It has been hypothesised that romantic partners are more similar than chance in relation to autistic traits. To test this theory, we recruited n = 105 heterosexual couples and examined within-couple correlations for autistic traits [measured using the Autism Spectrum Quotient (AQ)], empathizing [measured using the Empathy Quotient (EQ)], and systemizing [measured using the Systemizing Quotient-Revised (SQ-R)]. For a subsample that attended the lab (n = 58 couples), we also investigated theory of mind via facial expressions using the Reading the Mind in the Eyes Test (RMET) and attention to detail, a component within systemizing, using the Embedded Figures Task (EFT). Variable-centred analyses revealed positive within-couple correlations for all measures except EQ, although these effects were only statistically significant for unmarried couples and not for married/engaged couples. Follow-up analyses indicated that the observed couple similarity effects are likely consistent with people pairing with those more similar than chance (initial assortment) rather than becoming alike over time (convergence), and to seeking out self-resembling partners (active assortment) rather than pairing in this manner via social stratification processes (social homogamy). Additionally, a significant within-couple correlation for autistic traits was observed at the meta-analytic level. However, it should be noted that the meta-analytic effect size estimate was small (
Autism is characterised by unusually routine behaviours, narrow interests, sensory hyper-sensitivity, social and communicative problems, and difficulties in adjusting to unexpected change
While autistic people on average are less likely to marry or be in long-term relationships
Evidence for autism being subject to assortative mating comes from a study by Nordsletten et al.
Previous studies have examined autism as one entity and autistic traits as one set of traits. Here we consider autistic traits as a whole and as a combination of different traits. With excellent attention to detail, and the ability to remain focussed on pattern-based tasks, autistic people are more likely to work in Science, Technology, Engineering, and Mathematics (STEM)
The current study aims to increase our understanding of the processes that might underpin assortative mating as it relates to autism. More specifically, we examined within-couple correlations for quantitative self-report measures of autistic traits (the Autism Spectrum Quotient or AQ
Both male and female data were available for 105 couples. Most participants were self-reported White/Caucasian (males: n = 95, 90.48%; females: n = 95, 91.35% [one missing datapoint]) employed (males: n = 86, 81.90%; females: n = 89, 84.76%), and relatively few were currently students (males: n = 18, 17.14%; females, n = 27, 25.71%). The levels of educational attainment reported are as follows: no qualifications (males, n = 4, 3.81%; females, n = 1, 0.95%), GCSE or equivalent (males: n = 21, 20.00%; females, n = 15, 14.29%), A level or equivalent (males: n = 27, 25.71%; females: n = 35, 33.33%), Bachelor’s Degree (males: n = 38, 36.19%; females: n = 32, 30.48%), Master’s Degree (males: n = 8, 7.62%; females: n = 11, 10.48%), Doctorate Degree (males: n = 7, 6.67%; females: n = 11, 10.48%). One male (0.95%) and no females (0.00%) disclosed having an autism diagnosis, though 8 males (7.69%) and 3 females (2.86% [one missing datapoint]) suspected they might be autistic. Most participants reported living with their partner, although there was a slight discrepancy in that fewer males than females reported this (males: n = 73, 69.52%; females: n = 76, 72.38%). For marital status, there was no such discrepancy: 59 couples (56.19%) were not married, 8 (7.62%) were engaged, and 38 (36.19%) were married.
Although not specified as part of our pre-registration, intercorrelations between the autism-related variables are depicted in Fig. Correlation heatmap for autism-related variables. Descriptive statistics and sex differences for autism-related variables. Male Female Difference AQ 104 18.01 7.43 104 15.30 6.65 3.323 103 EQ 102 40.48 13.01 102 52.68 12.99 − 6.641 101 − SQ-R 103 60.16 21.50 103 46.04 17.16 6.040 102 D 101 0.039 0.102 101 − 0.086 0.103 9.662 100 RMET 55 23.85 4.96 55 24.84 4.95 − 1.386 54 0.172 − 0.198 EFT 58 32.24 21.24 58 41.34 27.60 − 2.587 57 − AQ, Autism Spectrum Quotient; EQ, Empathy Quotient; SQ-R, Systemizing Quotient-Revised; D, difference in standardized EQ (E) and SQ-R (S) scores; RMET, Reading the Mind in the Eyes Test; EFT, Embedded Figures Task. Effects in bold are statistically significant (
Partners’ ages were strongly positively correlated,
Although we preregistered an analysis based on Pearson’s correlations, some of the scale distributions were non-normal, and Spearman’s correlations were used instead. Six Spearman’s tests (Bonferroni-corrected required α level: Scatterplots with regression lines and 95% confidence intervals demonstrating within-couple correlations observed for AQ (
At the suggestion of a reviewer, we stratified the above analyses by marital status (unmarried or married/engaged). All outcome variables other than EQ exhibited statistically significant positive within-couple correlations for unmarried couples, whereas no significant correlations were observed for the married/engaged couples. We then used Fisher’s Within-couple correlations for autism-related variables stratified by marital status. Unmarried couples Married/engaged couples Difference AQ 59 45 0.058 0.704 EQ 59 0.095 0.476 43 − 0.106 0.497 0.97 0.332 SQ-R 58 45 0.083 0.588 1.93 0.054 D 58 43 0.104 0.507 1.31 0.190 RMET 35 20 0.046 0.847 1.62 0.105 EFT 37 21 0.196 0.394 1.29 0.197 AQ, Autism Spectrum Quotient; EQ, Empathy Quotient; SQ-R, Systemizing Quotient-Revised; D, difference in standardized EQ (E) and SQ-R (S) scores; RMET, Reading the Mind in the Eyes Test; EFT, Embedded Figures Task. Effects in bold are statistically significant at the conventional threshold of
To investigate whether the within-couple correlations for autism-related variables were explainable by couples being more similar than chance to begin with (initial assortment) or becoming more alike throughout the course of their relationship (convergence), we correlated the sex-standardised within-couple difference scores for autism-related variables with length of relationship. Essentially, if length or relationship is correlated with the difference score, it suggests that partners may become more similar (negative correlation) or more dissimilar (positive correlation) over the course of their relationship, and so provides evidence against initial assortment
Our pre-registration specified that we would use Pearson’s correlations for this analysis. However, we used linear regression instead, as this allowed for inclusion of marital status (unmarried vs. married/engaged) as a covariate. Relationship length was not associated with the within-couple differences for AQ,
We next aimed to investigate whether within-couple correlations for autism-related variables may be better explained by individuals consciously or unconsciously seeking out similar partners (active assortment) or by people tending to pair with others around them who, on average, are more similar than chance due to social stratification effects (social homogamy). We first correlated the within-couple sex-standardised difference scores for the autism-related variables with the couples’ absolute (i.e., unsigned) difference in age. The idea is that if similarity for age is positively correlated with similarity for an autism-related variable, then the within-couple correlation for that autism-related variable may be explained by social homogamy effects. Although we pre-registered to use Pearson’s correlations, we used linear regression analyses instead, as this allowed for marital status (unmarried or married/engaged) to be included as a covariate. We observed no correlation between absolute age difference and within-couple differences for AQ (
We then examined whether there were correlations between couple similarity for level of educational attainment and autism-related variables. We initially pre-registered Spearman’s correlations for this analysis, though again adopted the approach of using linear regression to control for marital status as a covariate. Couple similarity for educational attainment was not significantly associated with couple similarity for AQ (
Finally, we examined whether couples who study and/or work in STEM are more similar for autism-related variables than are couples for whom either one member or both members do not study or work in STEM. Of those participants who reported current employment and/or student status, there were 35 males and 33 females who worked or studied in STEM areas, and 57 males and 61 females who did not study or work in STEM areas. We compared the 20 couples for whom both members were in STEM with the 67 couples for whom either one or neither member was in STEM. These groups of couples did not differ in their levels of similarity on the AQ (
We next aimed to determine whether a participant’s scores for autism-related variables were more similar (i.e., numerically closer) to those of their partner than would be expected by chance. To do this, we first calculated sex-standardised absolute (unsigned) difference scores for each of the relevant variables. Next, we calculated the difference scores between any given male and all females that were not his partner (and, therefore, any given female and all males that were not her partner) and took the average. We then used six paired-samples Comparison of actual couple similarity for autism-related variables with the average difference scores calculated by pairing each participant with all other potential other-sex partners in the dataset. Actual couple difference Average difference of all other pairings Difference AQ 0.891 0.770 1.118 0.433 − 3.227 103 − EQ 1.109 0.861 1.111 0.384 − 0.025 101 0.980 − 0.003 SQ-R 0.928 0.779 1.128 0.410 − 2.761 102 D 0.973 0.788 1.092 0.433 − 1.817 100 0.072 − 0.170 RMET 0.837 0.634 1.101 0.408 − 2.817 54 EFT 0.788 0.722 1.071 0.520 − 2.847 57 AQ, Autism Spectrum Quotient; EQ, Empathy Quotient; SQ-R, Systemizing Quotient-Revised; D, difference in standardized EQ (E) and SQ-R (S) scores; RMET, Reading the Mind in the Eyes Test; EFT, Embedded Figures Task. Effects in bold are statistically significant (
Considering the exploratory nature of this analysis, we included all samples identified for which a within-couple correlation for an autistic traits measure was reported, as well as those for which the within-couple correlation could be calculated from available data (16 samples from 13 studies; see Table Forest plot for within-couple correlations for autistic traits. Note Baron-Cohen et al. (2014a) = parents of autistic children; Baron-Cohen et al. (2014b) = parents without autistic children; Connolly et al. (2019a) = Autism Genome Project sample; Connolly et al. (2019b) = Simons Simplex Collection sample; Richards et al. (2022a) = unmarried sample; Richards et al. (2022b) = married/engaged sample. Contour-enhanced funnel plot for studies for which within-couple correlations for autistic traits measures could be determined. Black circles represent individual studies; white circles represent the hypothesised missing studies imputed via the trim and fill procedure.
Due to the high level of heterogeneity observed ( Overview of studies included in the meta-analysis of within-couple correlations for autistic traits. References Year Journal Location Population Measurement Bishop et al. 2004 Australia Parents of autistic children (Western Australia Family Study of Autistic Spectrum Disorders) and control parents (recruited via advertisements sent to schools in the Perth Metropolitan Region) Presence/absence of Broader Autism Phenotype (determined via a principal components analysis of the AQ Social Skill and Communication subscales) 49 ϕ = − 0.16 0.459 Constantino and Todd 2005 USA Missouri Twin Study (parents of twins) SRS 285 < 0.05 Hoekstra et al. 2007 Netherlands Parents of twins Dutch AQ 128 0.59 Losh et al. 2008 USA Parents of autistic children (simplex families, n = 35; multiplex families, n = 23) Factor 1: Language Factor 2: Rigidity Factor 3: Anxiety Factor 4: Sociability 58 58 58 58 n.s n.s n.s n.s Virkud et al. 2009 USA Autism Genetic Resource Exchange (parents of autistic children [multiplex families]); parents of Washington University male siblings study (simplex families) SRS 99 < 0.01 Hoekstra et al. 2010 Netherlands Parents of twins Dutch AQ-Short 305 < 0.05? Schwichtenburg et al. 2010 USA Parents sampled by UCD (n = 115) and UCLA (n = 102) of at least one autistic child n = 135 or no autistic (but typically developing) child n = 82 SRS 217 β = 0.34 < 0.05? Pollmann et al. 2010 Netherlands Couples married for 10 months Modified Dutch AQ-Short 195 n.s Seidman et al. 2012 Israel Parents of autistic children BAPQ Total (Hebrew) (SR + IR) BAPQ Aloof (Hebrew) (SR + IR) BAPQ Rigid (Hebrew) (SR + IR) BAPQ Pragmatic Language (Hebrew) (SR + IR) 76 76 76 76 – – – 0.04 – 0.03 van Steijn et al. 2012 Netherlands Parents of autistic children (with or without ADHD) AQ (alternative scoring method) 121 n.s Baron-Cohen et al. 2014 UK Parents of autistic children in the Cambridge Autism Research Database (CARD) AQ 92 0.365 UK Parents without autistic children in the Cambridge Autism Research Database (CARD) AQ 29 0.693 Lau et al. 2014 Taiwan Parents of autistic children AQ-Chinese 491 ϕ = 0.10 0.032 Lyall et al. 2014 USA Nurses’ Health Study II parents of autistic children SRS 1198* - USA Nurses’ Health Study II control parents SRS 1198* - Duvekot et al. 2016 Netherlands Social Spectrum Study parents of children referred for various mental health conditions; 159 families of children at risk for autism (parent-reported SRS-2 total score of ≥ 75) and 72 families of children considered not at risk for autism (parent-reported SRS-2 total score of < 75) SRS-2 231 < 0.001 Connolly et al. 2019 Europe/North America Autism Genome Project (parents of autistic children [simplex and multiplex families]) BAPQ Total (SR + IR) BAPQ Aloof (SR + IR) BAPQ Rigid (SR + IR) BAPQ Pragmatic Language (SR + IR) SRS-A Total (IR) SRS-A Awareness (IR) SRS-A Cognition (IR) SRS-A Mannerisms (IR) SRS-A Motivation (IR) SRS-A Communication (IR) 270 270 270 270 428 428 428 428 428 428 – – < 0.001 n.s < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 n.s < 0.001 North America Simons Simplex Collection (parents of autistic children [simplex families]) BAPQ Total (SR) BAPQ Aloof (SR) BAPQ Rigid (SR) BAPQ Pragmatic Language (SR) SRS-A Total (IR) SRS-A Awareness (IR) SRS-A Cognition (IR) SRS-A Mannerisms (IR) SRS-A Motivation (IR) SRS-A Communication (IR) 1945 1945 1945 1945 1953 1953 1953 1953 1953 1953 – – – – - - n.s n.s n.s n.s < 0.001 n.s < 0.001 < 0.001 n.s < 0.001 Richards et al. (current study) 2022 UK Unmarried heterosexual couples from the general population AQ 59 0.001 UK Married/engaged heterosexual couples from the general population AQ 45 0.704 AQ, Autism Spectrum Quotient; BAPQ, Broad Autism Phenotype Questionnaire; SRS, Social Responsiveness Scale. *The two samples of Lyall et al. were analysed together; n = 1198 couples were included, though it is unclear how many were parents of autistic children and how many were control parents.
The current study aimed to test whether autistic traits are correlated within heterosexual couples in the general population. The main finding was that quantitative measures of autistic traits (AQ), systemizing (SQ-R), empathizing relative to systemizing (D score), the ability to read emotions in the eye region (RMET), and spatial skills (EFT) (but not empathizing; EQ) were all positively correlated, albeit the effects observed for D and RMET were no longer statistically significant after Bonferroni correction. Interestingly, statistically significant effects were observed for the subsample of unmarried couples, but not for the subsample that was either married or engaged. The couple-similarity effects also appear to be better explained by active assortment than social homogamy, and by initial assortment rather than convergence. Additionally, when we compared within-couple sex-standardised absolute (i.e., unsigned) difference scores for these variables with sex-standardised absolute difference scores calculated as the average of all other possible male–female pairings within the dataset, we found that actual couples were more similar for AQ, SQ-R, RMET, and EFT than would be expected under the assumption of random mating. We also observed a small but statistically significant within-couple correlation for autistic traits in a meta-analysis of 16 samples (including the current study).
A number of investigations have examined within-couple correlations for autistic trait variables, the majority of which have focused on very specific samples such as the parents of autistic children
It remains unclear why autistic traits and systemizing are correlated within couples. It may also be a mistake to conflate higher levels of autistic traits with a generalised deficit in social communication skills
The finding that the within-couple correlations for AQ, SQ-R, RMET and EFT were not moderated by length of relationship or social homogamy variables suggests that these effects reflect initial and active assortment (i.e., people seek out similar individuals as partners and do not become more alike over the course of their relationship). Although Pollman et al.
It is noted that the within-couple correlations observed for autism-related variables in the current study were only statistically significant in unmarried couples. This was unexpected because couples that have been together for only a relatively short time (< six months)
It is relevant to note that similarity in people’s levels of autistic traits extends beyond romantic relationships, as the phenomenon has also been observed in friendship dyads. Wainer et al.
There are several limitations to the current research. First, the average level of educational attainment was fairly high (17.6% had a Master’s degree or higher); although arguably more representative than many studies in this area, it is still questionable exactly how generalisable our sample is in regard to the general population. Our study is also correlational in nature, meaning that it was not possible to assess the development of relationships over time to determine causality
Due to COVID-19-related restrictions, only a subsample of our study participants was administered the RMET and EFT, meaning that analyses relating to these variables achieved lower statistical power than those for the AQ, EQ, SQ-R, and D. Although we still demonstrated statistically significant effects relating to the RMET and EFT, replication and extension of these findings will be necessary for firmer conclusions to be drawn. For instance, future studies will be required to determine whether assortative mating processes apply specifically to these variables or whether such effects are explained by within-couple similarity for intelligence quotient (IQ) scores. We also acknowledge that, although it provides a useful indication of effect size for the within-couple correlation for autistic traits, our meta-analysis is limited in certain ways. For instance, due to its ad hoc nature (it was not part of our pre-registration plan), it falls short of PRISMA guidelines
In conclusion, the current study provides evidence that small-to-moderate levels of partner similarity likely exist for a range of traits associated with autism spectrum conditions. Although limitations of the analyses conducted should be considered, the observed effects appear more consistent with initial and active assortment than with convergence and social homogamy. Notably, we also demonstrate that systemizing is positively correlated between heterosexual partners, and that actual couples are more similar for this trait than would be expected under the assumption of a random pattern of mating.
We conducted an a priori power analysis using G*Power 3.1
Participants first reported their sex (Male, Female, Prefer not to say, Other [please specify below]), age (years), and ethnicity (White, Mixed / multiple ethnic groups, Asian / Asian British, Black / African / Caribbean / Black British, Other ethnic group). They were then asked questions regarding their relationship, specifically their cohabiting status (Living with partner, Not living with partner), length of relationship (years and months), and marital status (Not married, Engaged to be married, Married). They were also asked to confirm their educational level (No qualifications, Completed GCSE level [or equivalent], Completed A level, Access Course [or equivalent], Bachelor’s Degree, Master’s Degree, Doctorate Degree, Other, please specify), current student status (Yes, No; if Yes, then area and year of study were also recorded), whether they were employed (Yes, No; if Yes, then place of work and job role were also recorded), whether they had an autism diagnosis (Yes, No) or suspected they were autistic (Yes, No).
We used the 50-item self-report Autism Spectrum Quotient (AQ
We measured self-reported empathizing via the 40-item Empathy Quotient (EQ
In addition to the self-report questionnaires, two behavioural measures were administered. First, the Reading the Mind in the Eyes Test (RMET
The current study utilised a correlational design. Participants were invited to attend a lab session in which they and their partner would independently complete several measures relating to autism. Each couple was offered a £10.00 shopping voucher as an incentive to participate. The study protocol, hypotheses, and analysis plan were pre-registered on the Open Science Framework (osf.io/6jg8p). (Please note that the pre-registration was put in place after data collection had been completed but before data analysis began.) However, due to restrictions imposed by the Coronavirus Disease (COVID-19) pandemic, a minority of participants completed the study via an online survey hosted by Qualtrics. Digit ratio (2D:4D) was also measured for those participants that attended the lab, and results from that aspect of the study have been published elsewhere
We examined intercorrelations between the outcome measures via Pearson’s correlations, and tested for sex differences in the autism-related variables by using independent samples
We conducted variable-centred and couple-centred analyses
For the couple-centred analysis we calculated sex-standardised difference scores for autism-related variables for actual couples and used paired samples
Although not specified in our pre-registration document, we conducted a meta-analysis to provide a more precise effect size estimate for the within-couple correlation for autistic traits. We performed a random effects meta-analysis using the R package metafor
We considered
The authors thank each of the participants for making the research possible, as well as Robin Dunbar who provided feedback on an early version of the manuscript. Preliminary findings were presented by Richards et al.
G.R., J.G., and S.B.-C. designed the study; B.M., J.D., L.G., G.R., and J.G. collected the research data; G.R. and V.W. performed data analysis; G.R. wrote the first draft, and all authors read and approved the manuscript prior to submission.
G.R. was supported by the School of Psychology, Newcastle University, and S.B.-C. received funding from the Wellcome Trust 214322\Z\18\Z. For the purpose of Open Access, the authors have applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. Further to this S.B.-C. received funding from Innovative Medicines Initiative 2 Joint Undertaking (JU) under Grant Agreement No. 777394. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA and AUTISM SPEAKS, Autistica, SFARI. S.B.-C. also received funding from the Autism Research Trust, Autistica, the MRC and the NIHR Cambridge Biomedical Research Centre. The research was supported by the National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health Research and Care East of England at Cambridgeshire and Peterborough NHS Foundation Trust. V.W. was funded by the Autism Research Trust, the Wellcome Trust, Templeton World Charity Inc., and the NIHR Biomedical Research Centre in Cambridge, during the period of this work. The research was also supported by a Birmingham City University Small Development Grant (ML/ZM/GP; Ref: SDG 17-18-1.2) awarded to J.G. The views expressed are those of the authors, and not necessarily those of the NIHR, NHS, Department of Health and Social Care, or the other funders.
The datasets supporting the conclusions of this article are available in the Open Science Framework repository [
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.