You are on page 1of 3

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/273575324

Convergent validity: Use of WAIS-IV vs. SB5 in intellectual functioning


assessment.

Conference Paper · March 2014

CITATION READS

1 2,500

5 authors, including:

Karen L. Salekin Mary Elizabeth Wood


University of Alabama Vanderbilt University
32 PUBLICATIONS   287 CITATIONS    19 PUBLICATIONS   33 CITATIONS   

SEE PROFILE SEE PROFILE

Kevin McGrew Krystal anne Hedge


Institute of Applied Psychometrics University of Alabama
126 PUBLICATIONS   7,328 CITATIONS    5 PUBLICATIONS   41 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Flynn effect (norm obsolescence) reference project View project

IAP Applied Psychometric Series 101 View project

All content following this page was uploaded by Mary Elizabeth Wood on 16 April 2015.

The user has requested enhancement of the downloaded file.


Convergent Validity: Use of WAIS-IV vs. SB5 in Intellectual Functioning Assessment
Karen Salekin1, Mary E. Wood1, Kevin McGrew2, Debra Chen1, Krystal Hedge1
The University of Alabama1, Institute for Applied Psychometrics2

Convergent validity is an important method for evaluating how well a measure works at
identifying a construct of interest. From the psychometric perspective, convergent validity exists
when the scores on one instrument are in high agreement with those of another instrument that
purports to measure the same trait. When measuring the construct of intelligence, test developers
look at multiple indices of intellectual ability and evaluate the correlations between scales that
tap similar abilities. Of particular import in diagnosing intellectual disability is the concordance
between the composite scores, primarily the Full Scale Intelligence Quotient (FSIQ), since this
value is used to determine if the individual meets the first prong of the diagnostic criteria (i.e.,
FSIQ = 70 +/- 5).
Research regarding the convergent validity of the Wechsler and Stanford Binet scales has
primarily evaluated relations in samples of children and the vast majority of the studies were
completed with earlier versions of the test batteries. In general, these studies have produced
correlations that are indicative of convergence. The results of seven studies produced
correlations between .71 and .82 (Garred & Gilmore, 2009; Gerken & Hodapp, 1992; Carvajal,
Parks, Bays, Logan, Lujano, et al., 1991; McCrowell & Nagle, 1994; Rust & Lindstrom, 1996;
Prewett & Matavich, 1994; Simpson et al., 2002). Convergent validity between the current
versions of the Wechsler and Stanford Binet scales has not yet been assessed, but convergent
validity was supported for the current Stanford Binet (Fifth edition; Roid, 2003) and the previous
version of the Wechsler scale (WAIS-III). As noted in the Stanford Binet technical manual,
correlations between domain and composite scores were relatively strong (e.g., PIQ-NIQ = .76;
VIQ-VIQ = .81; FSIQ-FSIQ = .82) and mean FSIQ scores were similar (SB5 = 101.7; WASI-III
= 105.3).
Perhaps a more valuable method for evaluating convergent validity is to evaluate
difference scores. Using record review, Silverman and colleagues (2010) reported significant
differences between the Wechsler and Stanford Binet composiste IQ scores. The authors
examined the records of 74 adults who had been administered both tests at some point during
their lives. The test dates ranged from 1949 and 2005 with age at the time of assessment ranging
between 5 and 81 years. If different test scores within the same battery family were available for
one individual, then a median IQ score was utilized. The results indicated that convergence was
low and that similar results should not be expected when using these tests. Difference scores
ranged from 4 to 31 points (mean=16.7) and for 24.3% of the individuals, the difference between
FSIQ scores was more than 20 points. In all cases, the Wechsler scale produced the higher score.
Though valuable, the Silverman study is of limited utility given that the time between
administrations was often lengthy and comparisons were made between older versions of the test
batteries. Generalizing these findings to the SB5 and WAIS-IV fails account for improvements in
testing procedures and incorporation of new knowledge in the area of intelligence.
The current study evaluated the convergent validity between the most recent version of
the Stanford-Binet intelligence battery (SB5) and the most recent version of the Wechsler scale
(WAIS-IV) in a sample of individuals previously diagnosed with ID. The study utilized a within-
subject design in which participants completed both the SB5 and the WAIS-IV; administrations
were counterbalanced to control for order effects. The sample consisted of 28 individuals who
were registered in either a two-year transition program that promotes the development of
independent living skills or housed in a residential state hospital for individuals with
developmental disabilities.
The full scale IQ scores for the WAIS-IV and the SB-5 demonstrated a strong positive
correlation (i.e., Flynn and non-Flynn corrected, r = 0.90) suggesting that the two indices are
measuring the same construct. Comparison of the full scale IQ scores showed a mean difference
of 6.5 points with the SB5 (M = 52.25) being lower than the WAIS-IV (M = 58.79; t(27) = -
7.37, p < .000). This difference remained intact and was slightly larger (i.e., 8.01) when the
scores were adjusted for norm obsolescence (Flynn Effect) with the SB5 mean (M = 49.72) being
lower than the WAIS-IV mean (M = 7.74; t(27) = -9.16, p < .000). Moderate to strong
correlations were found between the composite scores from the SB5 and WAIS-IV (VCI-VIQ =
.83; PRI-NVIQ = .78). Significant differences ranging from 6 to 13 standard score points were
also found for five of six a priori planned composite score comparisons. These results are
summarized in Table 1.

Table 1. SB-5 and WAIS-IV Composite Score Comparisons

Comparison Mean Mean Standard t-score df p-value


Difference Deviation
Score
SB-NVIQ 55.143 13.393 7.315 -9.688 27 0.000
W-PRI 68.536
SB-VIQ 53.750 9.714 6.497 -7.912 27 0.000
W-VCI 63.464
SB-VS 62.250 6.286 10.139 -3.280 27 0.000
W-PRI 68.536
SB-FR 61.179 2.286 11.158 -1.084 27 0.288
W-VCI 63.464
SB-KN 55.679 7.786 5.672 -7.263 27 0.000
W-VCI 63.464
SB-MI 55.714 6.893 8.530 -4.276 27 0.000
W-WMI 62.607

The results of the current study, as well of and that of Silverman and colleagues, puts into
question the value of the IQ score. IQ scores have enjoyed great power in multiple contexts and
have affected the lives of many individuals. In this case we only investigated IQ scores for
individuals in the lower tail of the distribution—those where the provision of support services,
and in death penalty cases, their lives are dependent on an IQ score. Given the power wielded by
the full scale IQ score and the highly discrepant results, triers-of-fact must be educated about IQ
testing and they must be provided direction as to how to proceed in the face of such
discrepancies.

View publication stats

You might also like