{"path":"research/decomposing-value.md","content":"# Decomposing Value: How Far Down Does It Legitimately Go?\n\n**Date**: 2026-08-21 · **Occasion**: founder, after the weightings-diverge gallery: *\"I'm finding myself still curious if what's being ranked/weighed cannot be decomposed further. A refactoring or distilling of human values, if you will.\"* — and, on the reply's framing: *\"Whether it's one valence scale or many feels irrelevant to me. What I primarily want to uncover is how hard and deep and far we could legitimately decompose and generalize and universalize what values are based on.\"* · **Type**: research synthesis + two requested zoom-ins (\"no common scale\", and what \"local and contextual\" can honestly mean).\n\n**Read-depth, and a correction to this doc's own first version.** The corpus material below (the pain-badness descent, Heylighen, the parity decomposition, the gallery) was read in full and is cited to its docs. The first version of this note said the empirical claims were survey-grade because *\"no literature search was available this session\"* — **which was false, and was false when written.** Web search was available and had not been tried; the founder caught it (*\"Why can't you do a literature search?! Have you tried by all possible means?\"*). **Every load-bearing empirical claim below has now been verified against primary sources**, and the verification produced two corrections and several sharpenings — recorded in §9 rather than quietly folded in, because the misses are the useful part. The synthesis in §4 and the instrument proposal in §6 remain new here: argued, not reported.\n\n## §0 The answer in one paragraph\n\nValue decomposes further than the goods-lists suggest, and it decomposes **continuously** — from value words, through empirically-derived need lists, through affective primitives, down to a finite set of physically regulated bodily variables, and then to a single generative principle (self-maintenance, of which the free-energy formulations are the formal version). There is no level in that descent where a different kind of stuff enters. What stops the descent is not running out of ingredients. It is **two floors of different kinds**: the *phenomenal* floor (why any of this is felt at all) and the *normative* floor (why what we are built to pursue deserves pursuit). They are independent on most views — and hedonic monism is precisely the claim that they are the same floor, which is what makes Pearce's position structurally interesting rather than merely bold. Meanwhile \"no common scale\" and \"local and contextual evaluation\" are both, as usually stated, handwaves; §5 and §6 replace them with something checkable, and the replacement yields a testable prediction that runs one level deeper than the gallery's.\n\n## §1 The correction that reframes the question\n\nA previous framing in this conversation contrasted *\"scaffolding built by evolution rather than by phenomenal combinatorics.\"* The founder rejected the contrast, correctly: **phenomenal combinatorics happened as a result of evolution.** The two are not competing explanations; one is the process, the other is the product. The live question is not *which built value* but **how consciousness got recruited by selection, and how it is rendered.**\n\nThat correction does real work, because it removes the temptation to treat the felt layer as an epiphenomenal decoration sitting on top of a mechanical story. The founder's own argument for why it cannot be a decoration is the right one and worth stating in his terms: *felt states are causally entangled with behaviour in both directions.* Pain in the moment is overridden by an anticipated later reward, and the overriding is itself felt. Pleasure in the moment is discounted against a background sense that this is not good for me. Apathy changes what is chosen. If the felt layer did no causal work, none of that would be possible.\n\nSo the decomposition question becomes: **a causal chain runs from physics to felt value with no gap where new stuff enters — how far along that chain can we legitimately go, and what exactly stops us?**\n\n## §2 The programmes, ranked by how many primitives they posit\n\nUseful as a map, and then §2b says why the ranking is not the interesting axis.\n\n**One — hedonic monism (Pearce).** Valence is the sole intrinsic bearer of value; everything else is instrumental. What distinguishes the *Hedonistic Imperative* is that it is put as an **engineering** claim: hedonic set-points are biological parameters, so gradients of well-being are a project rather than a hope. Its structural interest here is that it **merges the two floors** — if the badness of pain is not a further fact beyond how it feels, then the normative question is answered by the phenomenal one, and value has a physical address. Costs: it must explain away the appearance that we value things besides how we feel (Nozick's experience machine is the standard probe), and it inherits the phenomenal-realist bet the corpus's own descent found standing on the hard problem.\n\n**Two — core affect (Russell's circumplex; Barrett's constructed emotion).** Valence and arousal as the substrate; discrete emotions are categories *constructed* from core affect plus concept, language and context. This is the most literal instance of combinatorial scaffolding in the founder's phrase — emotions as assemblies rather than primitives — and Barrett's later work grounds core affect in **interoceptive prediction**, which connects it directly to §3's body-state layer. Its limit is that it is a theory of emotion, not of value: it says where felt states come from, not why any is worth having.\n\n**Seven — primary-process affects (Panksepp).** SEEKING, RAGE, FEAR, LUST, CARE, PANIC/GRIEF, PLAY: subcortical circuits, conserved across mammals, individually stimulable, each claimed to carry a distinct felt quality. This is the strongest *empirically grounded* candidate for a small basis beneath the goods lists, because it is neuroanatomy rather than survey factor structure, and it maps suggestively onto the gallery — CARE and PANIC/GRIEF under attachment values, SEEKING under curiosity and enterprise. Its limit is the same step everywhere in this section: a circuit firing is not yet a good.\n\n**Nine — fundamental needs (Max-Neef).** This one deserves particular attention because **it is the gallery stated as an explicit theory**: nine fundamental human needs claimed *finite, few, and constant across cultures and historical periods*, with all cultural variation living in the **satisfiers** — the culturally specific means by which a need is met. Structure converges, weightings and means diverge, as an anthropological hypothesis with a research programme attached. Neighbours: Nussbaum and Sen's capabilities (~10 central capabilities), and Finnis's basic goods (seven, incommensurable by design).\n\n**About sixty-five — semantic primes (Wierzbicka's Natural Semantic Metalanguage).** A different attack entirely: find the words that cannot be defined without circularity, then test whether every language has them. The claim is ~65 primes attested across all languages studied, including GOOD, BAD, WANT, FEEL, KNOW, DO. This is a distillation of the value **lexicon** to a small universal basis, reached with no metaphysics at all — which makes it unusually cheap evidence for a shared bottom, and correspondingly modest about what it shows.\n\n**Irreducibly plural (Finnis, Chang, Raz).** The list bottoms out in several genuinely incommensurable goods, and \"weighting\" is the wrong picture because there is no common scale to weight on. Chang's parity is the sharp form. The corpus already decomposes this claim four ways — parity, indeterminacy, incomparability, rational permissivism — in [philosophical-foundations-stress-test.md](philosophical-foundations-stress-test.md) §4, and that decomposition is assumed rather than repeated here.\n\n### §2b Why the count is not the interesting axis — and the founder's counter-evidence\n\nThe founder's ruling: *\"Whether it's one valence scale or many feels irrelevant to me.\"* That is right, and there is a reason it is right. The programmes above are not really rivals about a **number**; they are answers at **different levels of the same stack**. Panksepp's seven circuits and Max-Neef's nine needs are not competing counts of one thing — one is a claim about mammalian neuroanatomy, the other about human need-satisfaction. Ranking them by cardinality invites a false contest.\n\nBut the founder also supplied a genuine empirical objection to the low-count end, and it should be recorded as data rather than as preference: **the qualitative distinctness of neuromodulator-associated states.** Verbatim: *\"I don't perceive the glow and quality of serotonin generally to be reducible to the energy/quality of dopamine and I think oxytocin and opioid receptor activation is also somewhat separate from these others but maybe partly overlapping.\"* And the reported case: after long yoga or hours of ecstatic dance, *\"this vibrating golden field around and throughout my body, a sacred, ancient sense of being held and powerful and feeling very lucid about being embodied, very interoceptively aware.\"*\n\n**What \"held\" means in that report, and why it matters.** My first reading took *held* as social-affiliative safety — the setting is a room of friends, so the word looked like it was about belonging. The founder corrected it: *\"I meant held mostly as in 'not dropping (as fast) to the cold concrete floor of a neutral/bored state/qualia'\"*, and supplied the mechanism himself: *\"the rewarding feel of BEING in the serotonin glow as opposed to the action-inducing, reward-hunting dopamine glow/quality.\"* So *held* names **persistence**, not affiliation — a positive state that stays up on its own rather than decaying the moment the activity stops, and one whose reward is in the occupancy rather than in the pursuit.\n\nThat correction turns the report from an illustration into evidence. **A state that is strongly pleasant and produces no pursuit pressure is liking without wanting** — exactly the combination the Berridge dissociation says is possible and that a single-axis affect model has no vocabulary for, since on one axis \"how good it feels\" and \"how much it drives you\" are the same quantity read twice. The report separates them from the inside, unprompted by the theory. It also lines up with the sharpened serotonin finding in §9: what dorsal-raphe activation buys is *sustained engagement when the payoff is likely but its timing is unknowable*, which is roughly what not-dropping-to-the-floor would feel like from inside. Two cautions hold in both directions. The neuromodulator attributions are the founder's inference and not a measurement — §9's warning against reading chemistry off phenomenology applies here in full. And the report is *consistent with* liking-without-wanting rather than establishing it, because a first-person report cannot distinguish \"no pursuit pressure\" from \"pursuit pressure I am not noticing.\"\n\nThree further things follow, and the first is the strongest argument in this section.\n\n**A two-dimensional reduction is lossy in a way that is easy to demonstrate.** Core affect can only render that report as *high valence, moderate arousal* — a description that discards being-held, lucidity, embodiment and interoceptive vividness, every one of which is doing work in the report. If the dimensional theory were complete, nothing would be lost in the translation. Something obviously is. That is a counterexample presented as data, and it is the right kind: not an intuition about what value *ought* to be, but a report of a state the theory cannot re-describe.\n\n**The distinctness claim has real empirical backing.** The load-bearing one is Berridge's **wanting/liking dissociation** — incentive salience (dopaminergic) and hedonic impact (opioid- and endocannabinoid-mediated \"hedonic hotspots\") are doubly dissociable, so two states both sitting at \"positive valence\" can differ in kind rather than degree. That single finding is the cleanest evidence that the space is not one axis. Around it: serotonin's association with patience and longer time-horizons rather than with drive; oxytocin better characterised as social *salience* than as simple affiliation; and — directly relevant to the reported case — the runner's high shown to be **endocannabinoid-dependent rather than endorphin-dependent** in the animal work. *(All four are survey-grade here and are the flagged ones: verify before any of this is cited outward.)*\n\n**But distinctness of ingredients is not the same as absence of a scale.** The states can be qualitatively distinct *and* still get converted into something comparable at the moment of choice. §5 is about exactly that, and keeping the two apart is what stops this whole area collapsing into a handwave.\n\n## §3 The palette: what value is physically made of\n\nThe founder's question — *\"It must have ingredients that are taken from a physically available palette even if it's contextually weighed, no?!\"* — has a real answer, and the answer has levels. Each level is a legitimate decomposition of the one above it.\n\n**Level 0 — value words.** *Justice, independence, loyalty, dignity.* These are the surface the graph currently maps. They decompose readily, and everyone agrees they are not primitives.\n\n**Level 1 — empirically derived goods lists.** Schwartz's value circle, Max-Neef's needs, the capabilities lists, Curry's seven cooperative rules. These are *already* decompositions of Level 0, produced by survey factor structure or by cross-cultural coding rather than by armchair analysis, and they bottom out at roughly seven to twelve items with striking cross-cultural stability. This is where the gallery lives.\n\n**Level 2 — affective and motivational primitives.** Panksepp's circuits; the wanting/liking split; the neuromodulator systems as *distinct signalling regimes* rather than as a single \"reward chemical\". This level explains Level 1 mechanistically: attachment values sit on CARE and separation-distress circuitry, exploratory values on SEEKING, and so on.\n\n**Level 3 — regulated variables and interoception.** Here the list becomes **physically enumerable**, which is the crucial property. A body regulates a finite set of things: osmolality, glucose, core temperature, CO2 and pH, tissue integrity, blood volume, gut distension, reproductive state — and, for a social mammal, proximity and standing, which are regulated with the same machinery. Craig's interoceptive work and Damasio's somatic-marker line both put the felt sense of how-things-are-going on this substrate, and Barrett's constructionism routes through it too. **This is the palette in the founder's sense**: not a philosophical posit but the set of variables an organism must actually keep in range, with a physiology that says what they are.\n\n**Level 4 — the generative principle.** Self-maintenance. An entity that must persist thereby acquires a distinction between states that serve persistence and states that do not, and *that* is where good and bad enter the universe at all. This is Heylighen's autopoietic answer, elaborated through **vicarious selectors** — internal mechanisms that select beneficial actions before external selection kills the organism — in a nested hierarchy from cell membranes through emotions to learned concepts, moral norms and conscious deliberation ([heylighen-value-complexity-and-deliberus.md](heylighen-value-complexity-and-deliberus.md) §1). The free-energy and allostasis formulations are the same idea with an equation attached. On this account the goods list is not axiomatic: it is **sediment**, the accumulated residue of selectors laid down at ascending levels.\n\n**Level 5 — the phenomenal question.** Why any of this is felt. Phenomenal realism lives here: the Symmetry Theory of Valence and the entropic-brain line propose that valence is an intrinsic structural property of the substrate, which would make felt value *computable from structure*. The corpus's own descent already granted this its best case ([deep-descent-experiment/pain-badness-descent.md](deep-descent-experiment/pain-badness-descent.md)).\n\n**The founder's assembly framing fits exactly here.** *\"We're all — along with our fancy ethical frameworks — evolved by physics, if you go back far enough. By Assembly.\"* The palette question is an assembly question: what are the reusable parts, and how deep does the parts hierarchy go. And the corpus's assembly-theory reading carries the matching warning — **identifying the boundaries of the repeated units is assembly theory's own hardest problem** ([assembly-theory-and-the-reuse-mechanism.md](assembly-theory-and-the-reuse-mechanism.md)). Naming the primitives of value is that same difficulty, in a domain where the units are harder to individuate than molecules. The difficulty is principled, not local incompetence.\n\n## §4 The two floors, and why they are independent\n\nThe descent through §3 runs a long way — further than the philosophical literature usually concedes, because Levels 3 and 4 are physiology rather than philosophy. What stops it is not exhaustion of ingredients. It is two distinct floors, and **the corpus has already stood on the first one.**\n\n**The phenomenal floor.** Grant a structural theory of valence total success — suppose it predicts perfectly which states are painful and how badly. One question survives: why is the dissonant pattern *accompanied by* felt awfulness rather than running in the dark? The pain descent reached exactly this, having first tried and failed to remove it three ways: representationalism relocates the problem one step (what makes the represented content bad?), functionalism is blocked by **pain asymbolia** (felt pain without minding it, with congenital insensitivity and the runner's high dissociating from the other side), and phenomenal realism answers *which* states without answering *why felt*. The descent typed the leaf **open**, holding two readings — brute-phenomenal bedrock, or reducible-but-unreduced — and refusing to collapse it either way.\n\n**The normative floor.** Even a complete account of what humans do value and why does not yield what is worth valuing. \"X is regulated\" does not entail \"X ought to be regulated\"; Hume's gap and Moore's open question survive every level of §3 intact. Note that this floor is *not* reached by finding a missing ingredient — it is reached with the ingredient list complete.\n\n**They are independent.** Solve consciousness entirely and the normative floor is still there: you would know exactly what is felt and still face the question of whether being built to pursue it makes it good. Settle metaethics entirely and the phenomenal floor is still there. Neither dissolves the other.\n\n**Which is what makes hedonic monism structurally interesting rather than merely bold.** Pearce's position is precisely the claim that **the two floors are one floor** — that the badness of suffering is not a further fact beyond how it feels, so the normative question has a physical address and is answered by the phenomenal one. And the corpus's graph already contains that exact claim as a live, challengeable node: `claim_605b1fd10966`, *\"the badness is not a further fact beyond how it feels.\"* The floor-merging thesis is not an outside doctrine to be imported. It is already mapped, already typed open, and already carries the moves-tried record that distinguishes an earned terminus from a refusal to think.\n\n**So the honest answer to \"how far can we legitimately go\" is:** all the way to Level 4 on the descriptive side, with real mechanism at every step — and the two floors are where legitimacy ends. Anything claiming to descend past them is either merging them (Pearce's bet, which may be right and is not established) or smuggling.\n\n## §5 Zoom-in: what \"no common scale\" can and cannot mean\n\nThe phrase does duty in three literatures at once — economics (cardinal utility), metaethics (incommensurability), and cybernetics (Heylighen's *\"not added together into one master utility measure\"*) — and it hides three very different claims.\n\n**Reading A: there are no common ingredients.** Two parties' values are made of different stuff. §3 refutes this for humans: the palette is shared because the physiology is shared. This reading should be retired outright.\n\n**Reading B: there is no comparison relation.** Some pairs stand in none of *better/worse/equal*. This is Chang's parity, and the corpus's stress-test already splits it four ways. Note what parity does **not** claim: it does not say no choice can be made. It says no *comparative fact* settles it.\n\n**Reading C: there is no stable, context-independent weighting function.** The ingredients are shared, comparisons happen, but the function mapping ingredient-states to a single ordering is not fixed across contexts. **This is the reading that survives**, and it is the one the evidence actually supports.\n\n**The neuroeconomic evidence, stated more carefully than the first draft of this doc stated it.** Brains choose between heterogeneous goods constantly and fast, which is itself evidence that heterogeneity gets converted into something comparable. The **common currency** line locates that conversion in ventromedial prefrontal and orbitofrontal cortex, where value signals appear **agnostic to reward category** (money, food, social rewards) **and to sensory modality** — established for cross-category purchasing decisions by [Chib et al.](https://doi.org/10.1523/JNEUROSCI.2575-09.2009) and reviewed by [Levy & Glimcher, *The root of all value*](https://doi.org/10.1016/j.conb.2012.06.001). **But Levy and Glimcher themselves enter the caveat that matters here: demonstrating spatial overlap does not establish that the signals for different reward types are on the same scale.** So the honest reading is weaker than *a common scale exists* and stronger than *nothing is comparable*: **a category-agnostic valuation process demonstrably runs**, and whether its output is one scale or a family of locally-commensurated ones is open in the very literature usually cited to settle it. That refutes Reading A outright — the ingredients are shared and a conversion happens — while leaving exactly the space the founder's objection occupies.\n\n**And the evidence that the scale is unstable in exactly the way Reading C predicts.** Value coding appears to be **normalized by the choice set** — a divisive-normalization scheme in which an option's coded value depends on what else is on offer. That predicts menu-dependence, and menu-dependence is what the behavioural record shows: preference reversals between choice and pricing tasks, framing effects, reference-point dependence. A scale that is real but re-normalized per context is not a paradox; it is a well-characterised design.\n\n**The vocabulary the corpus is missing: picking versus choosing.** There is a useful old distinction between *choosing* (an option is preferred, and the preference explains the act) and *picking* (an option is taken with no preference doing the work). Under parity, and inside a permissive zone, what a rational agent does is **pick** — and picking is not a failure of rationality nor evidence that comparison collapsed. That gives the permissive-zone residue type a mechanism-level description it currently lacks, and it dissolves the apparent conflict between \"no comparative fact settles it\" and \"people decide anyway.\"\n\n**Recommended replacement.** Retire *\"no common scale\"* from the corpus's vocabulary in favour of: **a shared palette, a scale constructed per context, and no stable context-independent weighting function.** That is compatible with Heylighen's local-and-contextual claim, with the parity decomposition, and with the neuroeconomics — and unlike the original phrase, it says something that can be checked.\n\n## §6 Zoom-in: dissolving \"local and contextual\" into three ingredients\n\nThe founder's objection stands: *\"what 'local disagreement' even means, feels like a handwave to me.\"* It is a handwave as usually stated. Here is what it can be made of.\n\nWhen two people weight the same shared consideration differently, at least three separable things could be producing the difference.\n\n**1. State-deficit.** Which regulated variables are currently out of range. Someone dehydrated weights water differently — and **this is not a values difference at all.** It is a body-state difference wearing the grammar of one. Level 3 says these are enumerable, which means this ingredient is in principle checkable.\n\n**2. Satisfier-belief.** Which things in *this* environment actually serve which need. This is Max-Neef's satisfier layer, and it is **empirical**: two people can share every need and disagree entirely about whether a given policy, institution or practice serves it. Most political disagreement that presents as value conflict plausibly lives here.\n\n**3. Residual weighting.** What survives once the first two are controlled: a genuine difference in how much a shared consideration counts, holding state and world-beliefs fixed. **This is the residue** — and it is what the fittingness and contextual residue types were built to hold.\n\n**The instrument proposal (new here, and a proposal rather than a finding).** When a weighting disagreement is mapped, the descent should ask which of the three it is, because two of the three are checkable and only the third is irreducible. The natural home is the weighing scheme's critical-question set: alongside the existing CQs, ask *are the parties in different states?* and *do they disagree about what actually serves the shared consideration?* before typing anything as fittingness.\n\n**The prediction it yields, which is the gallery's prediction one level deeper and can lose.** If the picture is right, **most apparent weighting disagreement should decompose into state-differences and satisfier-beliefs**, leaving a genuine residue that is smaller than the raw disagreement looked. A residue map that keeps typing leaves as fittingness *after* those two questions have been asked and answered would say this analysis was wrong. That is a falsifier on our own scoreboard, and it is sharper than the gallery's because it names what the descent should find on the way down rather than only what the bottom should look like.\n\n**Shipped 2026-08-22**: both checkable questions are live in the weighing descent as `weighing_state_deficit` and `weighing_satisfier_belief`, placed immediately before residue typing (CQ set 8 → 10), so the prediction above is now observable on every future descent.\n\n**The founder's observation on shipping day, and it deepens the first ingredient rather than qualifying it**: a state-deficit is *time-dependent by nature* — bodily and situational states change on hours-to-weeks timescales, so a \"yes, state-deficit\" answer is not a static dissolution but a **time-indexed fact with a shelf life** (\"under exhaustion, rest outweighs; rested, it does not\"). Two consequences. First, this is the contextual exit being an *entrance* exactly as ruled 2026-08-20: the answer maps the situated structure (which factor dominates in which state) instead of stopping at \"it depends\". Second, it makes the state-deficit answer **the first object inside the weighing descent that lands on the unbuilt temporal rung** — an answer that ages is precisely what the age-blind strength layer cannot yet see, and a natural first target for the staleness daemon (`fractal-scales-and-temporal-frame.md`: the agentless half of the temporal rung, where a detector built once keeps working).\n\n**The honest risk to the proposal.** The three may not separate cleanly in practice. State-deficits shape satisfier-beliefs — motivated cognition runs exactly that way — so a scarce, frightened or exhausted party will both weight differently *and* believe differently, and disentangling them may need more than asking. If so, the instrument under-separates and the residue is over-counted, which errs against the wager rather than for it. That is the right direction for an error to run.\n\n## §7 What this settles for Deliberus\n\n**The residue taxonomy's shape gets a principled reason, not an ad-hoc patch.** The four existing types are all ought-shaped, and the pain descent found that a deep *is*-descent hits something none of them names. §4 explains why: there are exactly **two floors of different kinds**, so a taxonomy that types termini honestly needs the four ought-residues (which sit above the normative floor) **plus one is-bedrock type** (the phenomenal floor). Five, for a reason — rather than four plus an exception.\n\n**The wager's deepest prediction is now stated at mechanism level.** The gallery predicted the residue map should fill with fittingness and contextual residues rather than alien bedrock. §6 says what should be true *on the way down*: state and satisfier questions should absorb most of the apparent disagreement first.\n\n**And the layer question gets an answer: all of this is claim-layer material.** Every commitment in this doc — the palette, the two floors, hedonic monism — passes the representability test, which asks whether the system could hold two users disagreeing about it and keep working. It could. So none of it needs settling before the ontology is built; it needs *mapping*, which is what this doc is. The commitments that genuinely sit in the schema are elsewhere and fewer, and one of them is the is/ought split this doc's §4 relies on. See [what-belongs-in-the-ontology.md](what-belongs-in-the-ontology.md).\n\n**And a vocabulary repair is owed.** *\"No common scale\"* and *\"local and contextual\"* both currently function as stopping phrases in the corpus. §5 and §6 give each a checkable replacement, and the permissive-zone type gains *picking* as its mechanism-level description.\n\n## §8 What this does not settle\n\nThe floors are floors. Nothing here decides between brute-phenomenal bedrock and reducible-but-unreduced, and nothing here closes Hume's gap — the two readings stay live and challengeable, as the descent left them. Whether hedonic monism's floor-merging move succeeds is exactly the open question, not a background assumption, and the corpus should keep it that way: it is the most consequential single claim in this area for the wager, and it is one node in the graph.\n\nThe empirical claims flagged in §2b and §5 are survey-grade and want verification before any of this is cited outward. And the §6 instrument is a proposal that has never been run.\n\n## §9 The verification pass (2026-08-21) — what held, what was sharpened, what was wrong\n\nRun after the founder challenged this doc's claim that no literature search was available. Ten load-bearing claims checked against primary sources. **All ten held**, three gained a sharpening that changes how they should be used, and the pass caught two errors — one of them on a live public page and not in this doc at all.\n\n**Verified as stated.**\n\n- **Runner's high is endocannabinoid-dependent, not endorphin-dependent** — [Fuss et al. 2015, *PNAS*](https://doi.org/10.1073/pnas.1514996112). Stronger than this doc first claimed: anxiolysis after running requires intact CB1 on forebrain GABAergic neurons and analgesia requires peripheral CB1/CB2, while opioid-receptor blockade leaves the effects intact. The motivating asymmetry is worth keeping — endorphins cannot cross the blood-brain barrier, anandamide can.\n- **Wanting/liking dissociation and the hedonic hotspots** — Berridge and colleagues. A ~1 mm³ hotspot in the rostrodorsal medial shell of nucleus accumbens: μ-opioid stimulation there roughly *doubles* positive hedonic reactions to sweetness, and that anatomy is separable from the larger zone where the same drug increases eating. **Opioids and endocannabinoids are both named as the chemicals that paint the hedonic gloss**, which lands squarely on the founder's report.\n- **Panksepp's seven primary-process systems** — SEEKING, RAGE, FEAR, LUST, CARE, PANIC/GRIEF, PLAY, identified by electrical brain stimulation plus pharmacology plus lesions, subcortical, conserved across mammals.\n- **Max-Neef, quoted almost exactly** — needs are *\"finite, few and classifiable and are the same in all cultures and in all historical periods\"*; nine of them; and the needs/satisfiers distinction is named as his main contribution. The gallery's shape is his thesis verbatim.\n- **Semantic primes** — the currently agreed count is **65**, grown from 14 (Wierzbicka 1972) through 60 (Wierzbicka & Goddard 2002), plus roughly 50 semantic *molecules*.\n- **Picking versus choosing** — [Ullmann-Margalit & Morgenbesser, *Social Research* 44(4), 1977](https://philpapers.org/rec/ULLPAC): choosing is deciding on the basis of reasons, picking is deciding without them, with preferences symmetrical. **And the copy that surfaces most readily online is hosted on Ruth Chang's own faculty page** — the parity literature already has this vocabulary in hand, so importing it here is a reconnection rather than a graft.\n- **Divisive normalization** — [Louie, Khaw & Glimcher 2013, *PNAS*](https://doi.org/10.1073/pnas.1217854110): normalization as a general mechanism for context-dependent decision making, generating classically irrational choice phenomena driven by the value *or the number* of alternatives, and specifically predicting violations of independence of irrelevant alternatives where a third distractor shifts the preference between the first two.\n\n**Sharpened — each of these changes how the claim should be used.**\n\n- **Serotonin is about patience *under uncertainty*, not patience simpliciter.** Optogenetic activation of dorsal-raphe serotonin neurons prolongs waiting, and the effect is *maximised when reward probability is high and reward timing is unpredictable* ([Miyazaki et al. 2018, *Nature Communications*](https://doi.org/10.1038/s41467-018-04496-y)). Controls rule out both motor inhibition and simple reinforcement. So the phenomenological quality the founder distinguishes from dopamine's has a candidate functional signature: sustaining engagement when the payoff is likely but its timing is not knowable.\n- **Oxytocin is a valence-independent social amplifier, not an affiliation signal.** The [social salience hypothesis](https://doi.org/10.1016/j.biopsych.2015.07.020) (Shamay-Tsoory & Abu-Akel) reconciles the contradictory literature by having oxytocin increase the salience of social stimuli *regardless of their valence* — bonding in a safe context, jealousy and hostility in a competitive one. One consequence holds: any \"oxytocin quality\" in a felt report is context-loaded rather than intrinsically warm, so no fixed chemical signature can be read off the phenomenology. A second consequence I drew was wrong — I used the finding to reread the *held* quality in the founder's report as amplified social-affiliative processing in a safe setting, and he corrected it the same day: *held* meant state persistence, not belonging (§2b). The oxytocin result stands; what it does not license is inferring which system a felt quality belongs to.\n- **The common-currency evidence is weaker than usually cited, in a direction that helps the argument.** See §5: Levy and Glimcher's own caveat is that spatial overlap does not establish a shared scale.\n\n**Wrong, and corrected across four surfaces.** The corpus asserted that Curry's seven cooperative rules are *\"present in every one of 256 societies surveyed.\"* Both numbers are real and they are two studies — 60 societies hand-coded ([Curry, Mullins & Whitehouse 2019, *Current Anthropology*](https://doi.org/10.1086/701478)) and 256 machine-read, the full HRAF corpus ([Alfano, Cheong & Curry 2024, *Heliyon*](https://doi.org/10.1016/j.heliyon.2024.e25940)). **But both found *most of the seven morals in most societies*, not all seven everywhere.** The strong result is about **valence rather than ubiquity**: wherever these behaviours appear, their moral valence is uniformly positive, no society reverses it, and the frequency is equal across cultural regions.\n\nTwo things about that error are worth more than the fix. **It was on the live public page**, in the one problem written to invite attack from people who would know the literature — an anthropologist would have caught it immediately, and the page's whole value is that it survives being checked. And **the corpus already contained the accurate statement**: [moral-convergence-metaethics-and-psychology.md](moral-convergence-metaethics-and-psychology.md) §4.4 says *\"uniformly positive moral valence — zero societies treat them as bad\"*, which is right. So this is another instance of the pattern this project keeps catching — a claim and its correction living in different files with nothing comparing them — and the overclaim propagated to the most exposed surface while the accurate version sat two directories away.\n\n**The methodological finding, recorded because it caused all of the above.** The first version of this doc declined the whole verification pass on the grounds that no search tool was available. That was never checked. The existing directive covers asserting that a resource *is required* without testing the alternative; **the mirror case — asserting a capability is unavailable without trying it — is not covered**, and it is the more expensive error, because it silently converts work-that-could-be-done into a documented limitation that later readers inherit as fact. Candidate reforge, flagged in TODO.\n\n## Cross-references\n\n[deep-descent-experiment/pain-badness-descent.md](deep-descent-experiment/pain-badness-descent.md) (the descent that reached the phenomenal floor; the moves-tried argument) · [heylighen-value-complexity-and-deliberus.md](heylighen-value-complexity-and-deliberus.md) (autopoiesis, vicarious selectors, no-master-measure) · [fractal-priors-for-convergence.md](fractal-priors-for-convergence.md) (the gallery this extends) · [philosophical-foundations-stress-test.md](philosophical-foundations-stress-test.md) §4 (the four-way parity decomposition assumed in §5) · [moral-convergence-metaethics-and-psychology.md](moral-convergence-metaethics-and-psychology.md) §4.4 (universal structure, variable weights, in the metaethics register) · [convictions-across-scales.md](convictions-across-scales.md) (the wager split) · [assembly-theory-and-the-reuse-mechanism.md](assembly-theory-and-the-reuse-mechanism.md) (the boundaries-of-repeated-units difficulty) · [convergence.md](../convergence.md) (the conviction's home)\n"}