Demographically Corrected Norms For The California Verbal Learning Test

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Journal of Clinical and Experimental Neuropsychology

ISSN: 1380-3395 (Print) 1744-411X (Online) Journal homepage: http://www.tandfonline.com/loi/ncen20

Demographically Corrected Norms for the


California Verbal Learning Test

Marc A. Norman , Jovier D. Evans , Walden S. Miller & Robert K. Heaton

To cite this article: Marc A. Norman , Jovier D. Evans , Walden S. Miller & Robert K. Heaton
(2000) Demographically Corrected Norms for the California Verbal Learning Test, Journal of
Clinical and Experimental Neuropsychology, 22:1, 80-94

To link to this article: http://dx.doi.org/10.1076/1380-3395(200002)22:1;1-8;FT080

Published online: 09 Aug 2010.

Submit your article to this journal

Article views: 424

View related articles

Citing articles: 61 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=ncen20

Download by: [Deakin University Library] Date: 30 October 2015, At: 16:08
Journal of Clinical and Experimental Neuropsychology 1380-3395/00/2201-080$15.00
2000, Vol. 22, No. 1, pp. 80-94 © Swets & Zeitlinger

Demographically Corrected Norms for the California Verbal


Learning Test*
Marc A. Norman1, Jovier D. Evans2,3, S. Walden Miller2, and Robert K. Heaton2
Departments of 1Neurosciences and 2 Psychiatry , University of California, San Diego,
and 3 Indiana University Purdue University, Indianapolis

ABSTRACT
Downloaded by [Deakin University Library] at 16:08 30 October 2015

The California Verbal Learning Test (CVLT) is designed to quantify components of verbal learning, reten-
tion and retrieval. The present study used multiple regression analyses to correct for demographic charac-
teristics on CVLT performance measures. There were 906 subjects, of whom 549 were Caucasians (61%)
and 357 were African Americans (39%). Age, education, ethnicity, and gender were found to be significant
predictors of performance on several CVLT indices, including Total Words Recalled, Trial 1, Trial 5, List
B, Short Delay Free Recall (SDFR), and Long Delay Free Recall (LDFR). Demographically corrected T-
scores were calculated for a base sample of 672 subjects and cross-validated on 234 separate subjects.
Tables and regression equations are offered to convert raw scores into T-scores corrected for age, gender,
education, and ethnicity. Demographically corrected Recognition Discriminability cutoff scores were
calculated for age and education levels. In order to provide some indices of important memory processes,
we also computed indices of retrieval, Short-Delay forgetting and Long-Delay forgetting and present
normative information for them.

Memory impairment is a common neurobeha- with each of the four categories of words in List
vioral abnormality associated with brain dys- A; this is the Short Delay Cued Recall Trial
function, so memory testing is an especially im- (SDCR). After a 20–min delay, Long Delay
portant aspect of most neuropsychological eval- Free– (LDFR) and Long Delay Cued Recall
uations (Delis, Cullum, Butters, Cairns, & Pri- (LDCR) Trials are administered. Finally, the
fitera, 1988; Wiens, Tindall, & Crossen, 1994). subject is read a list of 40 shopping items and is
The California Verbal Learning Test (CVLT; asked to indicate whether or not each item was
Delis, Kramer, Kaplan, & Ober, 1987) is an ex- on List A, providing a measure of Recognition
amination of verbal list acquisition and memory, memory.
composed of 16 shopping items from four cate- The CVLT was developed not only to quan-
gories. The 16 shopping items on List A are read tify learning and memory abilities, but also to
to the subject over five sequential trials, with the qualitatively characterize components of learn-
subject recalling as many items as possible after ing and memory. The scoring system allows for
each trial. After Trial 5, the subject is presented analysis of overall learning, as measured by the
with List B, an interference trial using a differ- total words recalled across Trials 1–5, as well as
ent list of 16 items. Following a recall trial of enabling potentially important comparisons in-
List B, the subject is again asked to recall the volving Trial 1 versus Trial 5 and List B, SDFR
words from List A; this is the Short Delay Free versus Trial 5 and SDCR, Trial 5 and LDFR,
Recall Trial (SDFR). The subject is then cued and SDFR and LDFR. The subject’s Learning

*
This work was funded in part by National Institute of Mental Health Grant MH49550–01A1.
Address correspondence to: Marc A. Norman, Ph.D., Psychology service (116B), VA Medical Center, 3350 La
Jolla Village Drive, San Diego, CA 921226, USA. E-mail: mnorman@ucsd.edu
Accepted for publication: August 11, 1999.
CVLT NORMS 81

Characteristics are further clarified in indices of not provide data regarding performances of the
Semantic Clustering, Serial Clustering, Percent separate ethnic groups, although analyses of
Primacy Recall, Percent Middle Recall, Percent possible ethnicity effects reportedly did not
Recency Recall, Learning Slope, and Recall yield significant results. Limitations of this
Consistency. Also scored are error types includ- study are that the subjects were relatively young
ing the number of Perseverations and Intrusions (M = 29.1), well educated (M = 14.5 years), and
on the Free Recall and Cued Recall trials. Fi- mostly male (81%), and that quite small num-
nally, the number of Recognition Hits, Dis- bers of subjects were included within individual
criminability, False Positive responses, and Re- ethnic minority groups. Only 6% of the popula-
sponse Bias characterize Recognition memory. tion were over the age of 40. This study did not
Comparisons of these measures with LDFR re- offer comprehensively presented corrections for
flect the presence/absence and degree of diffi- age and gender, which were found to signifi-
culty with retrieval processes. cantly contribute to CVLT performance.
Downloaded by [Deakin University Library] at 16:08 30 October 2015

The commonly used norms presented in the To extend the CVLT norms for older individ-
CVLT manual (Delis et al., 1987) consisted of a uals, Paolo, Troster, and Ryan (1997) used over-
combination of several independently collected lapping age ranges to provide data on the 26
subject samples. It included 273 (104 male, 169 CVLT indices for persons between 53 and 83. In
female) neurologically intact subjects. These contrast to the Wiens et al. (1994) study, Paolo
norms have been described as limited because et al. presented clinically useful norms tables.
they are based upon a relatively small sample However, this normative sample is even more
size with less than desirable representativeness, limited regarding inclusion of ethnic minorities
and thus caution is advised in their clinical use as only 4 of the 212 subjects were minorities (2
(Delis et al., 1987; Elwood, 1995). Information African Americans and 2 Hispanics).
regarding ethnicity was not analyzed in the In general, previous CVLT normative studies
CVLT norming process, but ethnic minorities have yielded consistent findings with regard to
are believed to be poorly represented in this gender. Gender has been reported to signifi-
sample (Kramer, personal communication, No- cantly contribute to CVLT performance, with
vember, 1998). The lack of information regard- females generally performing better on immedi-
ing the contribution of ethnicity may lead to ate and delayed free recall trials, as well as on
misclassification of memory impairment, espe- semantic clustering (Delis et al., 1987; Kramer,
cially within ethnic minority populations. Delis, & Daniel, 1988). Delis et al. (1987) and
Most existing studies utilizing the CVLT Kramer et al. (1988) found that males and fe-
compare cerebrally compromised populations males did not differ significantly on Recognition
and normal controls, reporting means and stan- Hits. In contrast, Paolo (1997) and Wiens (1994)
dard deviations; however, they fail to analyze found gender differences favoring females on
demographic contributors to CVLT performan- Recognition Hits, False Positive errors, and
ce. Without knowledge of possible demographic Discriminability, as well as on free recall trials.
contributors, subjects may appear ‘‘impaired’’ In addition to gender, age consistently has
when they are not. In an attempt to expand the been found to significantly influence CVLT per-
normative data, Wiens, Tindall, and Crossen formance. Delis et al. (1987) and Wiens et al.
(1994) analyzed CVLT performance of 700 (568 (1994) found that subjects’ CVLT performances
male, 132 female) successive job applicants for generally worsened with advancing age. How-
the Civil Service. This sample, including 596 ever, not all measures of performance may be
Caucasians (85%), 39 African Americans influenced by advancing age. For example, Kra-
(5.6%), 30 Asians (4.3%), 24 Hispanics (3.4%), mer, Blusewicz and Preston (1989) found that
and 11 Native Americans (1.6%), performed performances on Trial 1 and recognition were
slightly lower on the number of words they re- essentially unrelated to age.
called on the learning trials when compared to Reports on the effects of education on CVLT
the Delis et al. (1987) sample. The authors did performance have been less consistent. Paolo et
82 MARC A. NORMAN ET AL.

al. (1997) found that education affected only cal populations. All participants were screened to
14% of the CVLT indices and always contrib- exclude those with any history of significant
uted the least in a multiple regression paradigm. trauma or disease involving the brain. Specifically,
in all contributing studies exclusion criteria in-
Delis et al (1987) and Wiens et al. (1994) did
cluded (a) meeting DSM-III-R criteria for sub-
not address the effects of education on CVLT stance dependence within the last year; (b) current
performance. therapeutic use of psychoactive drugs; (c) history
Although ethnicity may influence neuropsy- of serious mental illness (i.e., schizophrenia, bipo-
chological test performance, little effort has lar disorder, or major depression); (d) history of
been put forth to characterize neuropsychologi- head trauma with loss of consciousness greater
cal test performance for minority groups (Manly than 5-minutes and/or any head injury which re-
sulted in persisting neurological signs; and (e) his-
et al., 1998). Roberts and Hamsher (1984), using tory of diagnosed brain disease or brain surgery.
a standard cutoff, found that 22% of African Two hundred and eighty six subjects from the
Americans would have been misclassified as
Downloaded by [Deakin University Library] at 16:08 30 October 2015

original normative sample for the CVLT were used


impaired on their naming performance. Ethnic in the present analysis (32% of the current study’s
group differences have been demonstrated on total sample). Some of these subjects were ex-
the WAIS-R (Kaufman, McLean, and Reynolds, cluded from the original normative analysis be-
1988), Category Test (Bernard, 1989), and tasks cause they were especially young (<17 years) or
old (>80). 339 African American volunteers who
of simple and choice reaction time (Miller, were seen specifically to generate normative data
Bing, Selnes, Wesch, and Becker, 1993). Pre- on an expanded Halstead-Reitan neuropsychologi-
liminary data (Norman, Evans, Miller, Delis, cal test battery were also included in the sample
and Heaton, 1997; Evans et al., 1999) showed (37%). Subjects in this normative study were se-
that African American subjects performed out- lected from the San Diego, California community
side of the standards set by Caucasian normative using a census based survey method designed to
select a representative sample of neurologically
subjects on many of the CVLT indices. The
normal subjects from the area. In addition, normal
presence and severity of neurocognitive deficits comparison subjects were used from ongoing re-
may be overestimated within minority popula- search studies of various neurological and psychi-
tions due to nonrepresentative normative data. atric disorders at the University of California, San
The present large scale normative study was de- Diego School of Medicine Geriatric Psychiatry
signed to assess the contribution of influences of Clinical Research Center (n = 141; 16%), Alzhei-
ethnicity as well as other demographic factors mer’s Disease Research Center (n = 59; 7%), and
the HIV Neurobehavioral Research Center (n = 81;
on CVLT performance. Additionally, predicted
9%). All subjects were able to speak and under-
discriminability and predicted Short-Delay re- stand both written and spoken English, and were
call and Long-Delay recall were calculated and paid for their participation in the component stud-
used to generate a Retrieval Index as well as ies. The combined sample of participants for the
indices of Short-Delay Forgetting and Long-De- current study totaled 906.
lay Forgetting. The subjects in the combined group ranged in
age from 17 to 101 (M 51.1, SD = 20.6). About
half (55.7%) were female. 61% (n = 549) were
non-Hispanic Caucasian, and 39% (n = 357) were
METHODS African American. Subjects in other ethnic minor-
ity groups were excluded because they were not
Subjects available in sufficient numbers to provide valid
Subjects included in the present investigation were estimates of performances of the minority popula-
normal, community dwelling volunteers recruited tions they represent. In all, 362 (40%) of our sub-
from five separate research studies. One of these jects were age 60 or older. Education ranged from
studies was specifically designed to develop neu- 1 to 20 years (M = 13.72, SD = 2.62). 29.8% of
ropsychological norms for African Americans subjects had a college education or better (16+
(Miller et al. in press); another was the original years), 29.5% had some college (13-15 years),
CVLT normative study (Delis et al., 1987), while 25.3% had a high school education (12 years), and
the remaining three studies collected normal con- 15.5% had fewer than 12 years of formal educa-
trol data for comparison with that of various clini- tion.
CVLT NORMS 83

Education and age distributions were different data from the Base sample (see Appendix). The
in our African American and Caucasian samples, dependent variables used were the measures de-
due to sampling differences between the various scribed above, transformed into normally-distrib-
studies from which these subjects were drawn. uted scaled scores (M = 10, SD = 3; see Table 2).
Specifically, Caucasian subjects were older (M = Stepwise multiple regressions, predicting the indi-
60.3, SD = 18.9, range = 17-101) than African vidual scaled score from demographic variables
American subjects (M = 36.9, SD =14.1, range = (age, education, gender, and ethnicity) as well as
17-65, p < .001), and reported a higher mean level non-linear components (age-squared, age-cubed,
of education (M = 14.1 years, SD = 2.7, range = 1- education-squared, education-cubed) and interac-
24) than the African American sample (M = 13.2 tion terms (age by education, age by gender, age
years, SD = 2.4, range = 9-20; p < .001). Although by ethnicity for example) were performed to screen
most participants had at least 12 years of educa- for variables that did not significantly and
tion, 24.9% of the African American sample had uniquely contribute to the prediction equation;
not graduated from high school, compared to only these were then removed. To be considered signifi-
Downloaded by [Deakin University Library] at 16:08 30 October 2015

9.3% of the Caucasian sample. cant, these variables had to contribute at least one
percent of the variance of the scaled scores in the
Measures base sample. Multiple regression analyses with all
remaining demographic variables were then re-en-
California Verbal Learning Test (CVLT) tered in a stepwise fashion to predict the various
Each subject completed the California Verbal CVLT scaled scores. This prediction equation was
Learning Test (CVLT), administered in the stan- then used to generate predicted CVLT scores for
dard manner by trained psychometrists as part of a each subject. These predicted scores were then
larger neuropsychological examination. The fol- subtracted from each subject’s actual score to cal-
lowing raw score measures were used to develop culate a residual score. Finally, these residual
demographically corrected T-scores: Monday List scores were transformed to T-scores (M = 50, SD =
Trial 1 recall, Monday List Trial 5 recall, Monday 10) according to the following formula: T-score =
List Total Trials 1-5, Tuesday List (List B) recall, {[(residual score/standard error of estimate for the
and Monday List Short Delay Free Recall and regression equation) x 10] + 50} (Heaton et al.,
Long Delay Free Recall. The distributional proper- 1991).
ties of the Recognition Discriminability measures Following the procedures outlined in Heaton et
were significantly skewed; however, impairment al. (1991) ages between 17 and 34 were coded as
cut-off scores were calculated with age and educa- 34 in the present analysis. This was due to the fact
tion corrections. that there were no apparent age effects seen in neu-
ropsychological performance in early adulthood,
Analyses and using actual age in the younger group might
attenuate the linear relationships seen between age
Base and Validation subject samples and test scores in the remaining (older) sample.
Generation of normative equations generally fol- The current study supported the hypothesis that no
lowed procedures described in Heaton, Grant, and linear or curvilinear age effects were noted prior to
Matthews (1991). First, the subject sample was age thirty-four.
split into Base (n = 672) and Validation (n = 234) We then applied the T-score conversions to the
subsamples, using a random selection procedure. data from the Validation sample. To determine
Due to the fact that relatively few African Ameri- whether the demographic influences in the Valida-
cans in the entire sample were over the age of 60 tion sample CVLT-scores had been removed, we
(n = 24), all of these older subjects were placed in computed bivariate correlations between the com-
the Base sample to improve the resulting predic- puted CVLT T-scores and demographic variables
tion equations’ generalizability with regard to age (age, education, gender, and ethnicity) in both the
effects on CVLT performance. To verify that the Base and Validation samples. In addition, to deter-
resulting groups were comparable on demographic mine whether these T-score conversions were
characteristics, simple t tests and Chi-square anal- functioning equivalently at various levels of the
yses compared the groups on demographic charac- demographic variables, we split the subjects into
teristics (age, education, and gender). three age groups (<40, 40 – 60, 60 +) and three
education groups (<12, 12-15, 16+), and computed
Generation and Validation of Prediction Equa- ANOVAs to look for significant main effects of
tions age, education, gender, or ethnicity, as well as sig-
Next, we generated prediction equations using the nificant interaction effects on CVLT T-scores
84 MARC A. NORMAN ET AL.

which would indicate that the demographic correc- Development of the Prediction Equation and
tions were not sufficient or were not equivalent Cross-Validation
across various levels of the demographic variables. The raw to scaled score conversions are pre-
In addition, the Kolmogorov-Smirnov procedure
sented in Table 2. Stepwise multiple regressions
was used to determine if the resulting T-score dis-
tributions differed significantly from normality. in the Base sample were used to predict the vari-
Measures of predicted recognition discrimina- ous CVLT scaled scores from demographic vari-
bility performance and rates of predicted SDFR ables (age, education, gender, and ethnicity).
and LDFR were then calculated. Predicted Recog- Nonlinear components (e.g., age squared, age
nition Discriminability performance was calcu- cubed, education squared, education cubed, and
lated in a regression formula with LDFR as the the interaction terms) were analyzed, but did not
predictor variable. The predicted rate of short de-
lay recall was calculated using the Trial 5 raw
significantly contribute to the prediction equa-
score as the predictor variable and SDFR raw score tions. The basic demographic variables all con-
tributed significantly to prediction of CVLT per-
Downloaded by [Deakin University Library] at 16:08 30 October 2015

as the dependent variable in a regression formula.


The same calculation was then applied with Trial 5 formance; together they accounted for 18.8% of
to predict LDFR. Next, discrepancy scores were the variance on Trial 1 scaled scores (F(4, 667)
obtained by subtracting predicted scores from the = 39.84, p <.0001); 29.3% for Trial 5 (F(4, 667)
obtained scores to estimate retrieval, Short-Delay = 70.6, p <.0001); 31.9% for Total Words Trials
forgetting, and Long-Delay forgetting. These dis-
crepancy scores were then converted to percentile 1-5 (F(4, 667) = 79.4, p <.0001); 17.2% for List
ranges. B (F(4, 667) = 35.8, p <.0001); 26.4% for Short
Delay Free Recall (F(4,667) = 61.3, p <.0001);
and 27.4% Long Delay Free Recall (F(4, 667) =
RESULTS 64.2, p <.0001). The prediction equations are
presented in Appendix.
Base and Validation Subject Samples and Application of the T-score formulae to the
CVLT Raw Scores Validation sample demonstrated the generali-
Table 1 summarizes the demographic variables zability of the derived equations. As expected,
and CVLT raw scores for the Base and Valida- the mean T-scores did not differ significantly
tion samples. The two samples did not differ on from 50 (see Table 3). Table 4 shows the fre-
age, education, or gender. However, there was a quency distributions of the T-scores on Total
slightly greater percentage of African American Words Recalled on Trials 1 to 5, for the Base,
subjects in the Base sample due to our sampling Validation, and Total Samples, as compared to
adjustment. In addition, there were no differ- what would be expected from a normal distribu-
ences noted in CVLT performance across the tion. Neither sample differed significantly from
two samples. predictions based on assumptions of a normal
In the combined group, there were significant distribution (Kolmogorov-Smirnov p > .05).
demographic influences on the raw score CVLT Similar results were obtained for all other CVLT
performances. A 3 (age group) × 3 (education T-score distributions; that is, none differed sig-
group) × 2 (ethnicity) × 2 (gender) analysis of nificantly form normal for the base, validation,
variance (ANOVA) revealed significant main or total samples.
effects of age, education, gender, and ethnicity Table 5 presents the proportion of all subjects
on the total words recalled for Trials 1-5, Trial within clinically relevant T-score ranges for To-
1, Trial 5, List B, Short Delay Free Recall, and tal Words Recalled on Trials 1-5, as suggested
Long Delay Free Recall, such that younger per- by Heaton et al. (1991). In that study, a one-
sons, those with more advanced education, Cau- standard deviation cutoff was used to define
casians, and females tended to perform better. In ‘‘impairment’’ which results in an expected
addition, a significant education by ethnicity 84.4% specificity rate in neurologically normal
interaction was found only on Long Delay Free subjects. The total sample had an ‘‘impairment’’
Recall (p = .03). rate (T-score less than 40) of 16.7%, which is
very close to what would be predicted on the
Downloaded by [Deakin University Library] at 16:08 30 October 2015

Table 1. CVLT Raw Scores.

Age n Education Total Words Trial 1 Trial 5 List B SDFR LDFR


(years) (years)

M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) M (SD)


Base Sample
< 40 222 13.38 (2.45) 56.13 1(9.45) 7.76 (2.06) 13.08 (2.16) 7.26 (2.16) 11.69 (2.69) 12.05 (2.78)
40-50 184 13.90 (2.94) 52.42 (10.20) 7.26 (2.05) 12.28 (2.37) 6.60 (2.15) 10.77 (3.08) 11.02 (3.09)
60+ 266 13.79 (2.56) 46.37 (11.11) 6.29 (2.13) 11.06 (2.69) 6.07 (2.07) 19.00 (3.12) 19.35 (3.25)
Total 672 13.69 (2.64) 51.25 (11.14) 7.04 (2.18) 12.06 (2.58) 6.61 (2.18) 10.38 (3.19) 10.70 (3.27)
Validation Sample
<40 190 13.74 (2.30) 55.47 1(8.75) 7.41 (1.94) 13.33 (1.85) 7.01 (2.01) 11.83 (2.76) 12.29 (2.80)
40-50 148 14.27 (2.20) 54.60 (10.72) 8.17 (2.18) 12.19 (2.55) 7.23 (2.24) 10.63 (3.27) 11.19 (3.24)
60+ 196 13.60 (2.96) 46.31 (12.14) 6.43 (2.39) 11.04 (2.67) 5.85 (2.22) 18.90 (3.64) 19.31 (3.89)
Total 234 13.79 (2.58) 51.53 (11.47) 7.16 (2.27) 12.16 (2.56) 6.58 (2.26) 10.38 (3.49) 10.84 (3.61)

Note. SDFR = Short Delay Free Recall; LDFR = Long Delay Free Recall.
CVLT NORMS
85
86 MARC A. NORMAN ET AL.

Table 2. Raw Score Conversion to Scale Score Conversions for CVLT Variables.

Scaled Score Total Words Trial 1 Trial 5 List B SDFR LDFR


19
18 >13
17 >71 12 >12
16 69 – 70 11 11 16 16
15 67 – 68 10 16 10 15 15
14 63 – 66 15 9 14
13 60 – 62 9 14 13 14
12 57 – 59 8 8 12 13
11 53 – 56 13 7 11 12
10 49 – 52 7 12 6 10 10 – 11
9 45 – 48 6 11 5 9 9
Downloaded by [Deakin University Library] at 16:08 30 October 2015

8 41 – 44 10 8 8
7 37 – 40 5 9 4 6–7 6–7
6 32 – 36 4 8 5 5
5 28 – 31 3 7 3 3–4 3–4
4 24 – 27 1–2 6 2
3 17 – 23 5 1–2 2
2 14 – 16 0 4 1 0–1
1 <13 0–3 0 0

Note. SDFR = Short Delay Free Recall; LDFR = Long Delay Free Recall.

basis of the theoretical normal distribution (i.e., Validation of the Prediction Equation
15.6%). The proportions of subjects within clini- In order to determine if the influence of demo-
cally relevant T-score ranges for Trial 1, Trial 5, graphic factors had been removed, and to assess
List B, SDFR, and LDFR are presented in Table whether the corrections were functioning equiv-
5. alently at different levels of the demographic
variables, a series of 3 (age group) × 3 (educa-
tion group) × 2 (gender) × 2 (ethnicity)

Table 3. T Scores and Standard Deviations for Base-, Validation-, and Total Samples.

Base Validation Total


(n = 672) (n = 234) (n = 906)

M (SD) M (SD) M (SD)


Age (years) 51.21 (20.31) 53.32 (18.28) 51.06 (20.63)
Education (years) 13.69 1(2.64) 13.79 1(2.58) 13.72 1(2.62)
Total Words 50.00 1(9.97) 49.66 (10.12) 50.07 (10.00)
Trial 1 50.00 1(9.97) 50.27 (10.12) 49.92 1(9.94)
Trial 5 50.00 1(9.97) 49.67 1(9.85) 49.91 (10.01)
List B 50.00 1(9.97) 49.36 (10.14) 49.84 (10.01)
SDFR 50.00 1(9.97) 49.48 (10.77) 49.86 (10.18)
LDFR 50.00 1(9.97) 50.14 (11.46) 50.04 (10.37)

% Female 55.5 56.4 55.7


% Caucasian 58.6 66.2 60.6
CVLT NORMS 87

Table 4. Predicted And Actual Percentages Of T Scores For Total Words Recalled For Trials 1-5 Within Seven
T Score Ranges.

T score range Predicted % Base Validation Total


0-24 10.62 10.3 10.9 10.4
25-34 16.05 16.5 17.7 16.8
34-44 24.15 23.8 22.2 23.4
45-54 38.29 39.6 42.3 40.3
55-64 24.20 23.4 19.7 22.4
65-74 16.07 15.5 16.8 15.8
75+ 10.62 10.9 10.4 10.8
Downloaded by [Deakin University Library] at 16:08 30 October 2015

ANOVAs were conducted on the T-scores in gender, education, and ethnicity. After the eth-
both the Validation and total samples. There nicity correction was applied to the equation,
were no significant main or interaction effects, 17.78% of the African American subjects were
indicating that the T-score conversions were classified as impaired, approximating what
similarly effective at all levels of age, education, would be expected within a neurologically nor-
gender, and ethnicity. mal population; that is, these results do not dif-
Finally, in order to more fully examine the fer significantly from the expected false positive
diagnostic accuracy of the demographically cor- error rate of 15.6% (P2 (1, N = 357) = 1.14 , p =
rected T-score formulae, we compared the spec- .29).
ificity rates in the African American subsample, Table 6 shows the results of these compari-
with and without ethnicity adjustments (i.e., sons for all six CVLT measures. In each case,
without adjustments means treating the African the percent of African Americans classified as
American subjects as if they were Caucasian). impaired differs significantly from expectations
First, when we applied a standard cutoff of less and is unacceptably high, when ethnicity correc-
than a T-score of 40 to the Delis et al. (1987) tions are not applied. After ethnicity corrections,
norms, 46% of the African American subjects however, the false positive error rates do not
were classified as impaired based on the total differ significantly from expectations based, on
number of words they recalled on Trials 1-5. a one standard deviation cutoff with a normal
This number is significantly greater than what distribution, on five of the six measures. The
would be expected (e.g. 15.6%) within the nor- only exception was on Short Delay Free recall.
mal distribution (P2 (1, N =357) = 249.57, p < Although the false positive rate for this measure
.001). In comparison, when we applied the Delis exceeded normal expectations by 5.7 percent,
et al. (1987) norms to our Caucasian sample that this still represents a substantial improvement
was not part of the original normative group, over the 15.7 percent with the uncorrected score.
only 16.2% were classified as impaired (P2 (1, N Recognition discriminability was signifi-
= 281) = 1.39, p = .24). cantly skewed; however demographic variables
When our own corrections for age, gender, (e.g. age, education, gender, and ethnicity) were
and education, but not ethnicity were applied, examined to assess their relative contribution to
36% of the African American subjects were discriminability performance. Gender (r =
classified as impaired. Again, this is signifi- –.094) and ethnicity (r = .033) accounted for
cantly greater than would be expected under the less than one percent of the variance, but age
normal curve (P2 (1, N = 357) = 108.18, p < and education accounted for 10% and 2 % of the
.001). These results clearly demonstrate the un- variance, respectively. Table 7 presents Recog-
acceptable false positive error rates for African nition Discriminability impairment cutoff
Americans on CVLT without correcting for age, scores, using one standard deviation, were cal-
88 MARC A. NORMAN ET AL.

Table 5. Predicted and actual percentages of T scores within clinically relevant ranges for Trials 1 – 5, Trial 1, Trial 5, List
B, SDFR, and LDFR.

T Score Range Classification Predicted Base Validation Total

Trials 1 – 5 0 – 19 Severe 0.13 0.1 0.0 0.1


20 – 24 Moderate to Severe 0.49 0.1 0.9 0.3
25 – 29 Moderate 1.60 1.8 0.9 1.5
30 – 34 Mild to Moderate 4.45 4.8 6.8 5.3
35 – 39 Mild 8.95 10.0 8.1 9.5
40 – 44 Below Average 15.20 13.8 14.1 13.9
45 – 54 Average 38.29 39.6 42.3 40.3
55 + Above Average 30.89 29.8 26.9 29.0
0 – 39 Total Impaired 15.60 16.8 16.7 16.7

Trial 1 0 – 19 Severe 0.13 0.0 0.0 0.0


20 – 24 Moderate to Severe 0.49 0.1 0.4 0.2
Downloaded by [Deakin University Library] at 16:08 30 October 2015

25 – 29 Moderate 1.60 0.7 0.4 0.7


30 – 34 Mild to Moderate 4.45 4.8 5.6 5.0
35 – 39 Mild 8.95 11.8 9.0 11.0
40 – 44 Below Average 15.20 16.2 17.1 16.4
45 – 54 Average 38.29 36.3 33.3 35.5
55 + Above Average 30.89 30.1 34.2 31.4
0 – 39 Total Impaired 15.60 15.6 17.4 16.9

Trial 5 0 – 19 Severe 0.13 0.1 0.0 0.1


20 – 24 Moderate to Severe 0.49 0.7 0.0 0.6
25 – 29 Moderate 1.60 1.3 2.6 1.7
30 – 34 Mild to Moderate 4.45 4.3 4.3 4.3
35 – 39 Mild 8.95 9.7 12.4 10.4
40 – 44 Below Average 15.20 14.0 12.0 13.5
45 – 54 Average 38.29 40.0 39.3 39.8
55 + Above Average 30.89 29.8 26.5 29.7
0 – 39 Total Impaired 15.60 15.6 19.3 17.1

List B 0 – 19 Severe 0.13 0.3 0.9 0.4


20 – 24 Moderate to Severe 0.49 0.3 0.4 0.3
25 – 29 Moderate 1.60 1.9 0.4 1.5
30 – 34 Mild to Moderate 4.45 4.6 7.3 5.3
35 – 39 Mild 8.95 7.6 6.8 7.4
40 – 44 Below Average 15.20 13.4 15.8 14.0
45 – 54 Average 38.29 42.1 41.5 41.9
55 + Above Average 30.89 29.8 26.9 29.0
0 – 39 Total Impaired 15.60 15.6 14.7 14.9

SDFR 0 – 19 Severe 0.13 0.1 0.4 0.2


20 – 24 Moderate to Severe 0.49 0.3 0.9 0.4
25 – 29 Moderate 1.60 1.6 1.7 1.7
30 – 34 Mild to Moderate 4.45 4.6 6.8 5.2
35 – 39 Mild 8.95 10.7 12.0 11.0
40 – 44 Below Average 15.20 14.7 10.7 13.7
45 – 54 Average 38.29 35.0 33.8 34.7
55 + Above Average 30.89 32.9 33.8 33.1
0 – 39 Total Impaired 15.60 17.3 21.8 18.5

LDFR 0 – 19 Severe 0.13 0.3 0.0 0.2


20 – 24 Moderate to Severe 0.49 0.3 0.9 0.4
25 – 29 Moderate 1.60 1.2 2.1 1.4
30 – 34 Mild to Moderate 4.45 4.0 7.7 5.0
35 – 39 Mild 8.95 10.9 9.8 10.6
40 – 44 Below Average 15.20 14.4 12.8 14.0
45 – 54 Average 38.29 39.1 33.8 37.7
55 + Above Average 30.89 29.8 32.9 30.6
0 – 39 Total Impaired 15.60 16.7 20.5 17.6
CVLT NORMS 89

culated correcting for age (<40, 40 –60, and formance. Our large subject sample (n = 906)
60+) and education (<12, 12 – 15, and 16+) (Ta- not only included approximately equal numbers
ble 7). of males and females at various levels of age
and education, but also a large number of Afri-
Retrieval Index, Short-Delay Forgetting In- can Americans. Multiple regression analyses
dex, and Long-Delay Forgetting Index using a Base sample of 672 subjects revealed
Formulae for calculating predicted Recognition significant contributions of age, gender, educa-
Discriminability, predicted SDFR, and predicted tion, and ethnicity on the total words recalled on
LDFR are presented in the Appendix. These pre- the CVLT learning trials, as well as on Trial 1,
dicted values were then compared to the sub- Trial 5, List B, Short Delay Free Recall, and
jects’ obtained Discriminability, SDFR, and Long Delay Free Recall. Predication equations
LDFR values, to yield indexes of Retrieval, were then calculated to generate demographi-
Short-Delay Forgetting, and Long-Delay Forget- cally corrected T-scores. These T-scores were
Downloaded by [Deakin University Library] at 16:08 30 October 2015

ting. For the three index scores, age accounted successfully applied to a Validation sample of
for 2% of the variance for Short-Delay Forget- 234 separate subjects. The successful validation
ting (r = –.141) and Long-Delay Forgetting (r = of these T-scores supports the generalizability of
–.129). All other demographics accounted for these corrected CVLT performances at the vari-
less than one percent of the variance. Raw ous levels of all demographic variables. In addi-
scores were categorized into five percentile dis- tion to the demographically corrected T-scores,
tributions (e.g., > 16, 16th though 11th, 10th predicted performances of discriminability,
through 6th, 5th through 2nd, and less than or SDFR, and LDFR were calculated. These pre-
equal to 1st ) (Tables 8 and 9). Although the dicted values were then compared to obtained
three discrepancy distributions were normally values to yield measures of verbal retrieval,
distributed, it was decided that providing per- Short-Delay forgetting, and Long-Delay forget-
centile ranges would accurately reflect ranges of ting.
impaired performance. Two findings within this sample population
were consistent with previous research reported
in the CVLT manual (Delis et al., 1987), as well
DISCUSSION as in Wiens et al. (1994), and Paolo et al.
(1997): females generally performed better than
This study examined the contributions of age, males, and performance decreased with age. In
gender, education, and ethnicity on CVLT per- addition, however, in the present study educa-

Table 6. Percentage of the total African American Sample Classified as Impaired Based Upon Six CVLT Mea-
sures, With and Without Correction for Ethnicity.

Without ethnicity correction With ethnicity correction

% ‘Impaired’ P2 % ‘Impaired’ P2
Expecteda 15.6 15.6
Trials 1-5 35.6 108.18** 17.7 1.14
Trial 1 39.5 154.83** 14.0 .69
Trial 5 36.1 114.33** 15.1 .06
List B 31.9 72.33** 14.4 .47
SDFR 31.3 67.45** 21.3 8.77*
LDFR 31.7 69.87** 18.8 1.47

a
Chi square compares obtained percent with the 15.6 percent expected using a one standard deviation cutoff and
assuming a normal distribution.
* p = .003; ** p < .001.
90 MARC A. NORMAN ET AL.

Table 7. Impairment Cutoffsa for Recognition Discriminability Corrected for Age and Education.

Education

<12 years 12 – 15 years 16+ years


Age
< 40 89% 91% 93%
(n = 64) (n = 166) (n = 82)
40 – 59 84% 89% 91%
(n = 35) (n = 122) (n = 74)
60 – 69 73% 89% 89%
(n = 16) (n = 97) (n = 44)
70 – 79 84% 84% 84%
(n = 13) (n = 76) (n = 37)
Downloaded by [Deakin University Library] at 16:08 30 October 2015

80+ 84% 77% 84%


(n = 11) (n = 33) (n = 30)

a
Impairment cutoffs are based on the 15.6 percent expected using one standard deviation cutoff.

tion was found to significantly contribute to fects in the present study are due to a broader
CVLT performance. range of our subjects’ education levels (e.g., 140
Overall, the contribution of education to subjects with less than high school and 267 with
CVLT performance remains unclear. Education at least a Bachelor’s degree).
effects were not reported in the CVLT manual An important factor in the present investiga-
(Delis et al., 1987). Weins et al. (1994) did not tion was the ability of these norms to contribute
find that education significantly contributed to to the correct classification of normal memory
CVLT performance and Paolo et al. (1997) did performance within the African American popu-
not find consistent education effects in their lation. The contribution of ethnicity to neuro-
sample of elderly persons. In the latter study, psychological performance in this study was
education was significantly related to only three consistent with previous studies of other neuro-
of the twenty-nine CVLT variables investigated. psychological measures (Kaufman et al., 1988;
Three of these variables (e.g. Total Words, Trial Roberts & Hamsher, 1984). However, no prior
5, and SDFR) were analyzed in the current study has adequately addressed possible ethnic-
study, and in all cases education was found to ity effects on the CVLT. Where previous studies
significantly contribute to CVLT performance. have consistently found age and gender to sig-
It is possible that the more robust education ef- nificantly contribute to CVLT performance (De-

Table 8. Raw Score Conversion to Percentile Range for Retrieval Index.

Retrieval Indexa Percentile Rangeb


< 3.48 > 16
3.48 to 4.97 11 – 16
4.98 to 6.61 6 – 10
6.62 to 12.85 2–5
> 16.62 <1

a
Retrieval Index = obtained Discriminability (in Discriminability percent units) – predicted Discriminability
(Formula in Appendix).
b
Percent of people in the normative sample that obtained this discrepancy.
CVLT NORMS 91

Table 9. Raw Score Conversion to Percentile Ranges for Short-Delay Forgetting and Long-Delay Forgetting

Short-Delay Forgettinga Long-Delay Forgettingb Percentile Rangec


> –1.71 > –1.50 >16
–1.86 to –1.71 –2.31 to –1.50 11 – 16
–2.76 to –1.87 –2.55 to –2.32 6 – 10
–3.85 to –2.77 –3.55 to –2.56 2–5
< –3.86 < –4.40 <1

a
Short-Delay Forgetting = obtained SDFR (in raw score units) – predicted SDFR (Formula in Appendix)
b
Long-Delay Forgetting = obtained LDFR (in raw score units) – predicted LDFR (Formula in Appendix)
c
Percent of people in the normative sample that obtained this discrepancy.
Downloaded by [Deakin University Library] at 16:08 30 October 2015

lis et al., 1987; Paolo et al., 1997; Wiens et al., vide appropriate normative standards for each
1994), this study demonstrated the importance group with which the tests will be used. Re-
of ethnicity on CVLT performance. Consistent searchers should make this a priority.
with the results of Roberts and Hamsher (1984), To expand the CVLT’s clinical and research
our findings indicate that correcting for ethnicity utility, three new indices were explored. The
lowers the rate of normal African Americans Retrieval Index compared obtained Recognition
classified as impaired. When removing the cor- Discriminability with performance predicted
rection on the learning trials, African Americans from results on contemporaneous assessment of
were categorized as impaired over twice as fre- free recall. By comparing the obtained and pre-
quently as were Caucasians, and this discrep- dicted Recognition Discriminability values, it is
ancy was noted on the delayed memory trials as believed that researchers and clinicians will be
well. False positive error rates of this magnitude better able to quantify verbal retrieval deficits.
are unacceptable in any clinical work, and repre- Higher positive discrepancies indicate that an
sent a major disservice to African Americans. individual performed better on Recognition
Present findings also suggest that Afri- Discriminability than would be expected given
can–American subjects may frequently be inac- their LDFR performance. The disproportionate
curately characterized as impaired when using ability to correctly identify the words on Recog-
the published CVLT norms, which reportedly nition testing may reflect a deficit in verbal re-
are based upon a primarily Caucasian sample. It trieval. By comparing predicted performance to
is important to note that these norms, as well as obtained values it is believed that clinicians and
our own, can only be confidently applied to the researchers will be able to better assess free-re-
ethnic groups on which they were derived. Clini- call relative to recognition performance.
cians and researchers should generally not use Patterns of rapid forgetting have proven to be
norms based on one ethnic group with other eth- important in differentiating subcortical and cor-
nic populations without crossvalidation; alterna- tical cognitive and memory profiles (e.g., Par-
tively, if appropriate norms do not exist, avail- kinson’s Disease and Alzheimer’s disease; But-
able norms based on other groups should be ters, Delis,& Lucas, 1995). To date, the litera-
used with caution, and the limitations of the ture has lacked normative data measuring CVLT
norms should be acknowledged whenever inter- rates of forgetting. The current study attempted
pretations are offered. This is not only true for to estimate both rates of short-delay and long-
just the CVLT, but for neuropsychological tests delay forgetting by comparing actual perfor-
generally (Heaton, Ryan, Grant, & Matthews, mance on short-delay and long-delay free recall
1996). Further research is needed to clarify the with performance that was predicted on the basis
nature and causes of ethnic differences on this on scores on the last learning trial (Trial 5). It is
and other neuropsychological tests, and to pro- expected that the difference between the ob-
92 MARC A. NORMAN ET AL.

tained delayed recall and predicted delayed re- vised and California Verbal Learning Test: Con-
call should not deviate substantially from zero. vergence and divergence. The Clinical Neuropsy-
chologist, 2, 188-196.
When subjects’ actual performance is substan-
Delis, D. C., Kramer, J., Kaplan, E., & Ober, B.
tially below predicted values, an atypically rapid (1987). The California Verbal Learning Test. New
rate of forgetting may be present. Thus, these York: Psychological Corporation.
normative data may assist clinicians and re- Elwood. R. (1995). The California Verbal Learning
searchers in measuring rates of forgetting. test: Psychometric characteristics and clinical ap-
plication. Neuropsychology Review, 5, 173-201.
Special consideration needs to be given when Evans, J.D., Norman, M.A., Miller, S.W., Kramer,
interpreting the Retrieval Index and measures of J.H., Delis, D.C., & Heaton, R.K. (1999). Age-,
Short-Delay and Long-Delay Forgetting. These education-, gender-, and race-corrected norms for
discrepancy scores in particular need to be inter- the California Verbal Learning Test. Journal of the
preted in light of the overall pattern of perfor- International Neuropsychological Society, 5, 110.
Heaton, R. K., Grant, I., & Matthews, C. G. (1991).
mance. For example, if both the obtained and
Downloaded by [Deakin University Library] at 16:08 30 October 2015

Comprehensive Norms for an Expanded Halstead-


predicted scores fall within the normal range, Reitan Battery: Demographic Corrections, Re-
but the retrieval index falls within the ‘‘im- search Findings, and Clinical Applications.
paired’’ range, it is questionable whether or not Odessa, FL: Psychological Assessment Resources.
even a subtle retrieval deficit exists. Similar Heaton, R. K., Ryan, L., Grant, I., & Matthews, C. G.
(1996). Demographic influences on neuropsycho-
caution should be used in interpreting the forget- logical test performance. In I. Grant & K Adams
ting indices. Finally, caution should be used in (Eds.) Neuropsychological assessment of neuro-
using the age and education corrected discri- psychiatric disorders, 2nd Ed. (pp 141-163). New
minability cutoff scores because of the small York: Oxford University Press.
subgroup sizes for subjects with less than 12 Kaufman, A. S., McLean, J. E., & Reynolds, C. R.
(1988). Sex, race, residence, region, and education
years of education and subjects with greater than differences on the 11 WAIS–R subtests. Journal of
59 years of age. Clinical Psychology, 44, 231–248.
In conclusion, we again emphasize that the Kramer, J. H., Blusewicz, M., & Preston, K. (1989).
lack of normative data on minority populations The premature aging hypothesis: Old before its
has broad implications in making clinical as well time? Journal of Consulting and Clinical Psychol-
ogy, 57, 257-262.
as research interpretations. Individuals may be Kramer, J. H., Delis, D. C., & Daniel, M. (1988). Sex
misdiagnosed, scientific conclusions may be differences in verbal learning. Journal of Clinical
erroneously drawn, and improper social policies Psychology, 44, 907-915.
may be implemented. Practitioners and research- Manly, J. J., Miller, S. W., Heaton, R. K., et al.
ers are urged to carefully consider addressing (1998). The effect of African-American accultura-
tion of neuropsychological test performance in nor-
these issues in working with minority individu- mal and HIV-positive individuals. Journal of the
als, especially given that normative data is lack- International Neuropsychological Society, 4, 291-
ing for these populations. 302.
Miller, E., Bing, E., Selnes, O., Wesch, J., & Becker,
J. (1993). The effects of sociodemographic factors
on reaction time and speed of information process-
REFERENCES ing. Journal of Clinical and Experimental Neuro-
psychology, 15, 66.
Bernard, L. C. (1989). Halstead–Reitan neuropsycho- Miller, S. W., Heaton, R. K., Grant, I., & White, (in
logical test performance of black, Hispanic, and press). African American Norms Supplement to
white young adult males from poor academic back- Heaton Comprehensive Halstead-Reitan Battery
grounds. Archives of Clinical Neuropsychology, 4, Norms. Miami, FL: Psychological Assessment Re-
267–274. sources.
Butters, N., Delis, D.C., & Lucas, J.A. (1995). Clini- Norman, M. A., Evans, J. D., Miller, S. W., Delis, D.
cal assessment of memory disorders in amnesia C., & Heaton, R. K. (1997). Normative data on an
and dementia. Annual Review of Psychology, 46, African American population on the California
493 – 523. Verbal Learning Test. Journal of the International
Delis, D. C., Cullum, C. M., Butters, N., Cairns, P., & Neuropsychological Society, 31.
Prifitera, A. (1988). Wechsler Memory Scale-Re-
CVLT NORMS 93

Paolo, A. M., Troster, A. I., & Ryan, J. J. (1997). Cal- Sundberg, N., & Gonzales, L. (1981). Cross-cultural
ifornia Verbal Learning Test: Normative Data for and cross-ethnic assessment: Overview and issues.
the Elderly. Journal of Clinical and Experimental In P. McReynolds (ed.) Advances in psychological
Neuropsychology, 19, 220-234. assessment. ( Vol. 5) San Francisco, CA: Jossey-
Roberts, R. J., & Hamsher, K. D. (1984). Effects of Bass Publishers.
minority status on facial recognition and naming Wiens, A. N., Tindall, A. A. G., & Crossen, J. R.
performance. Journal of Clinical Psychology, 40, (1994). California Verbal Learning Test: A norma-
539-545. tive study. The Clinical Neuropsychologist, 8, 75-
Sattler, J. (1970). Racial experimenter effects in ex- 90.
perimentation, testing, interviewing, and psycho-
therapy. Psychological Bulletin, 73, 137-160.
Downloaded by [Deakin University Library] at 16:08 30 October 2015
94 MARC A. NORMAN ET AL.

APPENDIX

CVLT Norms Formulae

Demographically corrected T-scores for CVLT indices can be calculated as follows:

Trial 1 = 3.58(Trial 1 SS) – 3.12 + .323(age) – .369(education) + 3.35(gender) + 5.65(ethnicity).

Trial 5 = 4.04(Trial 5 SS) – 4.24 + .396(age) –.998(education) + 4.90(gender) + 5.99(ethnicity).

Trials 1-5 = 4.07(Trials 1-5 SS) – 10.8 + .443(age) –.748(education) + 5.08(gender) + 6.77 (ethnic-
ity).
Downloaded by [Deakin University Library] at 16:08 30 October 2015

List B = 3.82(List B SS) – 5.03 + .294(age) – .386(education) + 4.17(gender) + 5.799(ethnicity).

Short Delay Free Recall = 3.955(SDFR SS) – 4.13 + .372(age) – .815(education) + 4.48(gender) +
4.11(ethnicity).

Long Delay Free Recall = 4.17(LDFR SS) – 4.16 + .375(age) – .959(education) + 4.50(gender) +
4.42(ethnicity).

Predicted Discriminability = 1.492(LDFR Raw Score) + 77.268.

Predicted SDFR = 1.048(Trial 5 Raw Score) – 2.288.

Predicted LDFR = 1.048(Trial 5 Raw Score) – 1.930.

Note: SS = Scaled Score (see Table 2).

Discriminability = Obtained Recognition Discriminability

Age = true age (except that if age is 20-34, age is coded as 34).

Education = years of education successfully completed.

Gender: Female = 0, Male = 1

Ethnicity: Caucasian = 0, African American = 1.

You might also like