Stability of Everyday Memory in Age-Associated … of Everyday Memory in Age-Associated Memory Impairment: A Longitudinal Study James R. Youngjohn and Thomas H. Crook III Everyday

Neuropsycholog)1993, Vol. 7. No. 3. 406-41ft

Copyright 1993 by the American Psychological Association, inc.0894-4! 05/93/53.00

Stability of Everyday Memory in Age-Associated MemoryImpairment: A Longitudinal Study

James R. Youngjohn and Thomas H. Crook III

Everyday memory performance was examined longitudinally in 2 groups of Ssmeeting the diagnostic criteria for age-associated memory impairment (AAMI). Onegroup of 157 participants in a drug trial for reversing memory loss in AAMI wastested over multiple sessions. The other group of 75 persons did not participate ina drug trial and thus was tested only twice. Both groups were retested for longitudinalfollow-up about 4 years after their initial session. Follow-up test performance re-mained fairly stable relative to initial performance in both groups. The drug studygroup showed large practice effects during the course of the drug studies, but theseeffects subsided after the drug studies' end. Implications regarding memory declinein the normal elderly and neuropsychological measurement issues are discussed.

It is widely recognized that many cognitive func-tions often decline in the later decades of adulthood,particularly learning and memory (Fozard, 1985;Poon, 1985; Youngjohn & Crook, in press). Althoughthere have been numerous cross-sectional investiga-tions confirming the deleterious effects of advancingage on memory and cognition, relatively few longitu-dinal studies have been completed. Noteworthy ex-ceptions include the Baltimore longitudinal study(Alder, Adam, & Arenberg, 1990), the Duke longitu-dinal study (Siepler, McCarty, & Logue, 1982), andthe Seattle longitudinal study (Cooney, Schaie, &Willis, 1988).

Longitudinal results have generally indicated thatthere are modest, gradual cognitive declines in oldersubjects, with the inflection point occurring some-where around age 60 (Hertzog & Schaie, 1988).Mitrushina and Satz (1991) found that most subjectsimproved over three 1-year testing intervals, excepttheir oldest subjects, who remained relatively stable.These results illustrate one of the hazards of conduct-ing longitudinal research: practice effect artifact.

Until recently, there has been no generally accepteddiagnostic classification for normal, healthy persons

James R. Youngjohn and Thomas H. Crook III, MemoryAssessment Clinics, Inc., Scottsdale, Arizona.

We would like to thank Lisa Burrows for her insightfulcomments on the article.

Correspondence concerning this article should be ad-dressed to James R. Youngjohn, 7434 East Stetson Drive,Suite 250, Scottsdale, Arizona 85251.

who experience declining memory and cognition intheir later years. Krai (1962, 1966) introduced the termbenign senescent forgetfulness to describe otherwisehealthy elderly individuals who experience cognitivedeclines relative to their age peers. More recent studieshave validated this concept (Larrabee, Levin, & High,1986). However, this nosological framework fails toaddress the larger number of individuals experiencingmemory loss associated with normal developmentalprocesses.

A work group sponsored by the National Institute ofMental Health proposed diagnostic criteria for the clas-sification of age-associated memory impairment(AAMI; Crook et al., 1986). This nosological categoryrefers to persons more than 50 years of age who bothcomplain of memory impairment in tasks of daily lifeand have objective memory test performance at leastone standard deviation below the mean established foryoung adults.

A recent longitudinal epidemiological study (Lane& Snowdon, 1989) reported relatively high preva-lence and incidence rates for AAMI (34.93% [SE =4.54%] and 6.63% [SE = 9.41%], respectively, perannum), as compared with prevalence and incidencerates for probable Alzheimer's disease (13.01% [SE= 7.11%] and 3.06% [SE = 9.79%], respectively,per annum). The high incidence and prevalence ratesfor AAMI emphasize the need for further under-standing of this area. Although the Lane and Snow-don (1989) study suggests that most subjects withAAMI do not progress to Alzheimer's disease, nolarge-scale longitudinal study has actually measuredthe cognitive functioning of AAMI patients over ex-tended periods of time.

406

LONGITUDINAL STUDY OF AAMI 407

The purpose of the present investigation was to bet-ter understand the progression of AAMI. It has beenargued that AAMI is the result of cognitive declineseen in all elderly persons (Bamford & Caine, 1988;Blackford & LaRue, 1989; Smith et al., 1991). Atpresent, the rate of cognitive decline in normal elderlypeople is unclear. Furthermore, the possibility thatAAMI may actually be the result of very early Alz-heimer's disease has not yet been conclusively ruledout (Youngjohn, Larrabee, & Crook, 1992a). Conse-quently, a longitudinal study of this population wouldbe of considerable interest.

Longitudinal studies that rely on psychometric vari-ables as outcome measures are complicated by thewell-described phenomenon of practice effects (e.g.,Youngjohn, Larrabee, & Crook, 1992b). Simply, per-formance on ability tests tends to improve on eachsubsequent administration because the subject has in-creasing familiarity with the material. Although prac-tice effects in memory testing can be minimized byusing alternate test forms of equivalent difficulty(Crook, Youngjohn, & Larrabee, 1992), they may notbe eliminated entirely because the format of the in-strument remains the same. Consequently, our studyused two experimental groups, one in which practiceeffects were minimized and one that involved multipleadministrations of parallel forms of the same tests.

Method

Subjects

Seven hundred seventy-two unpaid volunteers were re-cruited through newspaper and radio advertising. They wereinitially evaluated for participation in one of several trials,supervised by the U.S. Food and Drug Administration, ofcandidate pharmaceutical compounds for treating or revers-ing memory loss associated with normal aging. All subjectssatisfied the diagnostic criteria for AAMI (Crook et al.,1986). Specifically, participants had to be more than 50 yearsold, report gradual decline of everyday memory function,have at least one objective memory test performance fallingat least one standard deviation below the mean for youngadults (Wechsler Memory Scale [WMS; Wechsler, 1945]Logical Memory, 6 or less; WMS Paired Associate Learninghard pairs, 6 or less; and Benton Visual Retention Test[BVRT; Benton, 1974] correct, 7 or less), show evidence ofadequate intelligence as determined by a Wechsler Adult In-telligence Scale (WAIS; Wechsler, 1955) Vocabulary rawscore of at least 32, and show absence of dementia as de-termined by a score of 24 or higher on the Mini-Mental StateExamination (Folstein, Folstein, & McHugh, 1975). Exclu-sion criteria included self-report on a health history ques-

tionnaire of any medical, psychiatric, or neurologic disorderthat could produce cognitive deterioration. All subjects whowere actually enrolled in a drug study also received com-prehensive medical and neurologic workups, including com-puted tomography (CT) scans of the brain.

Of the original 772 patients who satisfied the criteria forAAMI, 331 elected to participate in one of a dozen drug trialsinvestigating treatment effects. The remaining 441 subjectsdeclined to participate because of personal choice, incom-patible medications, and so forth. For this study, we at-tempted to recontact the original 772 subjects by letter andtelephone.

Two hundred thirty-two subjects (82 men and 150 women)agreed to come in, yielding an overall longitudinal studyparticipation rate of 30%. The ages of these participants oninitial evaluation ranged from 50 to 77 (M = 62.08, SD =6.53). Years of education ranged from 6 to 26 (M = 15.41,SD = 2.69). The mean Affective Rating Scale score(Yesavage et al., 1983) was in the nondepressed range (6.36,SD = 4.87), with 90% of the participants obtaining a scoreof 10 or less.

We divided our participants into two subgroups. The firstgroup consisted of 75 persons (27 men and 48 women) whowere not enrolled in a drug study after their initial assess-ment. The longitudinal study response rate for these non-drug study participants was 17%. These subjects were testedonly two times (initial and long-term follow-up). Their agesat the beginning of the investigation ranged from 51 to 74 (M= 61.93, SD = 6.05). Years of education ranged from 12 to22 (M = 15.82, SD = 2.37).

The second group consisted of 157 individuals (55 menand 102 women) who were actually enrolled in a study andwere enlisted in either the drug or placebo condition. Theresponse rate of these drug study participants was 47%. Be-cause all studies required periodic monitoring of treatmenteffects, these participants had been administered differentforms of the same everyday memory tests across multipletrials (mean number of test administrations = 7.16, SD =3.67). These subjects ranged in age from 50 to 77 (M =62.15, SD = 6.77), and years of education ranged from 6 to26 (M = 15.21, SD = 2.81).

Memory Assessment Clinics (MAC) has investigated 12candidate pharmaceutical agents to date. Only one of theseagents, a phospholipid compound (Crook et al., 1991), hasshown a positive treatment effect when compared with theplacebo control group. Given the lack of a positive drugeffect, we combined the drug and placebo control groups forthe purposes of the present longitudinal investigation.

Apparatus

A number of computer-simulated everyday memory testshave been developed at MAC to measure ecologically mean-ingful treatment outcome. They are administered in a stan-dardized manner with a 19-in. (48.26-cm) color monitor witha touch screen, interfaced with a computer equipped with a

408 JAMES R. YOUNGJOHN AND THOMAS H. CROOK III

20-megabyte hard-disk drive, a laser-disk player, and cus-tomized computer graphics hardware. Participants do not usethe computer keyboard; rather, they respond verbally, withthe touch screen or with a touch-tone or rotary telephone. Thetester is present throughout the session and sits behind andgenerally out of view of the participant. Test administrationand scoring are controlled by the tester with a separate com-puter monitor.

Procedure

All subjects were tested in a controlled, clinical environ-ment. Each subsequent administration of a test involved adifferent parallel form, when available. There are now eightparallel forms of each of the everyday memory tests (Crooket al., 1992). The participants in the present study were con-tacted a number of years after the initial evaluation or thecompletion of the drug study and were asked to return forfollow-up testing. In the case of those who did not participatein drug studies, the average test-retest interval was 4.6 years(M = 1,683 days, SD = 292, range = 1,035 days to 2,130days).

Those subjects who participated in drug studies were en-rolled in either the drug or placebo condition, as noted ear-lier, and returned at frequent intervals for retesting so thattheir progress could be monitored for the duration of thestudy. The mean length of time that elapsed between theirinitial evaluation and their last drug study visit was slightlyless than a year (M = 348 days, SD = 283). The drugstudy participants were contacted for the long-termfollow-up assessment several years after the study hadbeen completed and they had discontinued the experimen-tal drug or placebo, or both. The average time that elapsedbetween the end of the drug study and the long-termfollow-up examination was 3 years (M = 1,085 days, SD= 461, range = 58 days to 1,996 days).

In the following paragraphs, we describe the dependentvariables, including the MAC computer-simulated everydaymemory tests and the traditional memory tests.

Name-Face Association. In this test, subjects are pre-sented with live recordings of individuals introducing them-selves by common first names. Recall is assessed by showingthe same individuals in a different order, stating the name ofthe city in which they reside, and asking the subject to pro-vide the name of each person. This study used two learningtrials in which 14 name-face pairs were presented and recallwas assessed. The total number of name-face pairs recalledwas summed over both learning trials. The test has beenshown to be a sensitive measure of age-related memory de-cline (Crook & West, 1990).

First-Last Names. This is a verbal learning test usingthe WMS paired associates paradigm. Subjects are pre-sented with a series of six first-last name pairs, followedby a presentation of the last name only; the task of subjectsis to recall the first name. There were two learning trials inthis investigation. The First-Last Names test is sensitive

to age effects and loads highly on a verbal memory factor(Youngjohn, Larrabee, & Crook, 1991).

Grocery List Selective Reminding Test (GLSRT). Thistest follows the standard selective reminding paradigm(Buschke, 1973), using common grocery-list items as thestimuli to be recalled. On the first trial, subjects are requestedto read aloud a list of 15 grocery items that appear on thevideo screen. After the recall attempt, the words that thesubjects have not recalled briefly reappear on the screen;subjects must then attempt to recall the entire list, havingbeen selectively reminded of the words that they omitted onthe previous trial. Our dependent variable in this study wasthe total number of words recalled on the first two learningtrials. The GLSRT has been demonstrated to be negativelyrelated to age and to load highly on a verbal memory factor(Youngjohn et al., 1991).

Telephone Dialing (with interference). This task is avariation of the standard digit repetition paradigm. Partici-pants are presented with a 10-digit (long distance) tele-phone number on the monitor screen and asked to read italoud. The number then disappears from the screen andthey are instructed to dial the number on a touch-tonephone interfaced with the computer. After the subjectshave completed dialing, they hear a busy signal. The in-struction "Please Redial" appears on the video screen, andthey are asked to redial. Credit is given for each digit di-aled in the correct position, regardless of errors made else-where in the sequence. West and Crook (1990) demon-strated that older subjects are particularly sensitive to theinterference condition of this test.

Recognition of Faces—Delayed, Nonmatching to Sample.Subjects are presented with facial photographs on the touch-screen monitor, with a new face being added to the array untilthere are 25 faces. Subjects are required to identify the newface added by touching it on the monitor. In the present study,we examined the number of correctly completed trials beforethe first error was committed. Previous factor-analytic stud-ies have demonstrated this test to be a relatively robust mea-sure of visual memory (Larrabee & Crook, 1989). An earlierversion was shown to discriminate between normal and de-mented elderly subjects (Flicker, Ferris, Crook, & Bartus,1987).

Misplaced objects. This is a test of object location recall.Subjects are required (using the touch screen) to place 20common objects into the schematic representation of a 12-room house (maximum of 2 objects per room). Delayed re-call is tested at 40 min by requesting subjects to press theroom in which they previously placed each of the objects. Asecond attempt is allowed if they miss on the first attempt,to closely simulate the common activity of looking for a lostarticle. Our dependent variable in the present investigationwas the total number of objects correctly located after bothattempts. Previous studies have suggested that the misplacedobjects test is a measure of verbal-visual associative memory(Crook, Youngjohn, & Larrabee, 1990). It is structurally sim-ilar to a paradigm used in animal model pharmacological


research (Bartus, Dean, & Beer, 1983; Bartus, Fleming, &Johnson, 1978; Bartus & Johnson, 1976).

Divided Attention Recall. This test combines two tasks,a variation on a standard reaction time paradigm in which thetask of driving a car and responding appropriately to chang-ing traffic signals is simulated and a variation of the WMSLogical Memory subtest. Subjects are instructed to changebrake and accelerator pedals appropriately and as quickly aspossible in response to traffic light changes.

Subjects listen to simulated radio weather and traffic re-ports while performing the simulated driving task. After eachbroadcast is presented, immediate recall is assessed in a man-ner similar to that of the WMS Logical Memory subtest; thescore is the number of ideas recalled, averaged between thetwo stories. Larrabee and Crook (1989) reported that per-formances on this test, the WMS Logical Memory subtest,and the WAIS Vocabulary subscale all loaded on a verbalmemory and intelligence factor.

Traditional tests. The following traditional instrumentswere administered in the standard format: the WMS LogicalMemory and Paired Associate Learning subtests (Wechsler,1945), the WAIS Vocabulary subtest (Wechsler, 1955), andthe BVRT (Benton, 1974).

Results

Long-Term Follow-Up Responders VersusNonresponders

Because of the relatively high rate of attrition in ourstudy, we compared participants who did not come infor long-term follow-up with those who did to assessthe potential for selective attrition bias. The ages of theresponders (M = 62.08 years, SD = 6.53) and non-responders (M = 61.36 years, SD = 6.86) were notsignificantly different (p > .05). There was a slightlyhigher ratio of women in the responding sample (65%)as opposed to the nonresponding group (56%; p < .05).Education level was slightly higher in the nonrespond-ing group (M = 16.09 years, SD = 2.95) than theresponding group (M = 15.41 years, SD = 2.69;p < .05).

To determine whether the effects of selective attri-tion were significantly biasing our results, we com-pared initial test performances in the responding andnonresponding groups. A multivariate analysis ofvariance (MANOVA) examining the seven everydaymemory test variables and the five traditional mem-ory test variables in the two groups was not signifi-cant (p > .05). Although the MANOVA did notreveal significant overall differences, independentunivariate analyses of variance (ANOVAs) suggestedthat the participants who came back for long-term

follow-up performed better than those who didnot on two of the seven everyday memory tests(GLSRT and telephone dialing; p < .05) and three ofthe five traditional measures (WMS Logical Mem-ory, BVRT correct, and WAIS Vocabulary). Initialperformances on name-face association, First-LastNames, Recognition of Faces—Delayed, Nonmatch-ing to Sample, Misplaced Objects, Divided AttentionRecall, WMS Paired Associate Learning, and BVRTerrors did not differ significantly between the twogroups (p > .05).

Long-Term Follow-Up in Non-Drug StudyParticipants

In our investigation of longitudinal effects on thoseparticipants who came in for long-term follow-up, wefirst looked at the subgroup in which practice effectswere minimized (i.e., those subjects who did not par-ticipate in drug studies). Table 1 presents mean per-formances on the initial and long-term follow-up ses-sions. We used two statistical techniques to analyze ourdata: repeated measures MANOVA/ANOVA and mul-tiple regression analysis.

The repeated measures MANOVA comparing ini-tial everyday memory and traditional test perfor-mances with long-term follow-up performances wassignificant, Hotelling's T = 7.12, approximate F(7,12) = 4.15, p < .05. A series of univariate repeatedmeasures ANOVAs was then conducted. Table 2 pre-sents F values, degrees of freedom, significance lev-els, and effect sizes (r)2). Inspection of Table 2shows that long-term follow-up performance signifi-cantly differed from initial performance on only fourof seven everyday memory tests (i.e., Name-FaceAssociation, GLSRT, Telephone Dialing, and Di-vided Attention Recall) at the .05 level. Marginallysignificant differences (p < .06) were noted for twotraditional measures, WMS Logical Memory andBVRT errors. Effect sizes were generally modest.Furthermore, these differences did not necessarilyconform to the expected pattern of decline over time.Inspection of Table 1 shows that test performanceactually improved on long-term follow-up for bothName-Face Association and BVRT errors.

Initial test performance and amount of elapsed timewere used to predict long-term follow-up everydaymemory and traditional test performance in the non-drug study participants in a series of multiple regres-sion analyses. Initial performance was entered into theequation first, and quantity of elapsed time was entered

410 JAMES R. YOUNGJOHN AND THOMAS H. CROOK HI

second. Table 3 presents squared multiple correlationsfor initial everyday and traditional memory test per-formance and the change in those values for elapsedtime, as well as significance levels.

As expected, subjects' initial performance on a mea-sure was a powerful predictor of their subsequent per-formance on that same test at follow-up. Initial testperformance was a significant predictor of follow-upperformance on five of seven everyday memory tests;the only exceptions were Recognition of Faces—Delayed, Nonmatching to Sample and Misplaced Ob-jects. Proportions of variance in follow-up perfor-mance accounted for by initial performance in thoseequations that were significant at the .05 level rangedfrom a high of 47% for Name-Face Association to alow of 13% for Divided Attention Recall.

In contrast, the amount of elapsed time between theinitial and follow-up sessions was a very poor predictorof follow-up performance. The addition of elapsedtime resulted in a significant increase in predictive ac-curacy at the .05 level on only three measures: theGLSRT, on which it accounted for an additional 6% ofperformance variance; WMS Logical Memory, onwhich elapsed time accounted for an additional 3% ofperformance variance; and WAIS Vocabulary, onwhich elapsed time accounted for 2% of additionalvariance.

Table 1 shows very slight decreases in performanceon long-term follow-up for GLSRT, telephone dialing,WMS Logical Memory, and WMS Paired AssociateLearning. A slight increase in performance was noted

on WAIS Vocabulary. Negative associations betweenfollow-up performance and elapsing time were con-firmed by negative standardized beta weights for tele-phone dialing (/3 = -.244), WMS Logical Memory (/3= -.196), and WMS Paired Associate Learning O =-.111). However, increasing elapsed time was actuallyassociated with improved long-term follow-up perfor-mance, as demonstrated by positive beta weights forGLSRT O = .239) and WAIS Vocabulary (J3 = .152).

Drug Study Participants

We examined the drug study participants in thesecond phase of the study. The effects of both timeand practice were considered in this stage of the in-vestigation. Table 4 presents means for initial testperformances, performances at the end of the study,and performances on long-term follow-up. Three re-peated measures MANOVAs were conducted: (a)Initial performances were compared with perfor-mances at the end of the drug study, (b) perfor-mances at the end of the drug study were comparedwith long-term follow-up performances, and (c) ini-tial performances were compared with long-termfollow-up performances.

We first compared initial test performances with per-formances at the end of the drug study, which involvedmultiple administrations of different forms of the sameinstruments. The repeated measures MANOVA wassignificant, Hotelling's T = 1.44, approximate F(6,51)= 12.27, p < .001.

Table 1Initial and Long-Term Follow-Up Performance on Everyday Memoryand Traditional Memory Tests in Non-Drug Study Participants

Visit 1

Test M SD

Note. WAIS = Wechsler Adult Intelligence Scale.

Foilow-up visit

M SD

Name-Face AssociationFirst-Last NamesGrocery List Selective Reminding TestTelephone DialingRecognition of Faces — Delayed,

Nonmatching to SampleMisplaced ObjectsDivided Attention RecallWechsler Memory Scale

Logical MemoryPaired Associate Learning

Benton Visual Retention TestCorrectError

WAIS Vocabulary

7250494545

4945

7173

737378

9.474.16

30.614.59

12.60

14.1220.16

8.8414.52

6.485.45

64.32

5.232.226.201.846.38

2.672.94

2.793.18

1.432.579.80

11.004.12

29.483.80

12.55

14.4419.21

8.2214.23

6.664.73

65.06

5.122.075.582.245.97

2.523.74

2.223.61

1.853.029.20


Table 2Repeated Measures Analysis of Variance SummaryData for Non-Drug Study Participants

Test

Name-Face AssociationFirst-Last NamesGrocery List Selective

Reminding TestTelephone DialingRecognition of Faces —

Delayed, Nonmatchingto Sample

Misplaced ObjectsDivided Attention RecallWechsler Memory Scale

Logical MemoryPaired Associate Learning

Benton VisualRetention Test

CorrectErrors

WAIS Vocabulary

F

10.760.254.10

7.080.75

0.005.24

3.730.49

0.633.671.66

df

1,711,491,49

1,431,45

1,581,40

1,701, 72

1,721,721,69

a

.002*

.662

.048*

.011*

.392

.967

.027*

.058

.488

.432

.059

.203

Tf

.13

.01

.08

.14

.02

.00

.12

.05

.01

.01

.05

.02Note. WAIS = Wechsler Adult Intelligence Scale.*p< .05.

Table 5 presents the results from the repeated mea-sured ANOVAs. Inspection of Table 5 reveals that sig-nificant improvements had occurred on all seven of theeveryday memory tests by the completion of the drugstudy (at the .05 level). Effect sizes ranged from a highof 44% of variance accounted for by practice in name-face association to a low of 2% for first-last names.

We further examined the influence of practice effectsby regressing initial performances against test perfor-mances at the end of the drug study, entering the num-ber of test administrations occurring between the initialevaluation and last drug study session, and, finally,entering the amount of intervening time between theinitial and last drug study session. Table 6 presentssquares multiple correlations for initial test perfor-mance, and the change in those values for number ofadministrations and elapsed time, as well as signifi-cance levels for each stage of the analyses.

As expected, subjects' initial performance on a mea-sure was a powerful predictor of their performance ona different form of the same test at the end of a drugstudy. Initial performance was significantly associatedwith performance at the end of the drug study, at the.05 level, on all seven everyday memory tests. Squaredmultiple correlations ranged from a high of .39 fortelephone dialing to a low of .05 for recognition offaces—delayed nonmatching-to-sample (see Table 6).

Adding the number of visits to the equation resultedin significant increases in predictive accuracy for threeof the seven everyday memory tests at the .05 level

(Name-Face Association, GLSRT, and Divided Atten-tion Recall). Additional proportions of performancevariance accounted for by adding the number of visitswere 9% for Name-Face Association, 6% for DividedAttention Recall, and 4% for GLSRT (see Table 6).Table 4 demonstrates improvements in performancefrom the initial visit to the last drug study visit on alleveryday memory tests, suggesting that increased ex-posure to the tests resulted in improved performance,as predicted. Beta weights for number of visits hadpositive values across all seven everyday memorytests, further demonstrating that increased exposure re-sulted in improving test performance.

Adding the amount of elapsed time to the equationpredicting performance at the end of the drug studyresulted in significant improvement in predictive ac-curacy on only one everyday memory test (Name-FaceAssociation) at the .05 level (see Table 6). Interest-ingly, a positive beta weight demonstrated that in-creases in elapsing time were actually associated withimprovements in performance on this measure (j3 =.144). This finding probably reflects additional influ-ence of practice effects not picked up by the numberof visits. However, the additional variance in Name-Face Association performance accounted for by elaps-ing time was minimal (A/f2 change = .02).

Table 3Regression Analyses for Non-DrugStudy Participants

Firstadministration

Test




Misplaced ObjectsDivided Attention RecallWechsler Memory Scale

Logical MemoryPaired Associate

Learning totalBenton Visual

Retention TestCorrectErrors

WAIS Vocabulary

R2

.47

.19

.34

.32

.00

.03

.13

.25

.16

.12

.18

.73

a

.001*

.002*

.001*

.001*

.800

.210

.021*

.001*

.001*

.001*

.001*

.001*

Elapsedtime

A/?2

.01

.00

.06

.06

.03

.00

.01

.03

.01

.01

.01

.02

a

.265

.710

.040*

.054

.221

.683

.584

.003*

.078

.162

.411

.001*Note. Initial test performance was entered first as a predictor, andamount of elapsed time between initial and follow-up sessions wasentered second. WAIS = Wechsler Adult Intelligence Scale.* p < .05.


Table 4Everyday Memory Test Performance in Drug Study Participants

Last drugVisit 1

Test

Name-Face associationFirst-Last NamesGrocery List Selective

Reminding TestTelephone DialingRecognition of Faces — Delayed,

Nonmatching to SampleMisplaced ObjectsDivided Attention Recall

n

155102131

119130

106126

M

9.684.04

29.95

4.0410.87

14.5815.63

SD

4.472.125.54

2.265.17

2.916.37

study

M

14.174.59

32.23

4.8213.72

16.2419.36

visit

SD

5.612.315.85

2.216.40

2.747.24

Follow-up

M

13.193.82

29.07

4.0514.64

14.3114.14

SD

5.082.186.31

2.156.18

2.566.41

The effects of elapsed time between the final drugstudy administration and the long-term follow-up ses-sion were then considered. The repeated measuresMANOVA was significant, Hotelling's T = 1.25, ap-proximate F(6, 56) = 11.67, p < .001.

The results from repeated measures ANOVAs com-paring performances at the end of the drug study withperformances on long-term follow-up are presented inTable 7. Inspection of Table 7 demonstrates that sig-nificant declines in performance between the end of thedrug study and the long-term follow-up evaluation oc-curred on all but one of the everyday memory tests(Recognition of Faces—Delayed, Nonmatching toSample) at the .05 level. Effect sizes ranged from ahigh of .41 for Divided Attention Recall to a low of .05for Name-Face Association.

Test performance at the end of the drug study andamount of elapsed time were used to predict long-termfollow-up test performance in the drug study partici-pants in an additional series of multiple regressionanalyses. As expected, the final drug study test scorewas a significant predictor of long-term follow-up per-formance on all seven everyday memory tests at the .05level. As shown in Table 8, R2 ranged from a high of.49 for Name-Face Association to a low of .12 forRecognition of Faces—Delayed, Nonmatching toSample.

However, the addition of the elapsed time betweenthe end of the drug study and the long-term follow-upevaluation significantly added to the predictive accu-racy of the equation for only one everyday memorytest, Divided Attention Recall, at the .05 level; on thistest, elapsed time accounted for an additional 6% of thevariance in performance (see Table 8). The negativebeta weight (/3 = -.238) confirms that increasingelapsed time was associated with declining recall per-formance on the Divided Attention task. The relative

lack of association between the amount of elapsed timebetween the end of the drug study and the long-termfollow-up session suggests that the significant declinesin performance noted earlier were the result of the at-tenuation of practice effects rather than an actual lon-gitudinal decline of cognitive ability.

In a final repeated measures MANOVA, we com-pared the drug study participants' initial test perfor-mances with their final test performances at long-termfollow-up. The repeated measures MANOVA wasagain significant, Hotelling's T = .81, approximateF(6, 87) = 11.78, p < .001.

The summary of the univariate repeated measuresANOVAs presented in Table 9 reveals significant im-provements on long-term follow-up on two everydaymemory measures, Name-Face Association and Rec-ognition of Faces—Delayed, Nonmatching to Sample;the effects of both intervening time and experienceaccounted for 38% of the variance in performance onthe former test and 19% of the variance in performanceon the latter test. A significant decline was noted on one

Table 5Repeated Measures Analysis of VarianceSummary Data Comparing Initial ScoresWith Scores at the End of the Drug Study

Test




Misplaced ObjectsDivided Attention Recall

F

122.9]5.51

21.63

21.9920.36

33.4550.29

df

1, 1541,951, 115

1. 1181, 121

1, 871, 113

a

.001*

.021*

.001*

.001*

.001*

.001*

.001*

•n-

.44

.02

.16

.16

.14

.28

.31* p < .05.


Table 6Regression Analyses of Drug Study Participants Predicting Final Drug Study Test Performance

Test

First visit No. of visits

A/?2

Elapsed time

Atf2


Reminding TestTelephone DialingRecognition of Faces — Delayed,

Nonmatching to SampleMisplaced ObjectsDivided Attention Recall

.24

.07

.35

.39

.05

.25

.25

.001*

.010*

.001*

.001*

.012*

.001*

.001*

.09

.01

.04

.01

.00

.01

.06

.001*

.630

.005*

.093

.439

.409

.003*

.02

.00

.00

.01

.01

.00

.00

.045*

.526

.875

.112

.198

.668

.517

Note. Visit scores were entered first, number of visits during the drug study was entered second, and elapsed time of the study wasentered third.* p < .05.

everyday memory test, the GLSRT, but the effect sizewas minimal (T)2 = .04).

Discussion

The most notable finding of the present investigationis the relative lack of decline over time in the everydaymemory performance in either of our experimentalgroups. Although we found a number of significantlongitudinal declines in performance, the proportionsof performance variance accounted for by elapsingtime were generally small. Furthermore, we occasion-ally found significant longitudinal effects that were inthe opposite direction of our predictions (i.e., increas-ing elapsed time between assessments was actually as-sociated with slightly improving, rather than declining,test performance).

These results present strong evidence that the con-dition of AAMI is not rapidly progressive but, rather,remains relatively stable. Consequently, AAMI does

Table 7Repeated Measures Analysis of Variance SummaryData Comparing Scores at the End of the DrugStudy With Scores at Long-Term Follow-Up

Test





F

8.6514.3748.66

19.281.15

40.0275.89

df

1, 1561,991, 115

1, 1201, 122

1, 871, 109

a

.004*

.001*

.001*

.001*

.286

.001*

.001*

T)2

.05

.13

.30

.14

.01

.32

.41* p < .05.

not appear to be caused by a progressive neuropatho-logic process such as very early Alzheimer's disease,because our subjects showed minimal deterioration onlong-term follow-up. Rather, AAMI is more likely tobe associated with relatively subtle memory and cog-nitive declines in the elderly that occur over decadesinstead of months or years.

This lack of longitudinal progression in AAMI sug-gests that in terms of memory and cognitive test per-formance, AAMI is essentially equivalent to the effectsof normal aging. A number of investigators have takenthe position that the only factor that differentiates nor-mal aging from AAMI is the patient's complaint thathis or her memory is not what it used to be (Bamford& Caine, 1988; Blackford & LaRue, 1989; Smith et al.,1991). Our results tend to support this position. Ofcourse, although cognitive and memory declines as-sociated with aging may be normal, no one would ar-gue that they are desirable, and few would fault elderlypersons for complaining of them.

We attribute the significant improvements that oc-curred over the course of the drug studies to practiceeffects. Although the reader might be tempted to spec-ulate that these improvements were actually the resultof drug treatment effects, we have seen positive effectsrelative to placebo controls in only one compound ofthe dozen that we have examined (Crook et al., 1991).Our finding of significant practice effects on everydaymemory tests across multiple administrations extendsresults of previous investigations (Youngjohn et al.,1992b) in which everyday memory test performanceimproved on a single readministration. Notably, strongpractice effects were observed in spite of the use ofmultiple parallel test forms. The use of parallel testforms of equivalent difficulty has been proposed as a


Table 8Regression Analyses Examining PredictiveInfluence of Last Drug Study Visit and AddingAmount of Elapsed Time to Predict Long-TermFollow-Up Performance

Last drugstudy visit

Test





R2

.49

.20

.42

.24

.12

.21

.24

a

.001*

.002*

.001*

.001*

.001*

.001*

.001*

Elapsedtime

Afl2

.00

.00

.00

.00

.00

.01

.06

a

.900

.710

.431

.643

.645

.437

.004

* p < .05.

means of attenuating practice effects when performingserial evaluations overtime (Crook et al., 1992). Serialevaluations are required to assess the treatment effi-cacy of pharmacologic therapies and rehabilitationprograms, as well as the rate of disease progression, indiverse conditions such as Alzheimer's disease, headinjury, stroke, and so forth.

The results of the present investigation clearly dem-onstrate that learning occurs with repeated exposure toa procedure of a particular memory test, even when theactual stimuli to be remembered change. Conse-quently, the use of parallel forms of equivalent diffi-culty does not appear to be sufficient to eliminate prac-tice effects entirely. Similar results have also beenreported by Goldstein (1991), who suggested thatdouble-blind placebo control groups continue to be es-sential when conducting pharmaceutical trials.

It is notable that when our drug study ended and thepractice stopped, subjects tended to revert to their ini-tial levels of performance on most, but not all, mea-sures on long-term follow-up evaluation. Chelune(1991) has suggested that there may be a critical periodof elapsed time in neuropsychological testing afterwhich the size of practice effects can be expected to beminimized. Our results support this hypothesis. Itwould be of considerable interest and clinical utility todetermine the specific amount of time that must passafter which a test could be readministered without con-cern regarding practice effect artifact. Our results sug-gest that, for many measures, it is possible to retestafter a period of 3 years and be fairly confident thatpractice effects will be minimal. Future investigatorsare encouraged to determine whether briefer test-retest

intervals with minimal practice effects are possible.There are several potential weaknesses in our study.

For example, our participants were highly educatedand may not be representative of the general popula-tion. In addition, the literature has demonstrated a con-sistent selective attrition bias in which participantswho remain in longitudinal studies tend to have higherinitial levels of cognitive functioning than those whodo not (Mitrushina & Satz, 1991). Furthermore, thosesubjects who drop out during the course of longitudinalstudies tend to be declining at faster rates than thosewho do not (Cooney et al., 1988; Siepler et al., 1982).

Although initial differences between the long-termfollow-up responders and nonresponders in our inves-tigation appear to have been small, it is certainly pos-sible that the participants who did not come in for re-testing had greater longitudinal declines in theircognitive function than those who did come in. Thelong-term follow-up response rate of the subjects whooriginally had declined to participate in one of the drugstudies was low (17%), raising the chances of selectiveattrition bias. On the other hand, the response rate forthe drug study participants was much better (47%),increasing confidence in the representativeness of thisparticular group. The increased response rate of thedrug study participants perhaps reflects their greaterinvolvement with and commitment to research. Ofcourse, the ideal longitudinal study would entailfollow-up testing of all subjects (in which the attritionrate would be minimized).

Finally, our study was limited by relatively brief lon-gitudinal follow-up intervals. It would be of consid-erable interest to measure the cognitive function andself-report of normal adults over much greater timespans. Ideally, initial measurement would occur priorto the onset of AAMI, perhaps when subjects are in

Table 9Repeated Measures Analysis of Variance SummaryData Comprising Initial Scores With Scores atLong-Term Follow-Up in Drug Study Participants

Test





F

94.641.054.71

0.0029.81

0.073.30

df\, 1541, 1011, 129

1, 1171, 125

1, 931, 118

a

.001*

.309

.032*

.961

.001*

.789

.072

T)2

.38

.01

.04

.00

.19

.00

.03.05.


their 40s, with periodic assessment of the incidenceand prevalence of AAMI occurring throughout the re-maining natural life span.

References

Alder, A. G., Adam, J., & Arenberg, D. (1990). Individual-differences assessment of the relationship between changein and initial level of adult cognitive functioning. Psy-chology and Aging, 5, 560-568.

Bamford, K. A., & Caine, E. D. (1988). Does "benignsenescent forgetfulness" exist? Clinics in Geriatric Med-icine, 4, 897-915.

Bartus, R. T., Dean, R. L., & Beer, B. (1983). An evaluationof drugs for improving memory in aged monkeys: Impli-cations for clinical trials in humans. PsychophamacologyBulletin, 19, 168-184.

Bartus, R. T., Fleming, D., & Johnson, H. R. (1978). Agingin the Rhesus monkey: Effects on short-term memory.Journal of Gerontology, 33, 858-871.

Bartus, R. T., & Johnson, H. R. (1976). Short-term memoryin the Rhesus monkey: Disruption from the anti-cholinergic scopolamine. Pharmacology, Biochemistryand Behavioral Science, 5, 39^4-6.

Benton, A. L. (1974). The revised Visual Retention Test (4thed.). New York: Psychological Corporation.

Blackford, R. C., & LaRue, A. (1989). Criteria for diagnos-ing age-associated memory impairment: Proposed im-provements from the field. Developmental Neuropsychol-ogy, 5, 295-306.

Buschke, H. (1973). Selective reminding for analysis ofmemory and learning. Journal of Verbal Learning andVerbal Behavior, 12, 543-550.

Chelune, G. J. (1991, August). Impact of confidence inter-vals and baserate data on clinical interpretation ofchange scores. Paper presented at the 99th Annual Con-vention of the American Psychological Association, SanFrancisco, CA.

Cooney, T. M., Schaie, K. W., & Willis, S. L. (1988). Therelationship between prior functioning on personality andcognitive dimensions and subject attrition in longitudinalresearch. Journal of Gerontology, 43, P12-P17.

Crook, T. H., Bartus, R. T, Ferris, S. H., Whitehouse, P.,Cohen, G. D., & Gershon, S. (1986). Age-associatedmemory impairment: Proposed diagnostic criteria andmeasures of clinical change—Report of a National Insti-tute of Mental Health workgroup. Developmental Neu-ropsychology, 2, 261-276.

Crook, T. H., Tinklenberg, J., Yesavage, J., Petrie, W.,Nunzi, M. G., & Massari, D. C. (1991). Effects of phos-phatidylserine in age-associated memory impairment.Neurology, 41, 644-649.

Crook, T. H., & West, R. L. (1990). Name recall perfor-mance across the adult life-span. British Journal of Psy-chology, 81, 335-349.

Crook, T. H., Youngjohn, J. R., & Larrabee, G. J. (1990).

The Misplaced Objects Test: A measure of everyday vi-sual memory. Journal of Clinical and Experimental Neu-ropsychology, 12, 819-833.

Crook, T. H., Youngjohn, J. R., & Larrabee, G. J. (1992).Multiple equivalent forms in a computerized, everydaymemory battery. Archives of Clinical Neuropsychology, 7,221-232.

Flicker, C., Ferris, S. H., Crook, T., & Bartus, R. T. (1987).A visual recognition memory test for the assessment ofcognitive function in aging and dementia. ExperimentalAging Research, 13, 127-132.

Folstein, M. P., Folstein, S. E., & McHugh, P. R. (1975)."Mini-mental state": A practical method for grading thecognitive state of patients for the clinician. Journal ofPsychiatric Research, 12, 189-198.

Fozard, J. L. (1985). Psychology of aging-normal andpathological age differences in memory. In J. C. Brock-lehurst (Ed.), Textbook of geriatric medicine and geron-tology (pp. 122-144). Edinburgh, Scotland: Churchill/Livingstone.

Goldstein, G. (1991, August). Practice effect phenomena ina national hypertension study. Paper presented at the 99thAnnual Convention of the American Psychological Asso-ciation, San Francisco, CA.

Hertzog, C., & Schaie, K. W. (1988). Stability and change inadult intelligence: Simultaneous analysis of longitudinalmeans and covariance structures. Psychology and Aging,3, 122-130.

Krai, V. A. (1962). Senescent forgetfulness: Benign andmalignant. Journal of the Canadian Medical Association,86, 257-260.

Krai, V. A. (1966). Memory loss in the aged. Diseases of theNervous System, 27, 51-54.

Lane, F., & Snowdon, J. (1989). Memory and dementia: Alongitudinal survey of suburban elderly. In P. Lovibond &P. Wilson (Eds.), Clinical and abnormal psychology (pp.365-376). New York: Elsevier Science.

Larrabee, G. J., & Crook, T. H. (1989). Dimensions ofeveryday memory in age-associated memory impairment.Psychological Assessment: A Journal of Consulting andClinical Psychology, 1, 92-97.

Larrabee, G. J., Levin, H. S., & High, W. M. (1986). Se-nescent forgetfulness: A quantitative study. Developmen-tal Neuropsychology, 2, 373-385.

Mitrushina, M., & Satz, P. (1991). Changes in cognitivefunctioning associated with normal aging. Archives ofClinical Neuropsychology, 6, 49-60.

Poon, L. W. (1985). Differences in human memory withaging: Nature, causes, and clinical implications. In J. E.Birren & K. W. Schaie (Eds.), Handbook of the psychol-ogy of aging (2nd ed., pp. 427^162). New York: VanNostrand Reinhold.

Siepler, I. C., McCarty, S. M., & Logue, P. E. (1982).Wechsler Memory Scale scores, selective attrition,and distance from death. Journal of Gerontology, 37,176-181.


Smith, G. E., Ivnik. R. J., Petersen, R. C., Malec, J. F.Kokmen, E., & Tangalos, E. (1991). Age-associated mem-ory impairment diagnoses: Problems of reliability, con-cerns for terminology. Psychology and Aging, 6, 551-558.

Wechsler, D. (1945). A standardized memory scale for clin-ical use. Journal of Ps\cholog\; 19, 87-95.

Wechsler, D. (1955). Wechsler Adult Intelligence Scale[manual]. New York: Psychological Corporation.

West, R. L., & Crook, T. H. (1990). Age differences ineveryday memory: Laboratory analogues of telephonenumber recall. Psychology and Aging, 5, 520-529.

Yesavage, J., Brink, T., Rose, T., Lum, O., Huang, O., Adey,V., & Leirer, V. (1983). Development and validation of ageriatric depression scale: A preliminary report. Journalof Psychiatric Research, 17, 37-49.

Youngjohn, J. R., & Crook, T. H. (in press). Learning,forgetting, and retrieval of everyday material across theadult life span. Journal of Clinical and ExperimentalNeuropsychology.

Youngjohn, J. R., Larrabee, G. J., & Crook, T. H. (1991).First-Last Names and the Grocery List Selective Remind-ing Tests: Two computerized measures of everyday verballearning. Archives of Clinical Neuropsvchologv, 6, 287-300.

Youngjohn, J. R., Larrabee, G. J., & Crook, T. H. (1992a).Discriminating age-associated memory impairmentand Alzheimer's disease. Psychological Assessment, 4,54-59.

Youngjohn, J. R., Larrabee, G. J., & Crook, T. H. (1992b).Test-retest reliability of computerized, everyday memorymeasures and traditional memory tests. The Clinical Neu-ropsychologist, 6, 276-286.

Received August 24, 1992Revision received December 14, 1992

Accepted December 15, 1992

Documents

Stability of Everyday Memory in Age-Associated … of Everyday Memory in Age-Associated Memory Impairment: A Longitudinal Study James R. Youngjohn and Thomas H. Crook III Everyday