The question is a ratio, not a binary — and stating it that way is most of the work, because both poles are strawmen nobody defends. Chris’s claim: “we are more the same than we are different.” Not that nothing changed (physiology has changed; what we eat changes things; measured intelligence has moved), but that ends drift far more slowly than means. The page’s central instrument is a checkable asymmetry rather than an intuition: across a 3,000–4,000 year gap, the ends-content of ancient texts is transparent and the means-content is opaque. If ends drifted at the rate means do, both would be equally alien. They are not. That asymmetry is what turns “human nature is basically constant” from an assumption into a claim with evidence — and it is also what lets the claim be weakened to the much more defensible form the rest of the vault actually needs.
Links: The Four Trunks, The Constitutive/Elective Distinction (the trunk this feeds — see §5), Constructed ≠ Arbitrary, Subsistence vs. Participation (the origin — §6c’s means/ends split, and the assumption this page repairs), Positional Goods, Civilizational Cycles, Opposing Forces, Reading Outcome Statistics, Is AI Making People Dumber? (Word War) (Chris’s prior position on IQ, §4b), Scope Confusion
Trunk: sits on a boundary, and load-bearingly rather than decoratively. Trunk 1 — Normative Ontology holds that normative facts are entailed by the structure of the relata. This page asks which properties of the relata are fixed — that is, it supplies Trunk 1’s input parameter. Read through Trunk 4 for how those fixed properties behave under changing environments.
Almost every public version of this argument is fought between two positions nobody holds:
Both are refuted by ordinary observation, which is why the fight is unproductive. The useful question is quantitative:
How much of what a human is stays fixed across time and environment, how much moves, and how would we tell?
Chris’s answer, and this page’s thesis:
Chris: “Are we the same as always before? Can we adapt? No, there are differences — we do adapt. How much has stayed the same? I think more than people give credit… different environments producing different results, and yet we still can empathise with those of yester-year. It seems to me that we are more the same than we are different.”
Note what this concedes up front, and why conceding is what makes it strong: physiology has changed, diet changes outcomes, measured intelligence has moved, and “can I say with certainty that I think and feel like a caveman would in their circumstances? I might even say we wouldn’t.” The claim is a ratio with an admitted numerator, not a constancy claim. That is a far smaller target and a far more defensible one.
Sorting by how fast each layer moves is what makes the ratio tractable:
| Layer | Example | Rate of drift |
|---|---|---|
| Physiology | eating, breathing, heartbeat, reproduction | essentially fixed on historical timescales; real but slow on evolutionary ones |
| Means of satisfying it | what we eat, how we reproduce, how we keep warm | fast — this is where nearly all visible change is |
| Neurology / affect | grief, ambition, disgust, pair-bonding, status-seeking | slow; the contested layer, and the one this page is about |
| Expression / norms | what counts as an insult, what is shameful, marriage form | fast — a means layer wearing psychological clothes |
The load-bearing move is that the middle two are routinely confused. Change in expression is read as change in affect. That marriage forms vary wildly across cultures is a fact about the means of pair-bonding, not evidence that the drive is constructed — and the inference from one to the other is made constantly in both directions.
Chris flags the honest complication: “I won’t even dismiss that what we eat changes things.” Right — the layers are not sealed. Nutrition affects development affects cognition; the means layer feeds back into the substrate. The model is a gradient of rates, not a stack of independent shelves.
This is the page’s contribution, and it is an empirical test rather than a sentiment.
Across a 3,000–4,000 year gap, the ends-content of ancient texts is transparent while the means-content is opaque.
Transparent — you read these as a participant, and the feeling lands without mediation:
Opaque — you can only read these as a historian, never as a participant: ancient physics, humoral medicine, celestial mechanics, metallurgy-as-theory. The technical content is genuinely alien. You do not feel the four humours; you study them.
The inference: both bodies of text crossed the same gap, in the same languages, through the same translators, with the same survival filter. If ends drifted at the rate means drift, both would be equally opaque. The ends-content is not. That is the asymmetry, and it discriminates — it could have come out the other way, and for the technical content it does.
Projection and survivorship. We may be reading our emotions into these texts; translation is itself an interpretive act, and the works that survived, got copied, and got translated are disproportionately the ones that resonated with later readers — a selection effect pointed straight at the conclusion.
The partial answer: translation and survivorship apply equally to the technical corpus, which was also copied, also translated, and also selected. If projection explained the transparency of the ends-content, it should also make ancient cosmology feel familiar. It does not. So the objection weakens the evidence without dissolving the asymmetry.
What would actually settle it: the transparency should hold for texts not selected for literary merit — administrative records, receipts, legal complaints, graffiti. The Ea-nāṣir tablet is exactly that kind of document, which is why it is the strongest single specimen here and not merely the most entertaining. Systematically checking the mundane corpus is the real test, and it has not been done.
The David Reimer case is the standard citation against strong blank-slatism, and correctly so: a boy whose circumcision was botched in infancy, reassigned and raised as a girl on John Money’s advice, who never identified as female, reverted at 14, and died by suicide in 2004. It is unusually informative for an n of 1 because there was an identical twin control.
Bound the claim properly. It refutes the strong blank-slate position on gender identity. It does not establish strong nativism about anything, and a single case — even a controlled one — cannot carry a general thesis about human nature.
The second lesson is the more useful one for this vault. Money reported the case as a success for over a decade, and it circulated in textbooks as evidence for social construction until Milton Diamond and Keith Sigmundson published the follow-up in 1997 and Reimer went public. So the most-cited proof of one position was, on examination, evidence for the opposite — and it took thirty years because nobody checked the primary case. That is the grounding discipline in the wild, and it belongs on this page as a caution about this page: the specimen that everyone cites is the one least likely to have been verified by the person citing it.
Measured IQ rose roughly 3 points per decade through the 20th century. Flynn’s own reading was “scientific spectacles” — a trained facility with abstract classification and hypothetical reasoning, not an increase in raw processing power.
Chris’s version of the question: “if IQ was measured on a full-historical scale, would Aristotle, Newton and DaVinci still stand out? A bit — but Aristotle might be 180 relative IQ in his day and only 160 today.” That is roughly the mainstream reading, and it is means-side: the ratchet running on cognitive tooling while the substrate sits still.
Two things make this support constancy rather than threaten it:
Chris: “pair-bonding still has much of embedded code in it — women still select for strength, size and competency and the like.”
The cross-cultural evidence: Buss’s 37-culture study (1989) found consistent sex differences in stated mate preferences across wildly different societies — the kind of invariance that is hard to get from socialization alone.
The social-role counter, and why it is weaker than it looks. Eagly and Wood (1999) reanalyzed Buss’s data and found sex differences on certain items (earning capacity, housekeeping) shrinking as gender equality rose — which is what social role theory predicts. Zentner and Mitura (2012) reported similarly.
But the largest test did not replicate it. Walter et al. (2020), Psychological Science — 45 countries, with Buss himself among the authors — found the sex differences robust, and gender equality “only partially corroborated” the earlier work: it predicted sex differences in the actual age of partners, but there was no evidence that gender equality predicted preferences.
That split is the important part, and it is not a hedge — it is precisely the prediction this page’s model makes. Preferences (the end) did not move with social structure. Realized outcomes (what people actually got) did. A fixed end pursued through variable means is not a compromise between the two camps here; it is what the largest study found.
Chris’s objection, which is correct and which this section originally had backwards:
Chris: “I thought the egalitarian study showed the opposite? As societies become more egalitarian, gender differences become more amplified. I can understand the ideas behind Eagly–Wood, but it seems it doesn’t go far enough.”
The gender-equality paradox is the finding that sex differences in personality and occupational choice are larger in more gender-equal countries — the opposite sign from what social role theory predicts. Personality is the domain where it appears most reliably: Big Five and Dark Triad traits, self-esteem, depression, personal values, and educational/occupational choice. Falk and Hermle (2018, Science) found gender differences in economic preferences — risk, altruism, trust, patience — larger in richer and more gender-equal countries. Stoet and Geary (2018) reported the same shape for STEM participation.
So Chris’s “doesn’t go far enough” is an understatement of his own objection. A theory that predicts convergence, in a domain that shows divergence, is not merely incomplete — it has the sign wrong. The usual evolutionary reading is unmasking: lift material constraint and innate dispositions express more freely, so equality reveals rather than creates the difference.
This page’s error was presenting a contested literature as settled. Correcting it in the other direction would repeat the mistake, so the live challenges belong here too:
Honest status: the mate-preference replication is clear and favours invariance; the broader paradox is well-replicated in personality but methodologically contested; and the interpretation — unmasking versus artifact versus shifting constraint — is genuinely open. What the section can support is the narrow claim: the convergence prediction has not held up, and the evidence sits closer to a fixed end than this page originally allowed. It cannot support “innateness proven,” and it should not be cited that way.
It repairs an assumption I overstated. Subsistence vs. Participation §6c rests on the ratchet running on means while ends stay fixed, and that section was flagged as depending on “human nature is constant” — its weakest load-bearing claim. That was an overstatement of the requirement. The ratchet argument does not need constancy. It needs only:
Ends drift far more slowly than means.
Which is weaker, is what Chris actually claims, and is what §3 supplies evidence for. The assumption should be downgraded, not defended — and the argument gets stronger for resting on less.
It supplies Trunk 1’s input parameter. Constitutive realism holds that normative facts are entailed by the structure of the relata — given agents with certain properties, the norms follow. That formulation is only as determinate as the property list. If the relata were indefinitely plastic, the entailed norms would be too, and the trunk would collapse into the constructivism it explicitly contrasts with. So Trunk 1 has a standing dependency on some version of this page, and it has been running on an unstated one.
It bounds Constructed ≠ Arbitrary. That page argues constructions are constrained at design time and run time, and therefore evaluable. This page names what they are constrained against at run time: a human whose ends did not move. A construction that stops serving those ends fails, however entrenched it is — which is the mechanism behind §6c’s observation that family and religion return under modern complexity.
The strongest candidates for genuine ends-drift are exactly the ones this framework owes an explanation for and has not paid:
The means-substitution reading is available for each — contraception decoupling sex from reproduction while opportunity costs rose; institutional substitutes absorbing what religion supplied; households restructuring around wage labor rather than production. But “available” is not “established,” and this page does not establish it. Each is a genuine test the thesis could fail, and treating them as obviously means-side would be the exact error §2 warns about, run in reverse.
The honest status: the ratio claim is well-supported for affect and poorly tested for motivation. Grief and irritation are demonstrably stable across four millennia. Whether what people want is equally stable is the open half, and fertility is the case that will decide it.