Content uploaded by Amer Hamdan
Author content
All content in this area was uploaded by Amer Hamdan on Jan 28, 2019
Content may be subject to copyright.
129
Clinical Neuropsychiatry (2012) 9, 3,
© 2012 Giovanni Fioriti Editore s.r.l.
SUBMITTED SEPTEMBER 2011, ACCEPTED JULY 2012
VALIDITY CONVERGENT AND RELIABILITY TEST-RETEST
OF THE REY AUDITORY VERBAL LEARNING TEST
Sabrina de Sousa Magalhães, Leandro Fernandes Malloy-Diniz, Amer Cavalheiro Hamdan
Abstract
Objective: This study provides evidence for the validity and reliability of the Rey Auditory Verbal Learning Test
(RAVLT).
Method: The reliability was measured by internal consistency and the test-retest method with a mean interval of
35 days. To determine the convergent and divergent validity, it was compared the performance on the RAVLT with the
Benton Visual Retention Test (BVRT) and the Trail Making Test (TMT), respectively. The test was been taken by 34
healthy participants of both genders, ages 17 to 40 and 11.2±0.7 years of education.
Results: All test-retest correlation coefficients achieved significance, ranging between 0.36 and 0.68. The A2
measure obtained the weakest correlation (r =0.28), and the sum of A1-A5 (r =0.68) was the strongest. The rest of the
measures obtained moderate correlations. The value of the Cronbach’s Alpha coefficient was 0.80. The two RAVLT
measures, the sum of A1-A5 and A7, did not significantly correlate with the TMT measures. In contrast, these measures
did exhibit significant but modest correlations with the measures of BVRT (ranging from 0.37 to 0.44).
Conclusions: The results had adequate divergent and convergent validity and good reliability in terms of internal
consistency. The evidence collected in this study indicates that RAVLT is a valid and reliable psychometric instrument
in neuropsychological assessment.
Key words: neuropsychological tests, psychometrics, reliability, validity of tests
Declaration of interest: none
Sabrina de Sousa Magalhães – Department of Psychology, Federal University of Paraná, Paraná, Brazil
Leandro Fernandes Malloy-Diniz – Laboratory of Neuropsychological Investigations – National Institute of Science and
Technology of Molecular Medicine, Federal University of Minas Gerais, Minas Gerais, Brazil and Department of
Psychology, Federal University of Minas Gerais, Minas Gerais, Brazil
Amer Cavalheiro Hamdan – Department of Psychology, Federal University of Paraná, Paraná, Brazil
Corresponding author:
Amer Cavalheiro Hamdan, UFPR – Department of Psychology
Praça Santos Andrade, 50, Mestrado em Psicologia CEP: 80020-240, Curitiba, Paraná, Brazil
E-mail address: achamdan@ufpr.br
129-137
The Rey Auditory Verbal Learning Test (RAVLT)
is an instrument used to evaluate episodic declarative
memory, immediate memory span, verbal learning,
susceptibility to proactive and retroactive interferences,
retention of information, recall and memory recognition
(Lezak et al. 2004, Strauss et al. 2006). In Brazil,
Malloy-Diniz et al. (2007) developed an RAVLT v e rs io n
composed of a list of high frequency, one or two-syllable
Portuguese substantives, which served as stimuli to
assess the performance of Brazilian adults and senior
citizens. The demographic variables that commonly
influence the performance on the RAVLT are gender,
educational level and age (Magalhães and Hamdan
2010). Normative data are available (Malloy-Diniz et
al. 2007, Magalhães and Hamdan 2010), however
evidence regarding its psychometric properties is still
preliminary.
There is evidence that the RAVLT has construct
validity when compared with other measures of verbal
learning and memory, such as the California Verbal
Introduction
In the context of neuropsychological assessment,
there are two fundamental psychometric properties to
legitimate the efficiency of an instrument: validity and
reliability. Validity refers to what the test measures,
specifically the degree to which it actually measures
what it allegedly measures, according to empirical
verifications. Reliability is not about what is being
measured but whether the measure given by the
instrument is consistent. In other words, if the test shows
constancy in the scores for the same individual, the test
corresponds to an estimative of measurement error
(Mitrushina et al. 2005). The factors of reliability are
listed as follows: consistency (the test provides equal
or similar scores for the same individual); the ability to
replicate (within a certain margin of error, the test can
be refused and repeated); and confidence (the test
presents the same results to the same individual) (Hogan
2007).
130
Sabrina de Sousa Magalhães et al.
Clinical Neuropsychiatry (2012) 9, 3
Learning Test (Stallings et al. 1995). Several studies
conducted in Brazil also presented evidence of RAVTL
construct validity. For example, Fichman et al. (2010)
showed that some of the RAVLT indices present
significant and positive correlations with a test of visual
learning. De Paula et al. (2012) demonstrated that the
RAVLT has a bifactorial structure, which is related to
the processes of verbal learning and retrieval.
Nonetheless, despite the existence of a few studies on
RAVLT construct validity in the elderly Brazilian
population, we found no studies that evaluated the
reliability properties or construct validity of this test in
other age groups.
The main objective of this study was to assess the
psychometric properties of the reliability and validity
of the RAVLT. The additional objectives are listed as
follows: (a) to evaluate the confidence of the RAVLT
scores by comparing the performance in two distinct
experimental situations (test-retest); (b) to analyze the
internal consistency reliability of the instrument; (c) to
appraise the criterion validity of the RAVLT by
comparing it with other neuropsychological tests to
confirm that the test evaluates memory and not another
cognitive function. This study aims to contribute
evidence for the applicability of RAVLT by determining
whether the Brazilian adaptation is a valid psychometric
instrument in neuropsychological assessment.
Method
1. Participants
The participants were 34 male and female
undergraduates between 17 to 40 years in age and were
recruited by requests for volunteer participation during
visits to undergraduate psychology classes. All
participants signed two copies of the Volunteer
Participation Consent and Confidentiality Terms. The
research was conducted according to the Helsinki
Declaration of ethical principles.
The research used the following exclusion criteria
to select only healthy individuals: a) a history or
presence of psychiatric disturbances, b) diabetes, c)
heart problems or any related pathologies, d) past or
current use of psychoactive drugs, especially those with
already related side effects that disturb mnemonic
functions, and e) abuse of illicit substances according
to the Brazilian law. These criteria were investigated
through the “Questionnaire of demographic data, health
and cultural conditions” and the Mini Mental State Exam
(MMSE). The inclusion criterion for participation in
the MMSE sample was a score of at least 24 points.
Table 1 presents the demographic data from the
sample. The study included 34 participants of bother
genders (35% men and 65% women) with an age range
between 17 and 40 (mean 20.7±4.5). The education
variable was computed using the number of years the
subject studied according to the Brazilian school system.
All participants had reached at least the high school
level, and the average education time varied from 11 to
15 years of formal study, including repetitions of the
same grade. The MMSE score varied from 25 to 30
points, with the mean score of 29.1±1.1 points.
2. Instruments
The purpose and procedures of the instruments are
listed as follows:
1) The “Questionnaire on demographic data, health
and cultural conditions” was used to gather general per-
sonal, cultural and health information about the
individual, thus gathering data for the sample
characterization.
2) The Mini Mental State Exam (MMSE)
screened for cognitive impairment. The MMSE is a
useful screening instrument to estimate the level of
performance of individuals aged 18 to 85+ (Strauss et
al. 2006) because it is highly specific (Bertolucci et
al. 1994, Hancock and Larner 2011, Milian et al. 2011)
and allows for reliable detection of cognitive
impairment among healthy individuals within a modest
margin of error, although it does not provide a
differential diagnosis. A cut-off of 24 points was used
for our Brazilian sample.
Table 1. Demographic characterization of the sample (n = 34)
Mean (SD) CI 95%
Minimum Maximum
Age (years) 20.7 (4.5) 19.1 22.3
Gender (Male/Female) 12 / 22 - -
%35.3 / 64.7 - -
Education level (years) 11.2 (0.7) 10.9 11.4
MMSE score 29.1 (1.1) 28.7 29.5
Note: SD = standard deviation; CI = confidence interval; MMSE = Mini Mental State Exam.
131
Rey Auditory Verbal Learning Test
Clinical Neuropsychiatry (2012) 9, 3
3) A Portuguese language version of the Rey
Auditory Verbal Learning Test (RAVLT) was used
(Malloy Diniz et al. 2007). The test consists of five
consecutive oral presentations of 15 concrete nouns (List
A), followed by a free-recall after each presentation
(trials A1, A2, A3, A4 and A5). An interference list (List
B) with 15 different concrete nouns is presented,
followed by an immediate recall of this new list (trial
B1). Next, the recall of List A post-interference takes
place (trial A6). After a 20-minute interval, a delayed
recall of List A is required (trial A7). The last trial
consists of an oral presentation of a 50-noun list, which
is composed by List A and B and 20 phonologically or
semantically similar words to the previous lists, and then
recognition of the words in List A. The score for each
trial corresponds to the number of correctly recalled
words (Strauss et al. 2006). The rates of proactive
interference (B1/A1) and retroactive interference (A6/
A5) are also calculated; the rates represent the
susceptibility to previous and later activities,
respectively, involving the presentation of the content
about to be recalled. Proactive interference can be
observed when the subject’s recall of List B is affected
by previously learning List A. Retroactive interference
occurs when the retention or recall of the post-distractor
list is affected by learning the distractor list, which is
assumed to confuse one’s memory of the learned list. If
the interference ratio is equal to 1, no interference effect
is observed; however, if it is less than 1, interference is
demonstrated. Facilitation instead of interference can
be verified if the interference ratio is greater than 1
(Geffen and Geffen 2000).
4) The Trail Making Test (TMT) evaluates
executive function and is composed by two activities.
In part A (TMTA), the participant is asked to connect,
in ascending order, 25 numbered circles randomly
arranged on the page. In part B (TMTB), the participant
is required to alternate circles of numbers and letters
(numbers in crescent order, letters alphabetically). The
score of the test reflects the time (in seconds) taken to
complete the cycle (Hamdan and Hamdan 2009, Strauss
et al. 2006).
5) The Benton Visual Retention Test (BVRT)
evaluates immediate and delayed visual memory. Using
form C, the test was composed of 10 cards, each
featuring a complex figure. On the first two cards, the
figure is composed of simple geometric forms; the eight
other cards features two large geometric forms with a
small geometric figure located in the cards’ peripheries
(Strauss et al. 2006). In the immediate recall, the
participant must draw a replica, as exact as possible, of
the figure on the card after a period of 10 seconds of
studying the card with no opportunity to consult it
afterward. The procedure is repeated for all 10 cards.
After a time interval usually of 10 minutes, the
participant is requested to draw all the cards he or she
still remembers, in any order, with no opportunity to
consult the cards. Each drawing is scored with 0, 1 or 2
points. The maximum score is given to an identical
reproduction; the score of 1 to those with at least two
correct components; and a score of zero to drawings
that do not have sufficient components from the original
card (and do not attempt to earn the intermediate score)
or contain more mistakes than what is acceptable.
Afterward, these scores were added, and the immediate
(Be1) and delayed (Be2) recall scores were obtained
(maximum of 20 points).
3. Design
All participants completed all instruments. Each
test was administered to the same group of participants
twice to evaluate the score stability over time (test-retest
reliability). The present study contained a mean interval
of 35 days±8.9 between the two test sessions. The
interval length was 31 days at minimum and 38 days at
maximum.
The equivalence of the test scores across two
administrations and between all tests was first evaluated
prior to the assessment of scale reliability. Dependent t
tests were performed for the RAVLT if the means
calculated from the same form from both testing sessions
were significantly different. Internal consistency
reliability was estimated by intraclass correlation
coefficients (ICC) and coefficient a. The k-sample
significance test for independent coefficients proposed
by Hakstian and Whalen (1976) was adopted to
statistically evaluate the effects of the number of items
on coefficient a. The probability of making a Type I
error on each test was constrained at .01.
To access criterion validity, the two most reliable
RAVLT measures were chosen a posteriori, which were
the Sum A1-A5 and A7. Convergent validity was
estimated comparing those RAVLT items with the scores
on the Benton Visual Retention Test (BVRT), which is
also an instrument for episodic memory assessment.
Although the BVRT was developed for visual memory
assessment, some authors have argued that this test
assesses both visual and verbal memory because its
items (geometrical forms) can be verbalized (Strauss et
al. 2006). For divergent validity, the same RAVLT items
were compared with two measures of the TMT,
specifically, the TMTA, which demands mainly
attentional resources and the TMTB, which includes
an executive function component.
The first data collection began after the
presentation, reading, and discussion of possible doubts
about the Volunteer Participation Consent and
Confidentiality Terms, which was followed by the
“Questionnaire on demographic data, health and cultural
conditions” and the MMSE. After these screening
instruments, the RAVLT was administered. During the
twenty-minute interval necessary to collect data for the
A7 trial and Recognition, two other neuropsychological
tests, the TMT and immediate recall of the BVRT, were
given. Both instruments are assumed to have no
influence on the RAVLT scores because they are
nonverbal tasks and do not demand semantic processing.
The main concern here was to provide distractor tasks
without confounding the stimulus and the processing
demand of each task, in particular.
Immediately after the Recognition trial of the
RAVLT, the late recall of the BVRT was administered.
In the retest session, the demographic questionnaire and
MMSE were not applied, and the session began with
the participant’s permission and the administration of
the RAVLT. All experimental sessions lasted
approximately 50 minutes, and all activities were
completed in a single session.
132
Sabrina de Sousa Magalhães et al.
Clinical Neuropsychiatry (2012) 9, 3
Results
Table 2 presents the performance on the RAVLT
(test and retest session), Intraclass Correlation
Coefficient (ICC) and standard error of the mean (SEM).
The correlations varied from 0.36 to 0.68. The weakest
correlation was found in the A2 measure, and the
strongest was found in the Sum A1-A5. The indexes
A1, A5, B1 and A7 obtained medium correlation,
varying between 0.30 and 0.49. The A3, A4, Sum A1-
A5, A6 and Recognition reached a high correlation rate.
The proactive and retroactive interference did not
present any significant correlation between the two
sessions of testing. The susceptibility to interference
was small, except for those related to the expressive
proactive interference in the retest.
The paired t-test was performed to compare the
differences between the mean scores in both situations
(test-retest) and verify the presence of any disparity
between them. The results are listed as follows: A1:
t(33)=-9.13, p<0.001, d=3.18; A2: t(33)=-6.16,
p<=0.001, d=2.15; A3: t(33)=-6.54, p=<0.001, d=2.28;
A4: t(33)=-3.01, p<0.001, d=1.05; A5: t(33)=-4.13,
p<0.001, d=1.43; Sum A1 to A5: t(33)=-10.48, p<0.001,
d=3.65; B1/A1: t(33)=6.83, p<0.001, d=2.38; A6/A1:
t(33)=-0.70, p=0.4875, d=0.24; B1: t(33)=0.58,
p=0.5641, d=0.20; A6: t(33)= -4.61, p<0.001, d= 1.60
and A7: t(33)= -3.14, p<0.001, d= 1.09. Measures B1
and Recognition were the only ones that did not differ
in both situations (p>0,01); all the others demonstrate
significant differences in their results from both
experimental situations. The d measures reached small
effect sizes (Cohen 1988).
The F test for ICC are listed as follows: A1: F(1,
33)=83.39, p<0.001; A2: F(1, 33)=37.99, p<0.001; A3:
F(1, 33)= 42.73, p<0.001; A4: F(1, 33)= 9.08,
p=0.0049; A5: F(1, 33)=17.03, p<0.001; Sum A1 to
A5: F(1, 33)= 109.76, p<0.001; B1/A1: 46.71, p<0.001;
A6/A1: F(1, 33)= 0.49, p=0.488; B1: F(1, 33)= 0.34,
p=0.564; A6: F(1, 33)= 21.21, p<0.001 and A7: F(1,
33)= 9.85, p=0.0035.
The RAVLT measures analyzed with the
Cronbach’s Alpha Coefficient Reliability, deleting each
item in turn, are indicated in table 3. The alpha reliability
of the RAVLT was 0.84, and the standardized alpha was
0.88. The test condition coefficients varied from 0.78
to 0.82, and the retests varied between 0.81 and 0.86.
Table 4 shows the Pearson’s correlation between the
Table 2. Performance on the RAVLT (test and retest session), Intraclass Correlation Coefficient (ICC) and SEM
Test
Mean (SD)
Retest
Mean (SD)
ICC SEM
A1 7,2 (1,5) 10,2 (2,1) 0.71 1.38
A2 10,2 (2,1) 12,5 (1,7) 0.52 1.56
A3 11,9 (1,5) 13,6 (1,4) 0.55 1
A4 12,9 (1,6) 13,5 (1,3) 0.19 0.89
A5 12,8 (1,5) 13,8 (1,2) 0.32 1.06
Sum 54,9 (6,2) 63,6 (5,9) 0.76 3.42
B1/A1 0,98 (0,3) 0,66 (0,2) 0.57 0.19
A6/A5 0,95 (0,1) 0,97 (0,1) -0.02 0.11
B1 6,8 (1,6) 6,6 (1,7) -0.02 1.25
A6 12,1 (1,8) 13,3 (1,5) 0.37 1.13
A7 12,2 (1,8) 13,1 (1,6) 0.21 1.20
Rec 14,6 (0,7) 14,6 (0,9) -0,01 0.52
Note: SD = standard deviation; ICC = Intraclass Correlation Coefficient; SEM = standard error of the mean; Sum =
sum of scores across the five acquisition trials (A1–A5); B1/A1 = proactive interference; A6/A5 = retroactive
interference; Rec = recognition trial.
133
Rey Auditory Verbal Learning Test
Clinical Neuropsychiatry (2012) 9, 3
Table 3. Cronbach’s Alpha Coefficient for the RAVLT measures in the two experimental conditions
Table 4. Validity coefficients between the RAVLT, Trail Making Test and Benton Visual Retention Test
Test Retest
Alpha Sd.Alpha r(item, total) Alpha Sd.Alpha r(item, total)
A1 0.79 0.83 0.67 0.82 0.88 0.61
A2 0.79 0.83 0.60 0.81 0.86 0.75
A3 0.78 0.81 0.78 0.81 0.85 0.85
A4 0.79 0.82 0.68 0.82 0.86 0.72
A5 0.79 0.83 0.66 0.82 0.87 0.66
Sum 0.82 0.80 0.94 0.86 0.85 0.96
B1/A1 0.82 0.88 -0.23 0.83 0.87 0.47
A6/A5 0.82 0.84 0.26 0.84 0.90 0.02
B1 0.81 0.83 0.32 0.85 0.89 0.15
A6 0.78 0.81 0.73 0.82 0.86 0.68
A7 0.79 0.82 0.62 0.81 0.86 0.76
Rec 0.81 0.84 0.35 0.83 0.87 0.61
Note: Sd.Alpha = Standard Alpha; Sum = sum of scores across the five acquisition trials (A1 – A5);; B1/A1 = proactive
interference; A6/A5 = retroactive interference; A7 = RAVLT delayed recall measure; Rec = recognition trial.
Sum A7 B1/A1 A6/A5 TTA TTB Be1
Sum - - - - -
A7 0,57** - - - -
B1/A1 -0,33 -0,18 - -
A6/A5 0,12 0,45** 0,24 - -
TTA -0,03 0,01 0,22 0,02 - - -
TTB -0,19 -0,15 0,22 0,01 0,68** - -
Be1 0,44** 0,27 -0,12 -0,04 -0,31 -0,58** -
Be2 0,37* 0,39* -0,23 0,02 0,01 -0,13 0,31
Note: Sum = sum of scores across the five acquisition trials (A1 – A5); A7 = RAVLT delayed recall measure; B1/A1 =
proactive interference; A6/A5 = retroactive interference; TTA = part A score from TMT; TTB = part B score from
TMT; Be1 = immediate recall score from BVRT; Be2 = delayed recall score from BVRT; * = p<0,05; ** = p<0,01.
134
Sabrina de Sousa Magalhães et al.
Clinical Neuropsychiatry (2012) 9, 3
RAVLT, the TMT and the BVRT; these indices represent
the measurement of divergent and convergent validity,
respectively. Once the BVRT provided the immediate
and delayed memory rates, they were compared with
equivalent measures on the RAVLT, specifically Sum
A1-A5 (the most reliable between the immediate recall
indices) and A7. The measures of susceptibility to
proactive and retroactive interferences usually correlate
with measures of executive function (Mitrushina et al.
2005) and attention; therefore, they were also included
in the analysis.
The validity coefficient between Sum A1-A5 and
A7 of the RAVLT and the two TMT measures
demonstrated very weak and non-significant
correlations, which is evidence that they do not evaluate
the same function. None of the interference measures
of the RAVLT reached significant correlation with the
TMT measures. In contrast, regarding the BVRT
measures, the RAVLT mostly exhibited validity
coefficients of medium magnitude. The immediate
memory measures of the tests (Sum A1-A5 and Be1)
achieved a 0.44 correlation; the delayed memory
measures of both tests (A7 and Be2) reached a
correlation of 0.39. The Sum A1-A5 also demonstrated
a positive correlation of 0.37 with the Be2 measure but
did not verify a significant correlation between the
delayed memory recall of RAVLT (A7) and the imme-
diate memory BVRT measure (Be1).
Discussion
This study provides evidence for the validity and
reliability of the RAVLT. Some factors and conditions
are described (Hogan 2007) as non-reliable sources that
affect the measurements of the test and bring it below a
reliable level. Among the numerous sources that result
in non-systematic variations in test scores, three are
highlighted: (a) different criteria of test correction
adopted by the person that grades the test; (b) changes
in test application procedure; (c) the personal condition
of the examinees, i.e., the participant’s temporary self-
state (characteristics not related to the ones under
evaluation, such as health and mood conditions,
willingness and level of engagement), which might
influence the participant’s performance. This study aims
to contribute evidence for the applicability of the
RAVLT by determining whether this Brazilian
adaptation is a valid psychometric instrument in
neuropsychological assessment.
The RAVLT includes clear and specific
instructions, which minimize possible variations
between different researchers during the application and
correction process. Thus, this study aimed to neutralize
this non-reliable source. Procedures were standardized,
and all examiners were trained in the test application
policies related to physical environment preparation,
instructions and test administration. The objective was
to keep all these factors constant to prevent them from
interfering with both test situations. It has been reported
that the examiner has an important role in participant
performance, and his/her preparation for the
administration of the neuropsychological instruments
should not be neglected (Van den Burg and Kingma
1999). Finally, the third source of non-systematic error
was verified through test-retest methodology.
The weakest reliability correlation was observed
in the A2 measure (r = 0.36), and the strongest was
found in the Sum A1-A5 (r = 0.68). Researchers who
have studied the psychometric properties of the RAVLT
in different language versions all find the Sum A1-A5
as the most reliable measure of the instrument, although
with different coefficients: r = 0.77 (Geffen and Geffen
2000), r = 0.79 (Knight et al. 2007) e r = 0.70 (Van den
Burg and Kingma 1999). This finding was expected
because the reliability is proportional to the test
extension. Nevertheless, these data are not unanimous.
In another study, although the Sum A1-A5 demonstrated
a significant correlation of r = 0.55, between the test
and retest session, the higher correlation rate was found
in the A1 trial (r = 0.69) (Rezvanfard et al. 2011).
The second highest correlation rate obtained in the
present study was for the A4 measure (r
= 0.64), followed by the Recognition (r = 0.59) and the
A3 measure (r = 0.53). Evidence indicates that
subsequent trials of the RAVLT are usually more reliable
than the first ones (Van den Burg & Kingma, 1999).
The exception found in this study was the A5 measure,
which was expected to reach a high correlation rate but
obtained an average score of (r = 0.41). In contrast,
another study obtained a 0.61 rate with this same trial
(Van den Burg and Kingma 1999).
The proactive and retroactive interference rates
were quantified by the ratio between the two analyzed
scores instead of subtraction. This method (Geffen and
Geffen 2000) takes into account the different perfor-
mance levels of the participants. The results showed no
correlation between these rates in both experimental
situations. Even when alternate forms of the test were
used, a correlation between those interference measures
was not reported (Rezvanfard et al. 2011). Thereby, the
data suggest that the measures of higher reliability and
stability levels are Sum A1-A5, A3, A4, A6 and
Recognition because they obtained high correlation
rates. Other measures pointed to moderate reliability.
The RAVLT exhibited two measures, B1 and
Recognition, which showed no significant differences
between both test situations. It is possible to infer that
the learning process did not influence these two
measures; therefore, their original results were
maintained. After List B is shown one time in the first
session, we can assume that it retains its novelty. In the
case of Recognition, this measure may be influenced
by the ceiling effect because the mean of the scores is
lower than one standard deviation of the maximum score
(Uttl 2005). The Recognition task may not serve as a
good measure of mnemonic skills of healthy
participants. This task is proposed to assist in identifying
persons with suspected retrieval problems that might
benefit from mnemonic hints.
The results indicate that in all other RAVLT trials,
the effect of the practice can be credited to two different
main factors: (a) retention of specific test material (in
this case, words that compose the lists); (b) a
metamemoric factor, which considers how the exposure
to a similar task may improve performance strategies
(Mitrushina et al. 2005). If one considers implementing
a different interval between test-retest sessions, this data
could be distinct. Knight et al. (2007), when performing
an analysis of the mean scores of the test-retest sessions
135
Rey Auditory Verbal Learning Test
Clinical Neuropsychiatry (2012) 9, 3
(with a 1 year interval between applications), found
significant differences only for the measures A5, Sum
A1-A5 and A7. In addition, if the samples are also
different, another data configuration might be expected.
The Cronbach’s Alpha Coefficient represents the
equivalence degree regarding to realized measurement
(Hogan 2007), it is a homogeneity index of the items.
The high value to this coefficient (r = 0.80) indicates a
good internal consistency of the items that compose the
RAVLT. Therefore, all the trials are reliable measures
for learning processes and verbal memory. This measure
has a value close to the found for the same Portuguese
version of the test (r = 0.85) (Malloy-Diniz et al. 2007).
The RAVLT correlates moderately with other
learning measures (Strauss et al. 2006, Kessels et al.
2006, Helmstaedter et al. 2009), verbal memory (Strauss
et al. 2006, Helmstaedter et al. 2009) and spacial
memory (Strauss et al. 2006, Kessels et al. 2006),
therefore, have moderate convergent validity. Medium
magnitude evidence has been ascertained to the
convergent validity of the RAVLT with the BVRT, what
means that it, indeed, evaluates episodic memory. High
and moderate magnitude correlations (minimum r =
0.47 e maximum r = 0.54) were obtained by Duff et al.
(2005) when comparing the measures A6 and A7 of the
RAVLT with immediate and delayed visual memory
instruments. Probably, the correlations between RAVLT
and BVRT were not of a higher rate due to the different
structures of the tests, as they measure distinct aspects
of episodic memory. Whereas RAVLT assess mainly
verbal memory, BVRT assess visual memory and is
related to other cognitive skills as psychomotor abilities
(Strauss et al. 2006). Nonetheless, is important to note
that the mnemonic variables are together, regardless of
being verbal or nonverbal measures (Strauss et al. 2006),
which justifies the choice of BVRT in the present study.
The RAVLT demonstrated divergent validity
compared with the TMT, which supports the evidence
that both measure different cognitive functions.
However, other studies have presented contradictory
evidence regarding this correlation. Duff et al.
(2005) found a high and significant correlation of 0.51
between A7 and the performance on part B of the TMT.
This finding suggests a mutual relationship between
executive functions and verbal memory, which was not
verified in the present study. Although the proactive and
retroactive interference measures of the RAVLT were
expected to correlate with the TMT because they are
well-known executive measures, they did not exhibit
significant correlation. Additionally, for the German
version of the RAVLT, Helmstaedter et al. (2009) did
not find any significant correlation between A7 and the
tests that measure attention and fluency, both aspects
of executive function. It is important to highlight the
methodology differences regarding the sample
composition, chosen instrument and data analysis of the
mentioned studies. The range in which the measures of
the RAVLT correlate with executive functions still
requires additional empirical evidence.
Conclusions
Learning and memory difficulties are the most
common complaints in patients with neurological
impairment and significantly impact the daily life
activities and social functionality of these patients
(Messinis et al. 2007). The RAVLT is widely known to
offer objective methods to help identify these deficits.
Furthermore, repeated evaluations are often frequent
and necessary in the clinical context to closely follow
the progress of degenerative conditions, the effect of
prescribed drugs or recovery after intervention
(Uchiyama et al. 1995).
It is valid to note that all measurements obtained
by a neuropsychological test are subject to the influence
of non-reliable sources. Each investigation method seeks
to deal with one or more of the non-systematic variations
of the test scores. Therefore, one can assert that once
each reliability measures encompasses a part of each
possible source of variation, a single reliability measure
does not exist for a test (Hogan 2007).
Among the resultant limits of the test-retest
methodology choice for reliability measurement are: (a)
the difficulty of applying the test due to the fact that
participants need to cooperate to reestablish the same
application context in a second round; and, more
importantly, (b) decisions concerning the effects that
the first application might have on the second one. The
learning effect is a very important point to consider
because of its proclivity to inflate the reliability
coefficient and thus overestimate the measured attribute.
There are vast descriptions in the literature about
the RAVLT’s sensitivity to the effect of learning in a
second application of the same version of the test. After
successive applications, a small, but significant
enhancement in the number of recalled words (1 or two
words mean per trial) is expected (Mitrushina et al.
2005). The effect of the practice is diminished when
the participants are not exposed to the same list, which
is why the use of alternative lists is suggested to
minimize this effect (Uchiyama et al. 1995, Strauss et
al. 2006, Van den Burg and Kingma 1999, Beglinger et
al. 2005, Mitrushina et al. 2005, Knight et al. 2007).
Those studies that have applied the test-retest
methodology utilizing alternate forms of the RAVLT
found a lesser effect of practice (Strauss et al. 2006,
Va n d en B u r g a nd K i n gm a 1 9 9 9, Mi t r us h i na et a l . 2 0 05 ,
Rezvanfard et al. 2011). However, these data are not
unanimous; Uchiyama et al. (1995) demonstrated a
significant effect of practice on two distinct versions of
the RAVLT, using a 1-year interval between trials. Those
researchers therefore suggested the need for longitudinal
normative data for interpretation of the retest data
(Uchiyama et al. 1995).
Strauss et al. (2006) observed that with intervals
greater than 1 year, the RAVLT demonstrates moderate
test-retest reliability. In this study, the one-month interval
between applications was demonstrated to be
acceptable. However, this interval time is inferior when
observed in clinical practice, where the monitoring and
successive evaluations are performed with longer time
intervals, which varying from 3 months to 1 year, for
example.
In the Portuguese language and Brazilian culture,
a second version that meets the RAVLT construction
criteria is unfortunately not yet available. This is a
problem for which Brazilian researchers will have to
find viable solutions. An alternative would be to use
the first list published by Malloy-Diniz et al. (2000).
136
Sabrina de Sousa Magalhães et al.
Clinical Neuropsychiatry (2012) 9, 3
However, this list is a direct translation of the English
version and demonstrates several limitations, such as
not taking into account the word frequency in the
Portuguese language and the number of syllables that
the newest version has presented (Malloy-Diniz et al.
2007).
The RAVLT evaluates the participant’s ability to
encode, consolidate, store and recall verbal information.
The most reliable measures in terms of stability and
low measurement error (greater reliability) were, in
regressive order, the Sum A1-A5, A4, Recognition, A6
and A3. The total score performance represented by Sum
A1-A5 is substantially more reliable than the items
considered individually.
Compared with other RAVLT psychometrical
property studies (Strauss et al. 2006, Van den Burg and
Kingma 1999, Geffen and Geffen 2000, Duff et al. 2005,
Kessels et al. 2006, Knight et al. 2007, Helmstaedter et
al. 2009), the current research achieves modest, although
relatively high, correlation values. The small sample
size might have been a contributing factor to the lower
reliability coefficient expression.
To establish validity, an integrated collection of
evidence is required for the appropriate interpretation
of the analyzed instrument’s score. The results indicate
that the verbal memory measured by the RAVLT
positively correlates with the visual memory evaluated
by BVRT but does not correspond to the executive
functions assessed by the TMT. Although a vast amount
of evidence suggests that memory and executive
functions correlate with each other (Tremont et al.
2000), this study did not aim to verify the magnitude of
their interaction. As such, the chosen methodology
effectively supported the proposed objectives and
verified that the RAVLT measures mnemonic attributes
rather than ones related to executive functions.
Reliability and validity are distinct but
complementary constructs and are important in the field
of psychometrics. A test can be valid without being
reliable; however, a test cannot be reliable if it is not
valid (Hogan 2007). Therefore, the relationship between
these two properties is a fundamental consideration
because the validity of the test depends considerably
on its degree of reliability.
We aimed to determine whether the RAVLT is a
safe source of information about the mnemonic
construct. The collected results suggest that the RAVLT
shows good reliability and validity, and therefore this
psychometric instrument can be considered to provide
valid measurement of episodic verbal memory in
neuropsychological evaluation. This study proposed to
contribute psychometrical evidence of the validity and
reliability of the RAVLT, but because of the relatively
small sample, the outcomes must be interpreted with
caution. New investigations are expected to supplement
these results and strengthen the data that has thus far
collected.
References
Beglinger LJ, Gaydos B, Tangphao-Daniels O, Duff K, Kareken
DA, Crawford J, Fastenau PS, Siemers ER (2005). Practice
effects and the use of alternate forms in serial
neuropsychological testing. Archives of Clinical
Neuropsychology 20, 517-29.
Bertolucci PH, Brucki SM, Campacci SR, Juliano Y (1994).
The Mini-Mental State Examination in a general
population: impact of education status. Arquivos de
Neuropsiquiatria 52, 1, 1-7.
Cohen J (1988). Statistical Power Analysis for the Behavioral
Sciences, 2nd ed. Lawrence Erlbaum Associates, New
Jersey.
De Paula JJ, Melo LPC, Nicolato R, Moraes EN, Bicalho MA,
Hamdan AC, Malloy-Diniz LF (2012). Reliability and
construct validity of the Rey-Auditory Verbal Learning Test
in Brazilian elders. Revista de Psiquiatria Clínica 39, 1,
19-23.
Duff K, Schoenberg MR, Scott JG, Adams, RL (2005). The
relationship between executive functioning and verbal and
visual learning and memory. Archives of Clinical
Neuropsychology 20, 111-22.
Fichman HC, Dias LBT, Fernandes CS, Lourenço R, Caramelli
P, Nitrini R (2010). Normative data and construct validity
of the Rey Auditory Verbal Learning Test in a Brazilian
Elderly population. Psychology & Neuroscience 3, 1, 79-
84.
Geffen G, Geffen L (2000). Auditory Verbal Learning Test
(AVLT): computerised scoring program and population
norms. Acer Press, Australia.
Hakstian AR, Whelan TE (1976). A k-sample significance test
for independent alpha coefficients. Psychometrika 41, 219-
231.
Hamdan, AC. Hamdan EMLR (2009). Effects of age and
education level on the Trail Making Test in A healthy
Brazilian sample. Psychology & Neuroscience 2, 2, 199-
203.
Hancock P, Larner AJ (2011). Test Your Memory test: diagnostic
utility in a memory clinic population. International Journal
of Geriatric Psychiatry 26, 976-80.
Helmstaedter C, Wietzke J, Lutz MT (2009). Unique and shared
validity of the ‘‘Wechsler logical memory test’’, the
‘‘California verbal learning test’’, and the ‘‘verbal learning
and memory test’’ in patients with epilepsy. Epilepsy
Research 87, 203-12.
Hogan TP (2007). Psychological testing: a practical introduction
(2nd ed.). Hoboken, NJ: John Wiley and Sons.
Kessels RPC, Nys GMS, Brands AMA, van den Berg E, van
Zandvoort MJE (2006). The modified Location Learning
Test: Norms for the assessment of spatial memory function
in neuropsychological patients. Archives of Clinical
Neuropsychology 21, 841-46.
Knight RG, McMahon J, Skeaff, CM, Green, TJ (2007). Reliable
Change Index scores for persons over the age of 65 tested
on alternate forms of the Rey AVLT. Archives of Clinical
Neuropsychology 22, 513-18.
Lezak MD, Howieson DB, Loring DW (2004).
Neuropsychological assessment, 4th ed. Oxford University
Press, New York.
Magalhães SS, Hamdan AC (2010). The Rey Auditory Verbal
Learning Test: normative data for the Brazilian population
and analysis of the influence of demographic variables.
Psychology & Neuroscience 3, 1, 85-91.
Malloy-Diniz LF, Lasmar VAP, Gazinelli LSR, Fuentes D,
Salgado, JV (2007). The Rey Auditory-verbal Learning
Test: applicability for the brazilian elderly population.
Revista Brasileira de Psiquiatria 29, 4, 324-29.
Malloy-Diniz LFM, Cruz MF, Torres VM, Cosenza RM (2000).
O teste de Aprendizagem Auditivo-Verbal de Rey: normas
para uma população Brasileira. Revista Brasileira de Neu-
rologia 36, 79-83.
Messinis L, Tsakona I, Malefaki S, Papathanasopoulos P (2007).
Normative data and discriminant validity of Rey’s Verbal
Learning Test for the Greek adult population. Archives of
Clinical Neuropsychology 22, 739-52.
Milian M, Leiherr AM, Straten G, Müller S, Leyhe T, Eschweiler
GW (2011). The Mini-Cog versus the Mini-Mental State
Examination and the Clock Drawing Test in daily clinical
137
Rey Auditory Verbal Learning Test
Clinical Neuropsychiatry (2012) 9, 3
practice: screening value in a German Memory Clinic.
International Psychogeriatrics 15, 1-9.
Mitrushina MN, Bone KB, D’Elia LF (2005). Handbook of
normative data for neuropsychological assessment, 2nd
ed. Oxford University Press, New York.
Rezvanfard M, Ekhtiari, H, Noroozian M, Rezvanifar A, Nilipour
R, Javan GK (2011). The Rey Auditory Verbal Learning
Test: alternate forms equivalency and reliability for the
Iranian adult population (Persian version). Archives of
Iranian Medicine 14, 2, 104-9.
Stallings GA, Boake C, Sherer M (1995). Comparison of the
California Verbal Learning Test and the Rey Auditory Verbal
Learning Test in head-injured patients. Journal of Clinical
and Experimental Neuropsychology 17, 5, 706-712.
Strauss E, Sherman EMS, Spreen O (2006). A compendium of
neuropsychological tests: administration, norms, and
commentary, 3rd ed. Oxford University Press, New York.
Tremont G, Halpert S, Javorsky DJ, Stern RA (2000). Differential
impact of executive dysfunction on verbal list learning and
story recall. Clinical Neuropsychology 14, 3, 295-302.
Uchiyama CL, D’Elia LF, Dellinger AM, Becker JT, Selnes OA,
Wesch JE, Chen BB, Satz P, can Gorp W, Miller EN (1995).
Alternate forms of the Auditory-Verbal Learning Test:
issues of test comparability, longitudinal reliability, and
moderating demographic variables. Archives of Clinical
Neuropsychology 10, 2, 133-45.
Uttl B (2005). Measurement of Individual Differences: lessons
from memory assessment in Research and Clinical
Practice. Psychological Science 16, 6, 460-67.
Van den Burg W, Kingma A (1999). Performance of 225 Dutch
School Children on Rey’s Auditory Verbal Learning Test
(AVLT): Parallel Test-Retest Reliabilities with an Interval
of 3 Months and Normative Data. Archives of Clinical
Neuropsychology 14, 6, 545-59.
