Psychological Assessment Copyright 2004 by the American Psychological Association

2004, Vol. 16, No. 2, 146 –154 1040-3590/04/$12.00 DOI: 10.1037/1040-3590.16.2.146

Education-Stratified Base-Rate Information on Discrepancy Scores Within

and Between the Wechsler Adult Intelligence Scale—Third Edition and the
Wechsler Memory Scale—Third Edition

Galit A. Dori Gordon J. Chelune

Cleveland Clinic Foundation and Veterans Affairs Medical Cleveland Clinic Foundation
Center, Cleveland, Ohio
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The Wechsler Adult Intelligence Scale—Third Edition (WAIS–III; D. Wechsler, 1997a) and the Wech-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

sler Memory Scale—Third Edition (WMS–III; D. Wechsler, 1997b) are 2 of the most frequently used
measures in psychology and neuropsychology. To facilitate the diagnostic use of these measures in the
clinical decision-making process, this article provides information on education-stratified, directional
prevalence rates (i.e., base rates) of discrepancy scores between the major index scores for the WAIS–III,
the WMS–III, and between the WAIS–III and WMS–III. To illustrate how such base-rate data can be
clinically used, this article reviews the relative risk (i.e., odds ratio) of empirically defined “rare”
cognitive deficits in 2 of the clinical samples presented in the WAIS–III—WMS–III Technical Manual
(The Psychological Corporation, 1997).

As healthcare organizations and providers seek to better manage
the cost effectiveness of patient diagnosis and treatment, evidence-
based approaches have become increasingly important. Within the
context of evidence-based medicine, treatment effectiveness is
maximized through the integration of three factors: empirical
research findings, clinical expertise, and patient values (Sackett,
Straus, Richardson, Rosenberg, & Haynes, 2000). Central to
evidence-based practice is the use of diagnostically valid tests (i.e.,
tests that have been demonstrated to facilitate clinical decision
making regarding whether a given individual does or does not have
a specified condition of interest; Chelune, 2002; Ivnik et al., 2001).
Most traditional tests in psychology and neuropsychology have
been validated using inferential statistics and the null-hypothesis
significance testing paradigm. These methods evaluate whether
observed differences between mean group performances are sta-
tistically significant in terms of reliability, but they cannot inform
individual diagnoses (Ivnik et al., 2000; Smith, Cerhan, & Ivnik,
Galit A. Dori, Department of Neurology, The Mellen Center, Cleveland
Clinic Foundation, and Psychology Service, Department of Veterans Af-
fairs Medical Center, Cleveland, Ohio; Gordon J. Chelune, Department of
Neurology, The Mellen Center, Cleveland Clinic Foundation.
Galit A. Dori is now at Psychological and Behavioral Consultants,
Beachwood, Ohio.
Portions of this article were presented at the 106th Annual Meeting of
the American Psychological Association, San Francisco, August 1998.
Data presented in this article were obtained during the Wechsler Adult
Intelligence Scale—III and Wechsler Memory Scale—III standardization
projects and were graciously provided with permission to reproduce from
The Psychological Corporation.
Correspondence concerning this article should be addressed to Gordon
J. Chelune, Department of Neurology, The Mellen Center (U-10), Cleve-
land Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195.
E-mail: cheluneg@ccf.org
146
2003). For example, at the level of the individual, if a 30-year-old
person obtains a Verbal IQ score of 95 and a Performance IQ score
of 105 on the Wechsler Adult Intelligence Scale—Third Edition
(WAIS–III; Wechsler 1997a), the 10-point difference between the
two scores would be statistically reliable at the .05 level. At this
level of probability, the null hypothesis of no difference can be
rejected, indicating that the observed level of difference between
Verbal and Performance IQs is a reliable and repeatable difference
and is not likely due to measurement error or chance fluctuations
in the scores. However, a review of WAIS–III base-rate data
indicates that a 10-point discrepancy between these two scores is
found in approximately 37% of the standardization sample
(Wechsler, 1997a), suggesting that a 10-point difference is a
relatively common finding in the general population and not likely
to be meaningful in terms of clinical decision making.
To increase the applicability of test findings on an individual
basis, base-rate information on prevalence is needed in addition to
reliability data. Whereas analyses of group mean differences yield
information on how much aggregated change has taken place,
base-rate analyses can determine how individual changes contrib-
ute to mean changes (Smith, 2002). Statistical methods that focus
on diagnostic validity and base rates of a test in terms of clinical
decision making are commonly used in epidemiology and
evidence-based medicine (Sackett, Haynes, Guyatt, & Tugwell,
1991). These statistics are base-rate dependent and involve such
constructs as sensitivity, specificity, and positive and negative
predictive power. Base-rate information also allows for statistical
comparisons between groups (e.g., odds ratios) that characterize
the differential risk of having a condition of interest in terms of
clinically relevant test parameters (Ivnik et al., 2001; Smith et al.,
2003). Although statistically reliable differences are necessary to
establish the consistency of findings, such findings may be of little
clinical significance if they are common in the general population.
In contrast, base-rate analysis can be applied in a clinically mean-
 DISCREPANCY-SCORE BASE RATES
ingful way by identifying sufficiently rare and potentially abnor-
mal scores (i.e., those that are more likely to be obtained in a
population that is external to the normative group) that may be of
diagnostic value (see Chelune, 2002; Matarazzo & Herman, 1984).
Base-rate analysis of discrepancy scores may be particularly
relevant in clinical populations because it allows for contrasts
between cognitive areas presumed to be compromised with areas
of functioning commonly spared for a given clinical group. Base-
rate analysis also controls for baseline differences among individ-
uals. Discrepancy analysis has a long tradition in psychological
and neuropsychological research and practice. Early on, clinicians
looked to the original Wechsler-Bellevue Scale (Wechsler, 1939;
see Matarazzo, 1972, for review) to test the idea that certain
intellectual abilities hold up with age whereas others do not (i.e.,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
“hold– don’t hold” deterioration ratio). Contemporary research
continues to explore the clinical use of discrepancy scores to
identify patients at risk for neurological conditions such as Alz-
heimer’s disease (e.g., Hawkins & Tulsky, 2001, 2003; Jacobson,
Delis, Bondi, & Salmon, 2002). Discrepancy analysis is particu-
larly useful in contrasting factors of intelligence to memory given
the expected cognitive discrepancies associated with neurologic
damage. Using aspects of verbal functioning that are least likely to
show decline in neurologic injury to estimate premorbid level (e.g.,
verbal comprehension, reading level, or sight reading of words)
and contrasting them to aspects of memory frequently compro-
mised can assist in clinical decision making (see Bornstein, Che-
lune, & Prifitera, 1989; Chelune, 1998; Hawkins, 1998; Manly,
Jacobs, Touradji, Small, & Stern, 2002; Milner, 1954; Taylor &
Heaton, 2001). Thus, information on base rates of discrepancy
scores can provide additional empirical evidence for the diagnostic
value of neuropsychological measures in patient settings, combin-
ing information on both statistical significance and clinical abnor-
mality (Kaufman, 1990).
The recent revisions of the WAIS–III and the Wechsler Memory
Scale—Third Edition (WMS–III; Wechsler, 1997b) are two of the
most commonly used assessment instruments in clinical settings. A
major addition to the manuals of these revised measures is the
inclusion of base-rate information on the prevalence of various
discrepancy scores in the standardization sample. The base-rate
discrepancy information for within-test variables (e.g., Verbal IQ
versus Performance IQ) is presented in the respective WAIS–III
and WMS–III manuals (Wechsler, 1997a, 1997b) and Tables
D1–D5 of the WAIS–III—WMS–III Technical Manual (The Psy-
chological Corporation, 1997, 2002). In addition, because the
WMS–III was co-normed with a subset of the WAIS–III standard-
ization sample, unique prevalence data for between-test discrep-
ancies (e.g., Full Scale IQ vs. General Memory) are presented in
Tables B7–B9 and C4 –C6 of both the original and updated ver-
sions of the WAIS–III—WMS–III Technical Manual. When using
these base-rate tables, the reader should make careful note that the
prevalence data for within-test discrepancy scores in the respective
manuals (Wechsler, 1997a, 1997b) and in the WAIS–III—WMS–III
Technical Manuals are presented as cumulative percentages of the
absolute discrepancy between the scores of interest, regardless of
the directionality of the difference score (e.g., |Verbal IQ–
Performance IQ|). In contrast, the base-rate information for dis-
crepancies derived from between-test comparisons (cf. Tables
B.7–B.9 and Tables C.4 –C.6, The Psychological Corporation,
1997, 2002) is unidirectional, with the WMS–III scores always
147
subtracted from the WAIS–III variables (e.g., Full Scale IQ minus
General Memory Index). The difference between bidirectional
(absolute) and unidirectional discrepancy information is similar to
that between two-tail and one-tail comparisons. Cumulative, ab-
solute discrepancy information is useful when simply looking for
unusual strengths or weakness, without any specific expectations
(Tulsky, Rolfhus, & Zhu, 2000). However, in clinical settings
where a priori hypotheses exist about the functioning of an indi-
vidual (i.e., one knows the direction of the patient’s discrepancy),
base-rate information regarding directional discrepancies is clini-
cally relevant. Recognizing that clinicians may not appreciate the
difference between unidirectional and absolute cumulative scores,
Tulsky and colleagues (2000) examined the raw frequency data of
discrepancy scores for the WAIS–III. They report that the simple
procedure of dividing in half the cumulative absolute frequencies
reported in either the WAIS–III Administration and Scoring Man-
ual (Wechsler, 1997a) or the WAIS–III—WMS–III Technical Man-
ual (The Psychological Corporation, 1997) will yield a reasonably
close approximation of the actual unidirectional base rates, partic-
ularly at extreme levels of discrepancy.
Another factor affecting the clinical use of simple difference
discrepancy scores as presented in the original test manuals is the
observation that discrepancy base rates are not equally distributed
across ability levels. Hawkins and Tulsky (2001) found that large
discrepancies between IQ and WMS–III General Memory were
more frequent among individuals with above average IQ scores but
less frequent among individuals with below average IQ. They
argued that use of simple difference discrepancy tables that are not
stratified by ability level could lead to erroneous conclusions
(Hawkins & Tulsky, 2001, 2003). Although the original WAIS–
III—WMS–III Technical Manual (The Psychological Corporation,
1997) did not provide any base-rate information on discrepancy
scores that was stratified by ability level, the WAIS–III—WMS–III
Technical Manual Update (The Psychological Corporation, 2002)
now provides tables on intratest WAIS–III discrepancies stratified
by Full Scale IQ (cf. Tables D.1–D.5, 2002); however, these tables
again reflect the base rates of absolute (bidirectional) discrepancy
scores. A separate table (cf. Table D.6) is provided for directional
WAIS–III discrepancies, but these are not stratified. Likewise,
directional, but nonstratified, discrepancy data are presented for
WMS–III discrepancies in Table D.7 of the WAIS–III—WMS–III
Technical Manual Update.
The present investigation expands on the recent work by Tulsky
et al. (2000), Hawkins and Tulsky (2001, 2003), and the discrep-
ancy data presented by The Psychological Corporation (1997,
2002). Specifically, we present educationally stratified, directional
prevalence rates of discrepancy scores between the major index
scores for the WAIS–III and the WMS–III and between the
WAIS–III and WMS–III that can be directly used by clinicians to
facilitate clinical decision making at the level of the individual. In
addition to providing new discrepancy base-rate data for the
WAIS–III and WMS–III and comparisons between the WAIS–III
and WMS–III, the current study also provides unidirectional cut
scores that define simple difference prevalence rates at decision
points relevant for both clinical practice and research purposes
(i.e., 5%, 10%, and 15% levels).
We elected to stratify our base-rate tables on the basis of
education rather than on current Full Scale IQ for two reasons.
First, education is highly correlated with Full Scale IQ (FSIQ) in
 148 DORI AND CHELUNE
the standardization sample (r ⫽ .51). Since the WAIS–III Index
scores contribute to FSIQ, stratifying on the basis of education
rather than FSIQ allows us to avoid confounding many of our
dependent measures (i.e., discrepancy scores within and across
WAIS–WMS comparisons) with our independent (stratification)
variable. Second, demographic variables such as education are
likely to be more useful for estimating premorbid cognitive ability
in clinical samples (Heaton, Taylor, & Manley, 2003) than mea-
sures of current IQ, which themselves may be affected by cerebral
dysfunction; that is, stratifying on the basis of a measure of current
cognitive ability may attenuate the sensitivity of the discrepancy
variables to the condition of interest, as has been observed for
many years (Heaton, Baade, & Johnson, 1978). Thus, use of
education to stratify discrepancy scores not only provides a good
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
analog for general ability level in neurologically normal samples
but is independent of the effects of the condition of interest to
which the discrepancy scores may be a sensitive marker among
clinical populations.
To illustrate the usefulness of base-rate analysis in providing
clinically meaningful data, we compare the base rates of selected
discrepancy scores between the WAIS–III—WMS–III standard-
ization samples and two of the clinical samples presented in
WAIS–III—WMS–III Technical Manual (The Psychological Cor-
poration, 1997, 2002), namely those with Alzheimer’s disease and
those with Huntington’s disease. By comparing estimated relative
risk for abnormal or rare discrepancies using odds ratio analysis,
we demonstrate how base-rate data can be applied to define
markers that inform the decision-making process on an individual
basis. Given the small size of the clinical samples, the clinical
illustration is intended only as an exercise in the method of
base-rate analysis, and the information presented here is not meant
to characterize these populations on a general basis.
Method
Participants
Data from the WAIS–III and WMS–III standardization samples and two
clinical samples acquired during the clinical field trials of the standardiza-
tion project were obtained with permission from The Psychological Cor-
poration (San Antonio, TX). Detailed descriptions of the demographic
characteristics of these samples are provided in the original and updated
versions of the WAIS–III—WMS–III Technical Manual (The Psychological
Corporation, 1997, 2002) and are only briefly described here.
Standardization Samples
The standardization samples for both the WAIS–III (Wechsler, 1997a)
and WMS–III (Wechsler, 1997b) were stratified by age, gender, race/
ethnicity, educational level, and geographical region corresponding to the
U.S. population census from 1995. Overall, the standardization samples
can be viewed as estimates of the general U.S. population. Individuals with
medical or psychiatric conditions that potentially could have affected
cognitive performance were excluded from the standardization sample (for
exclusionary criteria see Table 2.1, The Psychological Corporation, 1997,
2002).
WAIS–III. A total of 2,450 individuals from the general population
compose the WAIS–III standardization sample. For discrepancy scores
involving the WAIS–III Working Memory Index, only the 1,299 individ-
uals who completed the Letter–Number Sequencing subtest were included
in comparisons.
WMS–III. A weighted sample (Tulsky & Ledbetter, 2000) of 1,250
individuals makes up the WMS–III standardization sample. These individ-
uals were a subset of the standardization sample of the WAIS–III and were
administered the WMS–III as well as the WAIS–III.
WAIS–III—WMS–III. Base-rate information for discrepancy scores be-
tween the WMS–III and WAIS–III were drawn from the weighted sample
(Tulsky & Ledbetter, 2000) of 1,250 individuals who completed both tests.
Patient Groups
The patient groups consisted of two clinical samples tested during field
trials of the WAIS–III and WMS–III (see the WAIS–III—WMS–III Tech-
nical Manual for details; The Psychological Corporation, 1997, 2002).
Rather than using other possible clinical samples, clinical samples from
The Psychological Corporation’s field trials were selected to ensure that
the administration of the tests was standardized and consistent across both
the normal and clinical groups being compared. The samples selected for
examination in the current study include individuals with mild Alzheimer’s
disease (AD; n ⫽ 35) and Huntington’s disease (HD; n ⫽ 15). The
demographic data for these patient samples are presented in the WAIS–
III—WMS–III Technical Manual (The Psychological Corporation, 1997,
Table 4.40; 2002, Table 4.31).
Procedure
Construction of base-rate tables. Discrepancy-score base-rate tables
were stratified on the basis of four education levels: less than 12 years, 12
years, 13–15 years, and 16 or more years. Discrepancies between potential
indexes of interest within and between the WAIS–III and WMS–III were
manually constructed from computer-generated frequency tables. Discrep-
ancies involving FSIQ were not included because it is largely a composite
of the four WAIS–III factor scores. The Auditory Recognition Delay Index
was also not included because scores on this index are not normally
distributed, and it has been suggested that they are more appropriately
reported as percentile scores (see Tulsky, Chiaravalloti, Palmer, & Che-
lune, 2003).
Unidirectional (one-tail) base-rate information is reported for the 5%,
10%, and 15% levels (⫾1.64, ⫾1.28, and ⫾1.04 z scores, respectively,
under the normal curve). A conservative approach was taken, selecting
whole scores closest to the chosen percentage levels without exceeding the
specified percentage level. For example, if a percentile level of 5% was
selected and a 24-point discrepancy occurred in 5.2% of the sample and a
25-point discrepancy occurred in 4.7% of the sample, the 25-point discrep-
ancy was chosen and listed in the table under the 5% column.
Estimated relative risk analyses. Odds ratio analyses were used to
calculate estimated relative risk of selected cognitive deficits in the AD and
HD samples. The odds ratio (OR) represents the ratio of two odds: the odds
of having the condition of interest when a factor is present compared with
the odds of having the condition of interest when the factor is absent
(Sackett et al., 2000). An OR that is not significantly greater than 1.0
indicates no difference in probability or relative risk of having the condi-
tion of interest regardless of the presence of the factor or whether there is
a statistically significant mean group difference (Bieliauskas, Fastenau,
Lacy, & Roper, 1997). Similar to p values, there is a confidence band that
surrounds an OR and provides information on its reliability. Typically,
ORs are reported with a 95% confidence interval (CI). ORs that have CI
ranges in which the lower boundary does not include 1.0 are interpreted as
reliable (e.g., 95 out of 100 times at a CI of 95%), similar to group mean
differences that are significant at p ⬍ .05 (Sackett et al., 1991).
In calculating the ORs, the WAIS–III Verbal Comprehension Index
(VCI) was used as the reference score with which to contrast Immediate
Memory (IM), General Memory (GM), and Processing Speed Index (PSI)
scores. Because verbal abilities are often thought to remain relatively intact
even with cognitive decline in neurologic conditions (see Hawkins &
 DISCREPANCY-SCORE BASE RATES 149

Tulsky, 2003; Matarazzo, 1972), the other variables were always sub-
tracted from VCI. The IM, GM, and PSI indexes were selected as potential
markers of compromised functioning because both AD and HD are clini-
cally differentiated by their variable associations with memory impairment
and slowed processing speed (Fields, 1998).
For the purpose of our clinical illustrations, we defined cognitive im-
pairment as an unusually large or rare discrepancy found in ⱕ10% (z ⫽
–1.28) of the standardization sample. Taylor and Heaton (2001) recom-
mended using a z score of –1.00 (16th percentile) to define an impairment
to maximize both sensitivity and specificity in clinical settings. However,
we selected a more conservative 10% level (–1.28 z score) that would
increase specificity because of the small sample size of the current patient
groups. By comparing the prevalence of cognitive deficits in the patient
groups versus in the general population, we determined the estimated
relative risk of having a single cognitive deficit or a combination of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Tables 1 and 2 present within-test base-rate information for the
WAIS–III and WMS–III, respectively, whereas Table 3 presents
base-rate data for between-test discrepancy scores. To illustrate
how these tables can be used, assume that a clinician observed a
patient with12 years of education to have a VCI score 13 points
greater than PSI. Looking in Table B.1 of the WAIS–III Adminis-
tration and Scoring Manual (Wechsler, 1997a), the clinician
would find that a VCI–PSI discrepancy of this magnitude is
statistically reliable at p ⱕ .05 for all ages. However, examining
the unidirectional base-rate data for VCI–PSI presented in Table 1
of this article, where VCI (V1) is greater than PSI (V2), the
clinician would discover that this discrepancy is not especially rare
in the general population, as more than 15% of the standardization
sample with 12 years of education had discrepancies greater than

cognitive deficits given the presence of known neurologic conditions.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

or equal to 14 points. Alternately, had the patient manifested a

Multiple deficits were evaluated to exemplify that even if scores in one
area did not differ significantly between the clinical groups, a pattern of
scores could be used to inform a clinical decision. Estimated relative risk
for one, two, and three abnormal discrepancies were calculated only to
illustrate how patterns of test scores can aid the diagnostic process, not to
make general clinical statements regarding the cognitive characteristics of
the conditions of interest examined in this study.
Results
Base-Rate Tables
Education-stratified cut-off scores defining the prevalence rates
of various unidirectional discrepancy scores observed in the
WAIS–III and WMS–III standardization samples are presented in
Tables 1 to 3. As can be seen from review of these tables,
discrepancy scores that define chosen prevalence rates vary direc-
tionally, particularly at extreme educational levels (i.e., educa-
tion ⬍ 12 years and education ⱖ16 years). For example, in Table
3, the VIQ–GM discrepancy at the 5% level for education ⬍ 12 is
defined by a discrepancy score of 18 points for VIQ ⬎ GM and 28
points for VIQ ⬍ GM, whereas the VIQ–GM discrepancy at the
5% level for education ⱖ16 is defined by a discrepancy score of 38
points for VIQ ⬎ GM and 17 points for VIQ ⬍ GM.
25-point VCI ⬎ PSI discrepancy, this difference would not only be
statistically reliable but also relatively rare, being expected to
occur in 5% or less of the general population. Using data from
Tables 1, 2, and 3, researchers and clinicians can empirically
describe observed discrepancy scores at the level of the individual
in terms of their rarity, adding to the reliability information pro-
vided by The Psychological Corporation (1997, 2002).
Clinical Illustration Analyses
To illustrate how the educationally stratified base-rate informa-
tion presented in Tables 1, 2, and 3 can be applied in a clinically
meaningful manner, prevalence rates of impairments in the AD
and HD clinical samples tested during the WAIS–III and WMS–III
field trials were compared with prevalence rates of extreme (rare)
discrepancies in the standardization sample. Discrepancy scores
found in 10% or less of the standardization sample were used to
define clinically rare and potentially abnormal differences. As
noted earlier, because verbal abilities are often well preserved in
neurologic conditions affecting cognition, VCI was selected as the
reference score against which to assess whether IM, GM, or PSI
scores were unusually depressed. Frequencies and estimated rela-
tive risk (as determined by OR analysis) of rare discrepancies

Table 1
Unidirectional Base-Rate Information for WAIS–III Discrepancy

Education ⬍ 12a Education ⫽ 12b Education ⫽ 13–15c Education ⱖ 16d

WAIS–III
indexes V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2

(V1)–(V2) 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5%
VIQ–PIQ 16 12 10 13 15 18 17 13 10 13 16 19 21 17 13 12 15 19 23 18 15 9 11 15
VCI–POI 17 12 11 14 17 23 20 15 12 15 19 22 24 18 15 14 18 22 26 20 18 8 12 17
VCI–WMI 17 14 12 12 16 21 20 16 13 14 18 23 23 18 14 15 20 23 29 22 19 9 13 18
VCI–PSI 20 14 11 16 20 25 23 18 14 18 21 28 27 21 18 17 21 26 30 25 22 12 14 21
POI–WMI 25 18 14 13 15 18 25 18 15 14 18 23 21 16 12 17 21 28 23 16 13 14 17 21
POI–PSI 20 15 12 13 16 20 24 19 15 16 20 24 24 19 15 18 21 27 26 21 18 14 18 22
WMI–PSI 22 16 12 14 17 25 26 19 15 16 20 25 29 10 17 18 22 27 25 22 17 15 17 22

Note. WAIS–III ⫽ Wechsler Adult Intelligence Scale—Third Edition; V1 ⫽ first variable listed; V2 ⫽ second variable listed; VIQ ⫽ Verbal IQ; PIQ ⫽
Performance IQ; VCI ⫽ Verbal Comprehension Index; POI ⫽ Perceptual Organization Index; WMI ⫽ Working Memory Index; PSI ⫽ Processing Speed
Index. Copyright © 1997 by Harcourt Assessment, Inc. Data reproduced by permission. All rights reserved.
a
N ⫽ 570 for comparisons without WMI; N ⫽ 310 for comparisons with WMI. b N ⫽ 855 for comparisons without WMI; N ⫽ 459 for comparisons
with WMI. c N ⫽ 584 for comparisons without WMI; N ⫽ 289 for comparisons with WMI. d N ⫽ 441 for comparisons without WMI; N ⫽ 241 for
comparisons with WMI.
 150 DORI AND CHELUNE

Table 2
Unidirectional Base-Rate Information for WMS–III Discrepancy Scores

Education ⬍ 12a Education ⫽ 12b Education ⫽ 13–15c Education ⱖ 16d

WMS–III
indexes V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2

(V1)–(V2) 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5%
WM–IM 27 20 17 18 21 28 26 22 19 16 20 26 32 26 20 17 22 26 33 20 16 21 24 27
IM–GM 11 10 8 8 9 12 12 10 8 8 10 11 12 8 8 8 9 12 13 9 8 9 11 13
WM–GM 25 20 16 18 21 27 26 21 17 17 20 25 31 26 18 18 23 26 27 21 18 17 22 26
WM–AudI 29 23 18 18 22 27 29 23 18 16 21 27 29 24 20 16 22 30 27 20 17 19 22 26
WM–AudD 27 22 17 20 24 30 30 23 20 14 19 25 31 23 18 18 21 30 26 18 14 17 21 25
WM–VisI 28 22 18 22 26 36 28 22 18 19 22 28 36 31 25 17 20 25 37 25 21 20 25 35
WM–VisD 22 19 15 22 26 31 28 21 16 19 24 30 34 28 25 17 19 26 36 25 19 18 21 28
AudI–AudD 14 10 8 10 13 15 14 11 9 8 10 12 13 10 7 7 10 11 11 10 9 8 10 12
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

VisI–VisD 15 11 10 10 12 14 13 10 7 10 11 15 16 10 8 10 11 16 16 13 10 10 13 17
This document is copyrighted by the American Psychological Association or one of its allied publishers.

AudI–VisI 24 19 15 24 30 33 27 21 18 18 24 30 34 24 21 14 19 24 30 26 23 18 22 27
AudD–VisD 24 17 13 18 22 26 25 19 15 21 24 30 33 25 21 14 17 22 33 25 21 17 21 26

Note. WMS–III ⫽ Wechsler Memory Scale—Third Edition; V1 ⫽ first variable listed; V2 ⫽ second variable listed; WM ⫽ Working Memory; IM ⫽
Immediate Memory; GM ⫽ General Memory; AudI ⫽ Auditory Immediate; AudD ⫽ Auditory Delayed; VisI ⫽ Visual Immediate; VisD ⫽ Visual
Delayed. Copyright © 1997 by Harcourt Assessment, Inc. Data reproduced by permission. All rights reserved.
a
Weighted N ⫽ 309 for all comparisons. b Weighted N ⫽ 423 for all comparisons. c Weighted N ⫽ 294 for all comparisons. d Weighted N ⫽ 294
for all comparisons.

characterizing the AD and HD groups are reported in Tables 4, 5,
and 6.
Table 4 presents information on the relative risk of having one
of three specified cognitive deficits (VCI ⬎ IM, VCI ⬎ GM, and
VCI ⬎ PSI) given the presence of a known neurologic condition.
As seen in Table 4, an education-adjusted deficit in IM relative to
VCI was found in 71.4% of patients with AD compared with only
9.1% in the standardization sample (SS). Comparing the odds
between the AD and SS groups of having an impaired IM score,
the AD patients had a estimated relative risk of impairment that
was nearly 25 times greater than that of SS cases (OR ⫽ 24.91;
CI ⫽ 11.67–53.17). Likewise, the HD patients were also at sig-
nificantly greater risk of having impaired IM scores than the SS
group, with an estimated relative risk of nearly 20 (OR ⫽ 19.93;
CI ⫽ 6.70 –59.32). However, the AD and HD groups did not
significantly differ in the prevalence of impaired IM scores, ␹2(1,
N ⫽ 50) ⫽ 0.11, p ⫽ .49 (OR ⫽ 0.80; CI ⫽ 0.22–2.94), or
impaired GM scores ␹2(1, N ⫽ 50) ⫽ 2.12, p ⫽ .13 (OR ⫽ 0.40;
CI ⫽ 0.11–1.40). A different pattern of risk for the AD and HD
groups was noted for VCI–PSI discrepancies. Whereas both the
AD and HD groups were at significantly greater risk of having PSI
deficits compared with the general population, the relative risk was
much higher in the HD group (OR ⫽ 24.97; CI ⫽ 7.83–79.60) than
in the AD group (OR ⫽ 4.16; CI ⫽ 1.99 – 8.70). Contrasting the
odds for the two clinical groups having an impaired PSI score, the
HD patients were at significantly greater risk than the AD patients
(OR ⫽ 6.00; CI ⫽ 1.56 –23.11).
Using base-rate information to evaluate the relative risks of
cognitive impairment in this illustration, the AD and HD groups
were both at increased risk for impaired IM and PSI scores
compared with the general population but did not differ in their
risk of memory impairment on IM. However, the HD patients were
six times more likely than the AD patients to also have impaired
PSI scores. Alternately, we could take a sequential approach and
examine the base rates of PSI deficits among only those HD and
AD patients who were significantly impaired on IM. Calculating
the OR for PSI deficits among only those HD and AD patients with
IM deficits, we found that the HD group had an estimated relative
risk for PSI deficits that was seven times higher than in the AD
group (OR ⫽ 7.11; CI ⫽ 1.23– 40.98). In both approaches, it was
possible to identify a shared area of cognitive risk (i.e., memory)
while demonstrating a difference in a second neurocognitive area
(processing speed).
Tables 5 and 6 extend the basic information detailed in Table 4
by presenting OR data on the coprobability of having two or three
cognitive deficits, respectively. Coprobabilities are presented to
exemplify that even in situations where two groups being com-
pared share common deficits, multivariate patterns of probabilities
can capture meaningful differences that may not be apparent when
the deficits are evaluated individually. Because discrepancy scores
are correlated and not independent, the prevalence or coprobability
of having two statistically rare scores at the 10% level theoretically
ranges between 10% (perfect correlation) to 1% (perfect indepen-
dence: i.e., .10 ⫻ .10), and the coprobability of having three
statistically rare discrepancies at the 10% level ranges from 10% to
0.1%. For example, even though IM and GM were highly corre-
lated (r ⫽ .90) in the standardization sample (The Psychological
Corporation, 1997), the reader will observe in Table 5 that the
prevalence in standardization sample of having both IM and GM
lower than VCI at a 10% base rate occurred in only 6.6% of the
sample. As seen in Table 6, the observed conjoint prevalence in the
standardization sample of having IM, GM, and PSI lower than VCI
at the 10% base-rate level was 3.1%. The estimated relative risk
(OR) for the patient samples having two (see Table 5) or three (see
Table 6) abnormal discrepancies is presented to demonstrate the
use of conjoint prevalence rates to calculate risk estimates.
Reviewing the information in Table 5, one can see that both the
AD and HD groups were at significantly higher risk than the
standardization sample to present with a combination of IM and
GM impairments (OR ⫽ 23.79, CI ⫽ 11.57– 48.93 and OR ⫽
16.07, CI ⫽ 5.69 – 45.40, respectively). Both clinical groups were
also at a significantly greater risk of having a PSI deficit in
 DISCREPANCY-SCORE BASE RATES 151

Table 3
Unidirectional Base-Rate Information for WAIS–III and WMS–III Discrepancy Scores

WAIS–III and Education ⬍ 12a Education ⫽ 12b Education ⫽ 13–15c Education ⱖ 16d
WMS–III
indexes V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2 V1 ⬎ V2 V1 ⬍ V2

(V1)–(V2) 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5% 5% 10% 15% 15% 10% 5%
VIQ–WM 18 13 10 16 21 24 19 15 13 15 19 23 23 16 14 14 18 20 30 24 21 7 10 15
VIQ–IM 19 14 11 20 24 27 25 18 15 16 20 25 25 22 16 13 18 21 36 28 23 10 13 19
VIQ–GM 18 14 11 19 21 28 24 16 14 18 20 23 24 21 15 13 16 20 38 27 21 9 13 17
PIQ–WM 19 14 11 15 19 21 21 16 14 14 17 20 20 15 14 14 18 23 22 20 17 8 11 16
PIQ–IM 23 17 13 18 22 29 24 19 15 15 18 23 24 20 18 14 19 23 29 22 17 13 16 20
PIQ–GM 21 16 14 18 21 27 22 17 15 15 17 22 25 18 15 16 18 25 28 23 18 12 14 19
VCI–WM 28 14 12 13 16 21 20 16 14 15 18 21 25 20 16 15 20 22 24 21 19 10 13 17
VCI–IM 21 16 13 19 23 27 28 19 16 17 20 25 25 20 18 15 18 21 33 28 22 9 13 15
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

VCI–GM 19 16 11 19 22 27 24 19 15 18 21 24 24 20 16 14 17 20 37 27 23 9 12 15
This document is copyrighted by the American Psychological Association or one of its allied publishers.

POI–WM 24 19 12 12 15 17 22 16 15 13 16 21 22 19 13 17 23 29 20 14 12 14 18 21
POI–IM 27 20 16 17 24 31 26 20 16 15 18 24 26 20 15 18 20 24 27 23 19 16 20 24
POI–GM 25 19 15 18 22 28 25 19 16 16 19 24 27 20 15 17 22 27 30 21 17 15 17 22
WMI–WM 14 11 9 11 14 19 15 11 10 11 13 16 16 12 10 11 12 14 21 16 14 7 10 12
WMI–IM 27 20 17 18 21 28 25 21 18 16 20 26 32 26 20 17 22 26 33 20 16 21 24 27
WMI–GM 25 20 16 17 21 27 26 21 17 17 20 25 32 26 18 18 23 26 27 21 18 17 22 26
PSI–WM 24 17 14 12 14 20 26 20 16 14 17 27 28 23 19 17 19 29 20 16 14 19 23 28
PSI–IM 27 19 15 17 22 26 26 22 18 17 21 27 33 26 20 17 19 24 26 22 17 16 20 26
PSI–GM 25 18 14 17 20 27 25 20 18 17 21 27 28 25 18 17 20 25 25 21 18 17 20 25
VIQ–AudI 20 15 12 17 23 27 22 18 15 15 17 22 22 19 16 13 16 21 31 27 20 8 10 13
VIQ–AudD 23 15 12 19 24 29 24 18 15 16 19 23 24 19 15 14 18 23 33 25 21 7 10 12
PIQ–VisI 23 16 12 21 28 32 27 20 16 17 21 27 32 24 20 14 18 23 34 31 24 15 19 27
PIQ–VisD 21 16 12 22 25 28 23 20 17 18 21 26 29 22 19 16 18 23 32 28 23 14 18 22
VCI–AudI 20 16 13 18 21 28 25 18 16 15 19 23 23 18 14 14 16 18 29 26 22 7 9 12
VCI–AudD 24 16 13 18 25 29 26 20 15 16 20 22 24 19 14 15 17 21 33 26 22 6 10 14
POI–VisI 26 16 14 23 28 36 28 22 17 17 21 29 35 25 18 17 19 22 35 31 24 18 21 28
POI–VisD 23 17 12 21 26 30 26 21 18 19 21 28 30 25 21 16 19 25 34 27 22 15 20 26

Note. WAIS–III ⫽ Wechsler Adult Intelligence Scale—Third Edition; WMS–III ⫽ Wechsler Memory Scale—Third Edition; V1 ⫽ first variable listed;
V2 ⫽ second variable listed; VIQ ⫽ Verbal IQ; WM ⫽ Working Memory (from WMS); IM ⫽ Immediate Memory; GM ⫽ General Memory; PIQ ⫽
Performance IQ; VCI ⫽ Verbal Comprehension Index; POI ⫽ Perceptual Organization Index; WMI ⫽ Working Memory Index (from WAIS); PSI ⫽
Processing Speed Index; AudI ⫽ Auditory Immediate; AudD ⫽ Auditory Delayed; VisI ⫽ Visual Immediate; VisD ⫽ Visual Delayed. Copyright © 1997
by Harcourt Assessment, Inc. Data reproduced by permission. All rights reserved.
a
Weighted N ⫽ 309 for all comparisons. b Weighted N ⫽ 423 for all comparisons. c Weighted N ⫽ 294 for all comparisons. d Weighted N ⫽ 224
for all comparisons.

combination with one of the memory score deficits. Although the
prevalence of dual memory impairments was higher in the AD
group than in the HD group (62.9% vs. 53.3%), the prevalence of
a combination of memory and processing speed impairments was
higher in the HD group (53.3% and 40.0%) compared with the AD
group (25.7% and 28.6%). However, none of the ORs contrasting
the base rates of these dual deficit patterns was statistically sig-
nificant, although there was a trend for the HD patients to be at
somewhat greater risk for having a combination of IM and PSI
deficits than the AD patients, ␹2(1, N ⫽ 50) ⫽ 3.57, p ⫽ .06
(OR ⫽ 3.30, CI ⫽ 0.93–11.71).
When conjoint cognitive deficits in IM, GM, and PSI were
evaluated (see Table 6), the AD group was nearly 11 times more
likely to present with such a combination of deficits than the
standardization sample (OR ⫽ 10.7, CI ⫽ 4.72–24.46), whereas
the HD patients had a relative risk that was nearly 21 times that of
the standardization sample (OR ⫽ 20.7; CI ⫽ 7.02– 61.02). De-
spite the difference between the two clinical groups in prevalence
rates of being impaired on all three indexes (25.7% vs. 40.0%), the
OR between the groups was not statistically significant (OR ⫽
1.93; CI ⫽ 0.54 – 6.94).
Discussion
The current article presents base-rate information on education-
stratified discrepancy scores both within and between the WAIS–
III and WMS–III index scores in a manner that can be used by
clinicians on an individual basis. To illustrate how methods of
base-rate analysis can be applied to describe estimated relative risk
of an individual in a given diagnostic group, two patient groups
from the WAIS–III—WMS–III field trials were compared with the
standardization sample and with one another using ORs.
A review of the base-rate information in Tables 1, 2, and 3
suggests that the size of the discrepancies between each of the
variable pairs of interest is not the same at each of the specified
levels of prevalence and the patterns of unidirectional discrepan-
cies for a given variable pair vary by education level. Although
each of the WAIS–III and WMS–IIII index scores is normed to
have a mean score of 100 and a standard deviation of 15, the
magnitude of discrepancy between index scores differs for defin-
ing the percentage of the population at a given level, and it is
dependent on the correlation between the variable pairs. By defi-
nition, the stronger the correlation is, the closer two scores are
expected to covary, and so the smaller the expected discrepancy is.
 152 DORI AND CHELUNE

Table 4
Education-Adjusted Prevalence and Odds Ratios of a Single
Cognitive Deficit (i.e., Found in ⬍ 10% of the General
Population) for Select Clinical Groups Compared With the
Standardization Sample
Table 5
Education-Adjusted Prevalence and Odds Ratios of Two
Simultaneous Cognitive Deficits (i.e., Found in ⬍ 10% of the
General Population) for Select Clinical Groups Compared With
the Standardization Sample

Group Prevalence Odds ratio (CI) Group Prevalence Odds ratio (CI)

VCI–IM discrepancy (n ⫽ 1,250) VCI–IM and VCI–GM discrepancies (n ⫽ 1,250)

SS 9.1 SS 6.6
AD (mild) 71.4 24.91* (11.67–53.17) AD (mild) 62.9 23.79* (11.57–48.93)
HD 66.7 19.93* (6.70–59.32) HD 53.3 16.07* (5.69–45.40)

VCI–GM discrepancy (n ⫽ 1,250) VCI–IM and VCI–PSI discrepancies (n ⫽ 1,250)

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

SS 9.0 SS 3.7
This document is copyrighted by the American Psychological Association or one of its allied publishers.

AD (mild) 74.3 29.35* (13.42–64.19) AD (mild) 25.7 9.06* (4.02–20.43)

HD 53.3 11.61* (4.13–32.62) HD 53.3 29.91* (10.40–86.02)

VCI–PSI discrepancy (n ⫽ 2,450) VCI–GM and VCI–PSI discrepancies (n ⫽ 1,250)

SS 9.9 SS 3.8
AD (mild) 31.4 4.16* (1.99–8.70) AD (mild) 28.6 10.02* (4.56–22.03)
HD 73.3 24.97* (7.83–79.60) HD 40.0 16.69* (5.71–48.79)

Note. VCI ⫽ Verbal Comprehension Index; IM ⫽ Immediate Memory
Index; SS ⫽ standardization sample; AD ⫽ Alzheimer’s disease; HD ⫽
Huntington’s disease; GM ⫽ General Memory Index; PSI ⫽ Processing
Speed Index. Wechsler Adult Intelligence Scale—Third Edition. Wechsler
Memory Scale—Third Edition. Copyright © 1997 by Harcourt Assess-
ment, Inc. Data reproduced by permission. All rights reserved.
* Significant at 95% confidence interval (CI).
Note. VCI ⫽ Verbal Comprehension Index; IM ⫽ Immediate Memory
Index; GM ⫽ General Memory Index; SS ⫽ standardization sample;
AD ⫽ Alzheimer’s disease; HD ⫽ Huntington’s disease; PSI ⫽ Processing
Speed Index. Wechsler Adult Intelligence Scale—Third Edition. Wechsler
Memory Scale—Third Edition. Copyright © 1997 by Harcourt Assess-
ment, Inc. Data reproduced by permission. All rights reserved.
* Significant at 95% confidence interval (CI).

variable pairs (e.g., VCI–POI) than those individuals with higher

This implies that for highly correlated variables, even a relatively
small discrepancy may be rare in the general population and could
reflect a clinically meaningful finding. For moderately correlated
variables, a discrepancy of similar magnitude may be relatively
common. For example, a positive 11-point discrepancy between
the WMS–III Auditory Immediate Index (AudI) minus the Audi-
tory Delayed Index (AudD), where AudI and AudD are correlated
(r ⫽ .88; The Psychological Corporation, 1997, Table 4.14; 2002,
Table 4.3), occurs in less than 10% of the standardization sample
when education level equals 12 years. In contrast, a positive
21-point discrepancy is required between the AudI and Visual
Immediate (VisI) Memory Index to occur in less than 10% of the
population when education level equals 12 years because the
intercorrelation is only .38. Because the prevalence of discrepancy
scores differs as a function of the intercorrelations between vari-
able pairs, choosing a single discrepancy score to indicate an
abnormal finding across variable pairs on the WAIS–III and
WMS–III would be highly problematic. One should refer to the
appropriate tables to accurately determine at what level a discrep-
ancy finding can be considered clinically rare for a specific pair of
variables.
Patterns of discrepancy base rates for some variable pairs also
covary as a function of premorbid cognitive ability as estimated by
education level. Reviewing Tables 1–3, one can see that discrep-
ancies covary with education level, especially when the discrep-
ancy includes index scores that are highly correlated with educa-
tion (e.g., VIQ, VCI). For example, participants in the lowest
education group (and thus assumed to have low premorbid verbal
abilities), required smaller positive discrepancies and larger neg-
ative discrepancies to be considered “rare” on verbally oriented
educational levels. The converse holds true for those in the highest
education group: Larger positive and smaller negative discrepan-
cies are required to be considered meaningful compared with those
who have less education. Our findings support those by Hawkins
and Tulsky (2001), who suggested that failure to stratify discrep-
ancy data could lead to erroneous interpretations of individuals
with low or high intellectual abilities, particularly when acquired
memory deficits are being evaluated. For example, whereas a
positive 20-point discrepancy between VCI and GM would not be
particularly uncommon for an individual with 18 years of educa-
Table 6
Education-Adjusted Prevalence and Odds Ratios of Three
Simultaneous Cognitive Deficits (i.e., Found in ⬍ 10% of the
General Population) for Select Clinical Groups Compared With
the Standardization Sample
Group Prevalence Odds ratio (CI)
VCI–IM, VCI–GM, and VCI–PSI discrepancies (n ⫽ 1,250)
SS 3.1
AD (mild) 25.7 10.75* (4.72–24.46)
HD 40.0 20.70* (7.02–61.02)
Note. VCI ⫽ Verbal Comprehension Index; IM ⫽ Immediate Memory
Index; GM ⫽ General Memory Index; PSI ⫽ Processing Speed Index;
SS ⫽ standardization sample; AD ⫽ Alzheimer’s disease; HD ⫽ Hunting-
ton’s disease. Wechsler Adult Intelligence Scale—Third Edition. Wechsler
Memory Scale—Third Edition. Copyright © 1997 by Harcourt Assess-
ment, Inc. Data reproduced by permission. All rights reserved.
* Significant at 95% confidence interval (CI).
 DISCREPANCY-SCORE BASE RATES
tion, such a discrepancy for an individual with 10 years of educa-
tion would be relatively rare and of potential clinical significance.
Thus, both unidirectional and stratified tables, such as those pro-
vided here, are useful and necessary adjuncts in making evidence-
based determinations regarding the significance of observed dis-
crepancy scores.
Using the data in Tables 1–3, we illustrated how the method of
base-rate analysis can be applied to clinical populations in a
relevant manner. By examining the base rates of select discrepancy
scores among the patient groups presented in the WAIS–III—
WMS–III Technical Manual (The Psychological Corporation,
1997, 2002), we produced clinically interesting results, albeit
solely for illustrative purposes, as the base-rate estimates derived
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
from these small clinical samples may not be replicable. Even
This document is copyrighted by the American Psychological Association or one of its allied publishers.
without directly evaluating mean score differences, determination
of reliable and significant differences between the standardization
sample and clinical groups was possible. In addition, it was pos-
sible to demonstrate how such information could assist in the
diagnosis of an individual (i.e., relative risk).
In our current example, the frequency of statistically rare dis-
crepancies in processing speed and in short-term and long-term
aspects of memory relative to verbal comprehension were found to
be markedly higher in both of the neurological groups compared
with the general population. These significant differences between
the patient and standardization groups were reliable at the 95% CI,
even with the exceptionally small patient sample sizes compared
with the standardization sample size. Using OR analyses, it also
was possible to characterize the estimated relative risk among
these clinical groups for specific types of cognitive deficit, in both
a univariate and multivariate manner.
In the present illustration, the estimated risk of memory impair-
ment in IM and/or GM was higher, but not statistically different, in
the AD group compared with the HD group, yet the prevalence of
PSI deficits was statistically much more characteristic of HD.
Through an application of base-rate analyses, we were able to
replicate previous findings that memory deficits are more primary
to AD than HD and that slowed processing speed is more central
to HD than AD (Fields, 1998).
As suggested by Smith (2002), analysis of base-rate information
provides information about “how many,” which may be different
from analysis of group mean data addressing the question of “how
much.” Data presented in the WAIS–III—WMS–III Technical Man-
ual (The Psychological Corporation, 1997, Table 4.41; 2002, Ta-
ble 4.32) reveal lower mean IM and GM standard scores for the
AD patients (62.9 and 60.4, respectively) compared with the HD
group (70.9 and 76.4, respectively) and higher PSI scores (79.6 vs.
69.3). In our application of base-rate statistics to the AD and HD
clinical samples, there was no difference between the groups in the
prevalence of memory deficits or impairment in general. However,
the HD group was at a greater risk of processing speed deficits
than the AD group. These findings suggest to researchers and
clinicians that more than statistically reliable mean differences are
needed to characterize a population and the individuals within that
population. That is, base-rate approaches should be used to deter-
mine what is truly rare and abnormal or common and normal. In
addition, reliance on solely one score or score discrepancy may
restrict understanding of the clustering of cognitive deficits under-
lying a disorder or disease.
153
We recognize that the patient data presented here are limited by
their small sample size and are sample specific. They were used to
serve only as model samples, not to generalize to larger and more
carefully selected patient samples. However, they did provide a
basis for illustrating how base-rate information can be applied to
characterize clinical samples and how such information can be
presented in a manner that can be more directly used by practitio-
ners (Chelune, 2002). The exercise of examining base-rate infor-
mation concerning specific discrepancy scores in select neurologic
groups was helpful in demonstrating how such information can be
used in clinical decision making. In the current study, it was
possible to use constellations of discrepancies to distinguish the
neurologic conditions of interest in terms of the prevalence of
certain cognitive deficits and to express these differences in terms
of relative risk. As stated above, although the details of the present
findings are limited to the specific samples studied, similar ap-
proaches can be used in other clinical settings. Base-rate data also
can be applied in combination with other psychological, neuropsy-
chological, or medical results using a multivariate sign approach,
looking at patterns of rare findings.
Base-rate analyses using techniques such as the OR for inter-
preting prevalence information are consistent with evidence-based
and epidemiological approaches (Chelune, 2002; Ivnik et al.,
2001). They are pragmatic and cost effective in that a priori
hypotheses can be directly addressed and more readily applied
within the context of clinical expertise and presentation of an
individual patient (see Smith et al., 2003). Most important, base-
rate statistical information helps move beyond issues of reliability
associated with mean group differences to diagnostic validity on
an individual level, distinguishing between patients who do and do
not have a condition of interest (Bieliauskas et al., 1997; Chelune,
2002; Ivnik et al., 2000; Sackett et al., 2000).
References
Bieliauskas, L. A., Fastenau, P. S., Lacy, M. A., & Roper, B. L. (1997).
Use of the odds ratio to translate neuropsychological test scores into
real-world outcomes: From statistical significance to clinical signifi-
cance. Journal of Clinical and Experimental Neuropsychology, 19, 889 –
893.
Bornstein, R. A., Chelune, G. J., & Prifitera, A. (1989). IQ–memory
discrepancies in normal and clinical samples. Psychological Assessment,
1, 203–206.
Chelune, G. J. (1998). Use of WAIS–III—WMS–III discrepancy scores to
meet the challenges of evidence-based medicine. The Clinical Neuro-
psychologist, 12, 295.
Chelune, G. J. (2002). Making neuropsychological outcomes research
consumer friendly: A comment on Keith et al. (2002). Neuropsychology,
16, 422– 425.
Fields, R. B. (1998). The dementias. In P. J. Snyder & P. D. Nussbaum
(Eds.), Clinical neuropsychology: A pocket handbook for assessment
(pp. 211–239). Washington, DC: American Psychological Association.
Hawkins, K. A. (1998). Indicators of brain dysfunction derived from
graphic representations of the WAIS–III/WMS–III Technical Manual
clinical samples data: A preliminary approach to clinical utility. Clinical
Neuropsychology, 12, 535–551.
Hawkins, K. A., & Tulsky, D. S. (2001). The influence of IQ stratification
on WAIS–III/WMS–III FSIQ—General Memory Index discrepancy
base-rates in the standardization sample. Journal of the International
Neuropsychological Society, 7, 875– 880.
Hawkins, K. A. & Tulsky, D. S. (2003). WAIS–III and WMS–III discrep-
 154 DORI AND CHELUNE
ancy analysis: Six-factor model index discrepancy base rates, implica-
tions, and preliminary considerations of utility. In D. S. Tulsky, D. H.
Saklofske, G. J. Chelune, R. K. Heaton, R. J. Ivnik, R. Bornstein, A.
Prifitera, & M. Ledbetter (Eds.), Clinical interpretation of the WAIS–III
and WMS–III (pp. 211–272). San Diego, CA: Academic Press.
Heaton, R. K., Baade, L. E., & Johnson, K. L. (1978). Neuropsychological
test results associated with psychiatric disorders in adults. Psychological
Bulletin, 85, 141–162.
Heaton, R. K., Taylor, M. J., & Manly, J. (2003). Demographic effects and
use of demographically corrected norms with the WAIS–III and WMS–
III. In D. S. Tulsky, D. H. Saklofske, G. J. Chelune, R. K. Heaton, R. J.
Ivnik, R. Bornstein, A. Prifitera, & M. Ledbetter (Eds.), Clinical inter-
pretation of the WAIS–III and WMS–-III (pp. 181–210). San Diego, CA:
Academic Press.
Ivnik, R. J., Smith, G. E., Cerhan, J. H., Boeve, B. F., Tangalos, E. G., &
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Petersen, R. C. (2001). Understanding the diagnostic capabilities of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
cognitive tests. The Clinical Neuropsychologist, 15, 114 –124.
Ivnik, R. J., Smith, G. E., Petersen, R. C., Boeve, B. F., Kokmen, E., &
Tangalos, E. G. (2000). Diagnostic accuracy of four approaches to
interpreting neuropsychological test data. Neuropsychology, 14, 163–
177.
Jacobson, M. W., Delis, D. C., Bondi, M. W., & Salmon, D. P. (2002). Do
neuropsychological tests detect preclinical Alzheimer’s disease:
Individual-test versus cognitive-discrepancy score analyses. Neuropsy-
chology, 16, 132–139.
Kaufman, A. S. (1990). Assessing adolescent and adult intelligence. Bos-
ton: Allyn & Bacon.
Manly, J. J., Jacobs, D. M., Touradji, P., Small, S. A., & Stern, Y. (2002).
Reading level attenuates differences in neuropsychological test perfor-
mance between African American and White elders. Journal of the
International Neuropsychological Society, 8, 341–348.
Matarazzo, J. D. (1972). Wechsler’s measurement and appraisal of adult
intelligence. Baltimore, MD: Williams & Wilkins.
Matarazzo, J. D., & Herman, D. O. (1984). Base rate data for the WAIS–R:
Test–retest stability and VIQ–PIQ differences. Journal of Clinical Neu-
ropsychology, 6, 351–366.
Milner, J. B. (1954). Intellectual function of the temporal lobes. Psycho-
logical Bulletin, 51, 42– 62.
The Psychological Corporation (1997). The WAIS–III—WMS–III Technical
Manual. San Antonio, TX: The Psychological Corporation.
The Psychological Corporation (2002). The WAIS–III—WMS–III Technical
Manual (Updated version). San Antonio, TX: The Psychological
Corporation.
Sackett, D. L., Haynes, R. B., Guyatt, G. H., & Tugwell, P. (1991). Clinical
epidemiology: A basic science for clinical medicine (2nd ed.). Boston:
Little, Brown.
Sackett, D. L., Straus, S. E., Richardson, W. S., Rosenberg, W., & Haynes,
R. B. (2000). Evidence-based medicine: How to practice and teach EBM
(2nd ed.). London: Churchill Livingstone.
Smith, G. E. (2002). What is the outcome we seek? A commentary on
Keith et al. (2002). Neuropsychology, 16, 432– 433.
Smith, G. E., Cerhan, J. H., & Ivnik, R. J. (2003). Diagnostic validity. In
D. S. Tulsky, D. H. Saklofske, G. J. Chelune, R. K. Heaton, R. J. Ivnik,
R. Bornstein, A. Prifitera, & M. F. Ledbetter (Eds.), Clinical interpre-
tation of the WAIS–III and WMS–III (pp. 273–301). San Diego, CA:
Academic Press.
Taylor, M. J., & Heaton, R. K. (2001). Sensitivity and specificity of
WAIS–III/WMS–III demographically corrected factor scores in neuro-
psychological assessment. Journal of the International Neuropsycholog-
ical Society, 7, 867– 874.
Tulsky, D. S., Chiaravalloti, N. D., Palmer, B. W., & Chelune, G. J. (2003).
The Wechsler Memory Scale, third edition: A new perspective. In D. S.
Tulsky, D. H. Saklofske, G. J. Chelune, R. K. Heaton, R. J. Ivnik, R.
Bornstein, A. Prifitera, & M. F. Ledbetter (Eds.), Clinical interpretation
of the WAIS–III and WMS–III (pp. 93–139). San Diego, CA: Academic
Press.
Tulsky, D. S., & Ledbetter, M. F. (2000). Updating the WAIS–III and
WMS–III: Considerations for research and clinical practice. Psycholog-
ical Assessment, 12, 253–262.
Tulsky, D. S., Rolfhus, E. L., & Zhu, J. (2000). Two-tailed versus one-
tailed base rates of discrepancy scores in the WAIS–III. The Clinical
Neuropsychologist, 14, 451– 460.
Wechsler, D. (1939). The measurement of adult intelligence. Baltimore,
MD: Williams & Wilkins.
Wechsler, D. (1997a). The WAIS–III administration and scoring manual.
San Antonio, TX: The Psychological Corporation.
Wechsler, D. (1997b). The WMS–III administration and scoring manual.
San Antonio, TX: The Psychological Corporation.
Received May 1, 2003
Revision received December 3, 2003
Accepted December 15, 2003 䡲