Examining the Mathematical Abilities of Children and Adolescents with Autism Spectrum Disorders: A Meta-Analysis
In order to develop appropriate interventions, remediation and even enhancement activities, we need to be able to predict which individuals with ASD may have math strengths versus weaknesses. In order to make these predictions, we must investigate possible variables that have been previously shown to predict math success in other populations. Four variables that we propose to examine in relation to the math skills of the ASD population are: visual-spatial/motor ability (VSM), executive functioning (EF), general knowledge (Gk) and vocabulary knowledge (Vk).
Objectives: 1. To determine the size of the difference between individuals with ASD and their TD peers in two measures of math ability—arithmetic and math problem solving (Ps)—as well as the direction and the consistency of this effect. 2. To examine four predictors of math ability in the ASD population.
Methods: After a comprehensive database search, we collected data on 1056 children and adolescents with ASD across 35 studies. Using the metafor package in R (Viechtbauer, 2010), we derived the standardized mean differences (SMD; Hedge’s g) between the math scores of individuals with ASD and their TD peers along with meta-regression models examining predictors of their math ability.
Results: The models demonstrated that individuals with ASD performed significantly lower (g = -0.5) than their TD peers in both arithmetic and Ps and the resultant prediction intervals were -2.26 to 1.30SD and -2.76 to 1.95SD, respectively. Secondly, the SMDs across the two math areas were highly variable and this variability was not only due to random error. The meta-regression model for arithmetic demonstrated that EF and VSM were significant predictors of math ability (ß = 0.028, ß = 0.031, respectively; p < .001), yet Gk and Vk were not.
Conclusions: Our findings reject the stereotype of mathematical prowess among individuals with autism, as their average performance was lower than their TD peers. As well, the range of the prediction intervals were wide, but positively skewed, indicating that students with ASD were more likely to struggle in math than to excel. Finally, the outcome of the meta-regression model was generally in line with the Pathways model of numerical development, according to which mathematical ability is influenced in distinct ways by different neural pathways (LeFevre et al., 2010). The current analyses suggest that VSM and EF may be critically important to understanding how the autistic mind processes mathematical information.