A Proposed Model of Intelligence and Its Implications for Children with Autism Spectrum Disorder

Background: Although research indicates that IQ tests are reliable and valid measures of cognition when administered to individuals with autism spectrum disorder (ASD), it also reveals that such tests have unusual psychometric properties in this population, notably low inter-correlation among subtests, distinctive patterns of strengths and weaknesses, and over-prediction of everyday functioning. A satisfactory factor structure that explains the cognitive strengths and deficits of this population has not been identified. Nor has there been an explanation of why their adaptive behavior is not commensurate with their cognitive abilities.

Objectives: This study evaluates a new model of the structure of intelligence in the ASD population that, if confirmed, would help account for the differences. A three-domain model is proposed, consisting of language ability, perceptual ability, and social ability, with each ability further broken down into a fluid reasoning component and crystalized knowledge component. The proposed model is then compared to a conventional IQ test in regards to its ability to predict adaptive behavior.

Methods: Children with ASD, ages 6-12, with full scale IQ scores above 50, were assessed on a standardized IQ test (Stanford-Binet Intelligence Test, 5th Edition), a test of social skills (NEPSY-II), and parent-rated adaptive behavior (Vineland Adaptive Behavior Scales-II) and executive functioning (BRIEF). Factor scores were estimated from 6 variables (Verbal Knowledge, Verbal Fluid Reasoning, Nonverbal Knowledge, and Nonverbal Fluid Reasoning from the SB-5; Affect Recognition and Theory of Mind from the NEPSY-II), and a path analysis was conducted to test the fit of a model relating these factors to (1) constructs representing cognitive abilities and adaptive abilities, and (2) constructs representing the components of the new model proposed in this study (language abilities, perceptual abilities, and social abilities).

Results: Fifty-three children (44 males, 9 females) with ASD, a mean age of 9.6 (SD = 1.9), and a mean Full Scale IQ of 92.49 (SD = 19.46; range: 51-126) participated in the study. Overall, results supported the use of the SB-5 with school-age children with ASD with high inter-correlations between subtests and Full Scale IQ, r(51) = 0.73 - 0.89 and between domains and Full Scale IQ, r(51) = 0.88 - 0.92. The proposed model was partially supported with the inclusion of social skills and only 4 subtests of the SB-5 (χ2 (6) = 1.166, p = 0.979) and predicted adaptive behavior at least as well as the full IQ test (β = 0.42, p = 0.002). BRIEF scores did not contribute significantly to the model.

Conclusions:  In this study, adding measures of social skills to cognitive assessments reduced the number of variables needed to predict children’s use of skills in their everyday environment and resulted in predictions that were at least as accurate as a more detailed cognitive assessment. Alternatively, assessment batteries that include Full IQ tests and social skills assessment may provide significantly more information compared to conventional batteries, supporting the use of different assessment batteries than are typically used in practice.