Intra-individual Subtest Variability on the Dutch Wechsler Intelligence Scales for Children-- Epilepsy. Loretta van Iterson

Transcription

1 Published: van Iterson, L & A.S Kaufman (2009) Intra-individual Subtest Variability on the Dutch Wechsler Intelligence Scales for Children--Revised (WISC-R NL ) for children with Learning Disabilities, Psychiatric Disorders, and Epilepsy. Psychological Reports, 2009, 105, Running head: SUBTEST VARIABILITY ON THE WISC-R NL Intra-individual Subtest Variability on the Dutch Wechsler Intelligence Scales for Children-- Revised (WISC-R NL ) for children with Learning Disabilities, Psychiatric Disorders, and Epilepsy Loretta van Iterson SEIN Stichting Epilepsie Instellingen Nederland & School De Waterlelie, Cruquius Alan S. Kaufman Yale University School of Medicine Author note Please address correspondence to Loretta van Iterson, SEIN, Stichting Epilepsie Instellingen Nederland. Afdeling psychologie. Postbus AA Heemstede, The Netherlands. lviterson@sein.nl

2 Abstract It is common practice to look at disparities among subtest scores ( scatter ) on an intelligence test to establish if a score is deviant. However, it remains unclear whether subtest scatter reflects primarily normal variation within individuals or is clinically meaningful. The present study explored this issue based on data from 467 children with developmental disabilities tested on the Dutch WISC-R NL. Of these children, 132 had learning disabilities, 178 had psychiatric disorders, and 157 had epilepsy. Subtest scatter was defined as scaled-score range (highest minus lowest scaled score). When contrasted with normal scatter, the overall sample revealed higher ranges on the Performance Scale and Full Scale, although effect sizes were small. Analysis of the data for the three separate clinical samples revealed unusual scatter only for the sample of children with psychiatric disorders. When comparing the clinical samples, scaled-score range was larger for the sample of children with psychiatric disorders than for those with epilepsy. Two distinct subsamples revealed elevated ranges with moderate effect sizes: children with autistic spectrum disorders and children with left hemisphere seizures. These results suggest that elevated subtest scaled-score range might characterize specific clinical samples rather than denoting an overall sign of pathology. (199 words) Key-words: Wechsler Intelligence Scale for Children, subtest scaled-score range, scatter, intra-individual variability, inter-subtest variability, learning disabilities; psychiatric disorders; childhood epilepsy.

3 In child neuropsychology, the clinician frequently looks for strengths and weaknesses in the cognitive profile, often operationalized as a positive or negative difference of 1 or 2 standard deviations, in order to make a diagnosis of a developmental disorder (Sattler, 2001). This approach is based on the assumption that a child s profile should be uniform, and that undue inter-subtest or intra-individual variability (scatter) can be interpreted as pointing toward a specific strength or deficit. Two common indexes of scatter are subtest scaled-score range (the simple difference between the highest and the lowest score in a profile) and univariate scatter (the number of subtests deviating 1 SD from an individual s own mean). Kaufman (1976, 1979) showed that large intra-individual variability on these indexes, far from being unusual, was seen frequently in the standardization sample of the WISC-R. Later, Silverstein (1987) demonstrated that the empirically-derived moments (mean and SD) from Kaufman's data were a function of the psychometric qualities of the test and could be estimated from the average intercorrelations among the subtests comprising the scales. Both subtest scaled-score range and univariate scatter make use only of the extreme values in a profile. As a more sensitive measure of intra-individual variability, the Profile Variability Index was proposed which, like a standard deviation, uses information derived from all subtests (Matarazzo, Daniel, Prifitera, & Herman, 1988; McLean, Reynolds, & Kaufman, 1990). Interestingly, subtest scaled-score range was shown to correlate highly with Profile Variability Index (Matarazzo, Daniel, Prifitera, & Herman, 1988; Boone, 1993). The question of whether elevated intra-individual variability is a sign of pathology, or only a reflection of the psychometric properties of the test, remains unsettled. Some researchers have provided evidence for elevated variability in pathology (Zimet, Goodman Zimet, Farley, Shapiro Adler, & Zimmerman, 1994; Mayes, Calhoun, & Crowell, 1998; Greenway & Milne, 1999; Ryan, Tree, Morris, & Gontkovsky, 2006), while others ardently advocate against any use of measures based on inter-subtest variability (Watkins & Glutting,

4 2000; Watkins, Glutting, & Youngstrom, 2005). In the studies by Watkins and his colleagues, inter-subtest variability did not have a significant incremental validity in predicting academic achievement over and above Full-Scale IQ in samples of exceptional children, mainly children with learning disabilities. The authors argue that inter-subtest variability is of no use as a diagnostic indicator, and its use can be considered prescientific (Watkins, Glutting, & Youngstrom, 2005, p.263). In spite of this controversy, recent Wechsler test manuals have incorporated subtest scaled-score range in the form of base rate tables (e.g. for the WISC-III, Wechsler, 1992, and the WISC-IV, Wechsler, 2004b). For the Dutch, the adult test version includes subtest scaledscore range (Wechsler, 2004a), while the children s versions do not (van Haasen, de Bruyn, Pijl, Poortinga, lutje Spelberg, Vandersteene, et al., 1986; Wechsler, 2005). In order to be a sign of pathology, the intra-individual variability should be significantly different in clinical samples when compared to the standardization sample. Significant scatter should not only be interpreted as reliable scatter that is, genuine and not the effect of measurement error but also as uncommon, in the sense that the magnitude of occurrence within the normal population is small, e.g., 5% (Crawford & Allan, 1996). Because the Wechsler scales are well-standardized and well-researched, they are used, in the present paper, as a model to further evaluate whether intra-individual variability is a clinically meaningful measure of pathology. This study focused on subtest scaled-score range because it is practical and easily computed by clinicians and is included in most recent test manuals. Furthermore, it correlates highly with Profile Variability Index. Univariate scatter, on the other hand, was not included because its distribution is skewed, preventing parametrical analyses. The data refer to the Dutch WISC-R (for this purpose: WISC-R NL ; van Haasen et al., 1986) and were collected up to This version was in use in the Netherlands for a prolonged time, thus allowing the collection of larger samples. For the newer test

5 version, it will take some time before sufficiently large samples are accrued, but underlying notions about subtest scatter can be understood independent of test version. The expected mean values of subtest scaled-score range and the cut-off values for uncommonly large ranges for the WISC-R NL were estimated according to Silverstein (1987, 1989), aided by Owen s (1962) range statistics. These estimates draw on the averaged intercorrelations between the subtests, which came from the technical manual of the WISC-R NL (de Bruyn, Vandersteene, & van Haasen, 1986, p 139; from N = 1961 children). Based on WISC-R NL data on 467 children from three clinical samples, the aims of this study were (1) to study whether subtest scaled-score range in children with developmental disabilities shows differences compared to expected (normal) values; (2) to study whether there are differences among the clinical samples, and, if so, (3) to explore whether specific subsamples account for these differences; and (4), to report rates of individuals with uncommonly large subtest scaled-score ranges found in clinical samples. Methods Participants were N = 467 children, aged six to 16 years, with FS-IQs > 75. Overall, 353 (76%) were male. The children were entitled to benefit from distinct special school services in The Netherlands, according to national regulations (e.g., Resing, Evers, Koomen, Pameijer, Bleichrodt, & van Boxtel, 2002). These regulations describe the criteria for placement in different settings specialized in, respectively, (specific) learning disabilities, childhood psychiatric disorders, or childhood epilepsy. Generally, information from four sources is weighted by an independent committee. These sources are the family of the child, the present school, a psychologist who did the assessment (including the intelligence testing), and an educational, psychiatric or medical specialist. For learning disabilities, specified criteria in terms of academic failure must be met; for psychiatric disorders, a DSM-IV diagnosis from a psychiatrist or qualified mental health psychologist is required; for epilepsy, a

6 diagnosis from a neurologist is required. In all cases, the difficulties caused by the diagnosed condition must exceed the competencies of the regular school. Normal intellectual abilities were a further criterion for the schools of the first two samples, but not the third (epilepsy). As indicated, in this study FS-IQ was set to be above 75 for all. Co-morbidity is a common phenomenon in childhood developmental disabilities and the samples are diagnostically heterogeneous; the primary diagnosis as reflected by special school placement was the criterion for inclusion in a sample. Diagnostic group membership type of special school was the independent variable in this study. Demographic data are presented in Table 1 and data on the Wechsler scales are shown in Table 2. The first sample consisted of N = 132 children with (specific) learning disabilities and the second sample consisted of N= 178 children with psychiatric disorders. The latter group included children with neurodevelopmental disorders as well as children with behavioral and emotional disorders related to major life events (e.g. traumas). The main diagnoses of this sample were autism spectrum disorders (ASD), conduct disorders or oppositional defiant disorders, reactive attachment disorders, attention deficit and hyperactivity disorders, tic disorders, and depression. The subsamples are listed in Table 1. The percentages add up to over 100% due to psychiatric co-morbidity. The third sample consisted of N = 157 children with seizure disorders. Mean age at epilepsy onset was 5.6 years (ranging from the first day of life to age 15 years with SD = 3.2). Mean duration of epilepsy was 4.0 years (SD = 3.2). Seizure type classification, side of epilepsy onset, and information on medication and neuroimaging are presented in Table 1. For each participant, subtest scaled-score range was calculated for five Verbal, five Performance, and ten Full-Scale subtests. Verbal, Performance, and Full-Scale subtest scaledscore range was the dependent variable in the study. Mean scores are given in Table 3; z- converted means are depicted in Figure 1. Levene s testing for homogeneity of variances

7 showed no significant values for ANOVA or ANCOVA. ANOVA revealed differences in mean age (age was higher in learning disabilities compared to epilepsy) and mean P-IQ (but not V-IQ or FS-IQ), indicating higher P-IQ in the samples of children with learning disabilities and psychiatric disorders compared to the sample of children with epilepsy. Also, chi-square showed that boys and girls were unevenly distributed among the samples; significant differences were found, indicating that the rate of boys was higher in the sample with psychiatric disorders than the sample with epilepsy. Three separate ANCOVA s were undertaken (for Verbal, Performance and Full-Scale subtest scaled-score range), controlling for the pre-existing differences in P-IQ, age, and sex. With multiple, one-sided, one-sample t- tests, the observed clinical values were compared to the estimated expected values (Silverstein, 1987), and effect sizes were calculated accordingly (Cohen, 1988, p.45). Overall, alpha was set at.05 and Bonferroni corrections were used to control for family-wise errors. With chi-square, rates of children with uncommonly high subtest scaled-score range (Verbal Scale: 8 points; Performance Scale 10, and Full-Scale 11), expected in ~5% of the normal population (Silverstein, 1989), were compared to this value. As this value was seen as an approximation only, alpha was set to.001. Rates of uncommonly high subtest scaled-score range were also compared between the clinical samples. Results Verbal Scale. Comparison of means. Table 3 presents the expected mean subtest scaled-score range for the Verbal Scale (mean = 4.7, SD = 1.7) and the observed valued for the distinct samples, together with the results of the one sample t-tests, and Figure 1 depicts the z-converted values of subtest scaled-score range. No differences were found between the mean expected values and the observed values of the total sample or any of the distinct clinical samples (Table 3).

8 No differences were found between the means of the clinical samples (ANCOVA: F(462,2) = 0.138, p =.871, n.s). Rates of uncommonly large ranges. Large ranges ( 8 points) were found, respectively, in 8.3%, 12.9% and 15.3% of the children with learning disabilities, psychiatric disorders, and epilepsy. Compared to the expected rates, these values reached significance for psychiatric disorders (Χ 2 = 23.51, p <.001) and epilepsy (Χ 2 = 34.97, p <.001). Chi-square revealed no difference in the distributions of uncommonly large ranges between the clinical samples. There was an almost twofold rate (likelihood ratio 1.8, 95% Confidence Interval [CI]: ) of children with epilepsy versus children with learning disabilities. Performance Scale. Comparison of means. The estimated expected mean subtest scaled-score range was 5.8 (SD = 2.4). The total sample and the children with psychiatric disorders differed from the expected value (Table 3). Significant differences were also suggested among the clinical samples (ANCOVA: F(461,2) = 3.024, p =.050, partial η 2 =.01). However, pair-wise comparisons between the clinical samples did not yield significant results. Rates of uncommonly large ranges. Large ranges ( 10 points) were found in 7.6%, 12.9%, and 7.0% of the children with, respectively, learning disabilities, psychiatric disorders, and epilepsy. Compared to expected values, these percentages were elevated for psychiatric disorders only (Χ 2 = 23.51, p <.001). Chi-square revealed no significantly different rates among the samples. Likelihood ratios were 1.7 (95% CI: ) and 1.8 (95% CI: ) for children with psychiatric disorders versus, respectively, children with learning disabilities and children with epilepsy. Full Scale. Comparison of means. The expected mean subtest scaled score range was 7.3 (SD = 2.1). Values differing significantly from expected were found for the total sample and the

9 sample with psychiatric disorders (Table 3). Significant differences were found between clinical samples (ANCOVA: F(462,2) = 4.130; p =.017, partial η 2 =.02); specifically, subtest scaled-score range was higher in children with psychiatric disorders than in children with epilepsy. Rates of uncommonly large ranges. Larges ranges ( 11 points) were found, respectively, in 6.8%, 12.9%, and 10.8 % of the children with learning disabilities, psychiatric disorders and epilepsy. Compared to expected values, significant differences were found for psychiatric disorders (Χ 2 = 23.51, p <.001) and epilepsy (Χ 2 = 11.23, p =.001). Again, differences of the large ranges between samples showed a trend that did not reach statistical significance. Notably, however, there was almost a twofold rate (likelihood ratio 1.9, 95% CI ) for children with psychiatric disorders compared to children with learning disabilities. Subsamples. Although it is beyond the scope of this paper to enter into detail on all subsamples, two subsamples were identified as showing conspicuously elevated subtest scaled-score ranges relative to expected values: (a) from the sample with psychiatric disorders, the subsample with autistic spectrum disorders (ASD, N = 58) was identified; and (b) from the epilepsy sample, children with left focal onset seizures (LH; N = 33) were identified. Data on these subsamples are also included in Table 2, Table 3, and Figure 1. Table 3 shows that the ASD sample had a significantly larger subtest scaled-score range than the expected values on the Verbal Scale, Performance Scale and Full Scale, all with moderate effect sizes. When the ASD sample was compared to the other children with diagnoses of psychiatric disorders (N = 120) in the psychiatric sample, mean subtest scaledscore range was elevated for the ASD sample on the Verbal Scale (t(94.7,1) = 2.49, p =.014, ES = 0.5) and the Full Scale (t(176,1) = 2.44, p =.016, ES = 0.4). Uncommonly large ranges

10 were found for the Verbal Scale in 20.7% of these children, for the Performance Scale in 15.5%, and for the Full Scale in 17.2%. The percentages were significantly elevated when compared to expected values for the Verbal Scale (Χ 2 = 30.06, p <.001), Performance Scale (Χ 2 = 13.51, p <.001), and Full Scale (Χ 2 = 18.3, p <.001). Classificatory statistics revealed that when the ASD group was contrasted to the others children with psychiatric disorders, uncommonly large ranges had classificatory utility for the Verbal Scale: sensitivity was 21%, specificity was 91%, Positive Predictive Power (PPP) was 52%, Negative Predictive Power (NPP) was 70%, and likelihood ratio was 2.26 (95% CI ). These values indicated that when a child with psychiatric disorders is found to have a subtest scaled-score range of 8 or more points on the Verbal Scale, it will more likely belong to the group with autistic spectrum disorders. Within the sample of children with psychiatric disorders, none of the other subsamples showed elevated subtest scaled-score range consistently on all scales. However, two subsamples of children with neurocognitive developmental disorders showed elevations on one scale specifically, the subsample with conduct disorders had substantial scatter on the Performance Scale and the subsample with tic disorders had elevated scatter on the Verbal Scale. These data merely suggest hypotheses for future study, but are not presented here because many children had multiple diagnoses and the sample sizes were too small to permit meaningful analyses. Table 3 shows that mean scaled-score range was higher than expected in the sample of children with left hemisphere seizures. Significantly elevated values were seen for the Verbal Scale (small effect size) and the Full Scale (moderate effect size). Such elevations were not seen in the other epilepsy subsamples; planned comparisons indicated that values were significantly different compared to the right hemisphere seizure-group for the Verbal Scale (t(137,1) = 2.05, p =.042, ES = 0.3). Large range was found for the Verbal Scale in 21.2% of

11 these LH children, for the Performance Scale in 6.1%, and for the Full Scale in 15.2%. The percentages were significant when compared to expected values for the Verbal Scale only (Χ 2 = 18.26, p <.001). Classificatory statistics revealed that when the left hemisphere seizuregroup was contrasted to the right hemisphere seizure-group, there was a clear trend to find more children with uncommonly large ranges in the Verbal Scale in the left hemisphere seizure-group. The valued failed to reach significance due to lack of statistical power: sensitivity was 21%, specificity was 96%, PPP was 88%, NPP was 47%, and likelihood ratio was 6.19 (95% CI ). Conclusions and Discussion The assertion that intra-individual variability is elevated in pathology, taken for granted by some researchers and opposed by others, was the subject of analysis in this study, which focused on subtest scaled-score range for clinical samples of children with learning disabilities, psychiatric disorders, and epilepsy. Analyses were conducted at three levels. At the broadest level, 467 children from three categories of developmental disabilities (learning disabilities, psychiatric disorders, and epilepsy) were compared to the expected ( normal ) values. Significant elevations were found in the Performance and Full Scales with conspicuously small effect sizes. This finding suggested that the study was profiting from the effects of a relatively large sample size and also that possible meaningful information was being masked by focusing on the heterogeneous total group. At the second level of analysis, each of the clinical samples was compared to the expected values and to each other. The sample with psychiatric disorders showed significantly more than normal intra-individual variability on both the Performance and Full Scales. Also, the sample with psychiatric disorders showed more variability than the sample with epilepsy on the Full Scale. Effect sizes were larger than for the total clinical sample, but were still small. At the third and

12 most specific level of analysis, two homogeneous subsamples were subjected to further scrutiny. It appeared that the sample with ASD (within psychiatric disorders) and the sample with focal LH seizures (within epilepsy) showed elevated scatter, compared both to the expected values and to the other children in their respective clinical original samples. For ASD, this was true for all three scales (moderate effect sizes). For left hemisphere epilepsy, this was true for the Verbal Scale (small effect size) and the Full Scale (moderate effect size). When evaluating the percents of children with uncommonly large ranges, children with (specific) learning disabilities did not display any unusual elevations relative to groups of normal children. However, rates were increased in children with psychiatric disorders on all scales, specifically for the ADS subsample. Rates were also increased for the sample of children with epilepsy on the Verbal and Full Scales, more clearly so in the subsample of LH seizures. The fact that the Performance Scale and not the Verbal Scale yielded the significant differences in the primary samples of this study is consistent with a diverse body of neuropsychological literature that has shown Wechsler s Performance subtests to be more sensitive to brain injury and brain dysfunction than Verbal subtests (Kaufman & Lichtenberger, 2006, chapters 8 and 9). Nonetheless, the present study suggests that elevated subtest scaled-score range can also be seen on the Verbal Scale in specific samples. For children with (specific) learning disabilities, no elevations were found on any measure. These results are in line of earlier studies (Watkins, Glutting, & Youngstrom, 2005; Flanagan & Kaufman, 2009). For children with ASD, elevated intra-individual variability has been reported earlier (Joseph, Tager-Flusberg, & Lord, 2002). For children with epilepsy, to the authors knowledge, no such studies have been reported, but the results are in line with the large V-IQ > P-IQ discrepancies reported for children with unilateral focal onset epilepsy (van Iterson & Augustijn, 2006) regardless of side of seizure onset. The elevations in subtest

13 scaled-score range on the Verbal Scale in left hemisphere epilepsy may be the result of plasticity in the developing brain (Vicari, Albertoni, Chilosi, Cipiriani, Coni, & Bates, 2000). The results found for children with ASD and children with epilepsy are interesting in the light of recent research on the commonalities underlying both conditions and the findings of high rates of subclinical EEG abnormalities in children with ASDs even in the absence of manifest clinical seizures (Spence & Schneider, 2009). Effect sizes increased when the selected samples were more homogeneous, suggesting that specific samples of children with developmental disabilities may show elevated intraindividual variability while others may not, or may even show decreased variability. Thus, studies of subtest scaled-score range and their interpretation should take into account type of pathology. Scaled-score range was not found suitable for classification purposes between the large samples; 95% confidence intervals for likelihood ratios were non-significant, though some trends could be found. This is not surprising as scatter is a non-specific measure which does not provide an answer as to where the variability is coming from, or if it follows some specific pattern. Classificatory statistics applied on selected samples indicated that an uncommonly large range in the Verbal Scale was more likely to belong to a child with ASD and not a child with another diagnosis within psychiatric disorders. Also the data suggest that uncommonly large variability on the Verbal Scale may be characteristic of children with left, but not right, hemisphere onset seizures; likelihood ratio was not significant, probably due to small sample sizes. In line with the results of this study, and pertinent to the discussion of the interpretability of scores beyond the summed scores of a scale (Watkins, Glutting, & Youngstrom, 2005; Flanagan & Kaufman, 2009), it is worthy noting Saling s (2009)

14 perspective. Based on the results of research within a highly specific area of research in neuropathology epilepsy surgery Saling (2009), advocates against the use of scales of summed scores in neuropsychological assessment of memory functions and argues in favor of task specific measurement. Intra-individual variability was studied with the WISC-R NL -version which has now been replaced by the WISC-III NL, and by the WISC-IV US/UK in English speaking countries. The study of intra-individual variability is not confined to a specific version of the Wechsler scales, but is applicable to any of Wechsler's scales and, in principle, to subtest profiles yielded by different batteries as well. Flanagan and Kaufman (2009) discuss the issue of intersubtest variability within WISC-IV Factor Indexes. The appreciation of a true difference between subtest scores depends on knowledge of the relationship among the measures (i.e., the intercorrelations of the subtests) as well as the frequency of occurrence of differences in the population studied.

15 Boone, D. E. (1993) WAIS-R scatter with psychiatric inpatients: II. Intersubtest scatter. Psychological Reports, 73, Cohen, J. (1988) Statistical Power Analyses for the Behavioral Sciences (2 nd ed.). Mahwah: Lawrence Erlbaum. Crawford, J. R., & Allan, K. M. (1996) WAIS-R subtest scatter: base-rate data from a healthy UK sample. British Journal of Clinical Psychology, 35, de Bruyn, E. E. J., Vandersteene, G., & van Haasen, P. P. (1986) WISC-R. Wechsler Intelligence Scale for Children-Revised. Nederlandstalige uitgave. Verantwoording. Lisse, The Netherlands: Swets Test Services. Flanagan, D. P., & Kaufman, A. S. (2009) Essentials of WISC-IV assessment. (2 nd ed.) Hoboken, NJ: Wiley. Greenway, P., & Milne, L. (1999) Relationship between psychopathology, learning or both and WISC-III subtest scatter in adolescents. Psychology in the Schools, 36, Joseph, R. M., Tager-Flusberg, H., & Lord, C. (2002) Cognitive profiles and socialcommunicative functioning in children with autism spectrum disorder. Journal of Child Psychology and Psychiatry and Allied Disciplines, 43, Kaufman, A. S. (1976) A new approach to the interpretation of test scatter on the WISC-R. Journal of Learning Disabilities, 9, Kaufman, A. S. (1979) Intelligent testing with the WISC-R. New York: Wiley. Kaufman, A. S., & Lichtenberger, E. O. (2006) Assessing adolescent and adult intelligence (3 rd ed.). New York: Wiley. Matarazzo, J. D., Daniel, M. H., Prifitera, A., & Herman, D. O. (1988) Inter-subtest scatter in the WAIS-R standardization sample. Journal of Clinical Psychology, 44, Mayes, S. D., Calhoun, S. L., & Crowell, E. (1998) WISC-III profiles for children with and without learning disabilities. Psychology in the Schools, 35,

16 McLean, J. E., Reynolds, C. R., & Kaufman, A. S. (1990) WAIS-R subtest scatter using the Profile Variability Index. Psychological Assessessment, 2, Owen, D. B. (1962) Handbook of Statistical Tables. Palo Alto, CA: Addison-Wesley Resing, W. C. M., Evers, A., Koomen, H. M. Y., Pameijer, N. K., Bleichrodt, N., & van Boxtel, H. (2002) Indicatiestelling: condities en instrumentarium in het kader van de leerlinggebonden financiering. Amsterdam: Boom. Ryan, J. J., Tree, H. A., Morris, J., & Gontkovsky, S. T. (2006) Wechsler Adult Intelligence Scale-III inter-subtest scatter: a comparison of brain-damaged patients and normal controls. Journal of Clinical Psychology, 62, Saling, M. M. (2009) Verbal memory in mesial temporal lobe epilepsy: beyond material specificity. Brain, 132, Sattler, J. M. (2001) Assessment of Children. Cognitive Applications (4 th ed.) San Diego: author. Silverstein, A. B. (1987) Two indices of subtest scatter on Wechsler's Intelligence Scales: estimated vs. empirical values. Journal of Clinical Psychology, 43, Silverstein, A. B. (1989) Reliability and abnormality of scaled-score ranges. Journal of Clinical Psychology, 45, Spence, S. J., & Schneider, M. T. (2009) The role of epilepsy and epileptiform EEGs in Autism Spectrum Disorders. Pediatric Research, in press. van Haasen, P., de Bruyn, E., Pijl, Y., Poortinga, Y. H., lutje Spelberg, H. C., Vandersteene, G., Coestier, P., Spoelders-Claes, R. Stinissen, J. (1986) WISC-R. Wechsler Intelligence Scale for Children Revised. Nederlandstalige uitgave. Lisse, The Netherlands: Swets Test Services.

17 van Iterson, L., & Augustijn, P. B. (2006) Unilateral hemisphere involvement associated predominantly to VIQ > PIQ discrepancies in children with (refractory) epilepsy. Epilepsia, 47, Vicari, S., Albertoni, A., Chilosi, A. M., Cipiriani, P., Coni, G., & Bates, E. (2000) Plasticity and reorganization during language development in children with early brain injury. Cortex, 36, Watkins, M. W., & Glutting, J. J. (2000) Incremental validity of WISC-III profile elevation, scatter, and shape information for predicting reading and math achievement. Psychological Assessment, 12, Watkins, M. W., Glutting, J. J., & Youngstrom, E. A. (2005) Issues in Subtest Profile Analysis. In D. P. Flanagan & P. L. Harrison (Eds.), Contemporary Intellectual Assessment: Theories, tests and issues (2 nd ed., pp ) New York: Guilford. Wechsler, D. (1992) WISC-III UK. Wechsler Intelligence Scale for Children. UK- Manual (3 rd ed., UK) London: The Psychological Corporation. Harcourt. Wechsler, D. (2004a) WAIS-III. Nederlandstalige bewerking. Wechsler Adult Intelligence Scale - Derde editie. London: Harcourt. Wechsler, D. (2004b) WISC-IV UK Wechsler Intelligence Scale for Children (4 th ed., UK) London: Harcourt Assessment. Wechsler, D. (2005) WISC-III NL. Wechsler Intelligence Scale for Children. Derde editie NL. Handleiding en Verantwoording. London / Amsterdam: The Psychological Corporation / Nederlands Dienstencentrum Nederlands Instituut van Psychologen. Zimet, G., Goodman Zimet, S., Farley, G. K., Shapiro Adler, S., & Zimmerman, T. (1994) Intellectual competence of children who are beginning inpatient and day psychiatric treatment. Journal of Clinical Psychology, 50,

18 Table 1 Samples and Diagnoses Sample n % n % Learning Disabilities Psychiatric Disorders Autism Spectrum Disorders Conduct Disorders / Oppositional Defiant Disorders Attachment Disorders Attention Deficit and Hyperactivity Disorders Tic Disorders Depression Other Epilepsy Seizure Type: Focal Onset / Localisation Related Seizures Left Hemisphere 33 Right Hemisphere 24 Bilateral / Multifocal 32 Generalized Seizures Uncertain whether Focal or Generalized Unknown Anti-epileptic Drug: > n a MRI positive data Total

19 Table 2 Mean Age, Mean WISC-R NL IQs and Sex for Three Clinical Samples and Two Subsamples (Autism Spectrum Disorders and Left Hemisphere Onset Seizures) age (yrs) V-IQ P-IQ FS-IQ boys M SD M SD M SD M SD n n % & range & range & range & range Learning Disabilities a to to to to 124 Psychiatric Disorders a to to to to 127 Autism Spectrum Disorders to to to to 127 Epilepsy b to to to to 125 Left Hemisphere Seizures to to to to 108 Total to to to to 127 a significantly higher than sample with epilepsy b significantly lower than learning disabilities and psychiatric disorders

20 Fig 1 Mean z-converted Uncorrected Subtest Scaled-score Range Values and SEM-bars for Verbal (black), Performance (white) and Full Scales (patterned), for Three Clinical Samples and Two Subsamples 1,0 0,9 0,8 Verbal range Performance range Full Scale range 0,7 0,6 z-score 0,5 0,4 0,3 0,2 0,1 0,0 Total Learning Disabilities Psychiatric Disorders Autism Spectrum Disorders Epilepsy Left Hemisphere Seizures

21 SUBTEST VARIABILITY ON THE WISC-R NL 21 Table 3 Subtest Scaled-score Range: Means, SDs, t-values, Probabilities and Effect Sizes for Three Samples and Two Subsamples contrasted to Expected Values. Verbal Scale Performance Scale Full Scale Sample n M SD t a p ES M SD t a p ES M SD t a p ES Expected Value b Learning Disabilities ns ns ns - Psychiatric Disorders ns < < Autism Spectrum Disorders < Epilepsy ns ns ns - Left Hemisphere Seizures ns Total ns < a one-sample t-tests for comparisons against estimated (expected) value, one-tailed tests, d.f. = N - 1 in all cases. b estimated expected values according to Silverstein (1987)