Holocene Minds on Hard Mode: Schizophrenia, Bipolar Disorder, and the Birth of Consciousness

TL;DR

• Akbari & Reich’s ancient DNA time series show strong directional selection in West Eurasians: schizophrenia and bipolar polygenic scores fall, while intelligence, educational attainment, and income scores rise over the last ~10–14k years. ¹
• Treat those polygenic score (PGS) shifts as rough shifts in underlying liability and you get stark numbers: early Holocene Europeans plausibly had 5–8× more schizophrenia, ~4–5× more bipolar disorder, ~10–12 IQ points lower mean ability, ~2 fewer years of schooling, and substantially lower “income potential.”
• Stack those traits and you get a psychology very different from ours: a world with one adult in twenty sliding into psychosis, another one in fifteen into manic-depressive storms, and a fatter left tail of cognitive limitations.
• Push the same slopes back 20k, 40k, 80k years as a thought experiment and the curves explode: naive extrapolation quickly yields double-digit psychosis rates and IQ gaps measured in multiple standard deviations, hinting that pre-Holocene minds lived in a radically different cognitive ecology.
• All of this reads naturally as quantitative support for the Eve Theory of Consciousness (EToC): recursion and selfhood appear as a dangerous innovation, heavily selected over the Holocene as lineages with more stable, higher-bandwidth selves gradually replace fragile, god-ridden architectures.

Companion article: For a deeper dive into schizophrenia specifically and its connections to the Eve Theory, see “Ancient DNA Shows Schizophrenia Risk Purged Over 10,000 Years”.

“There is nothing more difficult to carry out, nor more doubtful of success, nor more dangerous to handle, than to initiate a new order of things.”
— Machiavelli, The Prince

1. Ancient DNA as a history of consciousness#

The Akbari–Reich manuscript does something new with ancient DNA: rather than hunting for one-off sweeps, it fits time trends in polygenic scores for modern traits—schizophrenia, bipolar disorder, intelligence, educational attainment, income—across ~8,000 ancient and ~6,500 present-day West Eurasians. ¹

The headline result is simple and brutal:

• Alleles that raise risk for schizophrenia and bipolar disorder have been systematically pruned over the Holocene.
• Alleles that improve scores on IQ tests, increase years of schooling, and raise household income have been systematically favored. ¹

On the face of it this is a story about health and socioeconomic status. But if you’ve read the Eve Theory of Consciousness (EToC), it’s hard not to see something deeper:

The Holocene wasn’t just about wheat and cities. It was a 10,000-year debugging session on the human self.

To make that vivid we need to translate polygenic slopes into intuitive numbers. That means picking a model and actually running it.

2. One piece of math, used ruthlessly#

We’ll use the standard liability–threshold model for psychiatric diagnoses.²

1. Assume there’s a latent trait (liability) distributed ~N(0,1) in the modern population.
2. A fixed threshold (T) defines diagnosis: cross it and you’re schizophrenic, bipolar, etc.
3. Modern lifetime prevalence pins down (T).
4. If the mean of the liability distribution shifts by (\mu) (in SD units), the probability mass above (T) changes accordingly.

Akbari et al. give us the Holocene slopes for polygenic scores (γ): roughly

• Schizophrenia: γ ≈ –0.84 SD (early Holocene genomes ≈ +0.84 SD riskier than modern).
• Bipolar disorder: γ ≈ –0.67 SD.
• Intelligence: γ ≈ +0.79 SD.
• Household income: γ ≈ +1.11 SD.
• Years of schooling: γ ≈ +0.61 SD. ¹

Instead of treating PGS as a small noisy slice of liability, we will take the blunt approach: use γ as a proxy for the actual shift in liability mean over ~10,000 years and see what sort of world drops out.

Once for each trait, then we’ll stack them.

3. Schizophrenia: living near the edge of the cliff

3.1 From 0.7% to ~5%: how much more psychosis?#

Modern lifetime prevalence of schizophrenia sits around 0.3–0.7%, with ~0.7% a widely cited figure. ³

In a standard normal:

• 0.7% in the upper tail corresponds to a threshold (T \approx 2.46) SD.

Now shift the mean liability backward by +0.84 SD at 10,000 years ago:

• Modern: mean 0, threshold 2.46 ⇒ 0.7% above threshold.
• Early Holocene: mean +0.84, effective threshold = 2.46 – 0.84 = 1.62 SD, which corresponds to ≈ 5.3% in the tail.

That’s a 7.5× jump in prevalence.

Round a bit and you get the easier mnemonic:

Today: ~1 in 150 people develops schizophrenia. Early Holocene West Eurasians: roughly 1 in 20.

In a village of 200 adults, that’s not one haunted kid on the edge of town; that’s ten adults, scattered across families and lineages, slipping into chronic hallucinations, delusions, and disorganized behavior.

3.2 Not just the tail#

Liability is continuous. Raise the mean and you don’t only push more people over the diagnostic line; you fatten the subclinical cloud below it:

• More people hearing fleeting voices, seeing patterns in clouds and shadows, feeling watched.
• More people living with weak reality testing, sliding into paranoid explanatory frameworks under stress.

Schizophrenia in modern psychiatry is the catastrophic version of something that exists in mild forms across the whole distribution. Raising the mean by nearly a standard deviation shifts the entire landscape towards internal voices and fractured agency.

From an EToC perspective, this is exactly what you’d expect from a mind that has just discovered it can think about itself and is now doing so badly: the system keeps confusing its own generated content with external agents.

4. Bipolar disorder: overclocking the recursive loop

4.1 Four or five times more mania#

Global lifetime prevalence of bipolar spectrum disorders today is around 1–2%, with ~1.5% a reasonable central estimate. ⁴

1.5% in the tail of N(0,1) gives a threshold (T \approx 2.17) SD.

Apply Akbari’s bipolar slope γ ≈ –0.67 SD:

• Modern: mean 0, threshold 2.17 ⇒ 1.5% prevalence.
• Early Holocene: mean +0.67, effective threshold = 2.17 – 0.67 = 1.50 SD ⇒ ≈ 6.7% prevalence.

So bipolar spectrum conditions go from 1–2% to ~7%: a 4–5× amplification.

In that same 200-adult village with its ten schizophrenic villagers, you now have:

• ~14 people who, over their lifetimes, tilt into true manic or hypomanic episodes plus recurrent depressions.

Some of them will be devastatingly impaired. Others will be the classic “successful hypomanic”: charismatic, high-energy, aggressive risk-takers who periodically crash.

4.2 How mania interacts with a fragile self#

Bipolar disorder is temperamentally different from schizophrenia, but the architecture relates:

• Mania is overclocked recursive processing: racing thoughts, inflated goals, decreased need for sleep, self-importance.
• Depression is the same system collapsing, turning its recursive power against the self.

In a world where consciousness is still half-externalized—where gods, ancestors, and spirits are the default explanation—manic states are easily read as possession or inspiration. They become:

• Prophets, war-leaders, charismatic founders of cults, booming with certainty.
• And, when they crash, living proof that the gods are dangerous.

Selection in such a world is not simply “bipolar bad”: some mix of risk-taking, creativity, and hyper-focus can be intensely adaptive, but only if the self doesn’t fly apart. The Reich slopes strongly suggest that, over the Holocene, lineages in which manic architecture usually did fly apart slowly lost ground.

5. Intelligence: a double-digit gap

5.1 How big is 0.79 SD in real terms?#

The intelligence PGS panel in Akbari et al. shows γ ≈ +0.79 SD over the Holocene. ¹

Standardize IQ with mean 100, SD 15. If we take that 0.79 SD seriously as a shift in underlying ability, that equates to:

[ 0.79 \times 15 \approx 12\ \text{IQ points}. ]

Call it 10–12 points.

So:

Average West Eurasians around 8–10k years ago were, on this approximation, roughly a two-thirds-sigma below modern descendants on tasks captured by IQ tests: pattern detection, working memory, abstract reasoning.

This is not an insult; it’s a statement about means. Populations overlap massively; the smartest farmer in 7000 BC still out-thinks plenty of moderns. But in aggregate:

• Complex hierarchies of inference are harder to maintain.
• Fewer people live comfortably above, say, IQ 120.
• The tail below IQ 70 is significantly thicker.

5.2 What happens to the tails?#

With a 0.8 SD downward shift:

• Modern fraction below –2 SD (IQ < 70): ~2.3%.
• Shift mean by –0.8 SD: effective threshold becomes –1.2 SD; tail ≈ 11.5%.

So the fraction of people with very severe difficulty handling abstraction and multi-step planning might jump from ~2–3% to ~10–12%.

In a small group, that matters. It constrains:

• Who can manage irrigation and storage plans.
• Who can keep rituals, genealogies, and obligations straight in their heads.
• Who can invent and decode the first scripts.

The Golden and Silver Ages in EToC become not just mythic labels but statistically different cognitive regimes.

6. Educational attainment: the will to endure symbols#

Educational attainment (EA) is a proxy for something like “how long can you stand to sit still in symbolic culture before bailing.”

The EA3 GWAS (Lee et al. 2018) found that one SD of EA PGS predicts about 3.6 years of schooling in contemporary cohorts. ⁵

Akbari’s EA panel gives γ ≈ +0.61 SD. ¹

Multiply:

[ 0.61 \times 3.6 \approx 2.2\ \text{years}. ]

So the genomes of early Holocene West Eurasians “wanted” about two fewer years of schooling than modern ones, holding institutions constant.

Of course, there were no schools. But if literacy, numeracy, or priestly training had been options, our model predicts:

• Fewer people willing to grind through long apprenticeships in symbol land.
• More people whose temperaments revolt against extended abstraction—boredom, restlessness, affective dysregulation.

The cultural byproduct is obvious: fewer scribes, fewer legalists, fewer full-time philosophers; more reliance on oral tradition, direct experience, and ritual.

7. Household income: climbing the social maze#

Household income PGS is noisy but not meaningless. Hill et al. (2019) show that it overlaps heavily with EA and IQ, and accounts for around 10–11% of variance in income in UK Biobank. ⁶

Akbari’s income panel shows γ ≈ +1.11 SD. ¹

Even without overfitting the exact units, a one-sigma shift is big. A rough interpretation:

• Modern genotypes are, on average, wired for substantially higher success in complex, formalized economic environments—roughly a doubling of “income capacity” if you map liability to log income.
• Ancient genotypes are more likely to stall out in narrow, local niches.

In EToC’s language, this is a shift from small-band opportunists to stable-role navigators: brains that can track institutional rules, future payoffs, and reputational games over long horizons.

8. Putting it together: the Holocene mental ecology#

Here’s a compact table summarizing the 10k-year contrast, using the model sketched above:

Now imagine a population where these all co-occur.

8.1 The three-layer structure#

You get something like a three-layer mental ecology:

1. Expanded low tail

• Many more individuals who struggle with anything beyond immediate, concrete tasks.
• Planning a multi-year irrigation project or complex caravan route is out of reach.
• Myth and ritual must be encoded in short, vivid, highly redundant forms.

2. Swollen unstable mid-tail

• Quite a lot of people smart enough to see beyond the village but with fragile self-models: psychotic episodes, mood storms, paranoid ideation.
• These are the shamans, seers, cult founders, war-prophets—and many of the casualties.

3. Thin but powerful high tail

• Very few individuals able to manage high-level abstraction and maintain stable mood and reality testing.
• When they appear, they become nodal points around which culture crystallizes: lawgivers, architect-priests, founders of lasting cults.

From the Eve-Theory perspective, this is exactly what a species looks like when it has:

• Recently discovered recursive self-awareness,
• Not yet fully stabilized the trait, and
• Is now undergoing an extended period of selection on ways to be a self.

The Holocene slopes are not about “health” vs “disease”; they’re about pruning architectures that break too easily and amplifying ones that can handle large-scale abstraction without melting down.

9. Pushing the line: 10k, 20k, 40k, 80k years ago#

Now for the fun, slightly irresponsible part.

Akbari’s trends are roughly linear over the sampled period. Treat γ as the change over ~10k years and extend the line backwards as a toy model just to see how quickly things blow up.

9.1 Schizophrenia through time#

Recall:

• Threshold (T_{sz} ≈ 2.46).
• γ_sz ≈ –0.84 per 10k years.
• Mean shift at time (t) (kyr BP) ≈ (\mu(t) = 0.084 \times t) SD.

Compute a few points:

By 20k BP, the naive line already claims one in five adults meets modern schizophrenia thresholds. By 40k, it says most people do; by 80k, essentially everyone.

Obviously that cannot literally be true. The point of the exercise is the opposite:

A simple linear extrapolation of the Holocene trend back into the Upper Paleolithic immediately saturates the trait, which tells you both that (a) selection must have already been acting earlier, and (b) the modern category “schizophrenia” stops making sense in deep time.

At 40k or 80k years ago, whatever was going on in human minds, it wasn’t “80% of people have the DSM-5 disorder.” It was something like:

• Inner voices, visions, and loosely bound agency are so common that they’re the default mode of cognition;
• Continuous, stable, single-self subjectivity (Eve’s insight) is rare and unstable when it appears.

The model is screaming that pre-Holocene minds lived in an environment where what we call psychosis was not a deviation but the background against which stable selves gradually emerged.

9.2 Bipolar through time#

Same drill for bipolar:

• Threshold (T_{bp} ≈ 2.17).
• γ_bp ≈ –0.67 per 10k ⇒ μ(t) ≈ 0.067 × t.

Again, absurd if taken literally—and that’s the point. Extend the Holocene trend far enough and it describes a world of almost universal mood instability. Somewhere before that, the curve must bend. Or, more interestingly: the architecture that later manifests as bipolar may have been something like the default motivational regime before long-term, role-based selves stabilized.

9.3 Intelligence and schooling back in time#

For IQ we have γ_IQ ≈ +0.79 per 10k. Interpreted as a shift in SD units of ability:

So if you play the same linear game, the average person 40k years ago would, relative to us, score around IQ 50 by our norms—not because they’re “defective” but because the tests are calibrated to a very different niche.

Again, this is not meant as literal psychometrics for Cro-Magnons. It’s a vivid way of saying:

The architectures that make modern abstract reasoning so easy for us have probably been under selection for a very long time, and the Holocene slice we can see is only the last steep section of that climb.

The same applies to EA and income: extend γ_EA and γ_income and you quickly hit worlds where almost nobody can tolerate long schooling or manage complex economic roles. That fits uncomfortably well with the archaeological record of small, loosely organized hunter-gatherer bands for most of Homo sapiens’ history.

10. Eve Theory: consciousness as a lethal innovation#

The Eve Theory of Consciousness argues, in cartoon form, that:

1. Recursive self-awareness is a late development in human evolution.
2. It first stabilizes in women (Eve), with men following later via selection on social and sexual success.
3. The initial emergence of the self is not a perk but a trauma: you acquire guilt, anxiety, and awareness of death before you have cultural tools to handle them.
4. Cultures respond with rituals and myths that both induce and contain self-awareness—snake cults, ordeal rites, initiation, mystery traditions.
5. Over millennia, lineages whose brains can host a self without disintegrating out-reproduce those whose brains cannot.

Akbari/Reich’s cognitive panels, interpreted the way we’ve just done, are a quantitative sketch of exactly this process:

• The psychosis and extreme mood tails are trimmed. You still see schizophrenia and bipolar today, but at a fraction of early Holocene rates. That’s what you’d expect if catastrophic failure modes of the self are strongly selected against.
• At the same time, intelligence, educational tenacity, and income-linked traits are pushed upward. These are proxies for the kind of stable abstraction and institutional navigation that a fully conscious self excels at.
• The joint distribution shifts from “many unstable prophets, few boring bureaucrats” to “few prophets, many boring bureaucrats.”

The Y-chromosome bottleneck around 5–7k years ago—when male-line diversity collapsed by ~90% in many regions—adds another layer: something pruned male lineages very hard, likely linked to success in highly competitive, patrilineal social structures. ⁶ If EToC is roughly right, the men who could keep their minds intact while wielding new levels of strategic violence, law, and mythology left disproportionately many sons.

On this reading:

The gods didn’t just “die” as secularization crept in; the brains that needed gods in their heads lost ground to brains that could run a self locally.

11. Modern psychiatry as archaeology#

If schizophrenia and bipolar disorder today are the attenuated remnants of once-common architectures, several consequences follow:

• Psychiatric disorders are evolutionary fossils. They’re not just “diseases” but windows into how minds used to be wired when inner voices and manic inspiration were adaptive more often than not.
• Religion retains a psychosis-shaped hole. Scriptural accounts of prophecy, possession, and revelation look less like fantasy and more like stylized descriptions of minds closer to the Holocene mean.
• The “mental health crisis” is partly a mismatch problem. We now live in hyper-symbolic, high-bandwidth environments that demand continuous, stable self-monitoring. People near the historical mean for psychosis liability—let alone above it—are under chronic strain.

For EToC, the moral is simple:

We are not the default humans. We are the survivors of a long, bloody selection gradient favoring minds that can stare into the mirror without shattering.

FAQ#

Q1. Are you claiming ancient people were “crazy” and “stupid”?
A. No. The claim is that, on average, early Holocene West Eurasians had higher risk for psychosis and mood disorders and lower scores on modern-style cognitive tasks. They were running older firmware in harsher conditions; the ones who could stabilize consciousness rebuilt the world.

Q2. How solid are these effect sizes?
A. The directions (psychosis down, cognition/EA/income up) are robust in the Akbari/Reich data and in complementary work in Eastern Eurasia. ¹ The exact multipliers depend on modeling choices; the numbers here are best read as vivid upper bounds, not precise epidemiology.

Q3. Doesn’t culture matter more than genes?
A. Culture matters enormously, but Akbari et al. explicitly detect allele frequency changes inconsistent with drift, which is genetic selection. Culture is the environment creating the fitness differences; the genome is where those differences are recorded.

Q4. What would falsify this picture?
A. If denser ancient sampling before 10k BP showed no further increase in psychosis PGS, or if non-European lineages showed orthogonal trends, the EToC-style reading would weaken. Likewise, if future GWAS radically changed estimates of PGS–liability mappings.

Q5. Why tie this to Eve Theory rather than generic “gene–culture coevolution”?
A. You don’t have to. EToC simply provides a sharpened narrative: that the specific trait under selection is not “intelligence” in the abstract but selfhood—the ability to run a continuous, recursively aware, autobiographical subject without slipping into gods or madness.

Sources#

1. Akbari, A., et al. “Pervasive findings of directional selection realize the promise of ancient DNA to elucidate human adaptation.” bioRxiv (2024). ¹
2. Saha, S., et al. “A Systematic Review of the Prevalence of Schizophrenia.” PLoS Medicine 2(5) (2005): e141. ⁸
3. Merikangas, K. R., et al. “Prevalence and Correlates of Bipolar Spectrum Disorder in the World Mental Health Survey Initiative.” Archives of General Psychiatry 68(3) (2011): 241–251. ⁴
4. Savage, J. E., et al. “Genome-wide association meta-analysis in 269,867 individuals identifies new genetic and functional links to intelligence.” Nature Genetics 50(7) (2018): 912–919. ⁹
5. Lee, J. J., et al. “Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals.” Nature Genetics 50(8) (2018): 1112–1121. ⁵
6. Hill, W. D., et al. “Molecular genetic contributions to social deprivation and household income in UK Biobank.” Nature Human Behaviour 3 (2019): 610–625. ⁶
7. Rahman, T., et al. “Schizophrenia: An Overview.” HSMHA Health Reports (2016). ¹⁰
8. Oxley, F. A. R., et al. “DNA and IQ: Big deal or much ado about nothing?” Intelligence 100 (2024): 101768. ¹¹
9. Piffer, D. “Directional Selection and Evolution of Polygenic Traits in Eastern Eurasia: Insights from Ancient DNA.” (2025). ⁶
10. Cutler, A. “Holocene Selection on Human Intelligence.” Snake Cult of Consciousness (2025). ¹²