The Autism Spectrum Quotient in Mathematicians

Background:

Previous studies have identified a higher number of autistic traits in mathematicians, and a higher rate of formal diagnosis of autism in mathematicians (Baron-Cohen and Wheelwright, 2007). The reason for this link between autistic traits/autism and mathematics may be because autistic traits/autism includes excellent attention to detail and a strong interest in patterns (“systemizing”). Here we measure the distribution of autistic traits using the Autism Spectrum Quotient (AQ) in maths students/graduates of mathematics with and without autism. We also tested a group of mathematicians who suspect they have autism but had not pursued a formal diagnosis.

Objectives:

Our study had the following three objectives: 1. To investigate if mathematicians score significantly higher on the AQ than the general population; 2. To investigate sex differences in AQ scores in mathematicians; 3. To investigate differences in AQ scores by diagnosis in mathematicians.

Methods:

1049 university students or graduates of mathematics from top mathematics departments worldwide completed an online version of the AQ. Participants were grouped as either mathematicians with ASC (n = 80 (63 male)), mathematicians with suspected ASC (n = 222 (158 male)), and mathematicians without ASC (n = 747 controls (433 male)). We included a general population sample for comparison (n=3466 (898 male)).

Results:

Male mathematicians without ASC scored significantly higher than general population males (p=0.00017), and female mathematicians without ASC also scored higher than general population females (p=4.36x10^-13). Unlike the general population (Ruzich et al., 2015), where males score higher than females (means 17.89 and 14.88 respectively), male mathematicians without ASC (mean = 21.35, SD = 7.94) did not score significantly higher than their female counterparts (mean = 20.46, SD = 7.90). Mathematicians with ASC and mathematicians with suspected ASC did not differ from each other, but both scored significantly higher than mathematicians without ASC (p < 2.2x10^-16). The subdomain scores of female mathematicians without ASC were lower than the mathematicians with ASC and the mathematicians with suspected ASC on all subdomains except ‘attention to detail’, where there was no difference. Male mathematicians without ASC scored significantly lower than mathematicians with ASC (p<0.038) and mathematicians with suspected ASC (p<0.013) on the ‘attention to detail’ subdomain. No sex differences were observed within groups for AQ or on any subdomain scores.

Conclusions:

AQ scores are higher in mathematicians with ASC and suspected ASC compared to mathematicians without ASC, irrespective of sex, reflected by significantly higher scores in the subdomains of ‘communication’, ‘social skills’, ‘imagination’, and ‘attention switching’. The typical sex difference in AQ in the general population is absent amongst all groups of mathematicians, just as it is absent or attenuated in people with ASC. The absence or attenuation of the typical sex difference may reflect biological factors such as prenatal sex steroids, which are elevated in those with ASC. This study confirms the strong association between autistic traits and mathematical talent. Ongoing research is exploring the basis of this association at a genetic level.