Abstract

Full Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV) administration can be time-consuming and may not be necessary when intelligence quotient estimates will suffice. Estimated Full Scale Intelligence Quotient (FSIQ) and General Ability Index (GAI) scores were derived from nine dyadic short forms using individual regression equations based on data from a clinical sample (n = 113) that was then cross validated in a separate clinical sample (n = 50). Derived scores accounted for 70%–83% of the variance in FSIQ and 77%–88% of the variance in GAI. Predicted FSIQs were strongly associated with actual FSIQ (rs = .73–.88), as were predicted and actual GAIs (rs = .80–.93). Each of the nine dyadic short forms of the WAIS-IV was a good predictor of FSIQ and GAI in the validation sample. These data support the validity of WAIS-IV short forms when time is limited or lengthier batteries cannot be tolerated by patients.

Introduction

The Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV; Wechsler, 2008) is a commonly employed measure of global intellectual and cognitive functioning. The WAIS-IV yields two measures of global functioning, the Full Scale Intelligence Quotient (FSIQ) and the General Ability Index (GAI). The FSIQ is a weighted index of verbal, perceptual reasoning, processing speed, and working memory abilities, whereas the GAI is a global measure that removes working memory and processing speed subtests. The GAI is new to the WAIS-IV and important in neuropsychological evaluations because it provides a summary score that is less sensitive than FSIQ to the influence of attention and/or processing speed deficits commonly seen in many neurologic populations.

Short forms of previous WAIS versions (Wechsler, 1955, 1987, 1997) have been proposed to estimate intelligence quotient (IQ) in instances where full IQ testing is too lengthy for patients to tolerate or where an estimate of IQ will suffice. According to Thompson, LoBello, Atkinson, Chisholm, and Ryan (2004), the Wechsler intelligence scales are the most frequently applied measures used to derive short-form IQ estimations. Previous estimation methods include administering as few as one and as many as eight WAIS subtests to derive estimates (Axelrod, Dingel, Ryan, & Ward, 2000; Christensen, Girard, & Bagby, 2007; Engelhart, Eisenstein, Johnson, & Losonczy, 1999; Jeyakumar, Warriner, Raval, & Ahmad, 2004; Kaufman, Ishikuma, & Kaufman-Packer, 1991; Mendella, McFadden, Regan, & Medlock, 2000; Pilgrim, Meyers, Bayless, & Whetstone, 1999; Ringe, Saine, Lacritz, Hynan, & Cullum, 2002; Satterfield, Martin, & Leiker, 1994; Schoenberg, Duff, Dorfman, & Adams, 2004; Schoenberg, Duff, Scott, & Adams, 2002; Schoenberg, Scott, Ruwe, Patton, & Adams, 2004; Silverstein, 1982; Ward, 1990). These studies suggest that as short forms lengthen in number of subtests used to derive estimates, correlations between estimated and actual scores improve. However, as the number of subtests increase, the amount of time needed to attain the estimated score also increases. When time limitations factor into an evaluation, longer four- or seven-subtest short forms may not serve as an ideal solution.

There are several instruments designed to quickly estimate intelligence (e.g., Wechsler Abbreviated Scale of Intelligence-Second Edition [Wechsler, 2011], Kaufman Brief Intelligence Test-Second Edition [Kaufman & Kaufman, 2004], Reynolds Intellectual Assessment Scales [Reynolds & Kamphaus, 2003], etc.). While these measures possess utility in estimating intelligence, using short forms of the Wechsler intelligence scales (e.g., WAIS-IV) designed to fully assess intelligence (i.e., FSIQ) holds several benefits. WAIS-IV short forms provide greater flexibility than brief stand-alone intelligence measures, as the core 10-subtest WAIS-IV allows various subtest combinations that can better meet the needs of individual circumstances. Subtests from the Wechsler intelligence scales are also used to assess domain-specific neuropsychological functioning. Utilizing short forms of the WAIS-IV to estimate intelligence eliminates the need for additional measures to assess intelligence, which saves time and resources.

Multiple short forms have been examined for the WAIS-R and WAIS-III; however, information about WAIS-IV short forms is more limited. Sattler and Ryan (2009) derived tables using the Tellegen and Briggs (1967) procedure enabling the estimation of FSIQ from combinations of various selected subtests. These short forms were calculated using normative data and have yet to be validated in clinical populations. In a more recent study, Girard, Axelrod, Patel, and Crawford (2014) evaluated all possible dyad combinations using the 10 core subtests from the WAIS-IV. They used a clinical sample of 482 subjects (mostly male and low-average FSIQ) from a VA Medical Center to assess the reliability and validity of the dyads based on data from the standardization sample derived from linear scaling of composite subtest scaled scores. They note that no dyad consistently ranked among the five highest values of the psychometric properties they examined, though dyads that included processing speed or working memory subtests performed highest when examining a composite measure of validity. The authors encouraged readers to consider their goals for short-form use when evaluating or selecting a dyad. However, FSIQ estimation tables based on their approach were not provided. Intelligence estimates derived from a clinical population, therefore, remain absent for the WAIS-IV. Additionally, estimation tables or regression equations do not presently exist for estimating GAI. The present study used a regression-based approach, derived from clinical data of a mixed neurological sample, to generate dyadic short form estimates of WAIS-IV FSIQ and GAI.

Methods

Participants

Subjects included 163 consecutive individuals with known or suspected neurological disorders who underwent neuropsychological evaluation as part of their standard clinical care in the Neuropsychology Service at the University of Texas Southwestern Medical Center between September 2009 and January 2011. Subjects gave consent for their clinical data to be used for research purposes. The initial 113 consecutive individuals who presented for neuropsychological evaluation were included in the test group, and the following 50 consecutive patients comprised the cross-validation group.

Procedures

The core 10-subtest WAIS-IV was administered as part of a larger battery for clinical reasons, and all subjects gave written consent for their results to be used for research purposes. Short forms for FSIQ and GAI were derived by summing age-scaled scores from subtests administered from the Verbal Comprehension Index (VCI) (Vocabulary [V], Similarities [S], and Information [I]) to those from the Perceptual Reasoning Index (Block Design [BD], Matrix Reasoning [MR], and Visual Puzzles [VP]). This resulted in nine dyads: Vocabulary/Block Design (V/BD), Vocabulary/Matrix Reasoning (V/MR), Vocabulary/Visual Puzzles (V/VP), Similarities/Block Design (S/BD), Similarities/Matrix Reasoning (S/MR), Similarities/Visual Puzzles (S/VP), Information/Block Design (I/BD), Information/Matrix Reasoning (I/MR), and Information/Visual Puzzles (I/VP). Within the test sample, the summations of the age-scaled scores for each dyad were entered into separate regression equations to predict FSIQ and GAI. Short form conversion tables were created from the regression equations, which were validated by comparing actual FSIQ and GAI scores of the test sample with estimated FSIQ and GAI scores acquired from the regression data. Composite reliability (rxx) and correlation coefficients (rsf) between the short form predicted FSIQs and actual FSIQ, as well as those between short form predicted GAIs and actual GAI, were calculated using the methods outlined by Levy (1967) to correct for inflated Pearson correlations. Results were cross-validated on a second mixed neurological sample in which corrected correlations (rsf) between actual and estimated FSIQ and GAI were obtained. The percentages of estimated scores within 5 and 10 points of actual FSIQ and GAI scores were also calculated.

Results

Test Sample

Demographic data for the test sample (N = 113) can be found in Table 1. Means and SDs of test sample FSIQ, GAI, and subtest performances can be seen in Table 2. The test sample WAIS-IV FSIQ scores ranged from 55 to 135, with a mean of 92.5 (SD = 15.9). GAI scores ranged from 59 to 130, with a mean of 95.7 (SD = 15.2). The regression equations for FSIQ and GAI short forms can be found in Table 3. All standard multiple regressions were significant (p < .001). The R2 values ranged from 70% to 83% for FSIQ and 77% to 88% for GAI. Means and SDs of dyadic FSIQ and GAI estimations derived from the test sample were similar across dyad estimates (FSIQ estimates range from 92.5 to 92.6; GAI estimates range from 95.6 to 95.8) and slightly higher for GAI than FSIQ, in keeping with subjects' actual FSIQ and GAI scores (see Table 4). Estimated FSIQ scores were derived for each dyad, based on the regression equations in Table 3, and are presented in Table 5. The estimated GAI scores were also derived for each dyad and can be found in Table 6. Due to a lack of actual scores at the low and high ends of the spectrum, some estimated FSIQ and GAI scores presented in Tables 5 and 6 are theoretical and not based on actual data. Correlations between estimated and actual FSIQ in the test sample were highly significant (p < .001) for all dyads, with corrected correlations ranging from .76 to .86 (see Table 7). Likewise, correlations between estimated and actual GAI in the test sample were also significant (p < .001), ranging from .82 to .91 (see Table 7).

Table 1.

Test sample and validation sample demographic data

Variable Test sample (n = 113)
 
Validation sample (n = 50)
 
M (SDRange M (SDRange 
Age 59.9 (13.0) 25–86 59.3 (14.2) 23–83 
Education 14.2 (3.2)  1–20 14.7 (2.4)  9–19 
Gender (%) 
 Female 48  62  
 Male 52  38  
Race (%) 
 African American 11  10  
 Asian  0   4  
 Hispanic  3  10  
 White 86  76  
Variable Test sample (n = 113)
 
Validation sample (n = 50)
 
M (SDRange M (SDRange 
Age 59.9 (13.0) 25–86 59.3 (14.2) 23–83 
Education 14.2 (3.2)  1–20 14.7 (2.4)  9–19 
Gender (%) 
 Female 48  62  
 Male 52  38  
Race (%) 
 African American 11  10  
 Asian  0   4  
 Hispanic  3  10  
 White 86  76  

Notes: Race was not equally distributed in the population, χ2 (3, N = 163) = 8.82, p < .05. Other demographic variables were not significantly different between groups.

Table 2.

Means and SDs of the Wechsler Adult Intelligence Scale-Fourth Edition variables

Variable Test sample (n = 113)
 
Validation sample (n = 50)
 
M (SD) Range M (SD) Range 
Actual Full Scale IQ 92.5 (15.9) 55–135 98.4 (14.5) 51–131 
Actual GAI 95.7 (15.2) 59–130 101.2 (14.8) 55–137 
Vocabulary 10.4 (2.8) 4–19 11.6 (3.0) 4–19 
Information 9.4 (3.0) 4–17 10.1 (2.9) 5–17 
Similarities 9.7 (3.0) 1–16 10.7 (2.5) 3–17 
Block Design 8.7 (3.2) 1–18 9.3 (3.5) 1–19 
Matrix Reasoning 9.2 (3.4) 3–18 10.0 (3.2) 1–16 
Visual Puzzles 8.8 (3.1) 2–17 9.5 (2.8) 4–17 
Variable Test sample (n = 113)
 
Validation sample (n = 50)
 
M (SD) Range M (SD) Range 
Actual Full Scale IQ 92.5 (15.9) 55–135 98.4 (14.5) 51–131 
Actual GAI 95.7 (15.2) 59–130 101.2 (14.8) 55–137 
Vocabulary 10.4 (2.8) 4–19 11.6 (3.0) 4–19 
Information 9.4 (3.0) 4–17 10.1 (2.9) 5–17 
Similarities 9.7 (3.0) 1–16 10.7 (2.5) 3–17 
Block Design 8.7 (3.2) 1–18 9.3 (3.5) 1–19 
Matrix Reasoning 9.2 (3.4) 3–18 10.0 (3.2) 1–16 
Visual Puzzles 8.8 (3.1) 2–17 9.5 (2.8) 4–17 
Table 3.

Regression analyses for dyads for predicting FSIQ and GAI

Dyad FSIQ (n = 113)
 
GAI (n = 113)
 
 F* R2 Predicted FSIQ equation F* R2 Predicted GAI equation 
V/BD 524.07 .83 (V + BD)2.75 + 39.94 848.04 .88 (V + BD)2.72 + 43.66 
V/MR 263.03 .70 (V + MR)2.43 + 45.03 419.93 .79 (V + MR)2.46 + 47.55 
V/VP 381.07 .77 (V + VP)2.90 + 36.95 638.98 .85 (V + VP)2.90 + 39.98 
I/BD 414.18 .79 (I + BD)2.61 + 45.13 587.46 .84 (I + BD)2.58 + 48.91 
I/MR 301.10 .73 (I + MR)2.50 + 45.94 497.50 .82 (I + MR)2.53 + 48.59 
I/VP 289.42 .72 (I + VP)2.69 + 43.64 419.78 .79 (I + VP)2.69 + 46.82 
S/BD 504.40 .82 (S + BD)2.67 + 43.37 613.77 .85 (S + BD)2.59 + 47.93 
S/MR 267.21 .71 (S + MR)2.38 + 47.61 361.37 .77 (S + MR)2.36 + 51.02 
S/VP 319.01 .74 (S + VP)2.70 + 42.57 401.35 .78 (S + VP)2.66 + 46.62 
Dyad FSIQ (n = 113)
 
GAI (n = 113)
 
 F* R2 Predicted FSIQ equation F* R2 Predicted GAI equation 
V/BD 524.07 .83 (V + BD)2.75 + 39.94 848.04 .88 (V + BD)2.72 + 43.66 
V/MR 263.03 .70 (V + MR)2.43 + 45.03 419.93 .79 (V + MR)2.46 + 47.55 
V/VP 381.07 .77 (V + VP)2.90 + 36.95 638.98 .85 (V + VP)2.90 + 39.98 
I/BD 414.18 .79 (I + BD)2.61 + 45.13 587.46 .84 (I + BD)2.58 + 48.91 
I/MR 301.10 .73 (I + MR)2.50 + 45.94 497.50 .82 (I + MR)2.53 + 48.59 
I/VP 289.42 .72 (I + VP)2.69 + 43.64 419.78 .79 (I + VP)2.69 + 46.82 
S/BD 504.40 .82 (S + BD)2.67 + 43.37 613.77 .85 (S + BD)2.59 + 47.93 
S/MR 267.21 .71 (S + MR)2.38 + 47.61 361.37 .77 (S + MR)2.36 + 51.02 
S/VP 319.01 .74 (S + VP)2.70 + 42.57 401.35 .78 (S + VP)2.66 + 46.62 

Notes: Dyads are sums of age-scaled scores; V/BD = Vocabulary/Block Design; V/MR = Vocabulary/Matrix Reasoning; V/VP = Vocabulary/Visual Puzzles; I/BD = Information/Block Design; I/MR = Information/Matrix Reasoning; I/VP = Information/Visual Puzzles; S/BD = Similarities/Block Design; S/MR = Similarities/Matrix Reasoning; S/VP = Similarities/Visual Puzzles.

*df = (1, 111); all significant, p < .001.

Table 4.

Means and SDs of Wechsler Adult Intelligence Scale-Fourth Edition dyadic estimations

 Test sample
 
Validation sample
 
M SD Range M SD Range 
Estimated FSIQ 
 V/BD 92.5 14.4 62–128 97.5 15.7 54–136 
 V/MR 92.6 13.4 67–123 97.6 12.6 57–128 
 V/VP 92.5 14.0 60–133 98.1 14.1 60–133 
 I/BD 92.5 14.1 61–126 95.9 14.6 61–139 
 I/MR 92.5 13.6 66–128 96.3 12.1 61–121 
 I/VP 92.6 13.5 62–130 96.4 12.3 68–135 
 S/BD 92.6 14.4 54–129 98.4 14.5 54–131 
 S/MR 92.6 13.4 62–121 97.0 11.6 57–117 
 S/VP 92.5 13.7 59–126 97.1 12.3 61–126 
Estimated GAI 
 V/BD 95.6 14.3 65–131 100.6 15.6 57–139 
 V/MR 95.7 13.5 70–126 100.7 12.7 60–131 
 V/VP 95.5 14.0 63–136 101.1 14.1 63–136 
 I/BD 95.7 13.9 64–129 99.1 14.4 64–142 
 I/MR 95.7 13.7 69–132 99.6 12.3 64–124 
 I/VP 95.7 13.5 66–133 99.6 12.3 71–138 
 S/BD 95.7 14.0 58–131 99.8 13.8 58–133 
 S/MR 95.7 13.3 65–124 100.0 11.5 60–119 
 S/VP 95.8 13.5 63–129 100.4 12.1 65–129 
 Test sample
 
Validation sample
 
M SD Range M SD Range 
Estimated FSIQ 
 V/BD 92.5 14.4 62–128 97.5 15.7 54–136 
 V/MR 92.6 13.4 67–123 97.6 12.6 57–128 
 V/VP 92.5 14.0 60–133 98.1 14.1 60–133 
 I/BD 92.5 14.1 61–126 95.9 14.6 61–139 
 I/MR 92.5 13.6 66–128 96.3 12.1 61–121 
 I/VP 92.6 13.5 62–130 96.4 12.3 68–135 
 S/BD 92.6 14.4 54–129 98.4 14.5 54–131 
 S/MR 92.6 13.4 62–121 97.0 11.6 57–117 
 S/VP 92.5 13.7 59–126 97.1 12.3 61–126 
Estimated GAI 
 V/BD 95.6 14.3 65–131 100.6 15.6 57–139 
 V/MR 95.7 13.5 70–126 100.7 12.7 60–131 
 V/VP 95.5 14.0 63–136 101.1 14.1 63–136 
 I/BD 95.7 13.9 64–129 99.1 14.4 64–142 
 I/MR 95.7 13.7 69–132 99.6 12.3 64–124 
 I/VP 95.7 13.5 66–133 99.6 12.3 71–138 
 S/BD 95.7 14.0 58–131 99.8 13.8 58–133 
 S/MR 95.7 13.3 65–124 100.0 11.5 60–119 
 S/VP 95.8 13.5 63–129 100.4 12.1 65–129 

Notes: V/BD = Vocabulary/Block Design; V/MR = Vocabulary/Matrix Reasoning; V/VP = Vocabulary/Visual Puzzles; I/BD = Information/Block Design; I/MR = Information/Matrix Reasoning; I/VP = Information/Visual Puzzles; S/BD = Similarities/Block Design; S/MR = Similarities/Matrix Reasoning; S/VP = Similarities/Visual Puzzles.

Table 5.

Estimated full scale IQ scores from V/BD, V/MR, V/VP, I/BD, I/MR, I/VP, S/BD, S/MR, and S/VP dyad age-scaled score sums

Sum of age-scaled scores V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
45 50 43 50 51 49 49 52 48 
48 52 46 53 53 52 51 55 51 
51 55 49 56 56 54 54 57 53 
54 57 51 58 58 57 57 60 56 
56 60 54 61 61 60 59 62 59 
59 62 57 63 63 62 62 64 61 
62 64 60 66 66 65 65 67 64 
65 67 63 69 68 68 67 69 67 
10 67 69 66 71 71 71 70 71 70 
11 70 72 69 74 73 73 73 74 72 
12 73 74 72 76 76 76 75 76 75 
13 76 77 75 79 78 79 78 79 78 
14 78 79 78 82 81 81 81 81 80 
15 81 81 80 84 83 84 83 83 83 
16 84 84 83 87 86 87 86 86 86 
17 87 86 86 90 88 89 89 88 88 
18 89 89 89 92 91 92 91 90 91 
19 92 91 92 95 93 95 94 93 94 
20 95 94 95 97 96 97 97 95 97 
21 98 96 98 100 98 100 99 98 99 
22 100 98 101 103 101 103 102 100 102 
23 103 101 104 105 103 106 105 102 105 
24 106 103 107 108 106 108 107 105 107 
25 109 106 109 110 108 111 110 107 110 
26 111 108 112 113 111 114 113 109 113 
27 114 111 115 116 113 116 115 112 115 
28 117 113 118 118 116 119 118 114 118 
29 120 116 121 121 118 122 121 117 121 
30 122 118 124 123 121 124 123 119 124 
31 125 120 127 126 123 127 126 121 126 
32 128 123 130 129 126 130 129 124 129 
33 131 125 133 131 128 132 131 126 132 
34 133 128 136 134 131 135 134 129 134 
35 136 130 138 136 133 138 137 131 137 
36 139 133 141 139 136 140 139 133 140 
37 142 135 144 142 138 143 142 136 142 
38 144 137 147 144 141 146 145 138 145 
Sum of age-scaled scores V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
45 50 43 50 51 49 49 52 48 
48 52 46 53 53 52 51 55 51 
51 55 49 56 56 54 54 57 53 
54 57 51 58 58 57 57 60 56 
56 60 54 61 61 60 59 62 59 
59 62 57 63 63 62 62 64 61 
62 64 60 66 66 65 65 67 64 
65 67 63 69 68 68 67 69 67 
10 67 69 66 71 71 71 70 71 70 
11 70 72 69 74 73 73 73 74 72 
12 73 74 72 76 76 76 75 76 75 
13 76 77 75 79 78 79 78 79 78 
14 78 79 78 82 81 81 81 81 80 
15 81 81 80 84 83 84 83 83 83 
16 84 84 83 87 86 87 86 86 86 
17 87 86 86 90 88 89 89 88 88 
18 89 89 89 92 91 92 91 90 91 
19 92 91 92 95 93 95 94 93 94 
20 95 94 95 97 96 97 97 95 97 
21 98 96 98 100 98 100 99 98 99 
22 100 98 101 103 101 103 102 100 102 
23 103 101 104 105 103 106 105 102 105 
24 106 103 107 108 106 108 107 105 107 
25 109 106 109 110 108 111 110 107 110 
26 111 108 112 113 111 114 113 109 113 
27 114 111 115 116 113 116 115 112 115 
28 117 113 118 118 116 119 118 114 118 
29 120 116 121 121 118 122 121 117 121 
30 122 118 124 123 121 124 123 119 124 
31 125 120 127 126 123 127 126 121 126 
32 128 123 130 129 126 130 129 124 129 
33 131 125 133 131 128 132 131 126 132 
34 133 128 136 134 131 135 134 129 134 
35 136 130 138 136 133 138 137 131 137 
36 139 133 141 139 136 140 139 133 140 
37 142 135 144 142 138 143 142 136 142 
38 144 137 147 144 141 146 145 138 145 

Notes: The italicized scores in the table were not derived from actual data, and thus the Full Scale IQ estimates outside of these boundaries are suspect; V/BD = Vocabulary/Block Design; V/MR = Vocabulary/Matrix Reasoning; V/VP = Vocabulary/Visual Puzzles; I/BD = Information/Block Design; I/MR = Information/Matrix Reasoning; I/VP = Information/Visual Puzzles; S/BD = Similarities/Block Design; S/MR = Similarities/Matrix Reasoning; S/VP = Similarities/Visual Puzzles.

Table 6.

Estimated General Ability Index scores from V/BD, V/MR, V/VP, I/BD, I/MR, I/VP, S/BD, S/MR, and S/VP dyad age-scaled score sums

Sum of age-scaled scores V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
49 52 46 54 54 52 53 56 52 
52 55 49 57 56 55 56 58 55 
54 57 52 59 59 58 58 60 57 
57 60 54 62 61 60 61 63 60 
60 62 57 64 64 63 63 65 63 
63 65 60 67 66 66 66 68 65 
65 67 63 70 69 68 69 70 68 
68 70 66 72 71 71 71 72 71 
10 71 72 69 75 74 74 74 75 73 
11 74 75 72 77 76 76 76 77 76 
12 76 77 75 80 79 79 79 79 79 
13 79 80 78 82 81 82 82 82 81 
14 82 82 81 85 84 84 84 84 84 
15 84 84 83 88 87 87 87 86 87 
16 87 87 86 90 89 90 89 89 89 
17 90 89 89 93 92 93 92 91 92 
18 93 92 92 95 94 95 95 94 95 
19 95 94 95 98 97 98 97 96 97 
20 98 97 98 101 99 101 100 98 100 
21 101 99 101 103 102 103 102 101 102 
22 103 102 104 106 104 106 105 103 105 
23 106 104 107 108 107 109 108 105 108 
24 109 107 110 111 109 111 110 108 110 
25 112 109 112 113 112 114 113 110 113 
26 114 112 115 116 114 117 115 112 116 
27 117 114 118 119 117 119 118 115 118 
28 120 116 121 121 119 122 120 117 121 
29 122 119 124 124 122 125 123 119 124 
30 125 121 127 126 124 128 126 122 126 
31 128 124 130 129 127 130 128 124 129 
32 131 126 133 131 130 133 131 127 132 
33 133 129 136 134 132 136 133 129 134 
34 136 131 139 137 135 138 136 131 137 
35 139 134 141 139 137 141 139 134 140 
36 142 136 144 142 140 144 141 136 142 
37 144 139 147 144 142 146 144 138 145 
38 147 141 150 147 145 149 146 141 148 
Sum of age-scaled scores V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
49 52 46 54 54 52 53 56 52 
52 55 49 57 56 55 56 58 55 
54 57 52 59 59 58 58 60 57 
57 60 54 62 61 60 61 63 60 
60 62 57 64 64 63 63 65 63 
63 65 60 67 66 66 66 68 65 
65 67 63 70 69 68 69 70 68 
68 70 66 72 71 71 71 72 71 
10 71 72 69 75 74 74 74 75 73 
11 74 75 72 77 76 76 76 77 76 
12 76 77 75 80 79 79 79 79 79 
13 79 80 78 82 81 82 82 82 81 
14 82 82 81 85 84 84 84 84 84 
15 84 84 83 88 87 87 87 86 87 
16 87 87 86 90 89 90 89 89 89 
17 90 89 89 93 92 93 92 91 92 
18 93 92 92 95 94 95 95 94 95 
19 95 94 95 98 97 98 97 96 97 
20 98 97 98 101 99 101 100 98 100 
21 101 99 101 103 102 103 102 101 102 
22 103 102 104 106 104 106 105 103 105 
23 106 104 107 108 107 109 108 105 108 
24 109 107 110 111 109 111 110 108 110 
25 112 109 112 113 112 114 113 110 113 
26 114 112 115 116 114 117 115 112 116 
27 117 114 118 119 117 119 118 115 118 
28 120 116 121 121 119 122 120 117 121 
29 122 119 124 124 122 125 123 119 124 
30 125 121 127 126 124 128 126 122 126 
31 128 124 130 129 127 130 128 124 129 
32 131 126 133 131 130 133 131 127 132 
33 133 129 136 134 132 136 133 129 134 
34 136 131 139 137 135 138 136 131 137 
35 139 134 141 139 137 141 139 134 140 
36 142 136 144 142 140 144 141 136 142 
37 144 139 147 144 142 146 144 138 145 
38 147 141 150 147 145 149 146 141 148 

Notes: The italicized scores in the table were not derived from actual data, and thus the General Ability Index estimates outside of these boundaries are suspect; V/BD = Vocabulary/Block Design; V/MR = Vocabulary/Matrix Reasoning; V/VP = Vocabulary/Visual Puzzles; I/BD = Information/Block Design; I/MR = Information/Matrix Reasoning; I/VP = Information/Visual Puzzles; S/BD = Similarities/Block Design; S/MR = Similarities/Matrix Reasoning; S/VP = Similarities/Visual Puzzles.

Table 7.

Corrected correlations between actual and predicted indices

 V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
FSIQ 
 Test sample .86 .76 .82 .84 .79 .77 .86 .76 .79 
 Validation sample .88 .82 .83 .86 .79 .80 .82 .73 .74 
GAI 
 Test sample .91 .84 .88 .88 .85 .84 .88 .82 .84 
 Validation sample .93 .87 .89 .86 .85 .85 .90 .80 .82 
 V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
FSIQ 
 Test sample .86 .76 .82 .84 .79 .77 .86 .76 .79 
 Validation sample .88 .82 .83 .86 .79 .80 .82 .73 .74 
GAI 
 Test sample .91 .84 .88 .88 .85 .84 .88 .82 .84 
 Validation sample .93 .87 .89 .86 .85 .85 .90 .80 .82 

Notes: Correlations corrected using procedure described by Levy (1967). All significant, p < .001.

Validation Sample

Demographic data for the validation sample (N = 50) can be found in Table 1. Validation sample WAIS-IV FSIQ scores ranged from 51 to 131, with a mean of 98.4 (SD = 14.5) and validation sample GAI scores ranged from 55 to 137, with a mean of 101.2 (SD = 14.8). Means and SDs of validation sample FSIQ, GAI, and subtest performance can be seen in Table 2. No significant differences were noted between the test and validation groups in terms of age [t(161) = .264, p = .792] and education [t(161) = −.977, p = .330]. However, mean FSIQ [t(101.969) = −2.308, p = .023; Cohen's d = 0.46] and GAI [t(161) = −2.163, p = .032; Cohen's d = 0.34] were significantly higher in the validation sample than in the test sample. No gender differences were found between the test and validation samples (χ2[1, N = 163] = 2.81, p = .094). Both the test and validation samples were primarily white (86% and 76%, respectively) (χ2 [3, N = 163] = 8.82, p = .03).

Using Tables 5 and 6, estimated FSIQ and GAI scores were, respectively, derived for the cross-validation sample. Means and SDs of dyadic FSIQ and GAI estimations derived from the validation sample can be found in Table 4 and have a slightly larger range across dyad estimates than seen in the test sample. Corrected correlations between estimated and actual FSIQ (rsfs = .73 to .88, p < .001) and GAI (rsfs = .80 to .93, p < .001) were highly significant in the cross-validation sample (see Table 7) and similar to correlations in the test sample.

Paired t-test analyses between actual and estimated FSIQ scores significantly differed only for I/BD [t(49) = 2.445, p = .018; Cohen's d = 0.70] in the cross-validation sample. Comparisons between actual and estimated FSIQ scores for the remaining dyads in the cross-validation sample were not significant. Similarly, paired t-test analyses revealed a significant difference between the actual and estimated GAI scores derived from I/BD (t[49] = 2.336, p = .024; Cohen's d = .67) in the cross-validation sample, while all other dyad estimates did not significantly differ from actual GAI scores.

Estimated FSIQ and GAI scores were compared with their actual counterparts to examine how often estimated scores fell within 5 and 10 points of actual FSIQ and GAI (see Table 8). For FSIQ, estimations were within five points of actual scores among 40%–54% of cases. V/BD, I/BD, and S/BD were the most accurate dyads, each falling within five points of actual scores in 54% of cases. FSIQ estimations were within 10 points of actual scores in 80%–94% of cases. V/BD was the most accurate dyad, as 94% of estimations were within 10 points of actual scores. For GAI, estimated scores were within five points of actual scores among 54%–76% of cases. V/BD was the most accurate dyad, with 76% of estimations within five points of actual scores. GAI estimations were within 10 points of actual scores in 84%–96% of cases. V/BD was again the most accurate dyad. Percentages of estimated scores that fell within 5 and 10 points of actual scores were also examined by IQ ranges (see Table 8). Discrepancies between actual and estimated scores were lower in the 90–100 FSIQ/GAI ranges; though, it was not common to have greater than a 10-point discrepancy in any range.

Table 8.

Percentage of estimated scores within 5 and 10 points of actual scores in the cross-validation sample by FSIQ/GAI range

  V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
Actual FSIQ 
 ≤89 (n = 8) ±5 63 50 25 38 50 63 50 25 25 
±10 88 88 88 88 100 88 75 88 88 
 90–100 (n = 22) ±5 55 59 59 59 55 59 50 64 55 
±10 96 96 100 86 86 86 82 86 96 
 ≥101 (n = 20) ±5 50 45 45 55 25 40 60 35 30 
±10 95 85 70 90 70 90 80 70 65 
 Total (N = 50) ±5 54 52 48 54 42 52 54 46 40 
±10 94 90 86 86 82 88 80 80 82 
Actual GAI 
 ≤89 (n = 8) ±5 88 88 63 50 75 75 63 63 38 
±10 100 100 88 100 100 88 88 100 100 
 90–100 (n = 18) ±5 83 72 61 61 83 78 72 61 72 
±10 100 94 100 94 94 89 94 89 94 
 ≥101 (n = 24) ±5 67 63 67 54 33 54 75 71 46 
±10 92 83 92 79 71 88 92 83 79 
 Total (N = 50) ±5 76 70 64 56 58 66 72 66 54 
±10 96 90 94 88 84 88 92 88 88 
  V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP 
Actual FSIQ 
 ≤89 (n = 8) ±5 63 50 25 38 50 63 50 25 25 
±10 88 88 88 88 100 88 75 88 88 
 90–100 (n = 22) ±5 55 59 59 59 55 59 50 64 55 
±10 96 96 100 86 86 86 82 86 96 
 ≥101 (n = 20) ±5 50 45 45 55 25 40 60 35 30 
±10 95 85 70 90 70 90 80 70 65 
 Total (N = 50) ±5 54 52 48 54 42 52 54 46 40 
±10 94 90 86 86 82 88 80 80 82 
Actual GAI 
 ≤89 (n = 8) ±5 88 88 63 50 75 75 63 63 38 
±10 100 100 88 100 100 88 88 100 100 
 90–100 (n = 18) ±5 83 72 61 61 83 78 72 61 72 
±10 100 94 100 94 94 89 94 89 94 
 ≥101 (n = 24) ±5 67 63 67 54 33 54 75 71 46 
±10 92 83 92 79 71 88 92 83 79 
 Total (N = 50) ±5 76 70 64 56 58 66 72 66 54 
±10 96 90 94 88 84 88 92 88 88 

In order to compare the current study regression technique to the Tellegen and Briggs technique, estimated FSIQ scores based on the V/BD dyad were generated from the Sattler and Ryan tables (2009, p. 245) in our validation sample. Paired-sample t-tests were used to compare means of estimated FSIQ to actual FSIQ scores using both approaches. The mean Sattler and Ryan estimated FSIQ (M = 102.68, SD = 16.78) was significantly higher than actual FSIQ ([M = 98.40, SD = 14.55], t [49] = 4.54, p < .001; Cohen's d = 1.30). In contrast, there was not a significant difference between the current study regression-based estimated FSIQ (M = 97.5, SD = 15.7) and actual FSIQ (M = 98.40, SD = 14.55).

Discussion

Short forms of the Wechsler intelligence scales have been used in research and clinical settings to acquire estimates of FSIQ. Sattler and Ryan (2009) provided data for various dyads of WAIS-IV subtests, but these were based on data from the standardization sample and have not been validated in a clinical sample. The present study examined nine WAIS-IV dyadic short forms, derived from a clinical sample, by entering subtest scores into regression equations to predict FSIQ and GAI. Estimated scores accounted for a significant amount of variance in each of the nine short forms used to predict FSIQ (70%–83%) and GAI (77%–88%). Each of the nine WAIS-IV dyadic short forms examined was a good predictor of FSIQ and GAI in a mixed diagnostic sample. The V/BD dyad yielded the best estimates in that it accounted for the greatest amount of variance in FSIQ and GAI scores, had the highest percentage of predicted scores within 5 (54% in FSIQ; 76% in GAI) and 10 (94% in FSIQ; 96% in GAI) points, and had the highest correlation with actual FSIQ (r ′sf = .88) and GAI (r ′sf = .93) in the validation sample. Overall, estimates correlated better with actual GAI than actual FSIQ. This is not surprising as processing speed and working memory subtests were not included in the dyads.

Dyads with BD had the three highest R2 values among the nine dyads for FSIQ (V/BD = .83; I/BD = .79; S/BD = .82) and three of the four highest R2 values for GAI (V/BD = .88; I/BD = .84; S/BD = .85). This suggests that regardless of the subtest used from the VCI, BD is the best single subtest for predicting global ability. However, when patients have motor limitations, BD may not be the optimal subtest because of the necessary physical manipulation of the blocks. Of the non-motor perceptual reasoning subtests, VP appears to be better than MR at predicting FSIQ and GAI with the V/VP dyad serving as a more accurate predictor (FSIQ R2 = .77; GAI R2 = .85) than I/VP, S/VP, V/MR, I/MR, or S/MR. Girard and colleagues' (2014) examination of V/BD and V/VP ranked these dyads lower than others based on their composite validity measure due to their tendency to overestimate FSIQ in their sample. This discrepancy may be due to differences in subject populations and Girard and colleagues's inclusion of working memory and processing speed subtests in their analyses.

Sattler and Ryan (2009) also identified V/BD as being among the 10 best two-subtest short forms, using the Tellegen and Briggs (1967) approach. Comparisons of Sattler and Ryan's estimated FSIQ tables (p. 245) using the V/BD dyad to actual FSIQ scores in our validation sample showed that mean Sattler and Ryan estimated FSIQ was significantly higher than actual FSIQ. In contrast, there was not a significant difference between the current study regression-based V/BD estimated FSIQ and actual FSIQ. These findings indicate that our V/BD regression-based estimates of FSIQ derived from a clinical/neurologic sample were more accurate in predicting actual FSIQ than the previously reported Tellegen and Briggs method that used WAIS-IV standardization data. While our dyadic regressions appeared to produce more reliable estimates when used within neurological populations, further investigation is needed to determine the generalizability of these findings in other populations.

Despite the high correlations between predicted and actual FSIQ and GAI scores, caution is recommended in interpreting overall levels of intelligence based on these estimations. A comparison of agreement between actual and estimated classification ranges (based on the WAIS-IV technical manual) in the cross-validation sample showed only 56%–70% agreement among the FSIQ dyads while the GAI dyads ranged between 64% and 86% agreement. Complementing this finding is the relatively low number of cases whose estimations fell within five points of actual FSIQ (40%–54%) and GAI (54%–76%) across all nine dyads. However, all nine dyads provided estimated FSIQs within 10 points of actual FSIQ in at least 80% of cases while GAI estimations were within 10 points of actual GAI in at least 84% of cases. It may be advisable to report estimated FSIQ and GAI scores as a range or classification level (e.g., low average, average, etc.) rather than a specific score.

An obvious advantage of using dyadic short forms to provide FSIQ or GAI estimates is the time savings that are inherent with abbreviated testing. While published data on WAIS-IV subtest administration times in clinical samples are lacking, such information is available for the WAIS-III, which suggest that among the subtests used in this study (V, S, I, BD, and MR), the V/BD dyad is the most time consuming (roughly averaging 26 min) but accounts for, on average, <30% of the time required for WAIS-III FSIQ. The I/MR dyad is the quickest (roughly averaging 14 min), accounting for only 16% of the time necessary for attainment of FSIQ (Ryan, Lopez, & Werth, 1998). The WAIS-IV was developed to allow for quicker subtest administration than its predecessors, suggesting that short forms presented in this study are likely speedier than what has been previously reported (Wechsler, 2008). While no such information about VP is available, clinical experience as well as knowledge of the mechanics of the subtest indicates that, on average, VP administration time is less than BD or MR. The I/VP dyad may, therefore, be the most efficient option among those included in this study.

Limitations to the generalizability of these findings may include a lack of broad diversity among some demographic factors. While gender was well balanced, the test sample was primarily white (86%), highly educated (mean = 14.2 years; median = 15), and at the upper end of middle age (mean = 59.9 years; median = 62). Also, it is less clear how accurate predicted scores are in individuals at the lower and higher ends of the IQ spectrum given the relatively few subjects with FSIQ scores <90 (n = 8) or >109 (n = 8) in the validation sample. Processing speed and working memory subtests were not used, and it remains to be seen if inclusion of these subtests appreciatively adds to these results. However, Girard and colleagues (2014) found that overall, the Coding and Information dyad had the strongest measures of reliability and validity in their sample. Their results also showed that 9 of the top 10 dyads included either a processing speed or working memory subtest. Thus, inclusion of processing speed and working memory subtests for estimating intelligence may be beneficial in some populations.

Overall, the nine dyadic short forms of the WAIS-IV examined in this investigation were highly correlated with actual FSIQ and GAI scores in test and cross-validation samples. This model also demonstrated greater accuracy at predicting FSIQ in a clinical sample than the Sattler and Ryan tables for the V/BD dyad. Despite these high correlations, caution is recommended when reporting predicted scores, as relative ranges may be more appropriate. While utilizing short forms to predict FSIQ and GAI has inherent limitations, these nine dyads appear to be appropriate for use in mixed neurological populations when an estimate of overall intellectual functioning will suffice.

Conflict of Interest

None declared.

References

Axelrod
B. N.
Dingel
J. D.
Ryan
J. J.
Ward
L. C.
(
2000
).
Estimation of the Wechsler Adult Intelligence Sacle-III with the 7-subtest short form in a clinical sample
.
Assessment
 ,
7
,
157
161
.
Christensen
B. K.
Girard
T. A.
Bagby
M. R.
(
2007
).
Wechsler Adult Intelligence Scale—Third Edition short form for index and IQ scores in a psychiatric population
.
Psychological Assessment
 ,
19
,
236
240
.
Engelhart
C. I.
Eisenstein
J.
Johnson
V.
Losonczy
M.
(
1999
).
Comparison of linear equating and prorated short forms for estimating WAIS-R FSIQ in a neuropsychological population
.
The Clinical Neuropsychologist
 ,
13
,
95
99
.
Girard
T. A.
Axelrod
B. N.
Patel
R.
Crawford
J. R.
(
2014
).
Wechsler Adult Intelligence Scale-IV dyads for estimating global intelligence
.
Assessment
 ,
1–8
, .
Jeyakumar
S. L. E.
Warriner
E. M.
Raval
V. V.
Ahmad
S. A.
(
2004
).
Balancing the need for reliability and time efficiency: Short forms of the Wechsler Adult Intelligence Scale-III
.
Educational and Psychological Measurement
 ,
64
,
71
87
.
Kaufman
A. S.
Ishikuma
T.
Kaufman-Packer
J. L.
(
1991
).
Amazingly short forms of the WAIS-R
.
Journal of Psychoeducational Assessment
 ,
9
,
4
15
.
Kaufman
A. S.
Kaufman
N. L.
(
2004
).
Kaufman Brief Intelligence Test Second Edition
 .
Circle Pines, MN
:
AGS Publishing
.
Levy
P.
(
1967
).
The correction for spurious correlation in the evaluation of short-form tests
.
Journal of Clinical Psychology
 ,
1
,
84
86
.
Mendella
P. D.
McFadden
L.
Regan
J.
Medlock
L.
(
2000
).
Short-form prediction of WAIS-R scores in a sample of individuals diagnoses with multiple sclerosis
.
Applied Neuropsychology
 ,
7
,
102
107
.
Pilgrim
B. M.
Meyers
J. E.
Bayless
J.
Whetstone
M. M.
(
1999
).
Validity of the Ward seven-subtest WAIS-III short form in a neuropsychological population
.
Applied Neuropsychology
 ,
6
,
243
246
.
Reynolds
C. R.
Kamphaus
R. W.
(
2003
).
Reynolds Intellectual Assessment Scales and Reynolds Intellectual Screening Test: Professional Manual.
 
Odessa, FL
:
PAR
.
Ringe
W. K.
Saine
K. C.
Lacritz
L. H.
Hynan
L. S.
Cullum
C. M.
(
2002
).
Dyadic short forms of the Wechsler Adult Intelligence Scale-III
.
Assessment
 ,
9
,
254
260
.
Ryan
J. J.
Lopez
S. J.
Werth
T. R.
(
1998
).
Administration time estimates for WAIS-III subtests, scales, and short forms in a clinical sample
.
Journal of Psychoeducational Assessment
 ,
16
,
315
323
.
Satterfield
W. A.
Martin
C. W.
Leiker
M.
(
1994
).
A comparison of four WAIS-R short forms in patients referred for psychological/neuropsychological assessments
.
Journal of Psychoeducational Assessment
 ,
12
,
364
371
.
Sattler
J. M.
Ryan
J. J.
(
2009
).
Assessment with the WAIS-IV
 .
San Antonio, TX
:
Psychological Corporation
.
Schoenberg
M. R.
Duff
K.
Dorfman
K.
Adams
R. L.
(
2004
).
Differential estimation of verbal intelligence and performance intelligence scores from combined performance and demographic variables: The OPIE-3 verbal and performance algorithms
.
Clinical Neuropsychology
 ,
16
,
266
276
.
Schoenberg
M. R.
Duff
K.
Scott
J. G.
Adams
R. L.
(
2002
).
An evaluation of the clinical utility of the OPIE-3 as an estimate of premorbid WAIS-III FSIQ
.
Clinical Neuropsychologist
 ,
17
,
308
321
.
Schoenberg
M. R.
Scott
J. G.
Ruwe
W.
Patton
D.
Adams
R. L.
(
2004
).
Assumptions that underlie predicting premorbid IQ: A comment on the “Evaluation of the accuracy of two regression-based methods for estimated premorbid IQ.”
Archives of Clinical Neuropsychology
 ,
19
,
1103
1106
.
Silverstein
A. B.
(
1982
).
Two- and four-subtest short forms of the Wechsler Adult Intelligence Scale-Revised
.
Journal of Consulting and Clinical Psychology
 ,
50
,
415
418
.
Tellegen
A.
Briggs
P. F.
(
1967
).
Old wine in new skins: Grouping Wechsler subtests into new scales
.
Journal of Consulting Psychology
 ,
31
,
504
.
Thompson
A. P.
LoBello
S. G.
Atkinson
L.
Chisholm
V.
Ryan
J. J.
(
2004
).
Brief intelligence testing in Australia, Canada, the United Kingdom, and the United States
.
Professional Psychology: Research and Practice
 ,
35
,
286
290
.
Ward
L. C.
(
1990
).
Prediction of verbal, performance, and Full Scale IQs from seven subtests of the WAIS-R
.
Journal of Clinical Psychology
 ,
46
,
436
440
.
Wechsler
D.
(
1955
).
Wechsler Adult Intelligence Scale
 .
New York
:
Psychological Corporation
.
Wechsler
D.
(
1987
).
Wechsler Adult Intelligence Scale-Revised
 .
San Antonio, TX
:
Psychological Corporation
.
Wechsler
D.
(
1997
).
Wechsler Adult Intelligence Scale-Third Edition
 .
San Antonio, TX
:
Psychological Corporation
.
Wechsler
D.
(
2008
).
Wechsler Adult Intelligence Scale-Fourth Edition administration and scoring manual
 .
San Antonio, TX
:
Psychological Corporation
.
Wechsler
D.
(
2011
).
Wechsler Abbreviated Scale of Intelligence
 
(2nd ed.). Bloomington, MN: Pearson
.