The short answer
Once you get past the headlines and read the actual paper, the story is much quieter.
A 2026 study looked at 796 adults in Poland. It used the AQ, a questionnaire where people rate their own autistic-like traits. It then compared those scores with the things people do to keep a partner, and with their attachment style.
On the simple numbers, a higher AQ score went with slightly fewer "give your partner something good" behaviors (correlation −0.11). It also went with slightly more "put a cost on your partner" behaviors (+0.17). The links between AQ and attachment anxiety (+0.28) and attachment avoidance (+0.33) were bigger.
But the picture changes when the analysis model changes or when age is taken into account. The authors also admit that the AQ does not measure only autistic traits. It overlaps with social isolation, anxiety, showing little emotion, and trouble communicating.
So the takeaway is not "autistic people are selfish and don't support their partners." The AQ includes questions about trouble with social contact, and the study compares it with social behavior in love. Some link between the two is almost expected. You have to read the result with that in mind.
What did the researchers actually do?
The group was 796 adults in Poland who had been in a romantic relationship for at least three months. The average age was 36.4, and 52% were women.
They used four main kinds of questionnaires:
- AQ: 50 questions about autistic-like traits.
- A short form of the ECR-R: 18 questions about attachment anxiety and attachment avoidance.
- MRI-SF: 38 questions about what people do to keep a partner in the relationship.
- Two short scales about effort put into short-term dating and effort put into caring for children.
The most important point is that this was not a study of people diagnosed with autism.
No doctor sorted the participants into "autistic" and "not autistic." The study simply looked at whether AQ scores were high or low among ordinary people.
What does "autistic traits" actually measure?
The AQ has 50 questions. They cover five areas: social skills, communication, switching attention, imagination, and attention to detail.
Example questions ask whether you prefer doing things with other people rather than alone, whether you like doing things the same way every time, and whether it is easy to picture the characters in a story.
So a high AQ score is not a medical statement that "this person is autistic."
If you convert it carelessly, you get this:
High AQ
→ autism
→ only thinks about themselves
→ doesn't value their partner
That is a four-step leap, and the study proves none of it.
It is the one kind of leap a science article should never make.
There are two kinds of "keeping a partner" behavior
The researchers split relationship-keeping behavior into two groups.
Giving benefits means doing good things for your partner: showing love, giving gifts, making them happy, making them more satisfied with the relationship.
Imposing costs means doing things that hurt or pressure your partner: watching what they do, controlling them, manipulating their feelings, or using jealousy and worry.
Something like "testing" a partner is close to the second group. Acting cold to see whether they come after you, hinting at a breakup to see how they react, or trying to make them jealous all fit.
But "imposing costs" and "testing a partner" are not the same thing. The first group is much wider.
The numbers are less dramatic than the headline
Here are the simple correlations:
| What was compared | Correlation |
|---|---|
| AQ and attachment anxiety | +0.28 |
| AQ and attachment avoidance | +0.33 |
| AQ and imposing costs | +0.17 |
| AQ and giving benefits | −0.11 |
−0.11 is quite small.
If you square the correlation, the shared part is only about 1%. A rank correlation like this is not really a measure of how much one thing "explains" the other. Still, these are far from numbers where AQ alone decides how someone behaves in love.
The results also shift depending on the method. In the structural model, AQ had a negative direct effect on giving benefits. For imposing costs, the simple correlation was positive, but in the model the direct effect was treated differently. When age was added, a positive link appeared again.
In short, it is not true that high autistic traits always lead to bad relationship behavior, and it is not true that they never do. The result changes with the model, so it is more complicated than that.
The most interesting part: "aren't these two things measuring something similar?"
This is the main point.
The AQ has questions that pick up trouble with social skills and communication.
On the love side, the study measures things like showing affection, pleasing a partner, and actively working to keep the relationship going. These are also social behaviors.
So part of the study compares:
a scale that is likely to pick up people who find social contact hard
with
how much a person reaches out socially to their partner.
In an exaggerated version, it sounds like this:
"We picked out people who have little social contact using a questionnaire."
↓
"These people showed less social behavior!"
That would not be a discovery. It would be "well, of course."
This study is not completely circular. The AQ also includes a taste for routine, switching attention, and attention to detail, which are not the same as communication in love.
Even so, the authors themselves say the AQ overlaps a lot with social isolation, anxiety, showing little emotion, and trouble communicating. They say the link with the AQ should be read with care.
On top of that, studies disagree on how the AQ is built. Some find five factors, some three, some six, and some two big ones: social interaction and attention to detail. It is not one clean block.
So if you throw everything into one big pot called "total AQ score" and label it "autistic traits," it becomes hard to see how much the social items are doing the work.
This is not a study about "people who only think of themselves"
This part matters too.
In real life, a pattern like this can happen:
They think things through alone
→ they don't share it with their partner
→ they don't ask for their partner's reaction
→ they don't update what their partner needs
→ they reach a conclusion on their own
→ one day they quietly pull away
The back-and-forth of information between two people is weak.
For the partner, it can feel like "I'm giving so much and getting nothing back."
But this study did not measure why someone does this.
Is it that they don't care about the partner? That they don't notice things easily? That closeness feels heavy? That they feel love but don't show it? That they don't think relationship upkeep is needed at all? This data cannot tell us.
So reading "autistic traits = selfish" from these results is a stretch.
Attachment is a clearer story
The links with AQ were stronger for attachment anxiety (+0.28) and attachment avoidance (+0.33) than for giving benefits (−0.11).
Anxiety and avoidance also push behavior in opposite directions.
People high in anxiety are afraid of being left. They may do more loving things and more costly, controlling things.
People high in avoidance keep distance from closeness, so they tend to give fewer benefits.
So it is closer to what the paper says to read it like this:
traits picked up by the AQ → several paths, such as attachment style → behavior in love
rather than a single straight line from "autistic traits" to "love behavior."
To find out what is truly specific to autistic traits, we need another study
A stronger study would not use only the total AQ score. It would split it up and look at each part:
- social skills and communication
- attention to detail
- liking routine and predictability
- imagination, and understanding what others think
Suppose only "social communication" predicts behavior in love. Then the result is mostly "well, of course."
But suppose that even after controlling for the social questions, things like a focus on detail or a liking for routine still predict behavior in love. Then, for the first time, it becomes interesting: a link tied to autistic traits themselves, beyond just social difficulty.
We would also want reports from partners, real recordings of conversations, studies that follow people over time, a comparison with people who have an autism diagnosis, and repeats in other countries.
Conclusion: don't say "autistic, so no affection." Look at what is being measured
What we can say from this study: in the general population, AQ scores had some links with relationship-keeping behavior, and especially with attachment.
What we cannot say:
"Autistic people only think about themselves."
"Autistic people don't support their partners."
"A high AQ means they test their partners."
"Autism means no affection."
These are leaps about character and cause.
The cleanest summary is this:
People who scored high on a scale that partly captures social-communication difficulty also behaved a little differently in love. But the difference was small, and the story changes with attachment, age, overlap between scales, and the way the data is analyzed.
...Yes.
After climbing over a mountain of ads, the final boss was not "autism."
It was construct validity: whether the scale really measures what it says it does.
References (4)
- Łukasik A, Łaba-Hornecka A. Autistic-Like Traits in Members of the General Population and Mate Retention Tactics: The Mediation of Attachment Styles and Reproductive Strategies. Evolutionary Psychology. First published online September 3, 2026. https://doi.org/10.1177/14747049261473640
- Baron-Cohen S, Wheelwright S, Skinner R, Martin J, Clubley E. The Autism-Spectrum Quotient (AQ). Journal of Autism and Developmental Disorders. 2001. https://doi.org/10.1023/A:1005653411471
- English MCW, Gignac GE, Visser TAW, Whitehouse AJO, Maybery MT. A comprehensive psychometric analysis of Autism-Spectrum Quotient factor models. Autism Research. 2020. https://doi.org/10.1002/aur.2198
- Murray AL, Booth T, McKenzie K, Kuenssberg R. What range of trait levels can the Autism-Spectrum Quotient measure reliably? Psychological Assessment. 2016. https://doi.org/10.1037/pas0000215
