New PhD research reframes ADHD and apathy as opposite poles of motivational dysregulation
A PhD thesis completed at Dublin City University in December 2025 by Nam Trinh, supervised by Prof. Tomas Ward and Dr. Gerard Derosiere, presents three studies on the neural mechanisms of motivated behaviour. The thesis applies machine learning to electroencephalography (EEG) and structural MRI data to investigate ADHD, effort and reward sensitivity, and apathy. The unifying argument is conceptual: ADHD and apathy are not separate disorders requiring separate explanations. They are opposite poles of motivational dysregulation, sharing partially overlapping neural substrates.
The framing matters. ADHD is conventionally described through inattention, hyperactivity, and impulsivity. Apathy is conventionally described through reduced goal-directed behaviour, diminished emotional response, and social withdrawal. The clinical languages do not overlap. The diagnostic categories do not overlap. The research literatures rarely cite each other. Trinh’s thesis argues this separation obscures a deeper structural similarity: both conditions involve dysregulation of the fronto-striatal and supplementary motor area (SMA)-connected circuits that compute the value of effort and reward. ADHD reflects motivational hyperactivity — heightened impulsivity, dysregulated reward seeking, fragmented sustained attention. Apathy reflects motivational hypoactivity — reduced engagement, blunted reward sensitivity, diminished initiation. The mechanism is shared. The direction differs.
This is a useful reframe. It cuts against the tendency to treat ADHD as a pure “attention” disorder, separate from anything to do with motivation or reward. The thesis’s empirical work supports the conceptual move with concrete findings about which neural systems are involved and how they vary between individuals.
Reduced gamma-band power during inhibitory control — a more substantive finding than resting-state EEG
The first study compared EEG-based ADHD classification across two recording conditions: resting state and during a stop-signal task that demands inhibitory control. Nine machine learning classifiers were tested on 52 ADHD adults versus 98 controls (resting state) and 24 ADHD adults versus 78 controls (stop-signal task).
Task-based EEG significantly outperformed resting-state EEG. The best resting-state classifier achieved an AUC of approximately 0.59 — barely above chance. The best task-based classifier (AdaBoost) achieved an AUC of 0.75, with sensitivity of 0.74 and specificity of 0.76. The difference is substantial and methodologically important. It suggests that ADHD-related neural alterations are not particularly visible at rest but emerge clearly when cognitive systems are engaged in inhibitory control demands.
Feature importance analysis identified gamma-band power (45–49 Hz) over fronto-central, temporal, and parietal sites as the dominant discriminating feature — and the gamma-band power was significantly reduced in ADHD participants compared to controls during the stop-signal task. Trinh interprets this as reflecting disrupted dopamine-mediated reward circuits. Gamma oscillations are involved in coordinating distributed neural assemblies for sustained attention and inhibitory control. Reduced gamma during inhibitory engagement aligns with reduced capacity to sustain the kind of focused, top-down cognitive control that the stop-signal task demands.
This is a different finding from the standard resting-state theta-beta ratio (TBR) literature that has dominated ADHD EEG research. TBR has been losing reliability in adult samples — Trinh notes explicitly that TBR-related features can appear influential within multivariate models, but their group-level differences are too subtle and inconsistent to support reliable classification. The thesis represents a methodological shift toward task-based recording and toward gamma-band rather than slow-wave biomarkers.
The finding also connects to my prior work on ADHD and cybernetic attention versus the waning human attention. Sustained attention requires gamma-band coordination of distributed neural assemblies. Attention that operates within continuously interrupted, externally stimulated digital environments may not exercise these gamma-coordination mechanisms in the same way. The relationship between environmental input and neural rhythm is bidirectional — and the systems Trinh’s research measures are not measured in environmental isolation.
Shared neural substrates: SMA-connected white matter and fronto-striatal circuits
The second study examined white matter integrity in 45 healthy participants using diffusion MRI, correlating microstructural properties with computational modelling parameters of effort and reward sensitivity derived from a decision-making task. The analysis used data-driven cluster analysis of fractional anisotropy and mean diffusivity to identify regions where microstructure predicted individual sensitivity parameters.
Tracts connected to the supplementary motor area emerged as central hubs for both effort sensitivity and reward sensitivity. The SMA is involved in initiation and integration of action-related and reward-related information. White matter integrity in SMA-connected pathways reliably predicted individual differences in how participants weighted effort costs against reward magnitudes during decision-making. Machine learning classifiers built on these microstructural features could predict individual sensitivity profiles above chance.
The third study examined grey matter volume and its relationship to effort sensitivity, reward sensitivity, and subclinical apathy in healthy controls. Effort sensitivity correlated with grey matter in dorsal insula, primary motor cortex, and superior parietal cortex. Reward sensitivity correlated with ventromedial putamen and posterior parietal cortex. Subclinical apathy was associated with reduced grey matter in medial prefrontal cortex, SMA, and ventral temporal regions.
The convergence is striking. The SMA appears across all three studies — connected to effort and reward sensitivity in healthy participants, structurally reduced in those with higher subclinical apathy, and part of the broader motivational network the thesis argues is dysregulated in ADHD. Combined with the fronto-striatal circuits that the ADHD literature has long implicated, the picture is of a shared motivational substrate that varies continuously across individuals and that breaks down in different directions to produce different clinical presentations.
This connects to my recent work on the transdiagnostic mechanisms across ADHD and autism. Categorical diagnostic frameworks treat ADHD, autism, apathy, and other conditions as separate entities. The neural data suggest substantial mechanistic overlap. Dimensional and transdiagnostic frameworks may better describe what is actually happening in the brain.
The clinical frame cannot see what is shaping the systems it measures
Trinh’s thesis is careful, well-conducted, and conceptually significant. It uses standard clinical language throughout — “neuropsychiatric disorders,” “motivational deficits,” “biomarkers,” “diagnostic accuracy.” It positions the findings as supporting “personalised and targeted neurotechnological interventions for individuals with neuropsychiatric disorders.” This is the standard frame for clinical neuroscience research, and Trinh works squarely within it.
But the frame has a structural blind spot.
The thesis treats the motivational circuits it measures as endogenous — variation between individuals in dopaminergic signalling, reward sensitivity, effort computation, and inhibitory control is treated as a property of those individuals. Clinical conditions are treated as disorders of those individuals’ systems. The intervention pathway is individual-level: identify the neural variation, develop targeted neurotechnology to address it, and deliver the intervention to the affected individual.
What the frame cannot see is the environment in which these systems formed and operate.
Dopaminergic motivation circuits do not develop in a vacuum. They are exquisitely sensitive to reinforcement schedules, novelty, intermittent reward, social feedback, and visual-auditory stimulation. They are, in fact, the precise neural substrates that contemporary digital platforms have been engineered to engage. Notification systems exploit reward prediction error signalling. Infinite scroll exploits intermittent reinforcement. Social media engagement metrics exploit social reward processing. Recommendation algorithms exploit novelty-seeking dopaminergic responses. Gaming reward structures exploit effort-reward integration in the SMA-connected circuits Trinh’s thesis identifies as central.
The generations now presenting with motivational dysregulation — heightened impulsivity, blunted sustained attention, fragmented engagement, apathy, withdrawal — developed and operate within continuous technological environments specifically designed to capture these circuits.
The thesis’s findings are real. The question the thesis cannot ask is: what proportion of the individual variation being measured reflects endogenous neural difference, and what proportion reflects environmental shaping of the very systems being measured?
This matters because the proposed solutions are downstream of the same systems that produced the dysregulation. I recently covered a paper proposing AI-powered “cognitive co-regulation” tools for neurodivergent professionals — productivity software designed to externally regulate executive function and motivation in workers showing dysregulation. I’ve also examined how technostress research has ignored neurodivergent workers, and how that research is starting to detect what the population has been experiencing for years. Earlier still, I covered research on virtual autism — autism-like presentations in young children with high screen exposure — and how screen exposure is creating real and virtual autism by shaping the developing attentional system.
The pattern across these pieces is consistent: digital environments shape the neural systems involved in attention, reward, and motivation across the lifespan.
Clinical research measures the resulting variation as if it were endogenous. Industry develops tools that further engage the same systems, marketed as solutions to the dysregulation those systems’ continuous engagement produced. Neurodivergent populations are positioned as the entry market because they show the dysregulation first and most clearly. The dysregulation is then framed as individual disorder, justifying individual-level neurotechnological intervention.
Trinh’s thesis does not make this argument. The thesis is descriptive, careful, and bounded. But the data the thesis produces — reduced gamma during inhibitory control, SMA-connected motivational circuits, fronto-striatal dysregulation across ADHD and apathy — are findings that can be read in multiple ways. The clinical reading, obviously, produces individual disorders requiring individual interventions. A wider reading recognises that the systems being measured are not isolated from the environments they developed within. The same neural data could support intervention into the environment as much as intervention into the individual. The choice of which intervention to pursue is not determined by the neuroscience. It is determined by the framing applied to the neuroscience. And that framing, as it stands, is currently doing a lot of the work, and that work should be made visible.
