← Back to the atlas · Plate II

Working Memory Fragility

A rate relationship, not a rate. Fragility obtains when the rate of thought generation outpaces the rate of buffer stabilization before the specious present resets.

Not a storage deficit. Not a synonym for ADHD. A property of the race between two processes, expressed differently under load, arousal, sleep, and scaffolding. The teaching analogy is blood pressure: a real, stable, consequential property that is also a relationship, varies minute to minute, and is fully compatible with a normal reading right now.

Two editions

The reader's edition 2,800 words · 12 min The essay. One definition doing all the work, then the four layers in plain language. The canonical paper v4 · 60 references Every section labeled by how we know it. Limitations, a bridge-versus-finding audit, and a dated changelog of 42 entries.

If the two diverge, the paper wins.

Fig. 1a · The parts, at rest read this first · generator, buffer, the frame called now

Top row, the generator. Each tick is one thought arriving. The rate here is ordinary, about 6.5 transitions a minute, the median measured in young adults. Nothing on this row is unusual, and it is the same for every brain in the figure.

Middle row, the buffer. Where a thought has to set before it counts. One that finishes setting becomes a solid circle and survives. One overwritten by the next arrival before it sets is displaced. One still setting when the frame resets is lost.

Bottom row, the frame called now. The specious present: the stretch of time experienced as a single now. It resets on a clock of its own, and whatever has not set by then is gone with it.

Generator
thoughts emitted

Buffer
thoughts that set

Specious present
the window called now

emitted set: survives the reset displaced by the next thought still setting at the reset: lost

Fig. 1a. Fifteen thoughts in, four survive. What fails is the ratio between the first two rows, not the thinker: newer threads overwrite older ones before they set, so the failure mode is displacement, not decay, and from outside the same architecture reads as "weak working memory." Fig. 1b runs the same three rows in time.

Fig. 1b · The mechanism, running same generator · change the buffer, add a scaffold

What you are watching. The rows from Fig. 1a, running. A generator emits thoughts at an ordinary, fixed rate. Each one drops into the buffer and starts to set: the ring around it fills. A thought that finishes setting moves down to the held shelf and survives.

How it fails. If the next thought arrives before the ring completes, the old one is displaced and drifts away as a dashed ghost. When the frame called now resets, anything still setting is lost, and the frame flashes.

What to try. Switch the buffer from fragile to typical: the generator does not change, only the settling time, and the outcome flips. Then turn the scaffold on: every thought is written down the instant it appears, so the buffer can keep losing and nothing is lost.

Buffer

Scaffold

emitted, trying to set set: survives the reset displaced or lost written down

Fig. 1b. The whole theory in one toy: change the ratio, not the thinker. Emission timing is randomized within a fixed range, so no two runs match; the counters keep score.

Plate I Four layers, five kinds of evidence

Where the science is, and where I am on my own.

Every part of the theory is labeled by how we know it. The tempting assumption is that the brain science carries the evidence and the philosophy carries the speculation. It is wrong in both directions: the newest ideas sit inside the brain layer, and the upper layers borrow well-established frameworks. The original material and the least-supported material are the same material, on purpose and in the open.

Fig. 2 · Layer × kind of evidence filled: present in that layer · blue: where the new ideas sit

Established

Original

Borrowed

Philosophy

One case

what it says

Part 0 · The definition

rate relationship · axes · two cuts

WMF as a ratio between generation and stabilization within one frame. Giftedness demoted from required axis to highest-yield cause.

Layer 1 · Neural architecture

§1.1–1.11

Subcortical compensation, procedural preservation, stochastic resonance, the load paradox, the Explorer phenotype, extended mind; then the buffer-generator mechanism, attention, QEEG, and one MRI.

Layer 2 · Phenomenology

James · Metzinger

Collapse of the specious present; the transparent self-model made visible; consciousness as reconstruction; the visibility gradient.

Layer 3 · Five dials

parameterized states

Frame rate, buffer reliability, attentional saturation, hedonic range, generation rate. One pharmacological history as the basis.

Layer 4 · Alterocentrism

Dabrowski, inverted

Externalized self-knowledge as the precondition for seeing others, not the stage before it. The load-bearing ethical claim.

Established replicated science, independent of me (the paper marks it D) Original new in this work; carries the contribution and the risk (G) Borrowed an analogy from a neighboring field, untested here (B) Philosophy explanatory, not empirical (P) One case my own self-report (n1)

Plate II Layer 1 · the neural architecture

An Explorer architecture, not a defective default brain.

What is clinically read as dysfunction is substantially a mismatch between this architecture and an environment demanding static, low-stimulation processing. Eleven sections. The dots on each are how sure I am: three for replicated science, two for reasoning that has not been tested here, one for my own case.

§1.1●●●Established

Subcortical compensation

Instead of focal prefrontal firing, diffuse recruitment of insula, basal ganglia, and medial PFC: thinking routed through action loops rather than storage loops. In adulthood the cerebellum joins in.

§1.2●●○Established, contested

Sequence learning holds, instructions slip

The predicted procedural deficit did not show up in sequence learning (SMD 0.02, CI −0.35 to 0.39, absence of evidence), and statistical learning agrees. The double dissociation is withdrawn; feedback-based learning and overnight consolidation show costs.

§1.3●●●Established

Stochastic resonance

White noise improves ADHD performance while it degrades controls. The direction is well supported; the "comparable to medication" magnitude claim is withdrawn.

§1.4●●●Established

The load paradox

High cognitive load taxes the buffer; high perceptual load clamps attention and the system locks in. Hyperfocus is a switching failure, not an attention failure. Rest is harder than difficulty.

§1.5●●○Established, contested

The Explorer phenotype

Complementary cognition: a minority variant optimized to scan the periphery. DRD4-7R stated with the dispute intact; the hunter-farmer story named as precursor, not evidence.

§1.6●●●Established

Distributed cognition

For this architecture the notebook is a prosthetic, not a convenience. Strategic offloading is accurate self-modeling, not weakness.

§1.7●○○Original · one case

The buffer-generator mechanism

Two dissociable processes that feed forward: a leaky buffer makes every thought feel urgent, which raises branch emission, which displaces what was loaded. The loop is the architecture.

§1.8●○○Original

Cause, conflict, compensation

Giftedness sits upstream (a fast generator) and downstream (recruited to recover) at once. Perseveration is the expensive internal fallback for what scaffolding does cheaply.

§1.9●●○Mixed: established to one case

Attentional architecture

Posner profile: tonic alertness impaired, orienting intact, executive contested; distributed hyperfocus from mode-lock, scanning, and catecholamine sustainment; salience tagging; a separable converter. Tabs are the physical signature.

§1.10●●○Established + borrowed

Electrophysiological signatures

Elevated theta/beta, central slow-wave excess, frontal slowing (contested), beta hyper-connectivity read as compensation. E/I framing borrowed from the autism literature.

§1.11●○○One case + borrowed

Structural brain age

Composite 43.7 against a jagged per-structure profile; the older-reading left belt is thicker and larger, so lateralization by construction, not wear. Never cite a composite as if uniform.

Plate III Layers 2 and 3 · what it is like, and along which dials

Consciousness as continuous reconstruction.

The specious present collapsesborrowed · James

The window the mind holds as now is shortened by buffer fragility and generator displacement. The reset is the unit of the theory.

The transparent self-model breaksborrowed · Metzinger

Ordinarily we look through the self-model and see a seamless self. WMF breaks transparency intermittently; the reconstruction becomes visible from inside.

Identity is re-derived each momentoriginal

Remembering who you are is closer to literal than metaphorical. External self-model infrastructure supplies consistent reconstruction inputs.

The visibility gradientoriginal

WMF does not create a different consciousness. It removes the smoothing that hides reconstruction from everyone else. The fragile case is an instrument.

Philosophy, not measurement. James and Metzinger are cited as frameworks, not as evidence. The layer earns its place by explanatory coherence, not by measurement.

Fig. 3 · Five dials Layer 3 · original · one case

Frame rate

How often the specious present resets

thalamocortical gating

Buffer reliability

How much context survives between frames

working memory · PFC

Attentional saturation

Bandwidth allocated to current experience

dopamine

Hedonic range

Distance between dimmest and brightest available experience

serotonin

Generation rate

How fast new branches are emitted

catecholamine tone · E/I gain

Fig. 3. Generation rate is the fifth dial, promoted because a theory about a ratio has to parameterize both sides of it. Positions are illustrative; the empirical basis is one pharmacological history, uncontrolled and unblinded. It generates predictions. It is not evidence.

Plate IV Layer 4 · alterocentrism, inverted

The received arc · Dabrowski's tradition

Self-awareness first; then, through inner work, transcend self-focus into other-centeredness.

Self-focused work is a phase you graduate from. The arc assumes a self-model stable enough to be taken for granted and then set aside.

The inversion · this paper's contribution

Externalized self-knowledge is a precondition for perceiving other people accurately, not a stage that precedes doing so.

For a fragile-buffer mind the self-model is the unstable thing. You cannot know which part of your reading of someone is you until your own interior is named. The conventional framing of self-work as a phase is covertly neurotypical.

Philosophy, resting on one case. Its central claim is untested and, as formulated, it is unclear what would test it. It appears because it is the load-bearing ethical claim of the framework, and because stating it exposes it to correction. This is where the theory hands off to practice: a mind of this kind must build its model externally, and doing so is an ethical capacity, not a coping mechanism. The system on the homepage is that externalization.

Plate V Boundaries, limitations, corrections

The honesty is the argument.

What WMF is not

Sluggish Cognitive Tempo. A processing-speed deficit in parietal systems; low energy, internalized. The Explorer model does not fit it and WMF should not be extended to it.

Low working memory capacity. Span is a property of the buffer. Fragility is a property of the race. The phenomenology persists in people whose span is unremarkable.

A synonym for ADHD. Four axes contribute: ADHD, autism, giftedness, neuroticism. Giftedness is the highest-yield cause of a fast generator, not a required axis.

Settled on autism. How autistic architecture contributes to the rate relationship is not established. The largest single gap in Layer 1, stated rather than buried.

Limitations, principled

Counting a thought is a thought. The instrument generates the data it measures. Generation rate, one half of the central relationship, is not independently measurable by introspection.

Claims inflate during compression. A documented failure mode: the original session gets it right, each restatement rounds it up. Two worked examples are in the paper.

n = 1 where it matters most. Layers 3 and 4, and the buffer-generator mechanism itself, rest on a single case. They generate predictions; they are not results.

Withdrawn or corrected in v4

restatedSanjeevan: SMD 0.02 stands, now with its CI of −0.35 to 0.39 and demoted from "strongest pillar."

withdrawnThe working-memory/procedural double dissociation: the paper cited for it has no procedural task. §1.2 narrows to spared sequence and statistical learning and is now contested.

withdrawn"White noise comparable to stimulant medication": a 20-child pilot with a null medication arm. Uninformative in both directions.

withdrawnThe "67 citations" claim for the December monograph.

correctedCOMT rs4680 is G/A intermediate, not slow. The medication-failure explanation is downgraded to a hypothesis.

correctedIntelligence PRS restated as 98–100th percentile.

cutHartmann's hunter-farmer hypothesis "finds rigorous support": an overclaim. Named as precursor only.

addedNeuroticism as the fourth axis; the master-compensator measurement artifact; reference 12 split so a model is no longer cited to a paper that cites it.

60 references, machine-verified in both directions · a dated changelog of 42 entries

Plate VI Where the theory shows up

Essays Essays on how this brain works The theory thread and its behavioral applications, from tabs to texting to rucking. writing → Evidence, n = 1 The brain plate MRI and QEEG: the electrophysiology in §1.10 and the structural caution in §1.11, live. Plate III → The recursion The system that holds the theory WMF says a mind like this cannot keep its own model resident. jonmick.ai is the externalization the theory predicted it would need. Plate I →