When an autistic adult with higher support needs refuses to move on to the next activity, the interpretation applied often comes down to control. The person will not stop, will not switch, will not comply. Framed that way, the difficulty belongs to inhibition, and the response is behavioural management. A study published in July in the journal Autism gives reason to think that frame is aimed at the wrong target. Across 486 autistic adults, difficulties with flexibility and with emotion regulation were significantly associated with symptoms of anxiety and depression. Difficulties with inhibitory control were not.

The work comes from Ethan Greenstein, Nancy Raitano Lee, Goldie McQuaid and Gregory Wallace, and it addresses a gap the authors state directly. The relationship between executive function and co-occurring depression and anxiety is well documented in autistic people, but that documentation is built overwhelmingly on children, and on people without intellectual disability. Autistic adults with higher support needs have been left out of it.

The Population Research Keeps Skipping

This exclusion is not incidental and it is not new. Autistic adults with intellectual disability are harder to recruit, harder to assess with standard instruments, and frequently unable to complete the self-report questionnaires on which most mental health research depends. The practical result is a literature that describes a portion of the autistic population and gets cited as though it described all of it.

The consequences are concrete. Anxiety and depression are common in this group. A representative study conducted across one Norwegian county documented psychiatric disorders in adolescents and adults with both autism and intellectual disability, and the rates were not low. But if the instruments that detect distress require a person to describe their internal states in writing, then people who cannot do that go uncounted, and their distress is likely to be interpreted as behaviour instead.

Recruitment for this study went through SPARK, the Simons Powering Autism Research for Knowledge cohort, using its Research Match service. That produced 486 participants aged 18 to 68, with a mean age of 31. For a population this consistently under-studied, that is a substantial sample.

What Was Measured, and by Whom

Caregivers, not the autistic adults themselves, supplied every measure in the study. That is the study’s defining methodological feature and it shapes what can be concluded, so it is worth being precise about what was collected.

Caregivers completed the Flexibility Scale, which asks about difficulty adapting to change. They completed two subscales of the Barkley Deficits in Executive Functioning Scale, covering inhibitory control and emotion regulation. And they completed the anxiety and depression subscales of the Anxiety, Depression, and Mood Scale, an instrument developed specifically because standard mood measures do not work well for people with intellectual disability. It asks an informant about observable indicators rather than asking the person to introspect.

The analysis used hierarchical linear regressions controlling for age, sex assigned at birth, likely cognitive impairment, and caregiver educational attainment. That last control is a thoughtful inclusion, since caregiver education plausibly affects how questionnaire items are interpreted and how readily particular language is endorsed.

Why Detecting Depression Here Is Genuinely Hard

The choice of the Anxiety, Depression, and Mood Scale deserves more than a passing mention, because it points at a problem this study is partly a response to.

Standard depression and anxiety measures ask a person to report on their inner life. Have you felt hopeless. Do you worry more than you would like. Have you lost interest in things you used to enjoy. Answering requires noticing an internal state, comparing it against a remembered baseline, matching it to a written description, and selecting a frequency. That is a demanding sequence, and for many autistic adults with intellectual disability it is not available.

Instruments built for this population work differently. They ask an informant about observable indicators rather than asking the person to introspect: changes in sleep, appetite, activity level, agitation, withdrawal, or interest in things that previously mattered. The trade is deliberate. You lose direct access to the person’s experience and gain a measure that can actually be completed.

The consequence is that distress in this population is easy to miss and easy to misattribute. A person who becomes withdrawn, or agitated, or stops engaging with a favourite activity is showing something. Whether that something gets recorded as a mood change or as a behaviour problem depends heavily on who is looking and what they are looking for. When distress is filed as behaviour, the response is a behaviour plan, and the underlying state goes unaddressed.

That is the backdrop against which the study’s central finding should be read.

The Null Result Is the Practical Finding

Two significant associations and one null. It is the null that has the most direct bearing on how people in this group are treated day to day.

Inhibitory control is the executive domain most readily invoked when someone does something they were asked not to do, or continues something they were asked to stop. It is the vocabulary of impulse, and it sits close to the vocabulary of compliance. A person described as having poor inhibitory control is a short step from being described as unwilling.

In this sample, inhibitory control was the one domain that did not track distress. Flexibility did, and emotion regulation did. Those describe something quite different: difficulty when circumstances change, and difficulty managing the feeling that arrives when they do.

Read that way, the same observable behaviour supports two very different accounts. A person who will not leave the room when asked may be failing to inhibit a preferred activity. Or they may be facing an unplanned transition they cannot easily accommodate, experiencing distress they cannot regulate, and expressing it in the only available form. The behaviour is identical from the outside. The appropriate response is not.

This study cannot say which account is correct in any individual case. What it can say is that among these 486 adults, distress tracked the second pattern and not the first.

Three Components, Not One Capacity

The design choice that made this finding possible was measuring executive function as separate components rather than as a single score, and it is worth pausing on how easily the result could have been missed.

Many instruments produce a global executive composite, a single number summarising performance across domains. If this team had used one, they would have had a sample in which overall executive difficulty was associated with anxiety and depression, which is true, unsurprising, and considerably less useful. The two significant associations would have been averaged together with the null one, and the distinction between them would have vanished into the composite.

What emerged instead is a specific pattern. Two domains moved with distress and one did not, and it happens to be the one whose everyday interpretation carries the most weight. A global score would have supported the conclusion that executive difficulty and distress go together. The component analysis supports a conclusion about which difficulties, which is the version a practitioner can act on.

This mirrors what a separate 2026 study of autistic adults found when measuring planning, inhibition, and generativity separately: performance differed across domains rather than uniformly, and the authors argued that executive functions must be differentiated because an autistic person can show full capability in some and difficulty in others. Two studies, different measures and different populations, arriving at the same methodological point.

Which Way Does It Run

The design is cross-sectional, so direction is genuinely open, and the honest reading holds several possibilities at once.

Difficulty adapting to change could produce anxiety. If unpredictability is reliably distressing and the world is unpredictable, anxiety is a reasonable consequence, and repeated experiences of being overwhelmed by ordinary change would plausibly accumulate into low mood.

Anxiety could produce rigidity. Anxious people narrow their range, prefer the known, and resist the unfamiliar, and this is well described across clinical populations. Insistence on sameness may in some cases be an anxiety response rather than an independent trait.

Both could be caused by something else, or could reinforce each other in a loop. Emotion regulation difficulty in particular is not cleanly separable from anxiety and depression as constructs. A meta-analytic review by Aldao, Nolen-Hoeksema and Schweizer established that emotion regulation strategies relate to psychopathology across diagnostic categories, and some overlap between a measure of emotion regulation difficulty and a measure of mood symptoms is expected rather than surprising.

The authors are appropriately restrained about this. They describe flexibility and emotion regulation as key correlates and as potential targets for future mechanistic and intervention research. Correlates, potential targets, and future research: that is the correct weight for what a cross-sectional design supports.

The Single-Informant Problem

Every variable in this study came from the same person, and that constrains interpretation in a specific way that is worth naming plainly.

A caregiver rated the executive function difficulties and the same caregiver rated the anxiety and depression. So what the study establishes is that a caregiver’s report of flexibility difficulty predicts that same caregiver’s report of distress. That could reflect a real pattern the caregiver is exceptionally well positioned to observe, since they may have known the person for decades and see them in every setting. It could also partly reflect a general impression of how the person is doing that colours every item.

This is the same structural constraint that appears throughout executive function research. It limits a study of preschoolers where parents rate both the executive function and the behavioural outcomes, and it limits a study of adult ADHD evaluations where the same person supplies both the rating scale and the impairment report. It is a feature of informant-based research rather than an oversight by any of these teams.

In this population it is also difficult to avoid. Independent verification would require either self-report from adults who often cannot provide it, or direct behavioural observation across enough settings and enough time to be meaningful, which is expensive and intrusive. Caregiver report is not a shortcut chosen for convenience here. It is frequently the only accessible window.

That is worth holding alongside a further point. Autistic adults’ own accounts of mental health services, synthesised in recent qualitative work, describe experiences that differ substantially from how those services describe themselves. Research about a group that never includes that group’s own voice carries a limitation no statistical control addresses. For adults who communicate in ways standard research instruments cannot capture, solving that is a methodological problem the field has not yet solved.

What the Study Cannot Support

Several claims do not follow from this paper and should not be attached to it.

It does not show that flexibility training reduces anxiety or depression. No intervention was delivered and no one was followed over time. The authors identify these domains as potential targets for intervention research, which is a proposal for future work rather than a result.

It does not show that inhibitory control is irrelevant to the lives of autistic adults with higher support needs. It shows that in this sample, with these measures, inhibitory control difficulty was not significantly associated with these two symptom clusters. A null result on one relationship is not a finding of general unimportance.

It does not generalise straightforwardly to all autistic adults with higher support needs. SPARK participants are families who enrolled in a research cohort and responded to a match service, which is a self-selected group. Cognitive impairment was estimated rather than formally assessed, described in the paper as likely cognitive impairment.

What Follows in Practice

The practical implication is a discipline of interpretation rather than a technique.

When an autistic adult with higher support needs resists, refuses, or escalates, several distinct things could be happening, and they call for different responses. There may be genuine difficulty adapting to a change in plan. There may be anxiety or low mood that has reached the point of needing clinical attention. There may be sensory or environmental overload with nothing to do with executive function at all. There may be a straightforward and legitimate preference not to do the thing being asked, which every adult is entitled to.

Those four are not distinguishable from the behaviour alone. Distinguishing them requires knowing what preceded the moment, whether the pattern is specific to transitions, whether it has changed recently, and what the environment was doing at the time. This study’s contribution is evidence that at least two of those four possibilities are statistically connected to distress in this population, which is an argument for checking them before defaulting to the compliance frame.

There is also a referral point embedded here. Anxiety and depression in adults with intellectual disability are frequently missed, partly because the instruments that detect them require self-report, and partly because distress in this group may present as behaviour that gets addressed behaviourally. A study finding that executive difficulties track internalising symptoms is, among other things, a reminder that internalising symptoms are present and often undetected. Where they are suspected, that is a matter for a clinician, and the fact that a person cannot easily describe their own mood does not mean their mood should go unexamined.

The broader argument the paper makes, and makes well, is that executive function should not be treated as a single faculty in this population any more than in any other. Three components were measured, two behaved one way and one behaved differently. Collapsing them into a global impression of executive difficulty would have hidden that entirely, and the distinction it reveals is precisely the one that matters for deciding what to do.