Wisconsin Card Sorting Test Performance in Above Average and Superior School Children: Relationship to Intelligence and Age

Archives of Clinical Neuropsychology, Vol. 13, No. 8, pp. 713 720, 1998 Copyright 1998 National Academy of Neuropsychology Printed in the USA. All rights reserved 0887-6177/98 $19.00.00 PII S0887-6177(98)00007-9 Wisconsin Card Sorting Test Performance in Above Average and Superior School Children: Relationship to Intelligence and Age Sharon Arffa Allegheny University of the Health Sciences Mark Lovell Henry Ford Hospital Kenneth Podell Medical College of Pennsylvania and Hahnemann University Elkhonon Goldberg New York University School of Medicine This study explores the relationship of intelligence and age to scores on the Wisconsin Card Sorting Test, a measure of executive function. A sample of 26 normal children with Wechsler Intelligence Scale for Children-III (WISC-III) Full-Scale IQS above 130 and 24 normal children with WISC-III Full-Scale IQS between 110 and 129 were administered the test. A comparison to published norms revealed that above average children outperformed the average 9- to 14-yearold child on every measure at every age. Multiple regression analyses statistically related the Wisconsin Card Sorting Test, perseverative, nonperseverative, total errors, and trials to the first category of intelligence. Intelligence proved to be a significant qualifier of age trends. Gender relationships were nonsignificant in a preliminary analysis. 1998 National Academy of Neuropsychology. Published by Elsevier Science Ltd The relationship of intelligence tests to other neuropsychological measures is a complex one that may vary as a function of the age of the participant, the specific neurocognitive function sampled, and the overall level of functioning of the individual. Although IQ This research was supported by a grant from the Allegheny Singer Research Foundation. Address correspondence to Sharon Arffa, Allegheny General Hospital, Department of Psychiatry, 320 East North Avenue, Pittsburgh, PA 15212. 713

714 S. Arffa et al. and neuropsychological measures are clearly related in adults, neuropsychological measures may be more sensitive to cerebral integrity. During childhood, on the other hand, intelligence measures may be some of the most sensitive indicators of cerebral compromise (Klonoff & Low, 1974; Leikliter & Matarazzo, 1989; Seidenberg, Giordani, Berent, & Boll, 1983; Reitan, 1959, 1985; Waldemann, Dickson, Monahan, & Kazelskis, 1992). Not all researchers have identified a correlation between IQ and neuropsychological tests. In some instances, this may have occurred because of ceilings on certain tests (e.g., Trail Making Test; Waldemann et al., 1992) or restricted IQ ranges (Wiens & Matazzaro, 1977). Seidenberg et al. (1983) noted that the type of neurocognitive test is important, with the strongest IQ relationships occurring with neuropsychological measures requiring more conceptual problem-solving ability, mental efficiency, and languagerelated skills. Simple motor, constructional, and perceptual tasks, on the other hand, were unrelated to group differences in IQ. This study is also relevant to the relationship of the overall level of intelligence and neurocognitive functioning. Very high IQ samples are rarely studied, but clear improvement on neuropsychological measures occurs as IQ increases into the superior ranges. For example, Parsons (1984) found better performance on the Category Test, Oral Word Fluency Test, and Stroop Color-Word Association Test in one group of adults with a mean IQ of 120 compared to a group with a mean IQ of 104. The research on the relationship of IQ to performance on executive function tests is sparse and inconsistent. Executive functions refer to a variety of abilities ranging from simple voluntary initiation and inhibition of behavior to those involving complex planning, problem solving, and insight. Parsons (1984) reported a relationship between intelligence tests and some executive control measures in an adult sample with average to superior IQ. However, Boone, Ghafferian, Lesser, Hill-Gutierrez, and Berman (1993) found the Wechsler Adult Intelligence Scale-Revised IQ to be unrelated to Wisconsin Card Sorting Test (WCST) performance in healthy older adults, although education significantly correlated with this test as in Heaton (1981). Riccio et al. (1994), studying children with attention deficit hyperactivity disorder, ages 6 to 16 years, found that Wechsler Intelligence Scale for Children-Revised (WISC-R) performance IQ correlated with WCST variables for those between the ages of 9 years and 11 years, 11 months, but not with the older children. Welsh, Pennington, and Grossier (1991) gave normal children, ages 6 to 12 years, several measures of executive function, including WCST perseverative errors, and they found no correlation with measures of IQ. WCST/IQ relationships were not reported in several norming efforts on pediatric samples (Chelune & Baer, 1986; Heaton, Chelune, Talley, Kay, & Curtiss, 1993), although the 1993 manual does provide a normalized distribution of scores at each age level. Chelune and Baer (1986), Levin et al. (1991), and Welsh et al. (1991) reported that children performed at levels typical of the average adult by 10 years of age. Research clearly shows that executive functions improve with age, but the relationship of intelligence test scores to executive function in children, particularly the WCST, is far from settled. The purpose of the present study is to determine the role of level of intellectual performance and age on this one executive function test, the WCST, in an above average sample. Gender relationships are also explored. It was hypothesized that this test, which appears to tap higher level or conceptual problem solving and mental efficiency, would be significantly related to a measure of intelligence (i.e., WISC-III IQ). A second purpose was to document the point at which the children s performance approximated the average adult level performance. If variation occurs as a function of IQ, then it would be reasonable to hypothesize that the average adult levels are achieved earlier in brighter children.

Wisconsin Card Sorting Test Performance 715 Participants METHOD Twenty-six children (ages 9 to 14 years) with WISC-III Full-Scale IQS above 130 and 24 children (ages 9 to 14 years) with WISC-III Full-Scale IQS between 110 and 129 were given the WCST (in total, there were 25 boys and 25 girls). Participants were normal and gifted school children from several school districts in a north-central metropolitan city. Forty-eight of the children were White, and 2 were Asian Americans. Procedure The children and their parents had agreed to participate in this study, which was part of a larger effort to develop norms and validation data on a new neuropsychological instrument. Children were carefully screened for neurological dysfunction, learning disability, emotional disturbance, and attention deficits through a telephone interview prior to testing. The WISC-III (Wechsler, 1991) and WCST (Heaton et al., 1993) were administered along with other neuropsychological measures by a graduate student according to standard instructions. The WCST was scored for total number of categories, failures to maintain set, perseverative errors, nonperseverative errors, total errors, and trials to first category according to guidelines in the 1993 manual (Heaton et al., 1993). The trials to first category condition represents the total number of trials needed to complete the first category successfully. RESULTS For each of the six ages (9- to 14-year-old children), mean scores were computed for number of categories completed; number of failures to maintain set; and number of perseverative, nonperseverative, and total errors on the WCST (see Table 1). The mean for the current sample was compared to the corresponding age mean in the WCST manual (Heaton et al., 1993) or the Chelune and Baer (1986) study and the average adult mean (Heaton et al., 1993). A visual inspection of Table 1 indicates that average and superior IQ samples performed better (more categories and fewer errors and failures to maintain set) at every age than the average reported in the statistical manuals. Even the youngest age sampled (9-year-old children) performed better than the overall adult mean. In addition to an overall adult mean, means for 11 different age groupings during adulthood are reported in the 1993 manual. Further statistical analyses were computed, including eight separate t tests, contrasting the sample mean with the adult mean reported in the manual. There are 11 adult means reported, and the best adult mean in each category was used for these analyses. For perseverative errors, performance significantly better than the best adult mean was revealed in the younger and older gifted students (t 6.14, df 21, p.001; t 5.78, df 4, p.01, respectively) and in the older above average students (t 3.64, df 14, p.01) but not for the younger above average students (t 1.09, df 7, ns). For nonperseverative errors, t tests contrasting the sample mean with the best adult mean were significant for the younger and older gifted students (t 9.25, df 21, p.001; t 3.2, df 3, p.01, respectively) and for the older above average students (t 9.25, df 21, p.001) but not for the younger above average students (t 0.34, df 7, ns). Six separate stepwise multiple regression analyses were pursued to assess the relationship of intelligence and age to six WCST variables (perseverative errors, nonperse-

716 S. Arffa et al. TABLE 1 Means for 9- to 14-Year-Old Children for WCST Variables and Comparison With Published Norms Age (Years) n Total No. of Categories Chelune & Baer (1986) No. of Perseverative 1993 WCST (50th Percentile) Corresponding % in WCST 9 14 6 4 7.1 17 90% 10 8 6 5.6 7.3 16 90% 11 10 6 5.58 7.1 15 90% 12 10 6 5.70 10 14 73% 13 4 6 6.3 14 93% 14 4 6 6.7 13 90% Adult level 5.4 11 Nonperseverative Total Age (Years) n Nonperseverative 1993 WCST (50th Percentile) Corresponding % in WCST Total 1993 WCST (50th Percentile) Corresponding % in WCST 9 14 5 15 95% 12.1 34 95% 10 8 6.4 15 90% 14 33 91% 11 10 5.7 14 90% 13.1 31 92% 12 10 4 14 94% 14 30 88% 13 4 3 13 93% 9.3 29 96% 14 3 6.3 13 82% 13 28 87% Adult level 11 24 Age (Years) n Failure to Maintain Set Chelune & Baer (1986) 9 14 0.3 1.75 10 8 0.2 1 11 10 0.1 1.17 12 10 0.5 0.7 13 4 0.25 14 3 0.3 Adult level 0.8 Note. Chelune and Baer (1986) data are reported on variables not individually recorded at each age in the 1993 WCST manual. This information was not available for trials to first category. WCST Wisconsin Card Sorting Test. verative errors, total errors, number of categories completed, failure to maintain set, and trials to first category). Raw scores of each of the six WCST variables were used, first using age then the WISC-III raw scores, in the stepwise formula. Age and WISC-III raw scores were entered as continuous variables. In a preliminary analysis, gender was not significantly related to WCST variables and was removed from further regression formulas. Because computation of separate multiple analyses raises the possibility of achieving significance by chance occurrence, a Bonferroni alpha adjustment was used as a more stringent test of F level. The stepwise regression for perseverative errors revealed a significant R 2 change, F(1, 41) 8.68, p.01. Intelligence accounted for 17% of the variance in WCST perseverative errors. The stepwise regression for nonperseverative errors revealed a significant R 2 change, F(1, 52) 10.29, p.01. Intelligence ac-

Wisconsin Card Sorting Test Performance 717 TABLE 2 Mean WCST Perserverative, Nonperseverative, and Total and Trials to First Category for Two Age Groupings and Two IQ Groupings Perseverative Nonperseverative Total Trials to First Category Younger a IQ1 M 122.8, n 8 10.25 10.9 20.7 13.6 IQ2 M 138.3, n 22 5.7 5.6 11.2 10.7 Older b IQ1 M 122.4, n 15 6.6 5.7 11.9 12 IQ2 M 138, n 5 6.8 6.7 14 11.2 Note. WCST 5 Wisconsin Card Sorting Test. a 9 years, 0 months to 11 years, 6 months. b 11 years, 7 months to 14 years, 1 month. counted for 17% of the variance in WCST nonperseverative errors. The stepwise regression for total errors revealed a significant R 2 change, F(1, 51) 7.9, p.005. Intelligence accounted for 18% of the variance in WCST total errors. The stepwise regression for trials to first category revealed a significant R 2 change, F(2, 51) 2.38, p.05. Intelligence accounted for 9% of the variance in WCST trials to first category. The stepwise regressions for failure to maintain set and total number of categories were nonsignificant. In Table 2, a breakdown of mean perseverative, nonperseverative, and total errors and trials to first category by two age groups (9 years, 0 months to 11 years, 6 months vs. 11 years, 7 months to 14 years, 11 months) and two IQ categories (Full-Scale IQ between 110 and 129 vs. Full-Scale IQ at 130 and above) is reported for perseverative errors. This particular grouping, created for the purpose of performing an analysis of variance, was developed because mean IQ scores were equivalent across groups and clustered into gifted and above average/superior ranges. As sample sizes were nonequivalent, only perseverative errors were chosen for this additional evaluation, and results should be interpreted cautiously. Intergroup differences in perseverative errors were assessed with an analysis of variance based on age, IQ, and gender (Table 3). A significant main effect for IQ (p.025) and a significant Age IQ interaction (p.025) were found. Participants with very high IQ ( 130) made fewer mean perseverative errors than participants with above average to superior IQ S (IQ S between 110 and 129). Post hoc analyses revealed TABLE 3 Analysis of Variance Source Table Comparing Age, IQ, and Gender Variables in WCST Perseverative Sum of Squares df ms F p Age 1782.2 1 0 0 ns Gender 1812.6 1 30.4 3.46 ns IQ 1833.1 1 50.57 5.76.025 Age Gender 0.3 1 0.27 0.03 ns Age IQ 2426.8 1 50.57 5.76.025 Gender IQ 7.2 1 7.23 0.82 ns Age IQ Gender 500.1 35.4 4.03 ns Note. WCST Wisconsin Cart Sorting Test.

718 S. Arffa et al. that differences in these IQ ranges are pronounced only in the younger age range. No gender differences were found. DISCUSSION WCST performance clearly improved with increasing intellectual functioning in this above average sample. Multiple regressions were significant for WCST perseverative errors, nonperseverative errors, total errors, and trials to first category and intelligence test scores. An analysis of variance for perseverative errors also indicated that children with very superior IQ S ( 130) significantly outperformed children with above average and superior IQ S (110 to 129). These findings are consistent with Klonoff and Low (1974) and Seidenberg et al. (1983), who found that IQ was significantly related to certain pediatric neuropsychological measures. Consistent with Parsons (1984), this study documents better performance in the upper ranges of intellectual ability. The current results are also consistent with previous findings relating IQ and executive function measures (Parsons, 1984). More specifically, Riccio et al. (1994) identified a relationship between IQ and the WCST in 9- to 11-year-old children diagnosed with attention deficit hyperactivity disorder. In this study, a similar relationship is recognized in normal bright children, with results most prominently noted in children between the ages of 9 years and 11 years, 6 months. However, the current findings are in contrast to Welsh et al. (1991), who found no relationship to IQ on several executive function tests, including WCST perseverative errors. Without knowledge of the IQ ranges sampled or IQ measures employed, it is difficult to account for the variance of the Welsh study with the present findings. Although one cannot easily extrapolate from pediatric to adult samples, the current results are in contrast with studies suggesting that IQ is unrelated to WCST performance in healthy adult samples (Boone et al., 1993; Heaton, 1981). However, education (which is correlated with IQ in adults) is significantly correlated with WCST performance. One possibility is that insufficient sampling of IQ ranges depressed an IQ-WCST relationship. In the current sample, a range of 110 to 155 may have offered the sufficient variance to test this relationship. Another possibility for the discrepancy with adult research is that the tests measure different things for children. That is, the WCST may tap higher level conceptual function in younger children, explaining its relationship to IQ measures during childhood but not during adulthood. An alternate explanation is that intelligence tests measure more basic neurocognitive functions such as mental efficiency and new learning to greater degrees in children than in adults. A significant Age IQ interaction in the analysis of variance on WCST perseverative errors further refines the pattern of IQ-WCST relationships. Gifted youngsters outperformed above average to superior children only in the younger age group. Older gifted children did no better than the younger gifted children, whereas the above average group showed better performance in the older group. This is best interpreted as a ceiling effect on the WCST. Participants earned only 6 to 10 perseverative errors (and similar levels of other errors), and some degree of error is necessary to shift categories on the WCST. In addition, it is possible that there is less improvement after 9 to 10 years of age, which is consistent with previous research on age trends (Chelune & Baer, 1986; Levin et al., 1991). Of note, there was a limited number of gifted children in the older age range, thereby reducing the reliability of this particular finding. In any case, these results indicate a need to evaluate age trends as a function of intellectual variability.

Wisconsin Card Sorting Test Performance 719 It was hypothesized that average adult levels of performance would be achieved earlier in brighter children. Visual inspection of the means suggested that even the youngest age group (i.e., 9-year-old children) performed better than the average adult on every measure. However, when the sample was compared across two IQ ranges (gifted and above average) and two age ranges (ages 9 to 11 years, 6 months and ages 11 years, 7 months to 14), as shown in Table 2, the level of average adult performance was closely approximated in the younger above average group. Gifted youngsters appeared to exceed adult levels even at younger ages. Gifted youngsters also appeared to exceed the best (least impaired) adult mean on every measure. This indicates that adult level performance is achieved by 9 to 11 years of age in children with above average to superior IQ S. The data further suggest that this level might be achieved even earlier in children with very superior IQ S. Further research to define the relationship of intelligence to executive functions in normal children is needed. Because executive functions include numerous functions ranging from simpler response inhibition to more complex problem solving, it is important to monitor developmental trends and assess the correlation with intelligence for each of these. Executive dysfunction is often found with damage to the prefrontal regions in both adults and children (Heaton et al., 1993). The relationship of executive functions to intelligence in brain-injured populations with varying rates of severity needs to be specifically evaluated. Significant IQ decrements do not appear to occur in many cases of frontal lobe injury in adults (Warrington, James, & Maciejewski, 1986) and probably in children (Grattan & Eslinger, 1991), but the relationship to age at injury and injury severity may be more complex in the latter group. The possibility that the intelligence test neuropsychological test relationship is stronger during childhood is a question for further research. Clinically, it becomes very difficult to evaluate individuals who have a very high level of premorbid functioning and who have had mild injuries (Beers, Morrow, Ryan, & Morgan, 1994). The present results signal a need to consider premorbid intellectual functioning when evaluating executive function deficits in children with central nervous system dysfunction (assuming that IQ neuropsychological test correlations in normals are equivalent in brain-injured populations). The 1993 WCST manual contains normalized scores at every age range and should be used when children are estimated to have above average premorbid intellectual function. REFERENCES Beers, S. R., Morrow, L. A., Ryan, C. R., & Morgan, M. C. (1994). Methodological issues evaluation of high functioning medically ill patients: A case study [Abstract]. Archives of Clinical Neuropsychology, 9(2), 109. Boone, K. B., Ghafferian, S., Lesser, I. M., Hill-Gutierrez, E., & Berman, N. G. (1993). Wisconsin Card Sorting Test performance in healthy, older adults: Relationship to age, sex, education, and IQ. Journal of Clinical Psychology, 49(1), 54 60. Chelune, G. J., & Baer, R. A. (1986). Developmental norms for the Wisconsin Card Sorting Test. Journal of Clinical and Experimental Neuropsychology, 8(3), 219 228. Grattan, L. M., & Eslinger, P. J. (1991). Frontal lobe damage in children and adults: A comparative review. Developmental Neuropsychology, 7(3), 283 325. Heaton, R. K. (1981). Wisconsin Card Sorting Test manual. Odessa, FL: Psychological Assessment Resources. Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G., & Curtiss, G. (1993). Wisconsin Card Sorting Test manual: Revised and expanded. Odessa, FL: Psychological Assessment Resources.

720 S. Arffa et al. Klonoff, H., & Low, M. (1974). Disordered brain function in young children and early adolescents: Neuropsychological and electroencephalographic correlates. In R. M. Reitan & L. A. Davison (Eds.), Clinical neuropsychology: Current status and applications. Washington, DC: Winston. Leikliter, I. N., & Matarazzo, J. D. (1989). The influence of age, education, IQ, gender, and alcohol abuse on Halstead Reitan Neuropsychological Test battery performance. Journal of Clinical Psychology, 45(4), 484 512. Levin, H. S., Culhane, K. A., Hartmann, J., Evanovich, K., Mattson, A. J., Harward, H., Ringholz, G., Ewing- Cobbs, L., & Fletcher, J. M. (1991). Developmental changes in performance on tests of purported frontal lobe functioning. Developmental Neuropsychology, 7(3), 377 395. Parsons, E. J. C. (1984). An investigation of appropriate neuropsychological assessment procedures with adults in the higher intelligence ranges. Dissertation Abstracts International, 45(5), 1592-B. Reitan, R. M. (1959). The comparative effects of brain damage on the Halstead Impairment Index and Wechsler Bellevue Scale. Journal of Clinical Psychology, 15, 281 285. Reitan, R. M. (1985). Relationships between measures of brain function and general intelligence. Journal of Clinical Psychology, 41, 243 253. Riccio, C. A., Hall, J., Morgan, A., Hynd, G. W., Gonzalez, J. J., & Marshall, R. M. (1994). Executive function and the Wisconsin Card Sorting Test: Relationship with behavioral ratings and cognitive ability. Developmental Neuropsychology, 10(3), 215 229. Seidenberg, M., Giordani, B., Berent, S., & Boll, T. (1983). IQ level and performance of the Halstead Reitan Neuropsychological Test battery for older children. Journal of Consulting and Clinical Psychology, 51(3), 406 413. Waldemann, B. W., Dickson, A. L., Monahan, M. C., & Kazelskis, R. (1992). The relationship between intellectual ability and adult performance on the trail making test and the symbol digit modalities test. Journal of Clinical Psychology, 48(3), 360 365. Warrington, E. K., James, M., & Maciejewski, C. (1986). The WAIS as a lateralizing and localizing diagnostic instrument: A study of 656 patients with unilateral cerebral lesions. Neuropsychologia, 24, 223 236. Wechsler, D. (1991). Wechsler Intelligence Scale for Children-III. San Antonio, TX: Psychological Corp. Welsh, M. C., Pennington, B. F., & Grossier, D. B. (1991). A normative-developmental study of executive function: A window on prefrontal function in children. Developmental Neuropsychology, 7(2), 131 149. Wiens, A. N., & Matazzaro, J. D. (1977). WAIS and MMPI correlates of the Halstead Reitan neuropsychological battery in normal male subjects. Journal of Nervous and Mental Diseases, 164, 112 123.