport ion (i.e., 92.5%) of the age-related variance in the Digit Symbol measure is shared with the Let ter Comparison and Pattern Comparison perceptual speed measures. Third, the greater at tenuation 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 al ternate 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

1 2 3

Factor

1 2 3 4 5

1. Age 2. Education 3. Health

Factor I Speed Factor 2 Memory Factor 3 Reasoning Factor 4 Fluency Factor 5 Space

- - - . 1 3 . 25*

- - - . 1 4

-.67* -.55* -.43* -.20* -.46* .33* .26* .37* .34* .31"

-.26* -.08 -.11 -.06 -.13 - - . 50* .49* . 4 7 * .55*

- - . 38* . 4 0 * .41"

- - . 3 4 * . 4 5 *

- - . 3 6 *

*p < .01.


O O ¢/)

1.5

1.2

0.9

0.6

0.3

0

-0.3

-0.6

-0.9

-1.2

-1.5 :~0

Figure 2.

"~', .,tl... II.. \'~ ..";';" %'"...

~" ' 4 , , . , , , - ~ _ ~ ~--,~.':.

I I

Chronological Age

F1 (Sepeed)

F2 (Memory)

F3 (Reasoning)

F4 (Ruency)

FS (space)

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 age- related 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 earl ier 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

.77

.56

Memory

-.24 / \ 1 / / ~ Reasoning

-.74

.42

.75

.88

Speed

Space

Ruency

.87

.75

.23

.13

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.


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 inter- preted 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 pro- posed 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.

R e f e r e n c e s

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.


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). Neuropsychologi- cal 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 neuropsychi- atric 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: Psychological Sciences, 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 Neuropsy- chologist, 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: Scien- tific 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-74- year 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: Aca- demic 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: Psychological Sciences, 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). Perfor- mance of four age groups of normal elderly on the Rey Auditory- Verbal 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. Scandina- vian Journal of Psychology, 30, 37-45.

Parkin, A. J., & Walter, B. M. (1991). Aging, short-term memory, and frontal dysfunction. Psychobiology, 19, 175-179.


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 Neuropsy- chology 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: Psychologi- cal Sciences, 47, P121-P128.

Salthouse, T. A. (1993a). Speed and knowledge as determinants of adult age differences in verbal tasks. Journal of Gerontology: Psycho- logical 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? Developmental Review, 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. Jour- nal 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. Neuropsychol- ogy, 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.

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). Neuropsychologi- cal processes associated with normal aging. Developmental Neuropsy- chology, 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 Wechsler Adult 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., Kritz- Silverstein, 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 December 5, 1994 Revision received May 3, 1995

Accepted September 8, 1995 •

how localized are age-related effects on neuropsychological … · 2002-03-18 · neuropsychology...

Documents