The NCI Framework
What NCI proposes on top of findings it did not produce: the architecture underneath personality, the trade-offs it makes, and how each claim could be shown wrong.
Core Premise
Much of modern work is organized for legibility: tasks that decompose, completion that can be verified, deadlines, milestones and long single-task blocks. Part 2 observes that this closely matches what one configuration, the Builder, needs to reach flow. The others must build those conditions for themselves: a difference in tax, not in ability.
The misreading that follows is predictable. A person who cannot sustain deep work is not thereby undisciplined. They may be following instructions built for a comparator they do not run, and the failure is then attributed to their character (Part 2, §7.1.5).
The observation can be tested. O*NET, the US Department of Labor’s occupational database, rates 1,016 occupations on descriptors such as automation, freedom to set priorities and repetition. Scored against the five comparators, with the crosswalk registered before the distribution is examined, Part 2 predicts it skews toward the Builder’s.
NCI’s starting point is that habits, struggles and strengths are not random. They follow from a configuration of admission and energy: not personality, but the architecture underneath it.
The Model
Part 2 organizes a person’s operating state around two quantities. That they can be measured separately is a requirement the framework sets itself, not a demonstrated fact.
Position: what the system treats as signal
A setpoint across five components, stable from one day to the next. It starts partly set by the genes and prenatal environment that build gating circuits, is tuned through childhood and adolescence, and is held in place by physical brakes by the mid-twenties. What those years adjust is the weighting of the gate, not learning.
Range: the budget left to run it
The energy left for discretionary work once the fixed cost of admission, the Tax, is paid. It moves over hours to days with sleep, allostatic load and metabolic demand, and it is withdrawn in a proposed order: flexibility first, sustained abstraction next, the filtering itself last.
Position says what the system admits. Cortical complexity says how much can be built from it: how many levels of abstraction, over how long a horizon. Part 1 introduced the separation as the Two-Axis model, and Part 2 treats complexity as causally independent of Position and as the multiplier on both capability and failure. The same profile can look very different in two people.
The Five Components
Part 1 described temperament along five continuous dimensions, found from behavior. Part 2 derives the same five from architecture: two are the poles of a cortical container, inward or outward, and three are admission tags that flag what gets through.
| Component | Part 1 dimension | Kind | What it does | What it leaves out |
|---|---|---|---|---|
| Completion | Completion Capacity | Container pole: outward (DAN) | Sustained closure of open loops | The exit signal |
| Depth | Processing Depth | Container pole: inward (DMN) | Recursive internal modeling | The clock: urgency |
| Initiation | Initiative Capacity | Admission tag: novelty | Low-threshold move from rest to action | The wake: downstream cost |
| Boundary | Boundary Enforcement | Admission tag: threat | Threat and violation detection | The benign reading |
| Resonance | Resonant Empathy | Admission tag: resonance | The states of other people | One's own state |
Under compression, raising the gain on one channel discards a complementary signal. It is still processed below awareness; it is simply not forwarded to the brief. The persistence that carries someone past discouragement is the setting that carries them past the point where stopping would be wise: one parameter read from two sides, not two traits. That weighting is zero-sum within a bounded gate is the model’s premise. Which content each strength leaves out is a set of hypotheses (Part 2, §4.1).
These aren't choices. These are design features. Your nervous system was optimized for a specific function, and these five dimensions reflect the trade-offs your biology made to excel at that function.
Part 1 predicts a one-to-one alignment between the five dimensions and five Big Five aspects (DeYoung et al., 2007), with correlations of about r = 0.4–0.7. Part 2 derives the same split from the aspect structure independently (§4.5).
The Five Profiles
Crossing the container (outward or inward) with the admission state gives eight cells. Five recurring profiles settle among them: first identified from behavior in Part 1, derived from the grid in Part 2.
| Uncaptured | Novelty | Threat | Resonance | |
|---|---|---|---|---|
| Outward (DAN) | The Builder | The Sprinter, outward phase | The Catalyst, outward phase | A blend: resonance held outward by threat |
| Inward (DMN) | The Visionary | The Sprinter, inward phase | The Catalyst, inward phase | The Mirror |
Uncaptured is not a fourth tag: it is the state in which no tag has seized the container and the system rests on its default. That the count is five follows from Part 2’s routing assumptions (§5.2.6), not from the network findings alone, and a demonstrably stable outward-resonance profile would require revising it.
The Builder
SPP · Sustained Persevering Phenotype
Optimized for completion · What's incomplete?
Comparator: Satisfaction on closing a loop
Grid: Outward container at rest (E0)
Leaves out: The exit signal
Cloninger (TCI): Persistence
The Sprinter
ENP · Exploratory Novelty-Seeking Phenotype
Optimized for exploration · What's new?
Comparator: Anticipation of something new
Grid: Novelty, moving between inward and outward (I-N ↔ E-N)
Leaves out: Downstream cost
Cloninger (TCI): Novelty Seeking
The Visionary
HIP · High-Dimensional Integrative Phenotype
Optimized for synthesis · What's connected?
Comparator: A pattern resolving into coherence
Grid: Inward container at rest (I0)
Leaves out: Urgency
Cloninger (TCI): No counterpart; Part 1 adds Processing Depth
The Catalyst
AIP · Adrenal-Sympathetic Impact Phenotype
Optimized for boundary modulation · What needs redirecting?
Comparator: A threat neutralized
Grid: Threat, moving between outward and inward (E-T ↔ I-T)
Leaves out: The benign reading
Cloninger (TCI): Part 1 splits Harm Avoidance; its threat side loads here
The Mirror
RIP · Resonant Interoceptive Phenotype
Optimized for attunement · Who's misaligned?
Comparator: Contact that is genuinely met
Grid: Resonance held inward (I-R)
Leaves out: One's own state
Cloninger (TCI): Reward Dependence
These are attractors in a continuous space, not categories. Every person is a blend, and blends carry their own costs: two demands on the same object multiply, two on different objects add (Part 2, §5.3.1).
On neurochemical language
Where neuromodulators are named in the monographs, the claim is about what a channel does — the signal it carries and the orientation it produces — not about how many neurons of a given type a person has. Variation is located in implementation parameters such as receptor density, clearance rate and transporter expression, which are reachable by many molecular routes. The same channel gain is reachable by too many routes to be identified with any one of them.
Part 1 describes each profile at the edges of its stable range with the Window of Tolerance, a term introduced by Daniel Siegel (1999): an optimal zone between hyper-arousal and hypo-arousal. In NCI’s terms, the edges are the unstable boundary regions of each profile’s basin, and each profile has its own: rigid routine or inertia for the Builder, overdrive or flatness for the Sprinter, abstraction without landing or fog for the Visionary, alarm or freeze for the Catalyst, and overload from other people’s states or numbness for the Mirror. They are behavioral correlates offered for testing, not diagnostic descriptions (Part 1, §5).
Change
Position is held in place in adulthood, but not sealed. Part 2 proposes that revising a settled default needs three conditions open at once, and that each alone produces a characteristic stall.
Budget
Enough Range to afford revision: the building budget is the first thing load takes away. Alone, it yields capacity with no target.
Structure
A reopened plasticity window: the perineuronal nets that close critical periods loosen or come off. Alone, it yields lability with no direction.
Attribution
Recognizing a percept as an output of the gate rather than as reality; a raw percept carries no marker that it could be revised. Alone, it yields doubt with no means to act.
The budget comes first. Releasing structure while the budget is shut is predicted to yield lability without direction, in the cells least able to survive it. Part 2 treats this as its clearest statement about what not to do, and its scope limit governs: it is not a clinical recommendation (§1.2).
The most consequential proposal. In animal studies, inflammation dissolves the nets, by at least two routes. NCI proposes that while they are off, a setting can drift toward its own extreme and be held there when they re-form. The test is within-person: Position measured before and after a documented inflammatory episode, against matched intervals without one. The evidence so far is from rodents, mostly visual cortex, and no human equivalent has been shown (Part 2, §7.6).
Positioning
NCI is not a replacement for existing models. It works at a different level, proposing a mechanism beneath what trait models describe, and it marks which of its claims are tested and which are not.
| NCI | Big Five (OCEAN) | Cloninger TCI | |
|---|---|---|---|
| Describes | Proposed parameters of admission and energy | Behavioral traits, from language and factor analysis | Temperament and character |
| Level | Neurobiological and metabolic | Psychological | Psychobiological |
| Question | “What does the system treat as signal, and what can it afford?” | “What do people tend to do?” | “What are the innate temperament dimensions?” |
| Output | A Position and a Range; no validated instrument yet | Five domain scores; ten aspect scores | Four temperament and three character scores |
| Status | Two preprints; ten predictions, none yet tested | Extensively validated across languages and cultures | A validated inventory in use for decades |
| Relationship | — | Part 1 predicts one aspect per NCI dimension, r ≈ 0.4–0.7 (DeYoung et al., 2007) | The closest precursor: Part 1 inherits Persistence, Novelty Seeking and Reward Dependence (Cloninger, 1987) |
Trait questionnaires remain the best available comparison. Any instrument built for NCI has to beat them on their own ground (Part 2, §9.6).
How it could be tested
Part 2 closes with ten predictions, each stated so that a specific result would count against the framework. They are NCI’s own claims, not established findings, and none has been tested yet.
Metabolic hierarchy
As the energy budget tightens, flexibility (switching between tasks) should fail first, sustained abstract reasoning next, and sensory filtering last. When filtering does fail, it should look like disinhibition, not slowing.
Test: Progressive load or sleep restriction, with EEG and standard gating, switching and reasoning tasks. These paradigms exist, so the ordering can be tested now.
Repertoire width
Under load, a person’s range of behavior should narrow more than their usual behavior shifts.
Test: Repeated experience sampling and tasks alongside autonomic and sleep measures.
Container × tag interaction
The same content should produce different responses depending on whether a task directs attention inward or outward, beyond the effects of the content or its difficulty.
Test: Factorial behavioral tasks with event-related fMRI or EEG.
Developmental metabolic tracking
A person’s metabolic trajectory through development should predict how stable their traits later become, beyond age and environment.
Test: Longitudinal cohorts with repeated personality and metabolic measures.
Three-lock interaction
Lasting revision of a settled default should need three conditions at once: enough energy budget, loosened structural brakes, and recognition that the belief is revisable. Supplying only one should fail in a characteristic way.
Test: Staged mechanistic work and ethically approved longitudinal studies.
Comparator dissociation
People whose baseline leans outward should respond strongly to classic inconsistency and weakly to unresolved incoherence; people who lean inward should show the reverse. An effect in one direction only, or none, would count against the account.
Test: Standard dissonance paradigms alongside matched incoherence tasks, preregistered.
Axis separability
Working-memory training should not change the processing profile, and autonomic training should not raise the working-memory ceiling. Raw scores under load may rise with training; only a ceiling shift would count as transfer.
Test: Randomized training arms with active and passive controls, measured under matched conditions.
Blend-specific load
Burnout should take different forms in work that holds two demands on one target at once and in work that alternates between them.
Test: Factor analysis and measurement-invariance testing of validated burnout scales across role-stratified cohorts.
Two brakes, separable by frequency
Adult plasticity is held by a structural brake (perineuronal nets) and a transcriptional one. Sensory stimulation at 60 Hz should reduce net density; at 40 Hz it should restore plasticity with net density unchanged.
Test: Rodent entrainment at 40 Hz, 60 Hz and sham, with the relevant molecular measures.
Allocation without a change in total turnover
Range is a rate, not a stored pool. Under sustained load, energy production should shift toward the engaged region while total brain turnover barely changes. After sleep deprivation the shift should flatten, and creatine should restore it only in that depleted state.
Test: Phosphorus magnetization-transfer spectroscopy in humans, within the same person, at rest and under load.
Each prediction’s full statement, test conditions and falsifier are in Part 2, §8. If you can run one of these studies, or see a better test, we would like to hear from you.
Go Deeper
The findings the framework builds on are other laboratories’ published work. The Science page sets them out in plain language, in order, with every source.