Can you be “a little bit” autistic?

The trait evidence is strong, and it’s older than most people think

The dimensional picture isn’t a recent reframing by the neurodiversity movement. It came out of quantitative genetics two decades ago.

A twin study published in 2003 used the Social Responsiveness Scale on 788 twin pairs aged 7 to 15 from a large epidemiological sample. Autistic traits were continuously distributed and moderately to highly heritable. The authors’ conclusion is the one that still gets quoted: given the continuous distribution of these traits, it may be arbitrary where cut-offs are made between research designations of being affected and unaffected (Constantino and Todd, 2003).

A second study from the same group looked at families. Using parent and spouse report across 285 twin pairs and their parents, it found correlations for social impairment between parents and children, and between spouses, of around 0.4. Where both parents scored in the upper quartile for social impairment, their children’s mean scores were substantially elevated. The authors concluded that heritable subthreshold autistic traits are measurable in adults and appear continuously distributed in the general population (Constantino and Todd, 2005).

That family clustering is what researchers call the broader autism phenotype: relatives of autistic people showing autistic traits below the diagnostic threshold. It’s the closest thing in the literature to a technical version of “a little bit autistic,” and it’s a real, measured phenomenon.

Both studies rest on the Social Responsiveness Scale, a questionnaire completed by a parent or spouse, and both draw on the same Missouri twin sample. That’s a narrow evidential base for a claim this widely repeated, and rater bias is a live concern when one family member scores another. The finding has held up in later work, but the caveat belongs with it.

Is autism the extreme end of that continuum?

This is the more interesting question, and the evidence points one way while stopping short of settling it.

The largest twin study of autistic traits at the time compared the causes of typical variation with those in extreme-scoring groups, using 5,968 twin pairs aged 12 from a nationally representative English sample. Heritability was moderate to high in the general population, 53% for females and 72% for males, and high in the extreme-scoring groups. There were no differences in heritability between the top 5%, 2.5%, and 1% groups, or between those groups and the general population. Co-twins of individuals scoring in the top 1% showed a continuous shift toward autistic traits (Robinson et al., 2011).

The authors concluded this supports conceptualising autism spectrum disorders as the quantitative extreme of a neurodevelopmental continuum.

A later Swedish study tested the same assumption with a different method, combining twin modelling with polygenic risk scores. Twin genetic correlations between psychiatric diagnoses and their matching population traits ranged from 0.31 to 0.69, and polygenic risk for autism was associated with autistic traits in the general population. Its conclusion was that many psychiatric disorders are likely continuous phenotypes and not the categorical entities diagnostic manuals define (Taylor et al., 2019).

Two caveats keep this from being the whole story. The autism association in that study was small, with a beta of 0.04 at age 9, considerably smaller than for ADHD traits. And all of these samples are children and adolescents assessed by questionnaire, mostly in Northern Europe, so they describe trait structure and not adult diagnosis.

Why a continuum still has a threshold

None of the above dissolves the diagnosis, and it’s worth being precise about why.

DSM-5-TR requires persistent differences in social communication and social interaction across multiple contexts, with all three listed features present; restricted or repetitive patterns of behaviour, interests, or activities, with at least two of four features; traits present in the early developmental period; and clinically significant impairment in social, occupational, or other important areas (American Psychiatric Association, 2022).

No single requirement does the separating on its own. A trait questionnaire measures how much of a characteristic someone reports; the criteria ask for a particular pattern across two domains, evidence that it was present in early development, and clinically significant impact. A high trait score satisfies none of those by itself, and impairment alone doesn’t either.

Impairment is the requirement people most often overlook, which is why it gets emphasised. Two people with similar trait scores can have different lives depending on environment, support, and what their circumstances demand. But a page that presents impairment as the only difference is describing the criteria inaccurately.

That’s also where the criterion is weakest. A 2025 analysis in the British Journal of Psychiatry argues that although impairment is required, diagnostic manuals give limited guidance on how to interpret it, which affects diagnostic rates and access to support. It notes that impairment often reflects a mismatch between a person and a non-autistic environment, so better accommodation could reduce measured impairment, and the same person might meet criteria in one context and not another (Hollingdale, Woodhouse and Deeley, 2025). That’s a short analysis piece calling for operational definitions, not a study producing a finding.

The upshot is uncomfortable and worth stating plainly: the threshold is real, it does necessary work, and where exactly it falls depends partly on the life someone is living.

This is the point most versions of the argument miss in one direction or the other. Treating the continuum as proof that the diagnosis is arbitrary ignores that impairment is a separate measurement from trait intensity. Treating the diagnosis as a clean biological category ignores twenty years of genetics. Both things hold at once, and a page that picks one is easier to read and less accurate.

What “subthreshold” means in practice

People sitting below the line aren’t a theoretical category. They turn up in clinics.

A narrative review written for frontline clinicians describes subthreshold autism and ADHD as a practical problem: people who don’t meet full criteria can still have needs that warrant support, and services are generally organised around diagnostic categories that they fall outside (Ogundele and Morton, 2025). It’s a brief narrative review, not a systematic one, so read it as a description of clinical opinion.

Three practical consequences follow.

Support often tracks the diagnosis more than the difficulty. How far that holds varies by service, country, and funding route, so it’s a general pattern and not a rule. Where it does hold, it’s an argument about how services are organised, not about whether the threshold is meaningful.

Traits below the line can still be worth naming. Understanding why social situations cost what they cost doesn’t require a diagnosis, and neither does adjusting your environment. Naming a trait is not the same as claiming a diagnosis, and the distinction is worth keeping visible in how you describe it.

Subthreshold may not be a permanent verdict. Because clinically significant impact is assessed against someone’s current circumstances, an assessment during a well-supported stretch may reach a different conclusion than one during a demanding period. That’s an interpretation of how the impact criterion works, not a directly measured finding, and it isn’t a route around an assessment.

None of this is a route around an assessment, and it shouldn’t be read as one. A clinician weighing developmental history and current functioning is better placed to judge where someone sits than a trait score or a self-estimate, and the fact that a threshold has fuzzy edges doesn’t mean anyone can position themselves on either side of it at will.

Where the phrase causes trouble

The friction over “a little bit autistic” is usually about who’s saying it and about whom.

Used about yourself to describe traits you recognise, it’s an informal shorthand and not a clinical statement, and the research above supports the underlying observation that traits vary. Plenty of people use it that way without any intent to claim a diagnosis, and the objection they meet is often aimed at a different use of the same words. “I have some autistic traits” says the same thing without borrowing a diagnostic term. Used about a diagnosed person to suggest they’re only mildly affected, it collides with the reason the threshold exists: the diagnosis marks impairment, so “a little bit” contradicts the thing being described.

There’s also the practical objection. Autism isn’t graded in the way the phrase implies. DSM-5-TR uses support-level specifiers instead of a severity dial, and support needs shift across settings and across a life, so a single position on a scale doesn’t describe anyone well for long.

If you recognise yourself in the trait description and you’re unsure whether it adds up to more, the question worth asking isn’t how autistic you are. It’s whether the traits are costing you enough to warrant assessment. Our guide to neurodivergent traits covers what these patterns look like day to day, what happens in an adult autism assessment covers the process, and whether assessment is worth pursuing is the decision that comes first. The NeuroDiversion AuDHD self-reflection is a structured way to gather your own examples. A trait score can describe your responses to a questionnaire. It can’t tell you by itself whether the full developmental criteria for autism are met.