A new peer-reviewed study tests proactive cognitive control in adult ADHD and finds it intact
A study published in the Journal of Cognition in May 2026 by Anna Längle at UMIT Tirol, Ulrich Ettinger at the University of Bonn, and Stefan Duschek at UMIT Tirol set out to test something the ADHD cognitive-control literature has been struggling to pin down for years. Reactive control — the ability to interrupt an automatic response after a stop signal — is reliably reduced in adult ADHD; the stop-signal task meta-analyses have shown that for over a decade. Proactive control — the ability to use contextual cues to prepare for upcoming cognitive demands before a stimulus arrives — has been a much messier picture, with some studies showing impairment and others showing none. The present study used a manual face-word Stroop task with emotional stimuli and a context manipulation to test it directly.
The design is methodologically careful. Fifty-one adults meeting DSM-5 criteria for ADHD via the SCID-5 clinical interview, fifty-one healthy controls, drawn from an initial screening pool of over one thousand. The ADHD group was unmedicated, with no psychiatric comorbidities — a rare combination in cognitive ADHD research, and the cleaner signal for it. Participants saw images of happy or anxious faces with the word HAPPY or ANXIOUS written across them, congruent or incongruent with the expression, and had to identify the facial expression while ignoring the word. The context manipulation altered the proportion of incongruent trials: a mostly-congruent block (75% congruent) where automatic processing is fostered, and a mostly-incongruent block (75% incongruent) where the context cues participants to prepare for conflict. The logic is clean. If proactive control is impaired in ADHD, the ADHD group should fail to use the mostly-incongruent context to prepare, and their performance on incongruent trials in that context should suffer disproportionately.
But that is not what the data showed. The ADHD group did show the expected overall differences— higher error rates and longer reaction times across the board, reflecting the well-documented baseline impairments in attention and processing speed. But the Congruency × Context interaction, which is what indexes the use of contextual cues to apply proactive control, did not depend on group. Bayesian analysis confirmed this directly: the Bayes factor for the three-way Group × Congruency × Context model on error rate was less than .001, meaning the data provide strong evidence that group membership has almost no influence on the context effect. The ADHD adults used the contextual cues to prepare for cognitive demands just as well as the controls. The authors land it plainly — “the ability to use contextual cues to facilitate proactive cognitive control is preserved in adult ADHD, despite marked impairments in attention and processing speed.”
The dissociation the field has been blurring — strategy vs implementation
The executive function model of ADHD has been the dominant cognitive framework since Russell Barkley’s work in the 1990s (notably, I would say: ADHD and the Nature of Self-control). The model treats ADHD as a deficit in the brain’s top-down regulatory machinery — the prefrontal apparatus responsible for inhibiting prepotent responses, holding goals in mind, allocating attention, switching tasks, planning and preparing for what’s coming.
The clinical implications follow directly: if the strategic machinery is broken, the interventions should aim at the strategic machinery. Executive function coaching. Working memory training. Cognitive behavioural therapy targeting planning and preparation skills. The framing is so pervasive in clinical and educational practice that “executive function deficit/problems/management” has become almost synonymous with “ADHD” in both popular discourse and mainstream understanding.
The Längle data invite a different reading. The study isolates two things the dominant framing has been blurring together. There is the strategic layer — the system that uses available context to prepare for upcoming demands, that biases information processing toward a goal before the stimulus arrives. And there is the implementation layer — the system that actually executes the prepared response with the appropriate speed and accuracy when the stimulus arrives. The Längle results show clear dissociation. The strategic layer in adult ADHD is intact. The implementation layer is not. The brain is preparing correctly; it just runs slower and makes more errors at baseline when the prepared response gets executed.
This matters because the dominant framing has been treating reduced performance as evidence of failed strategy, when it might be evidence of something else entirely. If a child or adult with ADHD is slower and more error-prone on cognitive tasks, the executive-function reading attributes this to a failure of cognitive control. The corrective is to train the control system harder.
The Längle reading suggests something different: the control system is doing its job correctly, but it’s operating on a processing substrate that has different baseline parameters. Training the strategic layer harder does not change the substrate. It may even produce diminishing returns and accumulating frustration in people whose strategic machinery was already working fine.
This is my coherence-first frame arriving as cognitive neuroscience verification. The ADHD brain is not failing to plan, prepare, or strategically allocate attention. It is, though, running on different baseline processing parameters — slower, more variable, more error-prone in the moment-to-moment execution — while deploying the same higher-order regulatory functions as everyone else on the same externally-imposed demands as everyone else. The deficit, it turns out, is not where the executive-function story has been locating it all this time. In fact, it sits one layer down, in the substrate the executive control is operating on, and that distinction changes what adjustments should actually target.
Emotional information was the stimulus, and the result still held
The stimulus choice is worth pausing on. Most Stroop-and-ADHD studies have used colour-word stimuli — the classical Stroop paradigm where the word RED is printed in blue ink and participants name the ink colour. Längle and colleagues chose emotional Stroop materials specifically: happy and anxious facial expressions with HAPPY or ANXIOUS overlaid.
The choice reflects an existing literature on emotion dysregulation in adult ADHD that has been treating control over emotional information as a particularly weak point. Emotional lability, low frustration tolerance, irritability, maladaptive emotion regulation strategies — these features are documented across the adult ADHD literature, and the standard interpretation is that cognitive control over emotional content is selectively impaired.
The Längle data make that interpretation harder to hold from here on out. Genuinely, this is that significant — at least as far as I can see. The adults with ADHD showed preserved proactive control specifically on emotional stimuli. When the context cued them to prepare for the conflict between facial expression and word meaning, they prepared (“strategy”). The contextual cues facilitated their performance to the same degree as controls. The emotional content of the stimuli did not disrupt the proactive control machinery.
Ultimately, whatever is generating the emotional dysregulation observed clinically in adult ADHD, it is not a failure of cognitive control over emotional information in the strategic sense. The implication, then, is that emotional dysregulation in adult ADHD needs a different mechanistic account than the currently-running “broken or poorly-managed — even if, at the very least, compared to the typical — executive control over internal information.”
The dysregulation may sit, I conjecture, in the same place the rest of the “deficit” sits — at the implementation layer, in the baseline processing characteristics that govern speed, intensity, and recovery of emotional responses once they’re initiated. The strategic layer that decides when emotional regulation is needed and prepares the system for it is working fine. What’s reduced is the baseline machinery that executes the regulation in real time. That’s a different problem with different implications for what therapeutic or environmental adjustments should look like — but stating it that way still leaves something unsaid. The implementation substrate only registers as deficient because the emotional regulation demands placed on it were calibrated for a different substrate’s recovery characteristics. The Längle study makes the substrate visible; it does not make the demand structure visible. The demand structure is the missing variable, and naming it is where the rest of this piece goes.
The missing variable = externally-imposed demands
The Längle dissociation does necessary work but doesn’t yet do enough. To say “strategy is intact, substrate is different” is to leave one variable still treated as neutral — the demand structure the strategy is operating on and the substrate is being measured against. Throughout the study, throughout the cognitive control literature, throughout the clinical apparatus that sits downstream of both, the demands themselves are treated as the fixed reference point. Reaction time is reaction time. Error rate is error rate. The task is the task. The substrate is the variable, and the substrate is what either meets the task or doesn’t.
That is the assumption that needs surfacing. The demand structure is not the fixed-yet-invisible reference point, it’s the other variable. The Längle study uses a particular set of task parameters — stimulus durations, response keys, congruence ratios, time pressure — that produce particular cognitive demands. Those demands are not biologically given. They are conventions selected by the researchers from a much larger possibility space, and they reflect what the field implicitly treats as the relevant cognitive territory. Lengthen the stimulus duration. Remove the time pressure. Change the response modality. Different demand structure, different performance differential.
This is true beyond the laboratory. Every “deficit” in the ADHD literature is a measurement of substrate performance against a demand structure that the field has chosen to call the relevant one. The reaction time deficit is reaction time under particular pacing. The accuracy deficit is accuracy under particular precision requirements. The attention deficit is sustained attention to particular kinds of content under particular conditions (typically cybernetic ones). None of these demand structures are neutral. But all of them were selected. The substrate is being measured against a built thing, not against a natural reference.
Once the demand structure is named as variable, the “deficit” dissolves into mismatch. The substrate is not deficient relative to some neutral biological standard of cognition. It is mismatched against a specific demand structure that was calibrated for a different substrate. The mismatch is real. The substrate’s reduced output against that demand structure is real. The label “deficit” is not — it is what gets applied when one variable in the equation is treated as the fixed reference point and the other is treated as the test subject.
The Längle data make this visible because they dissociate two layers that the field has been treating as one. The strategic machinery doesn’t show a deficit, which means it isn’t being measured the same way. The implementation substrate shows a deficit, which means it is. The difference is not in the brain. The difference is in what each layer is being measured against. The strategic machinery is being measured against itself — does it use cues, does it prepare. The implementation substrate is being measured against the demand structure — does it execute at the expected speed, with the expected accuracy. The deficit appears at the layer where the externally-imposed demand structure is doing the measuring.
The round hole has a history
The demand structure didn’t fall from the sky. It has a history, and that history matters for understanding what the Längle dissociation actually reveals. Pre-modern economic life accommodated a much wider range of human cognitive profiles than what came afterwards, because the work itself was differentiated, locally calibrated, and largely free of standardised time discipline. Craftspeople worked on their own clocks. Agricultural labour followed seasonal and weather rhythms. Knowledge work was tied to specific contexts. This is not to romanticise the pre-modern — it had its own brutality, obviously — but the demand structure imposed on human cognition was looser, lumpier, and far less standardised than what was about to be constructed and that we worship religiously today.
Post-feudal market formation began compressing this. Goods needed to be made to consistent specifications to trade across distance. Labour needed to be priced and exchanged through abstract monetary units. The first standardisation pressure on human cognitive output began appearing here — workers needed to be sufficiently interchangeable that one could be hired in place of another, sufficiently predictable that delivery times could be planned, sufficiently legible that contracts could be enforced. The demand structure narrowed.
Mercantile capital extended the compression. Trade across distance and time required predictability at scale. Workers in one location had to produce outputs that aligned with workers in another location, on schedules set by people they would never meet. The demand structure narrowed further, and it began to be standardised across populations rather than within local communities. What had been a loose ecological match between cognitive profile and economic role was now a standardised compliance expectation imposed species-wide.
Post-industrial standardisation completed the move. The factory required human cognitive output to align with machine output: predictable cycle times, low variability, sustained attention to repetitive stimuli, compliance with externally-set pacing. The demand structure was now explicitly calibrated to the human-as-unit engaging with machine-as-unit. The school system was rebuilt to produce workers fit for this demand structure. The clinical apparatus was rebuilt to diagnose those who couldn’t meet it. The economic logic shaped the cognitive territory, and the cognitive territory got naturalised as just the way cognition works.
This is what I named in my book as the round hole. It has a shape, and the shape is the cumulative outcome of several centuries of economic pressure on what counts as functional human cognitive output. The Längle task, like every other cognitive control task in the literature, sits inside that demand structure. The participants are responding to time-pressed stimulus identification, fixed-duration windows, response-key precision, sustained sequence engagement — all of which are direct descendants of the industrial demand structure rendered into laboratory conditions. The “deficit” the task measures is the substrate’s reduced output against demands the substrate was never built to meet, and the demands themselves were never neutral measurements of cognition. They were the operationalisation of a specific economic-historical compliance expectation, dressed up as cognitive science.
Neurotypical names compliance, not biology
The term “neurotypical” entered widespread use in the late 1990s, originally as a neutral descriptor for non-autistic neurology. It has since broadened to refer to a notional default cognitive profile against which neurodivergent profiles are contrasted. The framing assumes the term names something biological — that there is a typical kind of brain, and other kinds of brains depart from it in specific ways. The assumption is wrong, and the Längle data are part of why.
“Neurotypical” does not name a kind of brain. It names a kind of compliance behaviour. The “typical” being referenced is the typical performance against the externally-imposed demand structure described above. The demand structure rewards certain compliance behaviours: predictable pacing and collaboration with externally-imposed demands (usually from and through “authority”), sustained attention to externally-set stimuli, low output variability, recovery curves that align with externally-set schedules, and emotional reactivity within ranges that don’t disrupt demanded or expected flow. A substrate that delivers these compliance behaviours is called “neurotypical.” A substrate that doesn’t is called “neurodivergent” — diagnosed, pathologised, treated, depending on the specific compliance failure and its cumulative profile.
The Längle dissociation makes this visible because it shows that the strategic machinery for cognitive control is the same across both populations. What differs is the substrate’s compliance output — the speed, accuracy, and consistency of execution against the demand structure. “Neurotypical” names the substrate whose compliance output matches the demand structure. “ADHD” names a substrate whose compliance output doesn’t. The strategic machinery is identical. The substrates are different. The demand structure is what’s measuring the difference. We’ve been using all of these names, and now you know what you’re naming.
This recasts the ADHD apathy and motivational dysregulation argument in sharper terms. Apathy and motivational dysregulation in ADHD are not pathology of motivation in any general sense. They are the substrate’s coherent response to demands that exceed its sustainable compliance capacity. The substrate is doing what coherent substrates do when asked to produce compliance behaviours they were not built for over sustained periods — it disengages, it withdraws internally, it produces the apathetic surface that the clinical apparatus then catches as another symptom of the underlying disorder. The withdrawal is not the disorder. The withdrawal is the substrate’s intelligent response to a demand structure it cannot sustainably comply with. The disorder is in the demand structure, but the demand structure has been naturalised, so the substrate becomes the only place left for the diagnosis to attach.
This is the move that the Längle data make available. Strategy is intact. Substrate is different. Compliance output against a built demand structure is reduced. The “deficit” was always a measure of compliance shortfall against a conventional demand structure, not a measure of broken cognitive machinery. “Neurotypical” was always a description of the compliance shape the demand structure expects, not a description of brains. The Längle dissociation is the place where this becomes unmistakable in the peer-reviewed cognitive neuroscience itself.
What this means for adjustments, training, and the executive function story
The clinical and educational pipeline built around the executive function framing has produced an entire industry of interventions aimed at training the strategic layer. Working memory training programmes, executive function coaching, cognitive remediation protocols. The meta-analytic evidence for transfer of these programmes to real-world functioning has been consistently underwhelming, and the field has tended to attribute the disappointing results to insufficient training intensity, wrong age windows, or insufficiently rich transfer environments. The Längle data make a different reading available. The training programmes are aimed at machinery that was never the source of the difference. The strategic layer doesn’t need training. The implementation substrate can’t be trained in the same way — it reflects baseline processing parameters that are not amenable to top-down instruction.
But the deeper point isn’t even that the training is aimed at the wrong layer. The deeper point is that the entire framing assumes the demand structure is fixed and the substrate is the variable that needs adjustment. Training the substrate to better comply with the demand structure leaves the demand structure unexamined. Even when the training fails, the demand structure stays in place as the reference point, and the failure gets attributed to insufficient training rather than to the demand structure itself. The industry produces three decades of disappointing transfer evidence, and the field reaches for “we need better training” rather than “we have been measuring the wrong variable.”
“Environmental adjustments”, then, are not compensation for a deficient substrate. They are recalibration of a demand structure that was never neutral. Longer deadlines for the same quality of work, reduced time pressure on tasks requiring precision, fewer simultaneous demands competing for limited processing bandwidth — these are not concessions to a struggling worker. They are recognition that the demand structure was calibrated for a substrate that isn’t the only one in the room, and that the calibration was conventional rather than biological. Renaming the move clarifies what it is. The adjustments are not corrections to the worker. They are corrections to the demand structure.
The connection to the cybernetic-attention work developed here on The Neurodiversity Directory lands here too, and harder than any previous framing conveyed. Cybernetic environments — high-density digital multitasking, attention-extractive interfaces in any attention-fragmenting systems, sustained cognitive load without recovery — do not create a new problem. They intensify the existing demand structure. The same standardisation logic that built the industrial demand structure is now extending into the cybernetic demand structure, and the cybernetic version demands compliance behaviours at much higher density and frequency than the industrial version did. The substrate that was already struggling to comply with the industrial demand structure is now being asked to comply with a demand structure that is harder still. The “ADHD epidemic” the field has been documenting is the (HUMAN) substrate’s response to an intensifying demand structure, not the emergence of new pathology.
The executive function frame has been useful for thirty years for locating ADHD in the brain. The Längle data don’t just relocate the deficit one layer down. They surface that the question “what is the deficit” was the wrong question all along. The substrate is different. The demands are conventional. The “deficit” is the artefact of measuring an unrecognised substrate against demands calibrated for a different one. The interventions and environments designed around the wrong premise have been doing the wrong work, for thirty years, on populations whose substrates were never broken.
The coherence-first reading and the empirical hinge
The coherence-first reading of neurodivergence holds that [what we call] neurodivergent cognition is coherent optimisation under environmental mismatch — not deficit nor inherent pathology. This reading has been doing structural and lived-experience work across my corpora for years now. But what it has not had, until now, is a place where it lands inside peer-reviewed cognitive neuroscience as the natural reading of the data rather than as projected-commentary on the data.
The Längle dissociation is that place. Strategy intact, substrate different, performance reduced relative to externally-imposed demands. Every term in that sentence is verifiable inside the study, with Bayesian and frequentist statistics, in an unmedicated sample diagnosed via clinical interview, published in a peer-reviewed journal. And every term, read together, says exactly what my coherence-first reading has been saying: the brain is not broken; the substrate is doing coherent (inherently human) work; the “deficit” is what appears when that coherent substrate engages an externally-imposed demand structure not optimised for that substrate.
This is the empirical hinge. The square peg has not changed. The round hole has a history, a built shape, and a name — “neurotypicality” — that describes the compliance behaviours the round hole expects rather than the brains that produce them. The Längle data show what is actually happening in the brain when a square peg encounters a round hole and is told the mismatch is its fault. The strategic machinery prepares for the demand. The implementation substrate executes against the demand. The output falls short of what the round hole was calibrated to receive. The shortfall gets diagnosed.
The second edition of The Neurodiversity Book will have to carry this. The first edition built the coherence-first frame from structural argument commentary and lived experience. The second edition gets to anchor it in cognitive neuroscience that the field cannot wave away. The Längle dissociation, paired with the demand-structure-as-variable argument, paired with the historical genealogy of the round hole, paired with the recognition that “neurotypical” names compliance rather than biology — that is the architecture of the next stage of the argument. The corpus arrives at the cognitive neuroscience layer, and the cognitive neuroscience layer turns out to be saying what the corpus has been saying all along.
The executive function frame held for thirty years because it gave the field a place to put ADHD inside the brain. The frame is now cracking, and the crack runs along the seam between strategy and implementation. What lies beyond the crack is not a better location for the deficit. It is the recognition that the deficit was a category mistake — produced by a demand structure that was never neutral, against substrates that were never broken, by a clinical apparatus that has been pathologising mismatch and calling it disorder for as long as the diagnostic category has existed.
Citations
Längle, A., Ettinger, U., & Duschek, S. (2026) — Proactive Control of Emotional Information in Adult ADHD
Senkowski, D., Ziegler, T., Singh, M., Heinz, A., He, J., Silk, T., & Lorenz, R. C. (2024) — Assessing inhibitory control deficits in adult ADHD: A systematic review and meta-analysis of the stop-signal task
Soler-Gutiérrez, A. M., Pérez-González, J. C., & Mayas, J. (2023) — Evidence of emotion dysregulation as a core symptom of adult ADHD: A systematic review
