Assessment and Special Education

SPED Assessment

In 2002 the American Association on Mental Retardation (AAMR) made changes to their manuals regarding the assessment of mental retardation (MR). The revisions were designed to affect changes in professional practice regarding assessment of MR, public policy, and the science and understanding of MR. Key in this change was the attempted change from the MR term to a more politically correct term Intellectual Disability. Assessment was to consider both IQ scores and adaptive behavior (AB) which was to be termed "adaptive skills as well as the individual's cultural background and in the context of associated strengths. Instead of following a deficit model of explanation the goal was to follow a needs model. The definition of intellectual disability then includes three core criteria: significant impairment of intellectual functioning (defined by decreased IQ scores), significant impairment of adaptive/social functioning and, onset before adulthood. Polloway et al. (2009) looked at the impact of these changes on state guidelines for the assessment and treatment of MR. Interestingly, 27 states still formally used the MR term in their guidelines/definitions with only four using the recommended term. Most states still adopted older formal definitions on MR. With respect to IQ scores, the topic of this paper, only 12 states reported no specific IQ cutoff score was needed for the diagnosis, whereas those reporting a specific IQ score typically used the 70 or 70-75 cutoff score either by formal definition or maintaining a score of two standard deviations below the mean was the cutoff score. Forty-nine states required deficits in AB. Age guidelines and an classification system (e.g., designating mild MR, moderate MR, etc.) were variably employed. Polloway et al. (2009) pretty much leave the IQ assessment issue alone, suggesting that it is a cornerstone of the recognition and assessment of MR (note, the discussion in this paper is primarily on Full Scale IQ scores and their equivalents). And here is where this study, although primarily descriptive in nature, totally misses the point. In fact IQ scores as a method of assessing MR have severe limitations and certainly warrant more attention regarding the accurate assessment of MR than whether a state terms the condition MR, cognitive impairment, or intellectual disability.

One issue with IQ scores is their consistency over time, especially at the lower end of the continuum in IQ distributions. Changes in IQ scores across time are often explained as artifacts of random or systemic error. This is why it may often be best to report confidence intervals as opposed just too reporting point estimates of IQ. However, Whitaker (2008) performed a meta-analysis of those obtaining low IQ scores (less than a Full Scale IQ of 80) that were retested at a mean interval time of 2.8 years. Despite the reporting in most IQ manuals of 95% confidence intervals of about five IQ points either side of the obtained score, Whitaker found that 14% of the scores in the meta-analysis changed by 12.5 points or more. Moreover, stability at the lower ends of the subtest score distribution is questionable and floor effects exist. For instance, in the Wechsler IQ protocols a raw score of zero on a subtest often still yields a scaled score of one or higher in older individuals who are given the WAIS and not the WISC tests. This means that scaled scores will often overestimate the person's ability on that particular domain. When these overestimates occur there are some serious ramifications in the classification of MR individuals who need special education. If IQ scores cannot show constancy over time, then their utility is questionable as the mainstream form of assessment for classifying special needs students with MR or in cases where the death penalty is involved and a cognitive assessment is crucial to the life of a person (it is against the law to execute person with MR who is convicted of a crime).

There are no standardized or formalized recommendations of which IQ test should provide the best estimate intellectual disability as different tests will yield different results, even when given to the same person. This effect has been observed even in tests produced by the same manufacturer. For instance, Gordon, Duff, Davison, and Whitaker (2010) observed that in the past the WISC and WAIS IQ tests provided disparate Full Scale IQ scores when administered to the same subjects. As the WISC has a ceiling age of 16 and the WAIS can be administered to subjects as young as 16 years-old Gordon et al. administered both (the WISC-IV (UK) and the WAIS-III (UK)) to 17 students with a mean age of 16.2 years who were identified as having intellectual disabilities. A counterbalanced repeated measures design was used to avoid order effects. Despite finding significant correlations between the IQ scores for the two tests the mean WISC-III Full Scale IQ was 64, whereas the mean WISC-IV score was 53, a significant difference and one that would result in two different classifications of MR severity. All the participants scored lower on the WISC-IV than the WAIS-III with the smallest difference between any participant's scores being five points. The Index Scores, with the exception of the Working Memory Index, were also significantly higher on the WAIS-III, with the differences ranging between 9.50 and 12.58 points. The primary difference in Full Scale IQ scores was reflected in the verbal domain (the mean Verbal Comprehension Index score was 67.59 on the WAIS-III and 55.76 on the WISC-IV).

The situation becomes even more complicated. Silverman et al. (2010) compared the WAIS and the Stanford-Binet scores of 74 older adult patients with intellectual disabilities (the majority of the participants had been diagnosed with Down syndrome). There was also additional data on some of the participants such as scores on other, but not as popular measures of intelligence such as the Leiter International Performance Scale, the Slosson Intelligence Test and the Vineland Adaptive Behavior Scales. The mean WAIS IQ score for the group was 58.1, whereas the mean Stanford-Binet IQ score was 41.3, a difference greater than a standard deviation for both tests (the standard deviation for the WAIS tests is 15; the Binet tests it is 16). Moreover, the WAIS Full Scale IQ score was consistently higher than comparable measures of other tests. The Standard-Binet tests were more strongly related to other measures of intelligence than were the WAIS tests, indicating that the WAIS, the most popular instrument for measuring IQ, overestimated intellectual ability at the low end of functioning relative to the other measures.