Representational Similarity Between Non-Symbolic and Symbolic Numerical Stimuli in High Level Visual Areas Is Uniquely Related to Individual Differences in Arithmetic Skills in Children with Autism

Background:  Recent findings have shown that while solving arithmetic problems, children with Autism Spectrum Disorders (ASD) display a unique pattern of brain organization that differs significantly from their Typically Developing (TD) peers (Iuculano et al., 2014). However, proficient arithmetic skills develop from more fundamental abilities. These include representing and manipulating different types of numerical information, such as arrays of dots (i.e., non-symbolic format) or pairs of Arabic digits (i.e., symbolic format) (Butterworth, 2003). Currently, nothing is known about how children with ASD process these basic numerical stimuli. Crucially, successful mapping between non-symbolic and symbolic stimuli has been proposed to be a strong predictor of arithmetic skills during typical development (Hiniker et al., 2015; Ansari, 2008).

Objectives:  Here we investigate whether (i) children with ASD exhibit a unique pattern of neural mapping between non-symbolic and symbolic stimuli, compared to TD children; (ii) a higher degree of representational similarity between these two stimulus-types is related to individual differences in arithmetic skills in children with ASD.

Methods:  We assessed a group of children with ASD (N =25; Μ_age = 10.55, SD = 1.36; 23 males; IQ = 109.96) and a matched group of TD children (N = 25; Μ_age = 10.07, SD= 1.35; 21 males; IQ = 112.12). ASD diagnosis was determined using the Autism Diagnostic Interview- Revised (ADI-R) and/or the Autism Diagnostic Observation Schedule (ADOS). A subtest of the Woodcock-Johnson III Tests of Achievement (Woodcock et al., 2001) battery was used to assess math proficiency in each participant. All children participated in a functional Magnetic Resonance Imaging (fMRI) session where they were asked to select the larger of two quantities, which were presented either in a non-symbolic or symbolic format. We used Representational Similarity Analysis (RSA; Kriegeskorte et al., 2008) to examine the degree of similarity between multivariate patterns of activity for non-symbolic and symbolic stimuli in both groups.

Results:  At the behavioral level, performance did not differ between ASD and TD children during the non-symbolic or symbolic comparison tasks. At the neural level, children with ASD showed greater RS values between non-symbolic and symbolic stimuli than TD children in multiple brain regions that support successful numerical manipulations (Ansari, 2008). These included the dorsal and ventrolateral prefrontal cortices, the intraparietal sulcus, and the fusiform gyrus. TD children did not display any regions of greater RS values, compared to children with ASD. Importantly, higher representational similarity between non-symbolic and symbolic stimuli in the right fusiform gyrus significantly predicted better arithmetic skills in ASD (r  = 0.48, p = 0.02). This was not evident in TD children.

Conclusions:  These results suggest that children with ASD display a unique neural mapping between non-symbolic and symbolic stimuli, compared to TD peers. Critically, these findings support the idea that high level visual processing areas, particularly the fusiform gyrus, play a crucial role in the successful development of arithmetic skills in ASD.