The study, published in the Journal of Autism and Developmental Disorders, analyzed records from 727 children aged six to sixteen who completed the WISC-V as part of comprehensive developmental evaluations at a large pediatric hospital in the midwestern United States between December 2014 and May 2022. The sample was predominantly male, at 78 percent, and predominantly White, at 77 percent, with a mean Full-Scale IQ of 88.4 and a standard deviation of 17.7. Slightly more than half of the children, 53 percent, had received an autism spectrum disorder diagnosis, while the remainder carried other neurodevelopmental conditions. Notably, the sample spanned a wide range of intellectual functioning, including roughly one hundred children with IQ scores in the intellectual disability range, a group that is frequently excluded from cognitive research.

Autism spectrum disorder is defined by differences in social communication and by restricted, repetitive patterns of behavior or interest, but its presentation varies enormously from one child to the next. Since the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, folded former subcategories such as Asperger syndrome and Pervasive Developmental Disorder-Not Otherwise Specified into a single spectrum diagnosis in 2013, clinicians have relied on dimensional ‘levels of support’ to describe severity. These levels, however, are subjective and vary with a clinician’s training and experience. Many researchers have hoped that quantifiable cognitive measures, particularly intelligence testing, could offer a more objective and standardized way to carve the spectrum into biologically or clinically meaningful subtypes, especially because general IQ is among the strongest predictors of long-term outcomes in autistic individuals.

To test that hypothesis rigorously, the team employed latent profile analysis, a person-centered statistical technique that assigns each child a probability of belonging to underlying subpopulations rather than simply comparing group averages. The method assumes that if genuine subtypes exist, they will emerge as distinct clusters when the five WISC-V index scores, covering verbal comprehension, visual-spatial skill, fluid reasoning, working memory, and processing speed, are fed into the model. The researchers tested solutions ranging from one to ten profiles under multiple variance assumptions, evaluating them with fit indices including the Bayesian and Akaike information criteria and the Analytic Hierarchy Process. Their sample comfortably exceeded the commonly recommended minimum of 500 participants for this class of analysis, lending statistical power that most prior studies lacked.

The results were strikingly consistent across criteria. Although one metric, the Analytic Hierarchy Process, favored a three-profile solution, that solution showed poor classification accuracy, with suboptimal entropy of 0.62 and marginal probabilities of group membership. More tellingly, the three extracted profiles were essentially parallel to one another, a pattern statisticians call the ‘salsa effect,’ in which the algorithm separates groups that differ only in overall level rather than in shape. Two-thirds of the children fell into a middle profile, with the rest split fairly evenly above and below it. The authors concluded that these were not genuinely distinct cognitive types but slices of a single continuous distribution of general intelligence, the psychometric construct known as the g factor.

To probe the underlying structure further, the team fitted bifactor confirmatory factor models in which the ten WISC-V primary subtests load both onto a general intelligence factor and onto specific domain factors. They then compared the two clinical groups to the nationally representative WISC-V standardization sample using multigroup measurement invariance testing. The models showed excellent fit, and the test confirmed that intelligence is measured in fundamentally the same way across autistic children, children with other neurodevelopmental disabilities, and typically developing children. Two subtests, Coding and Block Design, showed intercept differences between the clinical and standardization samples, hinting at modest measurement bias, but these differences appeared in both clinical groups, meaning they reflect neurodevelopmental risk in general rather than autism specifically.

The bifactor analyses delivered the study’s most clinically consequential finding: after accounting for general intelligence, nearly every WISC-V index explained too little unique variance to be meaningfully interpreted on its own. Fluid reasoning was the most extreme case, with factor loadings so close to zero that it could not be separated from the general factor at all. The lone exception was processing speed, which demonstrated sufficient explained variance above and beyond Full-Scale IQ across the clinical and standardization groups. Yet even here, the pattern was transdiagnostic. Slower processing speed appeared across children with autism and those with other neurodevelopmental conditions alike, marking it as a sensitive but non-specific signal of neurodevelopmental complexity rather than an autism-specific signature.

These findings land squarely in a long-running theoretical debate. The influential Cattell-Horn-Carroll framework treats intelligence as a hierarchy in which broad domain abilities, such as verbal comprehension or working memory, carry distinct clinical information beyond the general factor, and this logic underpins how many practitioners interpret index-level scores. The new results suggest an important boundary condition: when test indicators are heavily saturated by g, as the WISC-V subtests are, domain-level scores may add little incremental value, and the Full-Scale IQ remains the most informative single summary of a child’s cognitive ability. The authors are careful to note that this does not invalidate the broader theory, but it does challenge the routine clinical practice of narrating detailed strengths-and-weaknesses profiles from WISC-V index scores in neurodevelopmental evaluations.

For clinicians, the practical message is clear. Index scores on the WISC-V alone cannot be used to assign a child a cognitive subtype, and efforts to individualize support should instead draw on a comprehensive battery that includes speech and language assessment, executive function measures, adaptive functioning, and input from caregivers and educators. The authors acknowledge limitations, including the absence of a neurotypical comparison group in the latent profile analysis, a clinical sample that was not nationally representative, and the exclusion of children too profoundly affected to complete standardized testing. They also stress that the null result is theoretically meaningful rather than inconclusive: it directly supports dimensional models of cognition in clinical populations. Future work, they argue, should explore constructs beyond IQ, such as executive function, metacognition, academic enablers, and adaptive behavior, and should track cognitive strengths and weaknesses developmentally over time. Until such measures mature, the quest for a characteristic autistic cognitive profile on the WISC-V appears to be over, replaced by a simpler, more honest picture of intelligence as a continuum.

Subject of Research: Cognitive subtyping in children with autism and neurodevelopmental disabilities using WISC-V intelligence testing and latent profile analysis

Article Title: In Search of Subtypes of Children With Developmental Disabilities Using the WISC-V: A Latent Profile Analysis

Article References: Esprit, J. L., Youngstrom, E. A., Kim, S. Y., Norris, M., Levine, A., Butter, E. M., & Stephenson, K. G. (2026). In Search of Subtypes of Children With Developmental Disabilities Using the WISC-V: A Latent Profile Analysis. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07506-x

Image Credits: AI Generated

DOI: 10.1007/s10803-026-07506-x

Keywords: autism spectrum disorder, WISC-V, latent profile analysis, neurodevelopmental disabilities, cognitive profiles, general intelligence, processing speed, Full-Scale IQ, bifactor model, measurement invariance, child psychology, psychological assessment