Age Effects on Subtest and Composite Scores of the Wechsler Abbreviated Scale of Intelligence–Second Edition

Method

The WASI-II adult standardization sample (N = 1,200) was stratified on the basis of age, sex, education, ethnicity/race, and geographic region of residence in proportions consistent with the 2008 U.S. Census. The scale norms are in increments of 3 years for those in the age range 17 to 19 years, 5 years for those 20 to 34 years, 10 years for those 35 to 64 years, and 5 years for those 65 to 90 years. Of the 12 age groups, seven (i.e., 17–19 years, 20–24 years, 25–29 years, 30–34 years, 35–44 years, 45–54 years, and 55–64 years) include 50 males and 50 females. The remaining groups have more female than male participants as follows: 65 to 69 years (M = 47, F = 53), 70 to 74 years (M = 45, F = 55), 75 to 79 years (M = 42, F = 58), 80 to 89 years (M = 37, F = 63), and 85 to 90 years (M = 34, F = 66).

Previous studies (e.g., Ryan et al., 2000; Sattler, 1982; Sattler & Ryan, 2009) have used a reference group of individuals in the age range 20 to 34 years (a combination of 20–24 years, 25–29 years, and 30–34 years) to determine age effects on the Wechsler subtests. Separate tables for these reference groups are provided in the WAIS manuals, but the WASI-II does not provide a similar reference group table. Therefore, the 20 to 24 years age range was defined as the reference group in the present study. The raw score on each subtest that equals a T-score of 50 in the reference group was found in Table A.1 of the WASI-II manual for each of the 12 age groups. Next, each T-score was converted to a scaled score (M = 10, SD = 3) using Table A.2. Finally, to ascertain the number of points above or below the average of the 20- to 24-year-old reference group, 10 was subtracted from the scaled score for each subtest. For ease of interpretation, when performance on a subtest exceeded the reference group, it was recorded as negative value.

To clarify age effects on the VCI, PRI, and FSIQ, raw scores on the four subtests that produce a T-score of 50 were obtained from Table A.1 for each of the 12 age groups in the standardization sample. Next, the raw scores for each subtest at each age level were converted to T-scores using the 20- to 24-year reference group norms in Table A.1. The T-scores were then combined to obtain sums for the VCI (VC + SI), PRI (BD + MR), and FSIQ (BD + VC + MR + SI). Scores for each of the 12 age groups were converted to composites with a mean of 100 and standard deviation of 15 using Tables A.3, A.4, and A.5 in the test manual. Because overall ability level may influence age-related changes (Baxendale, 2011) on the VCI, PRI, and FSIQ, the analyses were repeated using the 20- to 24-year-old reference group in the borderline (i.e., FSIQ = 75) and superior (i.e., FSIQ = 121) ranges of intelligence. Raw scores derived from T-scores of 36 and 66 on each subtest were used in the borderline and superior level analyses, respectively.

Results

Table 1 indicates that the VC and SI subtests were more stable across the age range than the BD and MR subtests. The VC subtest is clearly the most stable and demonstrates minimal change across the life span. For example, a scaled score of 10 in the reference group was also achieved by individuals 70 to 74 years of age. Even in the 85 to 90 years age range, achievement was only one scaled score point below that of the reference group. It is also noted that VC scores peaked in the 30 to 64 years age range, being one scaled score point above persons 20 to 24 years of age. SI is the second most stable subtest, with peak performance among individuals 30 to 54 years of age followed by steady, mild decline beginning in the 55- to 64-year cohort. For persons 80 years and older, the scaled score was a full standard deviation (i.e., 3 scaled score points) below the reference group.

OPEN IN VIEWER

Table 1.

Additional Scaled Score Points by Age Awarded When the Reference Group Receives a Scaled Score of 10.

Subtest	Age
	17–19	20–24	25–29	30–34	35–44	45–54	55–64	65–69	70–74	75–79	80–84	85–90
Block design	0	0	0	0	0	1	2	3	3	4	4	4
Vocabulary	0	0	0	–1	–1	–1	–1	0	0	1	1	1
Matrix reasoning	1	0	0	1	1	1	2	3	4	4	5	6
Similarities	0	0	0	–1	–1	–1	1	1	2	2	3	3

Both the BD and MR subtests showed substantial declines as one moved from the younger to older age groups. BD peaked in the age range 17 to 44 years and began a steady decline thereafter. For individuals 65 to 90 years, scaled scores were 1.0 to 1. 3 SDs below the reference group mean scaled score of 10. A raw score of 44 in the reference group produces a scaled of 10 but for persons 85 to 90 years, the associated scaled score is 6 (raw score = 21). MR peaked at the relatively young age of 20 to 29 years and started to decline in the 30- to 34-year cohort. The 85- to 90-year-old group earned a score that was 2.0 SDs below that of the reference group. A raw score of 21 yields a scaled score of 10 in the reference group but produces a scaled score of 4 (raw score = 11) among those in the 85- to 90-year cohort. Table 1 indicates more fully the declines in performance on the subtests by providing the scaled scores needed at each age level to obtain a scaled score of 10.

Age effects on the three composite values are provided in Table 2 in the FSIQ = 100 column. The VCI peaks in the 30 to 44 years age range and slowly declines thereafter with a maximum drop of 14 points for persons 85 to 90 years. The PRI peaks in the 20- to 29-year cohort and demonstrates a marked decline exceeding 1.0 SD below the mean at age 65 to 69 years. For individuals in the 85 to 90 years age range, a drop of almost 2.0 SDs (i.e., 29 points) below the mean is evident on the PRI. The FSIQ remains in the average range until age 65 years followed by a decline into the borderline range for individuals 85 to 90 years of age.

OPEN IN VIEWER

Table 2.

Changes in Composite Scores for the 12 Age Cohorts When the FSIQ Is in the Borderline, Average, and Superior Ranges of Ability.

Age group	FSIQ = 75			FSIQ = 100			FSIQ = 121
	VCI	PRI	FSIQ	VCI	PRI	FSIQ	VCI	PRI	FSIQ
17–19	77	77	75	99	97	98	116	116	118
20–24	77	77	75	100	100	100	117	119	121
25–29	77	77	75	100	100	100	116	118	119
30–34	77	75	74	103	98	101	118	114	119
35–44	77	75	74	103	97	100	126	113	123
45–54	75	72	71	100	93	96	126	108	120
55–64	73	65	69	100	87	93	119	103	113
65–69	70	62	64	96	83	88	116	99	108
70–74	67	58	61	95	81	86	113	95	105
75–79	64	56	58	91	78	82	110	92	101
80–84	62	53	55	90	75	80	108	89	99
85–90	60	51	52	89	71	78	108	86	97

Note. FSIQ = Full Scale IQ; VCI = Verbal Comprehension Index; PRI = Perceptual Reasoning Index.

In addition to cases with average global intelligence, Table 2 reports the VCI, PRI, and FSIQ values for persons in the borderline and superior levels of ability. The first impression gleaned from Table 2 is that regardless of ability level, the VCI is less vulnerable to age effects than is the PRI. Peak achievement on the VCI differs slightly across ability levels as persons in the borderline group scored best at younger ages (i.e., 17–44 years) compared with those with average (i.e., 30–44 years) or superior (i.e., 35–54 years) intellect. Defining substantial decline in ability as 1.0 SD below the peak score, VCI decline occurs at 80 to 84 years, 85 to 90 years, and 75 to 79 years in the borderline, average, and superior ability groups, respectively. For the PRI, peak achievement levels were similar for the three categories, while a 1.0 SD decline was at a younger age cohort for superior ability (55–64 years) individuals than for those with borderline or average FSIQ (65–69 years). Examination of FSIQ scores indicated that peak performance occurs in the age range 17 to 29 years for those with borderline ability, but at slightly more advanced ages for individuals with average (30–34 years) or superior (35–44 years) global ability. A 1.0 SD decline occurs in the age range 70 to 74 years for individuals with borderline or average general ability. However, those with superior intelligence started to demonstrate reduced test scores at a younger age (65–69 years).

Discussion

The findings of the present investigation are very similar to those reported for the W-B I, WAIS, WAIS-R, WAIS-III, WAIS-IV, and WASI-II (e.g., Baxendale, 2011; Kaufman, 1990; Kaufman & Lichtenberger, 2002, 2006; Lichtenberger & Kaufman, 2013; Miller et al., 2009; Ryan et al., 2000; Sattler, 1982; Sattler & Ryan, 2009). With respect to the individual subtests, VC was clearly the least sensitive to the effects of aging as a raw score of 38 yielded a scaled of 10 for persons 65 to 69 years as well as for individuals as young as 20 to 29 years of age. The SI subtest was more susceptible to age effects than was the VC subtest. For instance, in the 65 to 69 years age range to achieve a scaled score of 10 VC required no additional points, whereas five supplementary points on the SI subtest were needed. This observation is consistent with the position that SI represents a blend of crystallized and fluid abilities (Lichtenberger & Kaufman, 2013).

MR and BD showed peak performance in early adulthood (MR = 20–29 years; BD = 17–34 years), with steady declines during early middle age that continued unabated into the eighth decade of life. As Table 1 suggests, MR scores begin to decline in an earlier age cohort (30–34 years) than do scores on BD (45–54 years) and the magnitude of deterioration is more pronounced for MR relative to BD. Within the crystallized and fluid theoretical framework, it has been recognized for decades that the BD subtest is a mixture of fluid and visualization abilities. However, the MR subtest, which first appeared in the WAIS-III (Wechsler, 1997) and the WASI (Wechsler, 1999), is described by its publisher as a measure of fluid intelligence and reasoning. Researchers and practitioners have quietly accepted this assertion and utilized the subtest in numerous investigations (e.g., Bugg et al., 2006; Keage et al., 2015) and clinical situations where the assessment of fluid abilities is required. Unfortunately, research has not always supported the publisher’s expectations concerning the MR subtest and the measurement of fluid intelligence.

Concerns about the validity of the MR subtest come from different sources. Donders et al. (2001) assessed the criterion validity of the WAIS-III subtests in a sample of 100 individuals with traumatic brain injury (TBI). Results indicated that MR was the least sensitive of the 11 core subtests to TBI, regardless of injury severity. In terms of scaled score means, moderately to severely injured participants earned an MR score of 10.5 and healthy controls produced a mean of 10.0. Tranel et al. (2008) reported that the WAIS-III MR subtest was insensitive to the consequences of frontal lobe damage. MR was no more sensitive to the condition of the frontal lobes than was the VC subtest, a measure of crystallized intelligence. Patients with dorsolateral prefrontal cortical lesions earned average mean scaled scores on MR (10.4) and VC (10.6) as did cases with ventromedial prefrontal cortical lesions (MR = 10.7; VC = 11.1). These findings prompted Tranel et al. to question the construct validity of the WAIS-III version of MR.

Studies that utilized the WAIS-IV in TBI reported similar findings. Donders and Strong (2015) evaluated 100 patients with TBI and reported MR means of approximately 9.6 for uncomplicated mild, 11.0 for complicated mild, and 9.5 for moderate to severe injury. The control group of 100 individuals earned a mean of approximately 9.5. The insensitivity of the MR subtest to the neurological consequences of TBI was reinforced by Carlozzi et al. (2015). In their study of 100 TBI cases who completed the WAIS-IV, the normal control group MR scaled score mean was 10.7, while the means for patients with mild-moderate or severe injury were 10.1 and 9.4, respectively.

The sensitivity to brain injury of the WASI version of the MR subtest was evaluated by Ryan et al. (2005). MR scores of patients with TBI, stroke, or a dementia syndrome were compared using demographically based premorbid ability estimates for the combined groups. The WASI FSIQ was found to be strongly influenced by the presence and severity of neurological damage or disease. Relative to the premorbid estimates, an average loss in FSIQ of 15 points was observed. However, the TBI patients’ MR subtest mean IQ equivalent was essentially the same as the premorbid IQ estimate and the IQ estimated means for VC, BD, and SI were all significantly below the premorbid IQ estimates. The MR subtest was insensitive to the sequelae of TBI, while performance on the remaining subtests was clearly compromised by the presence of dementia or stroke.

From the foregoing it appears that the aging patterns characterizing the WASI-II subtests are generally consistent with crystallized and fluid intelligence theories. These findings are also in line with previous investigations that used the standardization sample of one or more of five WAIS editions to assess age effects on intelligence test performance. Any differences in peak scores and starting points of cognitive decline probably reflect differences between the WASI-II standardization sample and the various WAIS standardization samples, the use of different reference groups to derive scores (e.g., 20–24 years vs. 20–34 years), and the necessity of transforming T-scores (M = 50, SD = 10) on the WASI-II into scaled scores (M = 10, SD = 3) as on the traditional Wechsler scales. In spite of these positive outcomes, the extant literature suggests that the MR subtest does not fully adhere to theoretical predictions of crystallized and fluid intelligence theory and may in fact be only selectively sensitive to brain injury and normal aging. An additional concern about the MR subtest is a report that it is not interchangeable with Raven’s Progressive Matrices (Kohutek, 1999), a validated measure of fluid intelligence.

The aging patterns for the WASI-II, VCI, PRI, and FSIQ are very similar to those reported for the VIQ, PIQ, and FSIQ on the WAIS, WAIS-R, and WAIS-III as well as the VCI, PRI, and FSIQ on the WAIS-III and WAIS-IV (Baxendale, 2011; Kaufman & Lichtenberger, 2006; Lichtenberger & Kaufman, 2013). For persons in the average range of intelligence, the WASI-II verbal tasks (VCI) are maintained throughout most of the adult life span, with the difference between the optimally functioning group (i.e., 30–44 years) and the oldest individuals (i.e., 85–90 years) being less than 1.0 SD (14 points). Conversely, perceptual reasoning skills (PRI) peak at a relatively early age (i.e., 20–29 years) and decline quickly as one moves from the younger to the older age groups, the decline from the optimally functioning individuals to the most elderly individuals (i.e., 85–90 years) being almost 2.0 SDs (29 points). These results for the WASI-II conform to predictions based on crystallized and fluid intelligence theory. The former construct represented by the VCI resists the cognitive effects of aging until relatively late in life, whereas the latter reflected by the PRI is vulnerable to age-related changes starting in early middle age with relatively rapid decline thereafter.

Based on her analysis of the WAIS-IV standardization sample, Baxendale (2011) reported that peak performance and initial decline in cognitive ability occurs in different age cohorts for individuals with borderline (i.e., FSIQ range, 70–79) compared to superior FSIQ (i.e., ≥120) intelligence. Using the WASI-II standardization sample and defining meaningful cognitive decline as 1.0 SD below the peak level, similar findings emerged as those with borderline ability achieved their best composite scores earlier in life (VCI = 17–44 years; PRI = 17–29 years; FSIQ = 17–29 years) than did individuals with superior ability (VCI = 35–54 years; PRI = 20–24 years; FSIQ = 35–44 years). Moreover, the borderline individuals maintained their cognitive functions longer, showing decline at an older age range (VCI = 80–84 years; PRI = 65–69 years; FSIQ = 70–74 years) than did those with superior FSIQ (VCI = 75–79 years; PRI = 55–64 years; FSIQ = 65–69 years). Baxendale also reported that in the WAIS-IV standardization sample declines on PRI and FSIQ were delayed in high-functioning people, occurring in a later age cohort (35–44 years) than in people with average intellect (30–34 years). This finding was reversed in the WASI-II standardization sample, as individuals with average ability maintained their cognitive abilities longer (VCI = 85–90 years; PIQ = 60–69 years, FSIQ = 70–74 years) than did persons with superior FSIQ (VCI = 75–79 years; PRI = 55–64 years; FSIQ = 65–69 years).

Decreases in the magnitudes of the PRIs and FSIQs from adolescence to old age are substantial. As Table 2 indicates, declines can be as much as 27 points on the PRI and 20 points on the FSIQ. Similar reductions in cognitive ability have been reported for the WAIS, WAIS-R, WAIS-III, and WAIS-IV (e.g., Kaufman & Lichtenberger, 2002, 2006; Lichtenberger & Kaufman, 2013; Miller et al., 2009). With respect to the WASI-II, performance reductions occurred even though standardization participants were screened to exclude those taking medications that might influence cognitive performance and those with neurological conditions, head injury, alcoholism, and significant mood or anxiety disorders. Persons with a history of receiving chemotherapy or electroconvulsive therapy were also excluded. This screening procedure likely produced a normative sample of optimally healthy elderly. Nevertheless, because normal aging is associated with increasing degrees of brain atrophy, reduced cerebral blood flow, hypertension, diabetes mellitus, and sleep disorders, to name just a few common problems, it is highly likely that the declines seen in this study reflect compromised brain functions (Rinn, 1988).

Knowledge of the declines reported in the present investigation may have practical implications. In the real world, psychometric test batteries are often used to predict everyday behaviors, such as the ability to safely operate a motor vehicle. If the WASI-II is part of a driving assessment battery, it would be wise to avoid making age corrections before interpreting the resulting test scores. Safe driving requires the same level of cognitive functioning from everybody regardless of chronological age. When judging the driving capacity of an 85-year-old, it is best to compare scores to norms from an optimally functioning sample (i.e., 20–24 years) and to avoid age corrections. This procedure yields very different results as the individual earned age-corrected values of VCI = 99, PRI = 100, and FSIQ = 99. These scores suggest that the individual is grossly cognitively intact and solidly average for his age group. However, when the age corrections are removed, the VCI = 89, PRI = 71, and FSIQ = 78. The most important finding is that the PRI score dropped from average (50th percentile) to borderline (3rd percentile), suggesting that the individual is average for his age group but when compared to optimally functioning individuals he is clearly disabled. This conclusion is based on research indicating that BD scores along with other measures of visual spatial skills are significant predictors of safe driving ability (Dawson et al., 2010). This example is also consistent with the position that demographic adjustments of raw scores on neuropsychological tests may diminish predictive accuracy for everyday tasks such as safe driving (Barrash et al., 2010).

For the WASI-II, the current study provides evidence of construct validity because the subtest and composite scores (a) show evidence of development and decline as one moves from young adulthood to old age and (b) conform nicely to the theory of crystallized and fluid intelligence and its various iterations. Careful scrutiny of the age curves reported for the various Wechsler adult intelligence scales published in the past decades reveals some differences in terms of the age cohorts where peak performance and initial declines were evident. The observation that age effects on the WASI-II do not precisely replicate similar analyses using the standard Wechsler scales should be of little concern and come as no surprise. Any discrepancies between current and past results must consider the fact that the WASI-II is composed of four subtests, whereas the traditional Wechsler scales include numerous subtests that attempt to uniquely assess processing speed, fund of general information, spatial visual-perceptual processing, and working memory. When comparing the WASI-II with the WAIS-III or WAIS-IV, it should be kept in mind that the WASI-II VCI and PRI, each composed of two subtests, are being contrasted to indexes that each contain three subtests. The four subtest WASI-II IQ is contrasted with FSIQs from the WAIS-III and WAIS-IV, which are based on 11 and 10 subtests, respectively.

A factor that may have contributed to differences between the WASI-II and the Wechsler adult scales in terms of peak performance and initial point of cognitive decline for the 12 age cohorts is the type of score utilized. The WASI-II subtest scores are normally reported as T-scores (M = 50, SD = 10), while the Wechsler scales use scaled scores (M = 10, SD = 3). Although the WASI-II uses T-scores for the subtests, we elected to use scaled scores because this matrix has been used in all previous studies of age effects on the Wechsler scales. However, the conversion of raw scores to T-scores and then to scaled scores may have introduced unanticipated error variance. This hypothesis seems reasonable since the BD subtest peak and initial decline points differ for T-scores (peaks at 20–34 years; initial decline at 35–44 years) and scaled scores (peaks at 17–44 years; initial decline at 45–54 years). Disagreement also emerged on the VC subtest as the point of initial decline was at a younger age when T-scores (70–74 years) were used compared to scaled scores (75–79 years).

The present study has a number of potential limitations. Past research used a reference group aged 20 to 34 years to determine developmental changes (e.g., Kaufman & Lichtenberger, 2002, 2006; Lichtenberger & Kaufman, 2013; Miller et al., 2009; Ryan et al., 2000; Sattler, 1982; Sattler & Ryan, 2009) in the allotment of scaled score points and to provide a standard for age-to-age comparisons. However, this investigation used a reference group 20 to 24 years of age because the manual does not provide an analogous reference table to those presented in, for example, the WAIS-IV (Table A.2, page 219). Also, the structure of Table A.1 in the WASI-II manual precludes the creation of a 20- to 34-year reference group without access to raw data from the standardization sample. The impact of using a 20- to 24-year-old reference group is unknown, but it may have altered slightly the identified points of peak performance and decline.

The use of a cross-sectional design is a possible limitation of the current study. Cohort effects characterize different age groups (e.g., 25–29 years vs. 70–74 years) in the standardization sample because growing up in different historical periods results in markedly dissimilar cultural and experiential experiences. Perhaps the most important cohort effect is educational background since individuals born in the mid to late 1940s, for example, tended to complete fewer years of formal schooling than did those born in the 1990s. The importance of this difference is reflected in by the finding of Kaufman (1990) that education accounts for a higher percentage of the variance in FSIQ (45.5%) than does chronological age (13.4%).

To compensate for the education problem, Kaufman equated the age groups of the WAIS, WAIS-R, WAIS-III, and WAIS-IV standardization samples (Kaufman, 1990; Kaufman & Lichtenberger, 2002, 2006; Lichtenberger & Kaufman, 2009, 2013) for differences in educational attainment. A series of investigations indicated that verbal abilities were maintained well into the eighth decade of life and that equating for education practically eliminated age-related decline in this area. However, declines in nonverbal functioning remained substantial and steady well into old age (Lichtenberger & Kaufman, 2009, 2013) The studies’ authors pointed out that concerns over educational differences across age cohorts in standardization samples have been justified in the past. However, because of demographic changes in the U.S. population, the age groups in normative samples have become more uniform in terms of stratification variables, especially in the latest edition of tests such as the WAIS-IV, which was standardized in 2007.

Examination of the educational levels in the WAIS-IV standardization sample indicate that, without controlling for education, the composite means now provide good estimates of the declines associated with normal aging, especially among individuals 16 to 69 years of age (Lichtenberger & Kaufman, 2013). It seems reasonable to generalize this conclusion to the WASI-II because it was standardized in 2010 to 2011 and included an increased percentage of elderly high school graduates. In the age cohorts 80 to 84 years and 85 to 90 years, the percentages of individuals with ≥12 years of educational attainment are 79% and 75%, respectively. These percentages are larger than those for the WAIS-IV standardization participants in the same age groups (i.e., 80–84 years = 72%; 85–90 years = 64%).

Additional limitations of the study include relatively small sample sizes, as only 100 individuals were included at each age level. Similar research using the WAIS-III (Ryan et al., 2000) and WAIS-IV (Miller et al., 2009) utilized samples of 200 participants at each age range. Until proven otherwise, it is reasonable to assume that meaningful error was not introduced into the WASI-II normative data by reducing the size of each age sample (Wechsler, 2011). Another concern is the absence in the literature of any studies that address the equivalence of WASI-II scores across the age-differentiated groups in the normative sample. Future research is needed to determine whether the factor structures are the same across the 12 age levels from adolescence to adulthood.

In conclusion, this is the first investigation of age effects on the WASI or WASI-II. The findings are consistent with previous research using the five editions of the Wechsler adult intelligence scales since a definite pattern of retained and reduced abilities emerged across the adult life span in line with predictions from crystallized and fluid intelligence theory. The capacity to engage in verbal expression and reasoning abilities (VCI) is maintained at least until age 75 years, whereas visuospatial problem-solving and fluid reasoning (PRI) appears to show decline beginning around 55 to 64 years of age. Analysis of individual subtest scores showed that VC subtest scores peaked at 30 to 64 years, with no meaningful decline thereafter. Conversely, the MR subset scores peaked in early adulthood (i.e., 20–29 years) and declined steadily until age 90. It is important to keep in mind that these differences do not document changes in ability with age but only how younger adults compare to older adults. The results of this study should inform psychologists who use the WASI-II concerning typical fluctuations in intellectual functioning associated with normal aging, including the classical aging pattern of VCI > PRI (Botwinick, 1977) and the rank ordering of subtest scores from highest to lowest. The pattern of subtest scores on the WASI-II (i.e., VC > SI > BD > MR) is identical to that reported for the WAIS-III and WAIS-IV.