Richard J E James, Indu Dubey, Danielle Smith, Danielle Ropar, Richard J Tunney
6 min
Abstract
Autistic traits are widely thought to operate along a continuum. A taxometric analysis of Adult Autism Spectrum Quotient data was conducted to test this assumption, finding little support but identifying a high severity taxon. To understand this further, latent class and latent profile models were estimated that indicated the presence of six distinct subtypes: one with little probability of endorsing any autistic traits, one engaging in ‘systemising’ behaviours, three groups endorsing multiple components of Wing and Gould’s autistic triad, and a group similar in size and profile to the taxon previously identified. These analyses suggest the AQ (and potentially by extension autistic traits) have a categorical structure. These findings have important implications for the analysis and interpretation of AQ data.
Alex: They identified six distinct subtypes. One group showed very few autistic traits overall. Another was characterized by what researchers call "systemising" — a strong drive to analyze and build systems, like rules, patterns, or mechanisms. And several others matched what's known as the classic "autistic triad": patterns in social interaction, communication, and repetitive or focused behaviors.
Sam: So the "autistic triad" is just shorthand for those three core areas that researchers have traditionally focused on?
Alex: Correct. And the key point is that these six groups weren't just different amounts of the same thing. The data pointed to a structure where people fell into specific, qualitative categories — each with its own distinct pattern of responses.
Sam: So the "spectrum" might actually be a collection of different profiles, and collapsing all of that into a single number might be obscuring what's really going on.
Alex: That is the implication the paper draws. It suggests that the way we currently analyze these questionnaires may need to be more nuanced — that a single total score might not capture the underlying structure the data is actually showing.
Sam: There's also something in the paper about what they call "nuisance covariance." What is that, and why does it matter here?
Alex: Good question. When you ask someone a long list of questions, some of their answers will naturally correlate with each other — not because of the trait you're measuring, but simply because the same person is answering all of them. That background noise is what researchers call nuisance covariance. It's a bit like static on a radio: it's not the signal you want, but it can interfere with what you're trying to hear.
Sam: And if you don't filter out that static, your analysis might detect a pattern — or a "category" — that isn't really there?
Alex: Exactly. The researchers accounted for this in their models, which is important for trusting the results. If you don't control for it, you risk finding structure in the data that's just an artifact of the measurement process itself.
Sam: Did factors like gender play into which group someone fell into?
Alex: The paper did find that certain categories had a higher proportion of male participants, but the effect was described as small. The structure of the groups was primarily defined by behavioral and cognitive patterns — not by gender alone.
Sam: So the bigger takeaway isn't really about demographics. It's about whether our fundamental model of autism — this idea of a single, linear spectrum — is actually the right one.
Alex: That's a fair summary of what the paper is arguing. The researchers aren't saying the spectrum model is wrong in every sense — but they are suggesting that treating a total score as a simple measure of "how autistic" someone is may flatten out real differences that matter for understanding and supporting people. The data, they argue, points to something more structured than a single line.
Sam: And that has real implications — for how clinicians interpret assessments, how researchers design studies, and potentially how support is tailored to individuals.
Alex: It does. The paper is careful not to overstate its conclusions — this is one study, using one questionnaire, and the authors acknowledge that further work is needed. But it raises a meaningful question about whether the tools we use are actually reflecting the complexity of what they're trying to measure. That's worth taking seriously.
Sam: It's a good reminder that the way we frame something — even the name "spectrum" — can shape what we think we're looking at.
Alex: It really can. Thanks for listening to ResearchPod.