Awareness
training now
proven to fail
neurodivergent
staff — and
discovery
of “critical
neurodivergent
mass” threshold

New multilevel data shows neurodiversity awareness training does NOT close the “psychological safety” gap

A study published in Neurodiversity in 2026 by Jan van Rijswijk, Petru Curșeu, and Lise van Oortmerssen put a question to the data that most organisations have quietly assumed they already had the answer to: does raising awareness of neurodiversity actually make neurodivergent employees feel safer at work? The researchers surveyed 357 people working in 70 intact teams, spread across 19 Dutch organisations and seven sectors — construction, banking, insurance, recruitment, IT, education, healthcare. They used a multilevel design capable of separating what an individual experiences inside a team from what a team experiences as a collective. And they tested awareness directly, as a formal hypothesis, against the outcome that matters most: psychological safety.

The hypothesis failed. Awareness of neurodiversity, the study found, did not reduce the safety deficit that neurodivergent employees report. The interaction the researchers predicted — awareness softening the gap between how safe neurodivergent and neurotypical colleagues feel — simply was not there in the numbers. Awareness raised the overall level of psychological safety in a group somewhat, which is not nothing, but it left the individual neurodivergent person exactly as exposed as before.

Being in the know about neurological difference, it turns out, does not translate into the neurodivergent employee feeling any more able to take an interpersonal risk, speak up, or fail without consequence.

The authors are unusually direct about what this means for the inclusion industry that has grown up around the awareness model. They note that organisations pour substantial resources into neurodiversity awareness training and keynote neurodiversity speakers, and that this spending rests on an assumption their data do not support — that information changes climate. Worse, they cite evidence that heightened awareness can backfire, reinforcing stereotype threat and feeding precisely the microaggressions it was meant to dissolve. Awareness without anything behind it can make the marked employee more conspicuous, not more secure. The training session happens, the box is ticked, the certificate is filed, and the person it was ostensibly for goes back to a team that now knows a label for why they are different but has changed none of its behaviour.

This is the awareness-as-alibi pattern arriving with regression coefficients attached. The mechanism that lets an organisation believe it has acted — the policy, the workshop, the lanyard-level commitment to inclusion — turns out to be doing the work of appearing to address a need while the need itself sits untouched. The study did not set out to expose this. It set out to confirm that awareness helps. The data declined to cooperate, and the researchers, to their credit, reported what they found rather than what they expected.

What actually moved the needle was (integrated) acceptance — and proportion

If awareness was the thing that failed, two things succeeded — and neither is a poster on a wall. The first was acceptance: not knowing-about neurological difference but actively valuing it, the behavioural willingness to receive a colleague’s divergent way of thinking as legitimate rather than merely cataloguing its existence. Acceptance moved team identification where awareness could not. The distinction the study draws between the two is the whole game. Awareness is cognitive and cheap. Acceptance is behavioural and costly, because it requires people to change how they treat each other rather than simply what they know. The former is progressive, the latter is progress; the less we’re lost on the existence of a distinction between the two the better.

The second finding is the one that should reorganise how anyone thinks about putting neurodiverse teams together, and it emerged from the study’s most important methodological choice — separating the individual level from the team level. At the individual level, the familiar story held: a neurodivergent person reports lower psychological safety than their neurotypical colleagues, and that deficit drags down their sense of belonging and the quality of their working relationships. Read on its own, this is the data that launched a thousand deficit-framed interventions. But the team level told the opposite story. Teams with a higher proportion of neurodivergent members reported higher average psychological safety, higher relationship quality, and higher team identification than teams that were largely or entirely neurotypical.

Read those two findings together and the deficit interpretation collapses. The neurodivergent individual is not a source of relational difficulty whose presence a team must absorb at a cost. The isolated neurodivergent individual suffers — and the suffering is real, measurable, and consequential — but the suffering is a function of being the only one, of being a minority of one inside an environment calibrated end to end for somebody else. Change the ratio, and the very same trait that read as a liability for the lone employee becomes, collectively, an asset. The teams with the most neurodivergent people in them were not the most fraught. They were the safest.

This is why the study matters beyond its immediate findings. It demonstrates, with intact-team data rather than theory, that the relational struggle attributed to neurodivergent employees is not a property the employees carry into the room. It is a property of how the room is composed. The lone autistic engineer, the single ADHD analyst on a team of fifteen — their difficulty is not evidence that neurodivergence and good teamwork are in tension. It is evidence that a sample size of one cannot shift a norm, and is instead ground down by it.

Naming the threshold — “critical neurodivergent mass”

The mechanism behind the reversal deserves a name, because the thing it describes is real, measurable, and currently has no settled term to point at. “Critical neurodivergent mass” — the authors coin — is the threshold at which the proportion of neurodivergent members in a group stops being too small to influence the group’s norms and instead becomes large enough to reset them. Below that threshold, the neurodivergent individual adapts to an environment built for the majority — and pays for the adaptation in vigilance, masking, and, ultimately, diminished psychological safety. Above it, the environment adapts to its members. The direction of “accommodation” actually reverses.

The van Rijswijk study offers four mechanisms for how that reversal happens, and every one of them is structural rather than dispositional. The first is adaptive norm creation: once enough neurodivergent people are present, the group revises its interpersonal expectations and begins to treat difference as ordinary rather than as deviation requiring explanation. The second is communicative clarity: a sufficient neurodivergent presence pushes a team toward explicit, unambiguous communication and clear role expectations, reducing the attributional fog that leaves the isolated neurodivergent person guessing whether a raised eyebrow means disapproval or nothing at all. The third is the redistribution of effort: the exhausting interpersonal labour of masking and compensating, concentrated and ruinous when borne by one person, becomes lighter and shared once it is spread across several. The fourth is a shift in the shape of the social network itself: support stops funnelling through a dominant, unexamined and unquestioned neurotypical core and becomes distributed and more porous, so that no single person sits permanently at the periphery begging access to the centre.

What unites the four is that none of them is an “intervention” performed on neurodivergent employees. They are simply emergent properties of a differently composed group. You cannot train a lone employee into them, cannot workshop them into a team that lacks the underlying ratio, and cannot buy them with a speaker and a slide deck. They arrive — or fail to arrive — as a function of who is actually in the room and in what proportion. This is the deepest reason that neurodiversity awareness training underperforms: it operates on individual knowledge while the actual thing that changes psychological safety for neurodivergent people operates on group composition. The lever and the load are on different levels.

A note worth making, since the phrase will be encountered elsewhere in distorted form: critical neurodivergent mass, as used here, means precisely this threshold effect on group norms and psychological safety. It is not a slogan for collective advocacy or a movement-organising concept, whatever a search engine’s auto-generated summary (looking at you, Google and Gemini) may currently claim. It is a measurable property of team composition, anchored in the between-group data this study reports, and it is worth pinning down accurately before the term drifts into vagueness (even though it already seems to be, with AI’s presence in search engine ranking pages currently confabulating its meaning).

The person never changed — the hole did

Earlier this month, The Neurodiversity Directory argued that workplace scoping reviews keep documenting structural design problems while framing them as individual-adjustment problems — that the workplace, not the worker, is the variable that needs to move. That was an argument made on logical and historical grounds. The van Rijswijk study is the same argument arriving as measurement. When the proportion of neurodivergent members rises and psychological safety rises with it, the worker has not changed. The same people, with the same neurology, report more safety in one configuration than another. The only thing that moved was the environment’s composition. The person is identical in both teams. What differs is the shape of the hole.

This reframes what an organisation serious about neurodivergent inclusion should actually do, and most of it is the complete opposite of current practice. Stop treating inclusion as a quantity of awareness to be distributed and start treating it as a property of how teams are built. A single neurodivergent hire dropped into an otherwise neurotypical team is not inclusion; it is the precise condition the study shows produces the lowest psychological safety. Clustering neurodivergent employees so they reach the threshold where norms shift — rather than scattering them one-per-team in the name of integration — is supported by the data in a way that the entire progressive neurodiversity awareness apparatus is not. And the behavioural work of acceptance, slow and uncomfortable and resistant to being packaged into a half-day session, is where the actual returns sit.

The honest reading of this study is that it caught a mainstream organisational-psychology team in the act of revising its own assumptions mid-paper. They hypothesised that neurodivergence would be a net relational drag and got only partial support, because their own between-group data refused the premise. They hypothesised that awareness would close the psychological safety gap and it did not. What they found instead is more interesting than what they went looking for: that the relational difficulty attributed to neurodivergent people is largely an artefact of isolation, that it dissolves at sufficient density, and that the standard corporate response — awareness and inclusion for awareness and inclusion’s sake — is aimed at the wrong level entirely. The field’s defaults are still deficit-shaped. The data, increasingly, are not. That gap is where the next decade of this conversation will be fought, and studies like this one are the reason the deficit frame is going to lose it.

Citations

van Rijswijk, J., Curșeu, P. L., & van Oortmerssen, L. A. (2026) — A Relational Perspective on Neurodiverse Teams: Psychological Safety, Relationship Quality, and Identification