Log — Psychology Wiki

Append-only history. Entries start with ## [YYYY-MM-DD] <op> | <title> (split · ingest · query · lint).

[2026-07-28] split | psychology-wiki created from the hub _inbox (3 sources)

Spun out by the hub router when kahneman-focusing-illusion arrived with no spoke to take it and completed a cluster of mind-science strays: jung-on-creativity (psychology-creativity, parked 2026-07-23) and human-mouse-perceptual-maps (neuroscience, parked 2026-07-12).

Why these three cohere. Both parked records name the same problem — no spoke owns psychology — and both were declined by ../philosophy-wiki for the same reason: it owns philosophy proper, and a psychiatrist or an experimentalist is a different discipline. Depth psychology, judgment-and-decision research and psychophysics are three sub-fields of one academic subject; all three sources would sit in a psychology department, and one of them was published in Communications Psychology. That is a tighter shared subject than the tag names suggested.

Corroborating signal: the hub already had a dangling reference to this body of thought. ../defensive-security-wiki’s soc-fast-slow-thinking designs a two-tier SOC on Kahneman’s System 1/2, borrowing a framework no spoke owned. A wiki that cites a discipline it can’t link to is evidence the discipline needs a home.

Founding pages: 3 sources + 4 concepts (focusing-illusion, heuristics-and-biases, analytical-psychology, perceptual-learning) + 2 people (daniel-kahneman, carl-jung)

• spine.

Recorded as founding weaknesses, not discovered later. Two of the three sources are essayists’ second-hand accounts (one on a general-interest outlet), nothing has been read in the original, and the corpus spans three incompatible standards of evidence — which synthesis turns into the founding tension rather than hiding: claims here get more interesting and less checkable in the same direction. CLAUDE.md carries a standing evidence caveat about replication. Named next ingest: the 2006 Science income-happiness paper, primary.

avoid-ai-writing run over the founding prose. Site rebuilt + verified at the clean rung, as the hub requires for a split.

[2026-07-28] ingest | six sources, batched: Kahneman, Sapolsky ×2, free-will effects, Fromm, Freud

Six sources arrived within ~40 minutes of each other, all psychology and heavily interlinked, so they were ingested as one batch rather than six passes. Pre-routed by the hub. Sources: thinking-fast-and-slow (Wikipedia, T2) · determined (Wikipedia, T2) · the-trouble-with-testosterone (Google Books listing, T4) · free-will-belief-effects (The Conversation, T3) · erich-fromm (Wikipedia, T2) · sigmund-freud (Wikipedia, T2). New: 3 books + 1 article + 3 people (sigmund-freud, robert-sapolsky, erich-fromm) + 2 concepts (free-will, replication-crisis). Updated: heuristics-and-biases (two selves + the peak-end rule; the replication damage), focusing-illusion (why the happiness studies disagree), daniel-kahneman, carl-jung, analytical-psychology, synthesis (three new sections, two open questions), index.

Four things this batch establishes:

1. The founding synthesis’s top gap is partly closed, but not the way it asked. It named the primary literature as the highest-value next ingest; what arrived is Wikipedia about the primary literature. thinking-fast-and-slow does deliver the thing the corpus most needed — the experiencing vs remembering self — which explains why the income-happiness studies disagree: they query two different selves, not one quantity badly. The weakness is recorded rather than papered over: this wiki can describe the book and has read neither it nor one study in it.
2. A determinism corner, and the useful move is splitting the question. Does free will exist (metaphysics, philosophy-wiki’s) / what produced this behaviour (determined) / what happens to people who stop believing in it (free-will-belief-effects, the only measurable one). All three are really about blame — Sapolsky spends his second half on criminal justice, Walton measures punitiveness and shame.
3. The sources arrive arguing about who owns them. Two of determined‘s four reviewers say a neurobiologist answered a philosophical question (Riskin: scientism; Setiya: ignored the literature). Routed here on the method rule and the objection recorded, not settled — a reflexive check on this spoke’s own boundary with ../philosophy-wiki.
4. Evidence became a subject, not just a caveat. thinking-fast-and-slow‘s own citations fail (Ch. 4 replicability index 14; Kahneman: “I placed too much faith in underpowered studies”), and free-will-belief-effects uses non-replication to clear away the opposing literature while resting on two ~260-person experiments. replication-crisis names both the crisis and that asymmetry: failures get cited against other people’s work more readily than one’s own. Also: sigmund-freud and erich-fromm complete a depth-psychology line that started mid-way at carl-jung — Freud as origin, Jung deepening the unconscious, Fromm socializing it. Fromm adds social character (the psyche as something an economic order shapes), a genuinely new object here, and is the spoke’s hardest philosophy boundary case. Homonym flagged: Fromm’s “freedom” (the burden of being unbound) is not the metaphysical freedom of free-will. the-trouble-with-testosterone is the spoke’s first T4 and is labelled as such: a catalogue entry that establishes what the book covers and nothing about what it argues. The popular testosterone-and-aggression claim is explicitly not sourced from it. Verify deferred per hub policy (content-only). avoid-ai-writing run.

[2026-07-28] ingest | Why writing feels so hard (and how to make it easier) — Lit Hub (lithub.com)

Routed here by the hub (runner-up: none — no spoke owns writing craft, and the substance is psychology). T3: a book excerpt promoting the author’s own method (April Dávila, Sit Write Here, St. Martin’s Essentials 2026), self-interested by construction, but citing named primary research rather than gesturing at “studies show.” New: why-writing-feels-hard (source), flow (concept). Updated: synthesis (two new sections — the citation-chain pattern folded into “Evidence became a subject”, and a new section on creativity having two accounts), index. Entities deferred per this spoke’s precedent: Dávila and the six named researchers are on the page, not paged — same treatment as Popova, James and Walton.

Finding 1 — a named failure mode: the citation chain terminating in a practice. The essay cites Eisenberger (2004, social pain sharing substrate with physical pain), Baumeister (Bad is Stronger than Good), Chen (2008, social pain re-lived more intensely), Raichle (Default Mode Network), Csikszentmihalyi with Nakamura (flow; 120 bits per second), and Dietrich (transient hypofrontality). Every one is real work, cited accurately enough that spot-checking a reference finds a genuine paper. The argument lives entirely in the joins: Eisenberger’s neuroimaging overlap becomes “your brain is genuinely convinced that putting words on a page might kill you,” a survival-threat belief about a modern activity none of the cited work tests; the 120-bit figure — whose provenance runs through popular writing, not a measurement — is used as a constant; transient hypofrontality is treated as mechanism rather than the hypothesis it is. And the last join is widest: no cited study is evidence that ten minutes of insight meditation before writing does anything for writing. That’s harder to catch than “studies show,” because the individual references check out. Recorded as method, not dismissal — the practice may work, the essay just doesn’t show it. Also worth noting from inside the citation list: Baumeister is a central figure in the replication-crisis (ego depletion), which the essay doesn’t mention. The spoke’s standing caveat applies to a source’s own bibliography, not just to the source.

Finding 2 — creativity now has two accounts, and they’re a test of the spoke’s thesis. flow gives an empirical account of making things (a describable state, proposed neural correlates, trainable by practice) directly against jung-on-creativity‘s depth account (the visionary work arrives from the collective unconscious; the artist is its instrument). Both explain the same fact — good creative work doesn’t feel deliberately authored — and disagree about what could settle it. So the founding method spectrum now shows up within one topic rather than across different ones, with the subject held constant and only the method varying. Better test than the founding corpus gave. Weakness recorded on both: each arrived second-hand (an essayist on Jung, a book excerpt on flow) and the spoke still holds no primary literature on either — consistent with the standing note that nothing here has been read in the original. Verify deferred per hub policy (content-only). avoid-ai-writing run.

[2026-07-29] ingest | Four therapy books (Hollis · Yalom · Nagoski & Nagoski · Hayes)

Arrived together via Telegram as bare references, two of them in Russian (Джеймс Холлис «Перевал в середине пути» = The Middle Passage; Ирвин Ялом «Вглядываясь в солнце» = Staring at the Sun). None is held as a document — each was verified and summarized from a secondary record, and every page says which one and what that record couldn’t establish.

New pages: the-middle-passage (T3, via The Marginalian), staring-at-the-sun (T2, via Wikipedia), a-liberated-mind (T3), burnout-stress-cycle (T3, weakest of the four) · concepts existential-psychotherapy and acceptance-and-commitment-therapy · people irvin-yalom, james-hollis, steven-hayes.

The finding: “clinical” was never a position on the method spectrum. The founding table put clinical interpretation at the untestable end, below psychophysics and survey work. Three of these four books are by practising clinicians and they land at opposite ends of the corpus. Hollis and Yalom are where the table predicts. ACT carries 900+ randomized controlled trials, WHO distribution and APA/VA recognition — a larger evidence base than human-mouse-perceptual-maps, the spoke’s only experiment. What the third row was actually measuring is a tradition’s relationship to testing, not clinical work. Depth psychology declines the question and behaviour analysis is organised around it; both happen in a consulting room. Synthesis amended rather than appended — the founding table now carries the correction inline.

The old caution survives at a smaller scale. burnout-stress-cycle makes confident physiological claims with thin citation and sits on the same shelf as a-liberated-mind, whose therapy has 900 trials behind it and which cites none of them. Reach and evidence still trade off, at the level of the book rather than the discipline.

Two open questions added, and the second is uncomfortable: what is actually in ACT’s 900 trials? This spoke discounted thinking-fast-and-slow‘s citations on replication grounds and is now crediting ACT’s on trial-count grounds, working from a Wikipedia summary in both cases. Flagged as an asymmetry in our own handling, not just a gap. Also logged: the “complete the stress cycle” mechanism is specific enough to test and no source here names a study.

No page for Emily or Amelia Nagoski — per ../ENTITIES.md recursion discipline, one source and no cross-source role yet; both named on the book page. Kate Manne likewise (Human Giver Syndrome is attributed, she’s a philosopher and would page in ../philosophy-wiki if she pages anywhere). avoid-ai-writing run. Verify deferred per hub policy (content-only, no page moves).

[2026-07-29] ingest | Dopamine Nation (Anna Lembke, 2021)

Fifth book routed here today and the best-evidenced, at T2 — not because the source is stronger but because the criticism is documented. Verified from Lembke’s Wikipedia entry plus several critical secondary readings rather than from memory. New pages dopamine-nation, anna-lembke.

The split is unusually clean. Supported: chronic overstimulation degrading reward sensitivity, and abstinence restoring it, is established addiction neuroscience. Disputed: the pleasure-pain seesaw oversimplifies dopamine’s relationship to pleasure; “dopamine fasting” names something you cannot do to a neurotransmitter and implies days where desensitization recovery runs weeks; the anti-dopamine framing reads to neuroscientists as an argument that ADHD medication is overprescribed, against the evidence; and addiction specialists hold that the abstinence emphasis undervalues medication-assisted treatment — the one critique with direct consequences for what a patient is told to do. Lembke conceded the central simplification in the New York Times.

It sharpens this morning’s amendment a third time. The four therapy books forced “clinical” off the method spectrum as a position. Lembke moves the variable again: she is a Stanford professor with 100+ peer-reviewed papers on exactly this literature, so no credential or tradition axis explains the gap. The variable is register — the same author produced the peer-reviewed record and the trade book her field disputes, and the trade book is what a million readers act on. The founding note of this wiki (popular writing compounds the replication problem) is now visible with both steps inside one career. Also gives the corpus its sharpest evidential contrast: human-mouse-perceptual-maps, one experiment nobody outside the field has heard of, against a household phrase whose author walked back its core.

No concept page for dopamine or the pleasure-pain balance: one source, folded in per the ENTITIES recursion discipline, and flagged here — a second addiction/reward source makes both warranted. Cross-spoke note recorded on the page rather than split: the digital-overstimulation thesis is load-bearing for products in ../web-browsers-wiki (opera-air sells distraction-blocking), which is a commercial reason to want the claim correct. avoid-ai-writing run. Verify deferred per hub policy (content-only, no page moves).

[2026-07-29] route | Dopamine Nation (Lembke) → psychology-wiki (runner-up: none)

Fifth book to psychology-wiki today, and it lands in the same clinical corner the morning’s four opened. Lembke is a Stanford addiction psychiatrist, so the spoke’s method rule (clinicians and experimentalists here, philosophers next door) applies without argument. Routed rather than parked despite arriving as a bare title: verified against Wikipedia plus documented critical reviews, which made it the day’s best-evidenced book ingest — the criticism exists, unlike for the other four. It refined psychology-wiki’s method-spectrum amendment a third time, moving the variable from discipline to register.

[2026-07-29] ingest | The Body Keeps the Score (van der Kolk 2014)

Arrived via Telegram as a bare Wikipedia lede — note the “purported effects” wording, which is the encyclopedia hedging in the first sentence, and a fair preview of what the entry then documents. Fetched the book entry plus van der Kolk’s own; T2, and better evidenced on the critical side than any book routed here so far, because the criticism is peer-reviewed rather than commentary.

Three pages: the-body-keeps-the-score (source), bessel-van-der-kolk (Person), psychological-trauma — the spoke’s first clinical subject page, and the first entry under MedicalCondition. No separate page for EMDR or the somatic treatments: one source, folded into the trauma page per the ENTITIES recursion discipline. A second trauma-treatment source makes one warranted. McNally, Cox & Codd and Nietfeld are named but not paged, consistent with how the other books’ critics were handled.

Why it moved the synthesis rather than just adding to it: the register finding from dopamine-nation had impressions on both sides. This one has ~300 NYT bestseller weeks against a 2025 BJPsych Bulletin audit of 122 claims. It also breaks the assumption hiding inside the Lembke ingest — that criticism eventually produces a correction. McNally’s critique landed in 2003, the book followed in 2014, and the peer-reviewed objections since have moved the readership not at all. And the stakes changed: the named harm is readers steered off CBT, not effort wasted. Opened one question: developmental trauma disorder’s actual diagnostic status, which no source here gives and which changes how the whole book reads. avoid-ai-writing run. Verify deferred per hub policy (content-only, no page moves).

[2026-08-02] ingest | Death and How to Be More Alive: Carl Jung on Immortality and the Rebirth of Middle Age (The Marginalian)

Routed here by the hub from Telegram. T3 — maria-popova in the-marginalian, 2026-07-31, assembled from Jenny Yates’s Jung on Death and Immortality plus letters. Third Popova source in this spoke, second on carl-jung.

Pages. Source summary jung-on-death plus a concept page second-half-of-life. Updated carl-jung (death is his other long subject here), index and synthesis. Two entities created: maria-popova and the-marginalian — deliberately paged now rather than at first contact, because three sources through one essayist is a property of the corpus rather than a coincidence. entity-index matched neither. Wolfgang Pauli and Jenny Yates stay prose mentions: subjects of the material, not agents around it.

The provenance work is the substance of this ingest. Popova quotes at length, names her anthology and dates the material, which is why the wiki now has Jung’s own sentences. But the connective argument between the quotes is hers and it is literary: she reads Jung’s spaceless, timeless unconscious through his dinner with Einstein and the curvature of spacetime, and none of the Jung she quotes mentions spacetime. Recorded on the source page as the reason the tier is T3, and generalized in synthesis as a standing rule for reading her.

Two smaller flags. The essay is named “Death and the Soul” on first mention and “The Soul and Death” twice after — an inconsistency inside one piece, left unresolved here rather than silently picked. And synchronicity appears as context with no argument attached, so it stays a mention.

Two things it changed in the corpus.

The spoke got its first substantive disagreement between sources. irvin-yalom treats mortality as a given to be faced with no promise about what follows (staring-at-the-sun); Jung calls death the goal of a directed life-process and what follows it “unspeakably glorious.” Two clinicians, one fact, incompatible consolations, neither citing the other, and nothing in this corpus could settle it — which places it on the founding methodological tension one layer down.

A lineage arrived backwards. the-middle-passage (ingested 2026-07-29) is james-hollis writing in a Jungian series about a second adulthood; this source is Jung’s own version, parabola quote included. second-half-of-life holds both and says what their agreement is worth: a school agreeing with its founder, relayed by the same essayist, is one source wearing two hats.

Content-only, no page moves — verify deferred per the hub’s standing policy. avoid-ai-writing run over the new prose.

[2026-08-03] ingest | In the Absence of the Ordinary (Francis Weller, via The Marginalian)

Routed from the hub (Telegram). New: in-the-absence-of-the-ordinary (T3 source summary), francis-weller (Person), grief (DefinedTerm/clinical-subject). Updated: maria-popova + the-marginalian (three sources → four), existential-psychotherapy (the shared instruction), synthesis (new section “Welcome the symptom: three schools, one instruction” + sharpened selection-filter note + two open questions), index. Substance: uncertainty as initiation, grief as apprenticeship, “welcoming all of who we are into the story. Nothing excluded.” Third independent statement of the corpus’s welcome-the-symptom instruction after the-middle-passage and existential-psychotherapy — three lineages, one stance, no evidence from any of them. acceptance-and-commitment-therapy is the testable relative (900+ trials on a similar stance) and connecting the two is now an open question. New in kind: scale. Weller claims a collective initiation (“we are tumbling through a rough initiation”), where everything else here puts the structure inside a person. Only prior move that way is erich-fromm‘s social character — and Fromm named a mechanism. Evidence: the far edge of the spoke’s method spectrum. T3 rates Popova’s transmission, not the content — the quotes are reliable, the statements have no truth conditions (“soul navigates the twining trail between sovereignty and intimacy”). the-middle-passage at least had an untestable mechanism; this has an image. First source here on a subject with a large clinical literature it never mentions, so grief exists mainly to name the gap (prolonged grief disorder, the unsupported stage model, process models) — all explicitly marked as NOT from this source. Selection filter now hard to call subtle: four of the wiki’s sources come through one essayist and all four sit at the depth-psychological end. Recorded in synthesis with the correction — the next few sources should come from elsewhere. Entity: francis-weller paged (author). Verify deferred per hub policy (content-only). avoid-ai-writing run.

[2026-08-03] ingest | Feeling Good: The New Mood Therapy (David D. Burns, 1980)

Routed from the hub (Telegram, Amazon listing). New: feeling-good (T2 source summary), cognitive-behavioral-therapy (MedicalTherapy), david-d-burns (Person), aaron-beck (Person). Updated: acceptance-and-commitment-therapy (the tradition it departed from), sigmund-freud (Beck’s break as a consequence of the untestability objection), synthesis (two new sections + two open questions), index. Fetch: the routed Amazon page gave almost nothing through WebFetch and only carousel junk through Firecrawl; the product-details block yielded the edition facts (William Morrow Paperbacks reprint, 1999-10-01, 736pp, ISBN 978-0380810338, 4.5/5, rank #476,469) and the jacket copy. Argument, reception and the bibliotherapy studies come from the Wikipedia article, so the page is tiered T2 on that with the split stated in a Provenance section and a precedent note added to index. Substance: CBT was a total gap — no page, no Beck, despite the ACT page already contrasting itself with “the cognitive-therapy move”. Two findings. (1) First bestseller here that was tested as itself: a meta-analysis of six studies on self-directed bibliotherapy, eleven more on the guided version, Scogin et al. 1989 — reading the book was the intervention, which inverts the corpus’s bestseller-then-audit pattern (the-body-keeps-the-score, dopamine-nation, thinking-fast-and-slow). Caveats kept attached: small n, therapist contact in several, 1994 survey. (2) A real contradiction: CBT says correct the distorted thought; acceptance-and-commitment-therapy says disputing it is more entanglement. Same object, opposite operation, both descended from Beck’s break, both with large trial literatures and no comparison held here. Unlike Jung-vs-Yalom this one is not beyond evidence, which makes it the most useful contradiction the spoke has. It also gives the “welcome the symptom” instruction (logged yesterday) a rival rather than merely an absence of evidence. Register problem, sharpest instance yet: Burns attacks antidepressant trials for drug-favouring bias while his own jacket says depression can be “cured without drugs” and works “immediately”. Distinguished on the page from bessel-van-der-kolk — a modest true claim sold immodestly is not an unsupported claim. No study named here has been read; the meta-analysis has no author/year/journal in the source. Verify deferred per hub policy (content-only). avoid-ai-writing run.

[2026-08-04] ingest | Does Language Determine Thought? (TheCollector)

Routed here by the hub (runner-up: philosophy-wiki). T4 — the spoke’s second, and its index header demands the justification: record the study behind a claim, or say that a source gave none. This one gave none. It names Sapir and Whorf and no other researcher, offers colour terms and emotion vocabulary as its entire evidence, and its author’s stated credential is a BA. Ingested under the soft gate with the weakness recorded, never refused.

Opens a new subject: language and cognition. Flagged as such rather than forced into an existing thread, per the Ingest rule. New pages (2): does-language-determine-thought (source) and sapir-whorf-hypothesis (DefinedTerm). Sapir and Whorf themselves are deferred — the source gives names and a century and no biography, which is below the evidence-only bar; page them when a real source on either arrives.

Where it connects, without forcing it. perceptual-learning is this spoke’s account of structure you did not choose — internal maps that form the same way in humans and mice. Linguistic relativity proposes unchosen structure from a different source, one you acquired socially. And where heuristics-and-biases documents systematic error that is species-level equipment, a relativity effect would be systematic error that varies by language community — a category of finding the spoke holds nothing else in.

Filed against ../philosophy-wiki on method, the line both spokes were spun out to hold. The version of this question that survived is empirical: it is settled by testing whether speakers of different languages actually perceive or remember differently, not by arguing about what thought must be. Same call the hub made when it declined to route the Le Guin essay to philosophy for the mirror reason.

Folded into “The corpus’s own weak spot” rather than a new dated section. It sharpens that section into a stronger claim: the spoke’s problem was “we read about primary work instead of reading it,” and this is the first case of an entire subject entering the corpus without its evidence base being touched at all. New growth edge #4 — experimental work on language and categorization.

[2026-08-05] ingest | The Dunning-Kruger effect isn’t what everyone thinks it is (Silicon Canals)

Routed by the hub from Telegram. dunning-kruger-misread — Justin Brown, Silicon Canals, 5 August 2026. T3. Popular corrective on what Kruger & Dunning (1999) measured, what the internet turned it into, and the twenty-year argument over whether the effect is real.

New pages (4): dunning-kruger-misread (source, NewsArticle), dunning-kruger-effect (DefinedTerm, cognitive-bias), plus two entity nodes — silicon-canals (NewsMediaOrganization) and justin-brown (Person). Updated (5): heuristics-and-biases, replication-crisis, synthesis.md, index.md, this log.

Tier reasoning. Six studies named, three with authors. That is well above does-language-determine-thought (T4, cited nobody) and short of anything primary. The publisher’s own about page discloses AI-assisted drafting with human editorial judgment, sourced-and-linked claims and 48-hour corrections — a stronger stated process than the-marginalian, and it caps at T3 for the same reason: the review is of the writing, not of the science. The 2021 Nature replication, the strongest evidence in the piece, arrives with no authors and no title.

Why it matters more than a debunking post normally would. Two things the corpus did not have.

A fourth failure mode, and it is not weak evidence. The 2022 Frontiers in Psychology case is that noise on a bounded scale reproduces the Dunning-Kruger pattern in data with no effect in it. That is the opposite complaint from non-replication: the result never vanishes, and bigger samples sharpen the artifact instead of exposing it. replication-crisis got a new section for it, and it lands on heuristics-and-biases where the programme is most exposed — systematic rather than random was the claim that made it a science, and a floor plus a bounded scale supplies systematic direction free.

The first named finding that held. A 2021 Nature replication, ~4,000 participants in each of two studies, supporting the metacognitive reading. Open question 1 asked which heuristics-and-biases results survived; this is the first answer of any kind, and it arrives beside an argument that surviving would not settle it either way. Recorded as partial.

Also a new kind of distortion. Every overreach on file here has an author at the end of it — a jacket, a bestseller, a practice being sold. This one accumulated downstream: the paper has no time axis, measured performance and self-estimate in one sitting, and never contained the “Mount Stupid” curve. Synthesis gained “The distortion with no author behind it.”

Entities. Created 2 of a budget of 8. Dunning and Kruger deliberately not paged — ../ENTITIES.md discovers the agents around a source, not its subject, and they are the subject; the source also supplies no affiliation, biography or second work to page them from. Noted in index.md beside the standing Tversky deferral. entity-index returned two MATCH hits (david-d-burns for “David Dunning”, justin-le for “Justin Kruger”) — both fuzzy first/last-name collisions on different people, overridden by judgment as the tool’s output invites.

Growth edge 4 added: the Dunning-Kruger primary literature. The spoke now holds a contested finding, four studies arguing over it, and none of the papers — edges 1 and 2 in their sharpest form, and four of the five are identifiable from what this source gives.

Selection. Answers the standing note that sources should stop arriving via one essayist. A tech newsroom arguing about a specific published result is the experimental end the Popova run never got near.

[2026-08-05] ingest | Kruger & Dunning (1999) + Magnus & Peresetsky (2022) — both primaries

Curator instruction after the Silicon Canals route: “get the 1999 paper and the 2022 critique.” Both fetched and read in full. The spoke’s first two T1 sources and its first primary research of any kind — growth edge 1 (zero peer-reviewed sources) is closed.

New pages (7): kruger-dunning-1999 and magnus-peresetsky-2022 (sources, ScholarlyArticle, both T1); entity nodes david-dunning, justin-kruger, jan-magnus, anatoly-peresetsky (Person) and cornell-university (CollegeOrUniversity). Rewritten: dunning-kruger-effect — now written from the papers rather than from a summary of them. Updated: dunning-kruger-misread, replication-crisis, heuristics-and-biases, synthesis.md, index.md, this log.

Fetching. The 1999 PDF needed the firecrawl fallback after two dead mirrors (avc.edu 404s, atlas.org serves a viewer shell); the Michigan SASI copy scraped clean at 72 KB. The 2022 paper is open access at Frontiers. Citation and abstract for the 1999 paper cross-checked against PubMed (PMID 10626367).

The finding this ingest exists for. magnus-peresetsky-2022 argues the DK pattern is a statistical artifact — a three-parameter doubly-censored tobit model, noise plus a random floor and ceiling and no psychological term, fits 665 students’ self-predicted exam grades almost perfectly (“there is an effect, but it does not reflect human nature”). It never engages Study 4. kruger-dunning-1999 anticipated the regression objection in print, twice, and built Study 4 against it: ten minutes of logic training moved the trained bottom quartile from an estimated 55th percentile to the 44th and from “5.3 correct” to “1.0 correct,” while an untrained control did not move. Censoring cannot produce that. In a smaller slip pointing the same way, Magnus and Peresetsky state that top-performer underestimation was “not discussed in Kruger and Dunning (1999)” — it has a titled section there, The Burden of Expertise, with a false-consensus account and a phase of Study 3 built to test it.

No secondary source in this corpus reports that gap, and neither paper alone shows it. It is visible only from holding both, which is the clearest justification the spoke’s method has yet had. Filed as a new contradiction in synthesis.md (its first between two peer-reviewed sources) and as a new subsection on what reading primaries actually bought.

Self-correction. Working from dunning-kruger-misread this morning, the spoke wrote that the 2022 analysis showed the pattern arising “in data with no effect in it.” The model is fitted to real exam predictions and the claim is that the fit needs no psychological parameter. Corrected on dunning-kruger-effect with the change marked in place, per the never-silently-overwrite rule, and scored against the source on dunning-kruger-misread.

Entities. 5 new nodes, budget 8, depth 1. david-dunning and justin-kruger were correctly deferred on the morning’s ingest (subject of a source, which ../ENTITIES.md excludes) and qualify now as its authors — a clean worked example of the rule, noted in index.md. VU Amsterdam and HSE Moscow left as plain mentions: bounded and real, but unlikely to recur in this spoke, and cornell-university earns its node because the entire 1999 evidence base is one undergraduate pool in cells of 11–37.

Evidence caveat applied. Both papers’ limits recorded on their pages: Cornell undergraduates throughout for 1999 plus the humour study’s shaky inter-rater criterion, which the authors concede; for 2022, the authors’ own note that only 14% of observations exceed 60, plus the unstated and larger one that a single cross-sectional dataset is the whole basis.

Growth edges rewritten. Edge 1 closed. Edge 2 (second-hand accounts everywhere else) is now the top edge and the 2006 Science income-happiness paper is the named next target. Edge 4 narrowed to the literature still missing — Krueger & Mueller (2002) and Kruger & Dunning’s published reply first, to see whether the earlier round of this argument also passed Study 4 by.

[2026-08-09] ingest | Epictetus via Space Daily — a popular source that got Dunning-Kruger right

Arrived by Telegram; routed here by the hub over philosophy-wiki (logged there as runner-up). The Epictetus quote is the frame, but the substance is metacognition and the accuracy of a popular account of a bias this spoke already holds five pages on. Same author and series as kahneman-focusing-illusion. T3.

New pages: epictetus-learning-spacedaily (source), illusion-of-explanatory-depth (concept), mal-james (Person) and space-daily (Organization) — both entities paged on their second source, per ../ENTITIES.md.

Two things it added.

1. A mechanism the corpus lacked. Fisher, Goddu and Keil (Yale, 2015; nine experiments): searching the internet inflates felt understanding without changing knowledge. It gets merged with dunning-kruger-effect constantly and should not be — rank versus depth, arrived-with versus inducible. The distinction is now on its own page and in synthesis.
2. A counter-case to the spoke’s own drift. synthesis had been treating the runaway Dunning-Kruger myth as the thing a popular source relays by default. This one relays the correction, citing Eric Gaze in Scientific American, in the genre most exposed to the myth. The lesson recorded is that genre is not a proxy for accuracy: does-language-determine-thought is T4 for citing nobody while a daily quote column cites four researchers.

Weaknesses recorded. Every study at second hand — the Yale work gets three authors and a year but no title or venue; the Gaze critique gets no date. And the essay welds two different problems together without noticing: Epictetus describes someone who will not begin, Fisher’s subjects are willing and simply miscalibrated. Motivation against calibration, and the remedy differs.

[2026-08-10] ingest | Dysgraphia via dyslexia — the first developmental disorder, and an undisclosed interest

Routed from the hub (Telegram). Psychology Today, coral-ps-hoh, 2026-08-09. T3 — edited, reviewed and referenced, which is this spoke’s ceiling for popular writing.

New pages: dysgraphia-via-dyslexia (source), dysgraphia and dyslexia (MedicalCondition), coral-ps-hoh (Person), psychology-today (Organization; entity-index returned nothing ≥ threshold).

What it opens. The spoke’s first developmental learning disorder, and its first subject where the interesting question is an intervention rather than an explanation. The argument: dysgraphia is a language disorder the name points away from — a child who can’t retain a word copies it letter by letter, and the messy page is what’s left after that effort — so treat it with the dyslexia remediation evidence rather than with classroom accommodations.

Two things recorded against it.

The sourcing is unusually good and the load-bearing weld still isn’t tested. Four references with volumes and pages (Chung 2020; Han & Wang 2025; Berninger 2009, 2002) — better than anything else in the popular shelf here. But nothing cited compares corrective against compensatory outcomes, which is the claim the piece exists to make. Third instance of the citation chain terminating in a practice pattern, with the untested join moved from mechanism to comparison.

No disclosure. Hoh is CEO of EduNational LLC and the piece argues against the free thing (extra time, oral testing) and for the purchasable one. That doesn’t make her wrong — Berninger’s research predates her — but it is what a reader can’t see. Second instance after anna-lembke, and the venue’s model makes it structural rather than personal.

Also folded in: the corpus now has two accounts of why writing is hard — why-writing-feels-hard‘s social-threat story and this linguistic one — which don’t conflict because they address different populations. And dyslexia is paged deliberately thin: the whole argument transfers dyslexia’s evidence and this wiki holds none of it. A dyslexia remediation outcome study is now worth more here than a second dysgraphia source.

Prevalence flagged, not corrected: “an estimated 20% of the population,” cited to Chung et al. and identical to the dyslexia figure it’s compared to, from a review rather than a prevalence study, unread here.

[2026-08-12] ingest | Kahneman et al., the income–happiness line: primary and revised (quality cycle)

Growth edge 2 half closed. It had named the source to get: the 2006 Science income-happiness paper. Two arrived, because reading the first one forward found that its central quantity had been revised.

kahneman-income-happiness-2006 (T1, Science 312:1908–1910, read in full from the UCLA copy). focusing-illusion had been carrying this paper’s conclusion from an essay about it since the spoke was founded. The page now holds the figures: household income against life satisfaction 0.32, against share of day in a good mood 0.20, against duration-weighted momentary happiness 0.06 (n=740, Columbus DRM). The naming experiment is the better instrument — predicted bad-mood gap between the income bands 32 pp, actual 12.2 pp, and the same overshoot for marriage, supervision and health insurance at P < 0.001. Recorded as a Perspective, not a research report: part literature review, part first presentation of the authors’ own data.

income-happiness-conflict-resolved (T1, PNAS 120(10):e2208661120, 2023, plus the authors’ 2024 reply). Killingsworth, Kahneman and Mellers reanalysed Killingsworth’s experience sampling as an adversarial collaboration: the plateau holds only near the 15th percentile of happiness and only to about $100,000; above that, happiness keeps rising with log(income) and accelerates in the happiest group; the original flat line came from ~85% of high-income responses at the scale maximum. Below $100,000 the slope is steeper at unhappy quantiles, above it flatter, each P < 0.00001.

Access recorded on the page. PNAS returns 403 here for both the article and the PDF, so the 2023 paper is held from its PubMed abstract, with every number taken from the open-access 2024 reply that restates them. No press summary used.

Why this matters past the topic. The spoke’s standing question is which heuristics-and-biases findings survived. It now has three shapes rather than two: one that broke (thinking-fast-and-slow Ch. 4), one that held (dunning-kruger-misread‘s 2021 replication), and one corrected to a narrower claim by its own author. Written into synthesis as a contradictions entry — the corpus’s only settled dispute — with the caveat that the reconciliation is correlational by the authors’ own statement.

Entity: matthew-killingsworth (Person). daniel-kahneman already held.

Edge 2 rewritten to what is left of it: Jung is still entirely essayists and encyclopedias. Two dated section headings dropped from synthesis (hub quality-log.md Outstanding item).

[2026-08-12] ingest | Handwriting vs typing notes — the finding that shrank

Routed here from the hub the moment it arrived by Telegram. The subject is learning and memory measured experimentally, which is this spoke’s declared ground, and the piece is structurally a replication-crisis story — which is where most of its value ended up.

handwriting-vs-typing-notes (T3, silicon-canals). Three studies assembled against the popular “paper beats keyboard” claim: a 2024 Educational Psychology Review meta-analysis of 24 studies across 21 articles putting handwriting’s test-performance advantage at Hedges’ g = 0.248 and typing’s advantage on volume recorded at g = 0.919; a 2021 direct replication of Mueller & Oppenheimer (2014) with 74 laptop and 68 longhand note-takers finding “neither group performed better” on the immediate quiz; and a 2026 Frontiers in Education EEG study of 33 South Korean undergraduates over three ten-minute video lessons, also finding no significant difference.

note-taking-and-learning (new Thing page) holds the reconciliation. The three studies only conflict under the popular framing. Under the mechanism the article itself proposes — handwriting is too slow to transcribe, so it forces compression, and compression is what a later test rewards — the nulls are predicted: an immediate-recall test measures before the effect has anywhere to land. The mechanism also makes the claim transferable to any capture method that forces compression, which no source here tests. Recorded as a prediction, not a finding.

The synthesis gained a fifth form of evidence failure, and it is the most ordinary one. The four already there are non-replication, the citation chain, the artifact, and the distortion with no author. This is the finding that shrank — survived, at a quarter of a standard deviation. Argued in synthesis.md that this is probably the most common fate of a famous result and the hardest for this corpus to notice, because g = 0.248 cannot be said in a headline, so the 2014 version circulates unopposed without anyone defending it.

A publisher pattern is now confirmed rather than suspected. silicon-canals gave this spoke dunning-kruger-misread on 2026-08-05 with three of six citations lacking an author. This piece names no researchers and no institution for the meta-analysis carrying its headline number, and none for the 2021 replication. Two for two. The page now states the rule plainly — the outlet reports findings accurately and citations incompletely, so it is a good lead and a poor terminus — which matters because both its pieces are load-bearing on replication-crisis.

Gaps recorded, not filled. Mueller & Oppenheimer (2014) is discussed on two pages here and held nowhere; the 2024 meta-analysis and the 2021 replication cannot be looked up from what the article gives; and every study reported tests immediately, while the mechanism predicts the gap should grow with delay. Four pages touched: two new, plus replication-crisis and silicon-canals.