Professional Documents
Culture Documents
Convergent and Discriminant Validity of The CVLT (Dementia Version)
Convergent and Discriminant Validity of The CVLT (Dementia Version)
Convergent and Discriminant Validity of The CVLT (Dementia Version)
To cite this article: John L. Woodard , Felicia C. Goldstein , Vicky J. Roberts & Colleen McGuire
(1999) Convergent and Discriminant Validity of the CVLT (Dementia Version), Journal of Clinical
and Experimental Neuropsychology, 21:4, 553-558
Article views: 66
Download by: [University of California, San Diego] Date: 12 November 2015, At: 18:31
Journal of Clinical and Experimental Neuropsychology 1380-3395/99/2104-553$15.00
1999, Vol. 21, No. 4, pp. 553-558 © Swets & Zeitlinger
ABSTRACT
This study investigated the convergent and discriminant validity of the 9-item ‘‘dementia version’’ of the
California Verbal Learning Test (CVLT-9) in a sample of 130 geriatric patients evaluated for memory
complaints. Moderate correlations were observed between the CVLT-9 sum of words recalled for trials 1–5
(Trial 1–5 Recall) and Long-Delay Free Recall (LDFR) measures and the immediate and delayed Logical
Memory (LM I and LM II) and Visual Reproduction (VR I and VR II) subtests from the Wechsler Memory
Scale-Revised (WMS-R). However, the CVLT-9 Trial 1–5 Recall and VR I measures demonstrated signifi-
cant correlations with a number of additional measures of language and visuospatial ability. The CVLT-9
LDFR, and the WMS-R LM I, LM II, and VR II showed less overlap with non-episodic memory function-
ing. A principal components analysis yielded a three-component solution consisting of a general or ‘‘g’’
component, a specific memory component, and a mood component. The CVLT-9 Trial 1–5 Recall and VR
I loaded on both the ‘‘g’’ and the memory components, whereas LM I, LM II, and VR II loaded on only the
memory component. We conclude that the CVLT-9 Trial 1–5 Recall and VR I demonstrate low
discriminant validity, suggesting diminished specificity as memory measures.
The California Verbal Learning Test (CVLT; senting an overly long stimulus list. In this ini-
Delis, Kramer, Kaplan, & Ober, 1986) is a popu- tial validation study, a principal component
lar measure of verbal learning, delayed verbal analysis of the CVLT-9 summary measures pro-
recall and recognition, and conceptual ability, duced a three-factor solution, corresponding to
with a reported use of 36% in a sample of 250 immediate free recall, delayed recall and recog-
surveyed members of the International Neuro- nition, and the production of intrusion re-
psychological Society (Butler, Retzlaff, & Van- sponses. In addition, the CVLT-9 was able to
derploeg, 1991). Despite its frequent use in neu- differentiate patients with Alzheimer’s disease
ropsychological assessment, the standard 16- (AD) from those with ischemic vascular demen-
word list may overwhelm and frustrate some tia (IVD). Specifically, the IVD group made
patients, raising questions regarding the validity substantially fewer intrusion errors on both free
of this procedure in certain patient groups. A and cued recall test trials and fewer false posi-
nine-item ‘‘dementia version’’ of the CVLT tive responses during delayed recognition test-
(CVLT-9) consisting of three items from each of ing than the patients with AD. The IVD group
three semantic categories has been developed by also showed significantly higher scores on all
Libon and colleagues (Lezak, 1995; Libon et al., measures of delayed free and cued recall mem-
1996) in order to circumvent the problem of pre- ory, the saving score, and on the recognition
*
Portions of these results were presented at the 25th Annual Meeting of the International Neuropsychological
Society, Chicago, IL, USA, February, 1996. Colleen McGuire is at the Department of Psychology, University of
Maryland.
Address correspondence to: John L. Woodard, Georgia State University, Memory Assessment Clinic and Alzhei-
mer’s Disease Program, One Park Place South, Suite 801, Atlanta, GA 30303-3083, USA. E-mail:
jlwoodard@gsu.edu
Accepted for publication: February 9, 1999.
554 JOHN L. WOODARD ET AL.
discriminability index relative to the AD pa- factor, we expected to find a greater number of
tients, lending support to the clinical validity of significant correlations between this subtest and
this measure. other non-memory measures.
Aside from this initial study (Libon et al.,
1996), little information is available regarding
the psychometric properties of the CVLT-9, par- METHOD
ticularly with respect to convergent and discri-
minant validity. Validity studies of the standard Participants
CVLT have reported moderate to high correla- Participants were 130 patients seen for evaluation
tions between the CVLT and other measures of of memory complaints through a geriatric Memory
Assessment Clinic. The mean age of the sample
memory, suggesting generally good convergent
Downloaded by [University of California, San Diego] at 18:31 12 November 2015
the sample size varied depending on the number of magnitude of these correlations was smaller than
persons who were administered the supplementary was seen with the CVLT-9 Trial 1–5 Recall.
neuropsychological measure, each correlation is Among the WMS-R variables, as expected,
based on pairwise deletion of missing cases (i.e., if
VR I showed significant and large overlap with
there is missing data for one of the two measures
in the correlation, the entire case is deleted). In all non-episodic memory measures except GDS.
order to deal with missing variables in the princi- The magnitude of these correlations tended to be
pal component analysis, we substituted the grand as large or larger than the relationship between
mean of a given variable for its respective missing VR I and the two CVLT-9 variables (see Table
value and computed the correlation matrix based 1). With the exception of significant correlations
on this mean substitution method. We then per- between Visual Naming and LM I and between
formed a principal component analysis using a
JOLO and VR II, the WMS-R variables (LM I,
Downloaded by [University of California, San Diego] at 18:31 12 November 2015
Variable LM I LM II VR I VR II
CVLT-9 Trials 1–5 Sum 0.46 0.35 0.49 0.41
CVLT-9 Long-Delay Free Recall 0.41 0.49 0.33 0.60
Note. N = 130. CVLT-9 = California Verbal Learning Test (Dementia Version); WMS-R = Wechsler Memory
Scale-Revised; LM = Logical Memory; VR = Visual Reproduction. The correlation between CVLT-9 Trials 1-5
Sum and CVLT-9 Long-Delay Free Recall was .50. All correlations are significant (p < .001).
556 JOHN L. WOODARD ET AL.
Note. CVLT-9 = California Verbal Learning Test (Dementia Version); WMS-R = Wechsler Memory Scale-
Revised; LM = Logical Memory; VR = Visual Reproduction; MAE = Multilingual Aphasia Examination;
COWAT = Controlled Oral Word Association Test; JOLO = Judgment of Line Orientation Test; MDRS = Mattis
Dementia Rating Scale; GDS = Geriatric Depression Scale.
***
p < .001; ** p < .01; * p < .05.
Table 3. Component Loadings for Episodic and Non-Episodic Memory Measures Based on Rotated Component
Matrix.
Component
1 2 3
CVLT-9 Trials 1-5 *.575 *.518 .071
CVLT-9 LDFR *.691 .227 .141
LM I *.842 .044 –.187
LM II *.882 –.127 –.095
VR I *.490 *.599 .187
VR II *.763 .199 .155
MAE Visual Naming .027 *.732 –.101
COWAT –.040 *.745 –.191
Animal Fluency .331 *.663 –.066
JOLO .045 *.766 .138
MDRS Total .345 *.519 –.289
GDS .053 –.119 *.914
Note: CVLT-9 = California Verbal Learning Test (Dementia Version); LDFR = Long-Delay Free Recall; LM =
Logical Memory; VR = Visual Reproduction; MAE = Multilingual Aphasia Examination; COWAT = Controlled
Oral Word Association Test; JOLO = Judgment of Line Orientation Test; MDRS = Mattis Dementia Rating
Scale; GDS = Geriatric Depression Scale.
VALIDITY OF THE CVLT (DEMENTIA VERSION) 557
amount of overlap with non-episodic memory ment has reported that the Trial 1-5 Recall
measures, suggesting that it might be regarded shows significant relationships across a variety
as the ‘‘purest’’ measure of episodic memory of executive function measures, including the
functioning among the measures evaluated in Wisconsin Card Sorting Test, COWAT, Trail
this study. The number and magnitude of signif- Making Test, and Wechsler Adult Intelligence
icant correlations between the WMS-R measures Scale-Revised (WAIS-R) Digit Span, Similari-
(LM I, LM II, and VR II) and non-episodic ties, and Block Design (Vanderploeg, Schinka,
memory measures were diminished as compared & Retzlaff, 1994). In addition, this same study
to the corresponding correlations for VR I and found that Long-Delay Free Recall demon-
CVLT-9 Trial 1–5 Recall. strated significant correlations with only the
The significant correlations between Animal Trail Making Test and WAIS-R Similarities.
Downloaded by [University of California, San Diego] at 18:31 12 November 2015
Fluency, the MDRS, and all CVLT-9 and WMS- This pattern of relationships is suggestive of
R measures were not surprising, given that per- greater sensitivity to executive functioning tasks
formance on these measures and on episodic for CVLT Trial 1-5 Recall, possibly due to reli-
memory tasks is typically compromised in de- ance on multiple cognitive processes to perform
mentia (Woodard, Salthouse, Godsall, & Green, verbal learning across multiple trials, and dimin-
1996b). In contrast, given that the diagnostic ished sensitivity to executive functioning for the
entity of pseudodementia has been traditionally CVLT Long-Delay Free Recall.
associated with cognitive impairment, particu- The sample size of the present study is com-
larly with episodic free recall tasks and mea- parable to the initial validation study reported by
sures of verbal fluency (King & Caine, 1990), it Libon and colleagues (Libon et al., 1996), al-
is noteworthy that there was absolutely no rela- though the sample composition differed to some
tionship between GDS score and any cognitive extent in that we did not include a normal con-
measure. However, it is important to draw a dis- trol group and our sample included a small num-
tinction between an elevated score on the GDS, ber of individuals with more diverse dementia
which may not necessarily be diagnostic of an etiologies. Interestingly, using a principal com-
underlying depressive disorder, and actually ponent analysis including only CVLT-9 vari-
meeting specific criteria for a mood disorder, ables obtained from their non-demented sample,
which was the case for a very small proportion Libon and co-workers found that the CVLT-9
of our sample. On the other hand, the lack of sum of Trials 1-5 variable loaded on multiple
correlation between the cognitive and mood components tapping delayed recall and immedi-
measures may be taken as evidence of discri- ate free recall, whereas the CVLT-9 Long-Delay
minant validity. Free Recall Variable loaded only on a delayed
Taken together with the results of our prior recall component. In contrast, these two vari-
study (Woodard et al., 1996a), the results of this ables loaded on the same general memory com-
study suggest that VR I, and the CVLT-9 Trial ponent in their sample of demented patients.
1–5 Recall might best be regarded as measures Whether or to what extent our obtained compo-
sensitive to level of cognitive functioning rather nent structure would differ in a normal elderly
than as specific measures of memory. In con- population would be an important follow-up to
trast, LM II in particular appears to measure our present results in the clinical sample. The
memory functioning more specifically than as- ability of CVLT-9 and WMS-R measures to dif-
sessing more general neuropsychological func- ferentiate between multiple diagnostically ho-
tioning. LM I, VR II, and CVLT-9 Long-Delay mogenous groups is another important future
Free Recall demonstrated some degree of over- consideration in establishing the validity of
lap with non-episodic memory measures but these measures.
showed stronger relationships with other epi- It is important to note that other performance
sodic memory tasks. At least one other study measures from the CVLT-9 were not systemati-
using the standard CVLT in a sample of young cally examined in order to limit the number of
adult males seen for neuropsychological assess- correlations performed. The initial validation
558 JOHN L. WOODARD ET AL.
study of the CVLT-9 (Libon et al., 1996), along Goodglass, H., & Kaplan, E. (1983). The Assessment
with other detailed studies of individual vari- of Aphasia and Related Disorders. Philadelphia:
Lea and Febiger.
ables from the standard CVLT (Vanderploeg et
King, D.A., & Caine, E.D. (1990). Depression. In J.
al., 1994) suggest their potential utility for dif- L. Cummings (Ed.), Subcortical dementia (pp.
ferentiating between diagnostic groups and for 218-250). New York: Oxford.
performing detailed characterization of aspects Lezak, M. D. (1995). Neuropsychological assessment.
of memory functioning. Thus, the validity and (3rd ed.). New York: Oxford.
Libon, D.J., Mattson, R.E., Blosser, G., Kaplan, E.,
utility of other CVLT-9 measures should be ex- Malamut, B.L., Sands, L.P., Swenson, R., &
amined in future studies. Cloud, B. S. (1996). A nine-word dementia version
of the California Verbal Learning Test. The Clini-
cal Neuropsychologist, 10, 237-244.
Downloaded by [University of California, San Diego] at 18:31 12 November 2015