Psychometric concerns with the 10-item Autism-Spectrum Quotient (AQ10) as a measure of trait autism in the general population

Abstract

The 10-item Autism-Spectrum Quotient (AQ10) is a self-report questionnaire used in clinical and research settings as a diagnostic screening tool for autism in adults. The AQ10 is also increasingly being used to quantify trait autism along a unitary dimension and correlated against performance on other psychological/medical tasks. However, its psychometric properties have yet to be examined when used in this way. By analysing AQ10 data from a large non-clinical sample of adults (n = 6,595), we found that the AQ10 does not have a unifactorial factor structure, and instead appears to have several factors. The AQ10 also had poor internal reliability. Taken together, whilst the AQ10 has important clinical utility in screening for diagnosable autism, it may not be a psychometrically robust measure when administered in non-clinical samples from the general population. Therefore, we caution against its use as a measure of trait autism in future research.

Internal reliability, homogeneity, and factor structure of the ten-item Autism-Spectrum Quotient (AQ-10) with two additional response categories

“…The present study showed that the AQ-10 with six-point rather than four-point response scales gained somewhat more internal consistency and homogeneity. However, overall, the parameters were still unsatisfactory with McDonald's ω and Cronbach's α < .70 (Taylor et al, 2020) and the average interitem correlation was < .20 (Briggs & Cheek, 1986). Furthermore, as the intercorrelations of the subdimensions implied, no acceptable model fit was achieved when the subdimensions attention to detail and imagination were included in the CFAs.…”

“…However, recently, the AQ-10 has been demonstrated to be psychometrically insufficient (Taylor et al, 2020). Re-analysing a large sample of 6,595 participants recruited online, Taylor et al (2020) found that the AQ-10 showed poor internal reliability defined by metrics <.70. Moreover, there was no unifactorial structure, indicating that the AQ-10 partly consists of unrelated dimensions, which explains the low internal reliability.…”

“…In particular, its ten-item short form (AQ-10; Allison et al, 2012) can be an efficient measure for research purposes (e.g., Bertrams & Schlegel, 2020;Gollwitzer et al, 2019;Lundin et al, 2019). However, recently, the AQ-10 has been demonstrated to be psychometrically insufficient (Taylor et al, 2020). Re-analysing a large sample of 6,595 participants recruited online, Taylor et al (2020) found that the AQ-10 showed poor internal reliability defined by metrics <.70.…”

Greater autism knowledge and contact with autistic people are independently associated with favourable attitudes towards autistic people

“…Similarly, autistic traits were not significantly associated with attitudes to autism (see also, Gardiner & Iarocci, 2014). There are ongoing concerns with measuring autistic traits in the general population (e.g., Taylor et al, 2020), which is a potential limitation of the study and many others in our field. Nonetheless, our finding is of interest as it is often assumed that people with many autistic traits have favourable attitudes towards autistic people and would therefore be good advocates for them (e.g., Komeda, 2015).…”

“They Think I’m Amazing but Won’t Talk with Me.” Neurotypical People Are More Cognitively Than Behaviorally Positive Towards Their Autistic Friends

“…Murray and colleagues (Murray, Allison, et al, 2017; also see Agelink van Rentergem et al, 2019) suggest a degree of diagnostic bias in the AQ, and that the AQ reflects "maleness" in autism, introducing inconsistency among female and nonbinary respondents. Agelink van Rentergem et al (2019) suggested, as well, that the AQ items function differently for different groups of people, and other emerging, psychometric research indicates poor reliabilities are common with various versions of the AQ (Jia et al, 2019;Taylor et al, 2020). We included the AQ-10 in the final manuscript out of an abundance of transparency, since this sheds further light on its psychometric qualities and since our relatively large sample size and pattern of replicated results allow us to be confident in its use within this study.…”