Arch,w o/Clrnrcd Neurops.vcholog.v, Vol. I, pp. 147-156. 1986 0887.6177186 53.00 + .

OO
Pnnted I” the USA. All nghts reserved. Copyrighr c 1986 naknal Academy of Chmcal Neuropsychologlrls

Factor Structure of Benton’s Tests

of Visual Retention, Visual Construction,

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
and Visual Form Discrimination

James A. Moses, Jr.

Veterans Administration Medical Center
Palo Alto. California

The standard drawing and multiple-choice versions of the Benton Visual Reten-
tion Test (BVRT), the Visual Form Discrimination Test (VFDT), the Rey Audito-
ry Verbal Learning Test (AVLT), and the Wechsler Adult Intelligence Scale-Re-
vised Digit Span subtest were administered to a mixed sample of 97psychiatric.
neurologic, and control subjects. Factor analysis of the Benton test battery
measures alone showed that the standard and multiple-choice BVRT protocols
performed very similarly. Scoring of the standard B VRT for errors and number
of correct items was shown to be statistically redundant for this sample. Possible
simplification of standard B VRTscoring was suggested. Copying and immediate
visual-memory skills emerged as separable factors on the standard B VRlY When
the AVLT and Digit Span measures were factor analyzed with the Benton test
battery measures, however, simple factorial structure emerged only for the AVLT
measures (one factor) and the BVRT copying measures that were grouped with
the VFDT A verbal mediation component was thought to be associated with the
multifactorial B VRT and multiple-choice B VRT results in this analysis.

A recent trend in the neuropsychological literature has involved use of the

factor-analytic technique to study the construct validity of a number of
common clinical tests of short-term memory. Relationships among selected
Wechsler Memory Scale subtests, the Benton Visual Retention Test, the Rey
Auditory Verbal Learning Test, factorial memory measures derived from the
Luria-Nebraska Neuropsychological Battery, and marker variables from the
Wechsler Adult Intelligence Scale-Revised subtests have been used in various

Correspondence concerning this article should be directed to James A. Moses, Jr., Psychology
Service (116B), Veterans Administration Medical Center, 3801 Miranda Avenue, Palo Alto, CA
94304.

147
148 J. A. Moses, Jr.

studies of this kind to date (cf. Larrabee, Kane, & Schuck, 1983; Larrabee,
Kane, Schuck, & Francis, 1985; Ryan, Rosenberg, & Mittenberg, 1984). A
series of related measures of visual form reproduction and recognition that
lend themselves to study with this factorial methodology have been intro-
duced over the years by Benton and- his colleagues. The Revised Visual
Retention Test is one such measure. It is commonly referred to in the litera-
ture as the Benton Visual Retention Test (BVRT) and will be so labeled in

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
this paper for consistency with that convention.
Benton (1974) described the BVRT as “a clinical and research tool de-
signed to assess visual perception, visual memory, and visuoconstructive
abilities” (p. 1). Commonly, the BVRT is administered first as a measure of
immediate visual memory (design presentation followed by reproduction),
and thereafter the same designs are copied as a control for visuoconstructive
ability. If one notes errors on the immediate-recall administration, the defi-
cits could be related to a problem with visual memory, visual constructional
copying ability, or visual perception. Accurate recall reproduction of the
BVRT figures rules out visual-motor and perceptual difficulties, as these
skills are prerequisite to normal performance on the recall task. Adequate
copying in the presence of impaired recall suggests a short-term visual-
memory deficit. Impaired copying, however, might still be due to a perceptu-
al deficit. Benton recommended the dual scoring of the BVRT copy and
memory performances for total number of designs reproduced correctly and
also for total number and types of errors made. Although performance
must be accurate in all essential details to gain full credit on a design,
multiple errors can be made when an item is failed. The scoring system
(Benton, 1974) allows for identification of a wide variety of possible errors.
Many patients with neurologic disorder also have sensorimotor impair-
ment of the dominant hand and arm and are unable to draw as a result. To
evaluate the status of immediate memory and perceptual ability in such
persons, Benton, Hamsher, and Stone (1977) developed the Benton Visual
Retention Test: Multiple Choice Form I (MCBVRT). Originally, this test was
designed to be given as an immediate-memory measure or a perceptual-
matching measure. The usual BVRT three-figure array with two large central
geometric forms and one small peripheral figure is presented to the patient
for a period of 10 sec. Immediately thereafter the original design is removed,
and the patient is presented with another card that has four similar arrays.
Three of these alternatives are variations on the original array (major figure
distortion, major figure rotation, peripheral figure rotation or misplace-
ment), and one of them is identical to the original pattern. The task is to
identify the original pattern among the group of four alternatives. This
immediate-memory use of the multiple-choice version of the Visual Reten-
tion Test was not reported in the published clinical manual for Benton’s test
 Benton Battery Visual Test Factors 149

battery (Benton, Hamsher, Varney, & Spreen, 1983). The alternative admin-
istration method for the multiple-choice form of the BVRT suggested by
Benton et al. (1977) involved the simultaneous presentation of the original
array card and a four-alternative choice card. The task in this instance was to
compare the two sets of arrays and to identify the identical design patterns
by perceptual matching. This task was renamed the Visual Form Discrimina-

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
tion Test by Benton et al. (1983). They described it as being “designed as a
brief, convenient procedure to assess the capacity for complex visual form
discrimination” (p. 55).
As a result of the work of Benton and his colleagues, there are no less
than six scores from their battery that can be used to measure various
aspects of immediate visual memory, visual constructional drawing ability,
and visual perceptual recognition. These scores are the correct and error
scores for the BVRT immediate recall, the correct and error scores for the
BVRT design copying, the correct weighted score for the Visual Form Dis-
crimination Test, and the correct weighted score for the Multiple Choice
BVRT. In the last two tests, a correct response receives full credit, and a
peripheral error receives half credit because, in the case of a peripheral error,
the major figures are correctly identified but there is an error in detail.
Because there is only one response for each item on the multiple-choice
tasks, there can be a maximum of one error per item. Only the total number
of correct item points is scored as a result.
Another instrument that is of interest to study in conjunction with the
Benton test battery is the Rey Auditory Verbal Learning Test (see Lezak,
1983, for review). This measure has recently been studied by Ryan et al.
(1984) in conjunction with the Wechsler Memory Scale and the Wechsler
Adult Intelligence Scale-Revised. They noted that the Auditory Verbal
Learning Test (AVLT) did not load on measures of verbal intelligence or
attention span (WAIS-R Digit Span). It appeared to be contributing infor-
mation that was specific for “verbal learning and memory” (p. 239). Their
inclusion of Digit Span as a measure in mnemonic analysis of the AVLT is
p~ticularly important, because the ability to attend to task is necessary for
accurate initial registration and, hence, for recall. Just as adequate visual
perception is necessary for accurate BVRT memory and copying perfor-
mance, attention span is basic to all memorial performance. It remains to be
demonstrated how Benton’s carefully constructed psychometric measures of
visual perceptual processing and recall are related to each other as theoreti-
cal constructs. Also of interest is their relationship to a measure of attention
span and the AVLT measures of immediate verbal recall, verbal learning,
and verbal recognition. These variable interactions must be clarified before
these measures from Benton’s test battery and the other measures discussed
can be used parametrically with each other in research and clinical applica-
tions.
I.50 J. A. Moses, Jr.

STUDY 1: BENTON TEST GROUP CONSTRUCT VALIDATION

Method

Subjects. The sample consisted of 97 patients in a variety of neurologic or

psychiatric diagnostic categories. Three patients had suffered traumatic

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
brain injuries to the left hemisphere and had subdural hematomas that had
been treated. None of these cases was acute. Eleven patients had cerebrovas-
cular disorder; all of them were cortical syndromes. Two patients had right-
hemispheric infarctions, one in the frontoparietal area, one in the occipital
area. One patient had had bilateral strokes. Eight patients had had left-
hemispheric strokes; of these two were in the occipital lobe, and six were in
the frontoparietal distribution. Six patients had a history of degenerative
brain disease (one with early-onset Huntington’s disease, five with primary
degenerative dementia). Four patients had partial complex-seizure disorder.
Six patients had a history of alcohol dependence; none of these was amne-
sic. Twenty-six patients had Schizophrenic or Schizoaffective Disorders, 5 had
Bipolar Disorder, 2 had other psychotic disorder, 25 had major depression, 1
had posttraumatic stress disorder, 2 had personality disorder, and 6 were
normal controls who completed the tests as part of a research protocol.
Neurologic diagnoses were made by board-certified neurologists or neu-
rosurgeons by means of the neurologic examination findings, in addition to
one or more of the following measures, as clinically indicated: computerized
tomographic brain scan, electroencephalography, radioisotope brain scan,
angiography, or findings at brain surgery. Psychiatric diagnoses were made
independently by the author, who is a board-certified clinical psychologist;
by a board-certified psychiatrist; and by one or more psychiatric residents.
All psychiatric diagnosticians made use of the standard DSM-III manual
criteria (American Psychiatric Association, 1980).
The current sample was normatively middle aged (M= 49.53, SD= 15.46,
range=21-77) and high school educated (M= 12.67, SD=2.66, range=
4-l@, but the range values show much variation from these average values.
The patients were almost exclusively male (97.9%), which is an artifact of
sampling from a Veterans Administration population. The sample was also
predominantly right-handed (91.7%).

Procedure. The copy (Administration C) and lo-set exposure with immedi-

ate reproduction (Administration A) of the Revised Visual Retention Test,
Form C, were administered individually to each subject. Responses were
scored for number of correct responses and number of errors on each ad-
ministration, according to the standardized procedures in the test manual
(Benton, 1974). The MCBVRT was administered and scored according to
 Benton Battery Visual Test Factors I51

instructions provided by Benton et al. (1977). The VFDT was administered

and scored according to directions presented in Benton et al. (1983). The
Digit Span subtest of the WAIS-R was administered according to the stan-
dard directions in the test manual (Wechsler, 1981).

Statistical analysis rationale. A principal-components method of factor ex-

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
traction with orthogonal Equamax rotation to simple structure was em-
ployed for both studies in this paper. The choice of principal components
as a method of factor extraction follows the recommendation of Nunnally
(1978), who noted that this method “is ideal for selecting a method of
condensation prior to rotation that optimizes some function of the actual
data” (p. 362). Equamax rotation was chosen as a method of orthogonal
rotation to achieve simple factorial structure in accordance with the recom-
mendations of Sawicki and Golden (1984). They concluded that this factor
rotation method is superior to the Varimax method because the Equamax
method is able “to evenly distribute variance among factors” (p. 132). This
type of result more closely approximates Thurstone’s original simple-struc-
ture criteria for factor rotation than other orthogonal rotation methods in
most cases. An orthogonal rotation method was chosen to produce factorial
groupings that would be independent and, hence, not redundant. Factors
were extracted if the eigenvalue was greater than or equal to 1.0. The num-
ber of factors to be retained for interpretation was determined by Cattell’s
Scree Test and examination of the relative sizes of successive eigenvalues
produced for each factor in the series. These procedures were based on the
recommendations of Franzen and Golden (1984). Items that loaded at a
minimum value of 0.40 on a factor were selected for inclusion in that factor
grouping. This relatively conservative value helps to minimize chance varia-
tion as a selection factor among items included in factor groupings. As such
it should increase the internal consistency and reproducibility of the factors
composed.

Results
The number correct and number of errors scores for the BVRT copy and
memory administrations, the multiple-choice BVRT total correct score
(MCBVRT), and the VFDT total correct score were submitted to a principal-
components factor analysis with Equamax rotation to simple structure.
Results of the factor analysis for these six measures from Benton’s test
battery are presented in Table 1.
A two-factor simple-structure solution was achieved that accounted for
77.1% of the total variance. Factors one and two respectively accounted for
59.4% and 17.7% of the total variance. Factor one showed loadings above
the 0.40 minimum criterion value on the BVRT memory correct and error
I52 .J. A. Moses, Jr.

TABLE 1
Equamax Rotated Factor Matrix for Benton’s Tests

Test Score Factor 1 Factor 2

BVRT memory-correct .88478 .20642

BVRT memory-errors - .88202 - .25708
BVRT memory-multiple- .75388 .26083

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
choice total score (correct)
Visual Form Discrimination .57792 -
Test (match multiple choice)
BVRT copy-correct .26307 .95405
BVRT copy-errors - .27938 - .95024

Notes. BVRT = Benton Visual Retention Test. Factor load-

ings less than or equal to 0.20 are left blank in the factor
matrix to aid inspection of the loadings for those that explain
a significant amount of variance.

scores, the MCBVRT correct memory score, and the VFDT total correct
matching score. Factor two showed very high loadings on the accuracy and
error scores for the copying administration of the BVRT.
The factor loadings were nearly identical for the number of errors and
number of correct scores for both the copy and the memory administrations
of the BVRT. To investigate the relationship of these accuracy and error
scores, the corresponding measures were correlated with each other for the
BVRT copy and memory adminstrations. The correlation of the correct and
error scores for the BVRT copying administration was -0.96138 @< .OOOOO5).
The correlation of the correct and error scores for the BVRT memory ad-
ministration was - 0.89833 (I? < .OOOOOS).

Discussion
The current series of analyses suggests that information obtained from
scoring the BVRT for number of items correctly reproduced and for number
of errors is redundant. This conclusion appears to hold true for both the
copying and the immediate-memory administrations of the test. If these
findings could be replicated in independent, large, neuropsychiatric sam-
ples, one could consider simplifying the scoring of the BVRT to just identify
the number of items that are correctly copied or reproduced from immediate
memory. Current findings suggest that no additional predictive power is
gained by specifying the number and kind of BVRT errors that have been
made once it is clear that there is some storable error in the patient’s repro-
duction.
The factorial groupings reveal a differentiable dimension of immediate
visual memory that is exemplified equally well by the correct or the error
 Benton Battery Visual Test Factors 153

scores of the BVRT memory administration. The MCBVRT administration

also loads quite highly on this factor and suggests that it, too, is an exemplar
of the underlying visual-memory construct. Use of the MCBVRT as initially
advocated by Benton et al. (1977) appears to warrant further experimental
investigation. This line of reasoning will be pursued further in the second
experiment of this paper. Interestingly, the VFDT is factorially included with

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
the immediate visual-memory variable group, but not with the visual con-
struction variables that form the second factor. The VFDT appears to in-
volve more active perceptual processing and analysis, in common with the
memory measures, than passive recognition, which is required for compari-
son and form discrimination when copying. This is an unexpected result.
The basis of this grouping is also highlighted by findings from the second
study in this paper. The second factorial grouping for this first study ap-
pears to be a relatively pure measure of visual constructional copying ability.
Copying and immediate memory emerge from current analyses as tasks that
are differentiable and that should be evaluated independently, as recom-
mended by Benton (1974).

STUDY 2: BENTON TEST ANALYSES WITH MARKER VARIABLES

Method

Subjects. The same sample of subjects used in Study One was also used for
Experiment Two. Please refer to Study One for demographic details.

Procedure. Please also refer to Study One for a listing of all test procedures
administered to the sample. In Study Two the error scores for the copy and
memory administrations of the BVRT were eliminated from the analyses
because of their redundancy with the number of correct scores in each
administration. In the current series of analyses the age-corrected Digit
Span score from the WAIS-R was added to the analysis, in addition to three
scores from Rey’s Auditory Verbal Learning Test. Following the methodolo-
gy of Ryan et al. (1984), the numbers of words correctly recalled for all five
acquisition trials of the AVLT were summed, to produce a total free-recall
performance-level score. The postinterference trial for recall of the initial
AVLT word list (Trial 6) and the score for number of original list words
recognized in the AVLT story recognition task were also included in the
Study Two analyses.
The variables analyzed in Study Two thus consisted of the three AVLT
variables, the accuracy scores for BVRT copy and immediate memory, the
total correct score for the MCBVRT, the total correct score for the VFDT,
and the WAIS-R Digit Span scaled score. These variables were submitted to
154 J. A. Moses, Jr.

a principal-components factor analysis with Equamax rotation to simple

structure.

Results

Results of the factor analysis of the variables in Study Two are presented

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
in Table 2. A two-factor solution was obtained from the current analysis that
accounted for 64.4% of the total variance. Factors one and two, respective-
ly, accounted for 50.6% and 13.9% of the explained variance.
The factorial solution for the Study Two variables appears to be an
extension and partial replication of that found in Study One, with some
important modifications. In the current factorial solution, simple-structure
criteria for the first factor were met only by the three AVLT variables.
Simple-structure criteria for the second factor were met by the BVRT copy-
ing accuracy score, the Digit Span scaled score, and the VFDT perceptual-
matching accuracy score. The BVRT memory accuracy and MCBVRT mea-
sures load significantly on both factors.

Discussion
In Study Two the first factor appears to represent an underlying construct
of immediate memory. It is most clearly and unambiguously exemplified in
the current analyses by the verbal memory measures, but high loadings from
the nonverbal measures suggest that cross-modal retention capacity is being
measured. Although there is a contribution of attention span to this dimen-
sion, it is not a primary component of it. The second factor is exemplified
by the BVRT copying accuracy task, with component-related skills of atten-
tion span (WAIS-R Digit Span) and complex perceptual analytic ability
(VFDT). Accurate copying requires the ability to perceptually attend to and
evaluate details, and all of these processes require intact attention span to
remain task oriented. In the current variable set, VFDT is differentiated
from the verbal AVLT measures, and VFDT joins the nonverbal variable
grouping. The VFDT task does no? appear to involve the verbal mediation
component shared by BVRT and MCBVRT with the AVLT.
The complex visual-memory tasks require an additional component of
short-term memory retention in addition to the skills outlined above. To
perform accurately on the MCBVRT or the BVRT memory tasks requires all
of the skills needed for the BVRT copying task and the ability to remember
the array briefly before reproducing it. In my clinical experience, encoding
of the geometric forms involves verbal mediation for many if not most
patients, and often they can be heard to repeat the names of the shapes
aloud to themselves as they view the figures before they attempt to reproduce
them and, again, as they draw them. This behavior is typically absent during
copying. If this hypothesis is valid, then it is sensible that there would be a
 Benton Battery Visual Test Factors 155

TABLE 2
Equamax Rotated Factor Matrix for Benton’s Tests,
Rey’s Test and WAIS-R Digit Span

Test Score Factor 1 Factor 2

AVLT - Trial 6 .90928

AVLT - Trials l-5 Total .89294 .21707

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
AVLT - Story Recognition .75162 .24720
BVRT - copy, correct .89872
BVRT - multiple-choice S6140 .58433
total memory score (correct)
BVRT - memory, correct .53537 S7416
Visual Form Discrimination .37076 S3725
Test (match multiple choice)
WAIS-R Digit Span (age corrected) .35617 .45472

Note. AVLT = Rey Auditory Verbal Learning Test; BVRT = Benton

Visual Retention Test; WAIS-R = Wechsler Adult Intelligence Scale-
Revised. Factor loadings less than or equal to 0.20 are left blank in
the factor matrix to aid inspection of the loadings for those that ex-
plain a significant amount of variance.

factorial linkage between the rote verbal (AVLT) memory measures and
recall of the BVRT and MCBVRT measures. The BVRT memory task with a
graphomotor reproduction format and the MCBVRT memory task without
a drawing requirement show very similar factorial patterning in Studies One
and Two. It could be that the BVRT and MCBVRT performances are sensi-
tive to the ability to form and verbally encode concepts rather than to visual
memory alone.
It is important in this context to distinguish between visual memory and
nonverbal processes that are related to recall. Just because the task presents
a nonverbal stimulus to the subject does not prevent one from verbally
mediating the task and changing the supposed task demands from a nonver-
bal task to a verbal one. The “common underlying factor” across verbal and
visual memory tasks inferred by Larrabee et al. (1985) thus might simply be
verbal mediation. Failure to demonstrate separable, factorially simple verbal
and nonverbal memory components in these analyses suggests that the na-
ture of the patient’s strategy in approaching the task rather than the nature
of the stimulus materials per se is a critical feature to be analyzed in the
evaluation of visual-memory performance.

REFERENCES

American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disor-
ders (3rd ed.). Washington, DC: Author.
Benton, A. L. (1974). Revised Visual Retention Test: Clinical and experimental applications
(4th ed.). New York: Psychological Corporation.
156 J. A. Moses, Jr.

Benton, A. L., Hamsher, K. deS., & Stone, F. B. (1977). Visual Retention Test: Multiple Choice
Form I. Administration F: Visual form discrimination. Administration G: Mulriple
Choice-memory. Unpublished manuscript, University of Iowa School of Medicine, Divi-
sion of Behavioral Neurology, Iowa City.
Benton, A. L., Hamsher, K. deS., Varney, N. R., & Spreen, 0. (1983). Confriburions to
neuropsychological assessment: A clinical manual. New York: Oxford University Press.
Franzen, M. D., &Golden, C. .I. (1984). Multivariate statistical techniques in neuropsychology:
II. Comparison of number of factors rules with the motor scale of the Luria-Nebraska

Downloaded from http://acn.oxfordjournals.org/ at Penn State University (Paterno Lib) on February 18, 2016
Neuropsychological Battery. International Journal of Clinical Neuropsychology, 6,
165-171.
Larrabee, G. J., Kane, R. L., & Schuck, J. R. (1983). Factor analysis of the WAIS and Wechsler
Memory Scale: An analysis of the construct validity of the Wechsler Memory Scale. Journal
of Clinical Neuropsychology, 5, 159-168.
Larrabee, G. J., Kane, R. L., Schuck, J. R., & Francis, D. J. (1985). Construct validity of
various memory testing procedures. Journal of Clinical and Experimental Neuropsycholo-
gy, 7, 239-250.
Lezak, M. D. (1983). Neuropsychological assessment (2nd ed.). New York: Oxford University
Press.
Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York: McGraw-Hill.
Ryan, J. J., Rosenberg, S. J., & Mittenberg, W. (1984). Factor analysis of the Rey Auditory-
Verbal Learning Test. International Journal of Clinical Neuropsychology, 6, 239-241.
Sawicki, R. F., & Golden, C. J. (1984). Multivariate statistical techniques in neuropsychology:
I. Comparison of orthogonal rotation methods with the receptive scale of the Luria-Nebras-
ka Neuropsychological Battery. International Journal of Clinical Neuropsychology, 6,
126-134.
Wechsler, D. (198 1). W’echslerAdult Intelligence Scale-Revised manual. New York: Psychologi-
cal Corporation.