Despite decades of research and a growing toolkit of evidence-based therapies, mental health services face a stubborn paradox: between 40 and 60 per cent of people with complex emotional difficulties don’t get better under current treatment protocols. For those living with severe emotional dysregulation, relationship instability, identity struggles, and recurring self-harm — conditions often grouped under the term Complex Emotional Needs (CEN) — this is not an abstract statistic. It is the reality of revolving doors, service discharge without recovery, and the exhausting sense that nothing seems to fit.

A new perspective article submitted to Frontiers in Human Neuroscience argues that this therapeutic impasse is not primarily a failure of effort or skill — it is a failure of the underlying model. Categorical diagnosis, the dominant framework through which mental health problems are identified and treated, may simply be the wrong map for the territory it is trying to chart.

The Problem with Categories

Standard diagnostic systems classify mental health conditions as discrete named categories: borderline personality disorder, complex PTSD, emotionally unstable personality disorder, and so on. This approach has generated decades of treatment research, but rests on an assumption that is increasingly difficult to defend: that there are meaningful, stable boundaries between conditions, and that people within a given category are fundamentally similar to one another.

Research tells a different story. Diagnostic categories in mental health show substantial overlap, extensive within-category variation, and poor ability to predict treatment response. Two people with identical diagnoses can respond completely differently to the same therapy. Critics have long argued that current categories are, at best, administrative conveniences rather than windows into the nature of the problems they describe.

“Rather than understanding psychopathology as the presence of pathological elements requiring removal, we reframe mental health conditions as failures of flexible functional synchronisation across nested bio-psychosocial scales.”

A Different Way of Seeing: Synchronisation and Dynamics

The new framework starts from a different premise. Instead of asking “which diagnosis does this person have?”, it asks: “how is this person’s nervous system organising itself, and what kind of flexibility or rigidity does it show?”

The human nervous system — from brainstem circuits regulating basic threat responses, through autonomic networks governing heart rate and breathing, to cortical systems that make sense of experience and manage relationships — functions as an enormously complex network of coupled oscillators. The healthy state, known as metastability, sits between rigid lockstep synchronisation and complete independence. In Complex Emotional Needs, this balance breaks down.

The paper draws on three converging scientific traditions: affective neuroscience (particularly Jaak Panksepp’s work on evolutionarily conserved emotional circuits), predictive processing theory (Karl Friston’s framework in which the brain is a prediction machine calibrating its models against incoming signals), and complexity science (the mathematical study of how coupled dynamical systems self-organise). All three point toward the same insight: CEN presentations are nervous systems stuck in particular dynamic patterns — patterns that can be characterised, measured, and matched to appropriate interventions.

The Attractor Landscape: A New Vocabulary for Emotional Life

Central to the framework is the concept of an attractor landscape. Imagine the state of someone’s nervous system as a ball rolling across hilly terrain. Valleys represent stable states the system gravitates toward; hills and ridges represent the energy required to transition between states. The shape of this terrain — how steep the valleys are, how high the barriers between them, whether the landscape is stable or shifting — determines the person’s characteristic emotional and relational patterns.

Using mathematical equations from coordination dynamics and numerical simulation, the research team derived four distinct landscape topologies, each corresponding to a clinically recognisable pattern of emotional experience.

Phenotype 1 — Hypervigilant

Sharp, narrow valleys with low barriers between them. The system is tightly wound, prone to sudden catastrophic shifts. Characterised by hyperarousal, threat sensitivity, and intense emotional reactivity punctuated by crashes. Corresponds to anxious-preoccupied attachment.

Phenotype 2 — Collapsed

Flat, featureless terrain bounded by very high barriers. The system is stuck, requiring enormous energy for any movement. Characterised by emotional numbing, motivational paralysis, and disconnection from internal experience. Corresponds to dismissive-avoidant attachment.

Phenotype 3 — Disorganised

Unstable, shifting terrain with no reliable valleys. Characterised by fragmented, unpredictable emotional responses and profound relationship instability. Corresponds to disorganised attachment.

Phenotype 4 — Balanced

Multiple moderate valleys connected by traversable ridges — the optimal metastable landscape. The system can settle, shift flexibly, and return to stability. Corresponds to secure attachment and represents the therapeutic target for the other three phenotypes.

Grounded in Biology, Not Theory Alone

The paper demonstrates that the four phenotypes map onto converging evidence from multiple scientific disciplines simultaneously. At the brain circuit level, Panksepp’s research anchors the Hypervigilant phenotype in chronic upregulation of FEAR circuitry, and the Collapsed phenotype in the neurobiological sequelae of prolonged GRIEF/PANIC followed by learned suppression. At the developmental level, the phenotypes map directly onto attachment classifications independently identified through decades of child observation research.

Perhaps most striking is the epigenetic layer. Research by Michael Meaney and colleagues has shown that early caregiving quality literally programmes the stress response system through DNA methylation — chemical modifications stable across the lifespan but not permanently fixed. The framework uses this to explain both why CEN patterns are so persistent and why therapeutic change is genuinely possible: attractor landscapes are biologically encoded but remain responsive to sustained relational experience of sufficient quality and duration.

Clinical Evidence: The IDEAS Pilot Study

The framework was directly motivated by results from the IDEAS pilot study, an 8-week modular intervention delivered to young people aged 16–25 within youth mental health services (N=48). The study demonstrated large effect sizes for emotional dysregulation (Cohen’s d = 1.15), moderate-large effects for interpersonal functioning (d = 0.82), and a successful discharge rate of 68.7% — substantially exceeding the service’s baseline rate of approximately 42% for comparable presentations. Improvements were maintained at 3-month follow-up.

The IDEAS intervention did not apply a fixed protocol. Clinicians personalised which therapeutic modules each person received and in what sequence — effectively engaging in “inferential phenotyping.” This proved effective but depended on individual practitioner skill. The new theoretical framework is designed to formalise and scale precisely that clinical insight.

Matching Treatment to Landscape

One of the most practically significant implications is the framework’s account of why particular therapeutic ingredients work for particular presentations. For the Hypervigilant phenotype, the primary target is landscape flattening: mindfulness and distress tolerance skills work because they flatten rather than eliminate emotional responses. For the Collapsed phenotype, the barrier height is the problem — insight-oriented work tends to fail not because the person lacks capacity but because the mechanism of dysfunction lies upstream of thinking; behavioural activation works by supplying external energy to overcome the barriers. For the Disorganised phenotype, the priority is creating stable structure where none exists, building attractor basins before attempting flexibility work.

Measurement and the Road Ahead

The framework makes specific, falsifiable predictions testable in future research. Heart rate variability analysed using nonlinear methods should produce characteristic signatures for each phenotype. Physiological synchrony between patient and therapist, measured using surrogate statistical methods that distinguish genuine coupling from coincidence, should track with therapeutic progress.

The paper proposes a three-phase validation programme: establishing phenotype reliability and predictive validity; conducting randomised trials comparing phenotype-matched versus standard treatment (primary hypothesis: effect size advantage d = 0.3–0.5); and examining implementation at scale with attention to health equity.

“The challenge and opportunity before the field is to embrace complexity without abandoning rigour, to pursue precision without losing humanity, and to advance scientific understanding while remaining grounded in the lived experience of those seeking help.”

The paper is currently under preparation for submission to Frontiers in Human Neuroscience as a Perspective Article. Further updates, including trial registration and data repository details, will be posted here as they become available.