Ten Papers that Push the Eve Theory of Consciousness Toward Science

TL;DR

• Multiple GWAS show schizophrenia risk variants enriched in human-specific or recently evolved genomic regions, linking psychosis directly to recent brain evolution.^{1 2}
• Genome-dating and ancient DNA work indicate that alleles affecting cortical structure, intelligence, and education shifted after our split from Neanderthals and well into the Holocene.
• Polygenic time-series reconstructions in Europe and Eastern Eurasia show sustained directional selection on cognitive traits, away from psychosis risk and toward higher educational attainment.
• Human-accelerated regulatory regions and neuron types are now tied to language and autism, exactly the “edge-of-stability” phenotype Eve Theory of Consciousness (EToC) predicts. ³
• Recent work on schizophrenia genetics supports a “cliff-edge” by-product model: the same architecture that enabled recursive, language-saturated selves leaves a minority over the edge into psychosis. ⁴

Companion articles: For deeper dives into Akbari & Reich’s schizophrenia findings, see “Ancient DNA Shows Schizophrenia Risk Purged Over 10,000 Years” and “Holocene Minds on Hard Mode”. For broader exploration of schizophrenia as a normal ancestral state, see “Was Schizophrenia Once Normal?”.

“The gods were, in reality, voices.”
— Julian Jaynes, The Origin of Consciousness in the Breakdown of the Bicameral Mind (1976)

Why Go Looking for Evidence in the Genome#

Eve Theory of Consciousness (EToC) makes a rude claim about our species: selfhood is a late, fragile adaptation riding on top of a much older control system, and the transition from “Golden Man” to fully self-aware humans was biologically violent. Psychosis, voices, myth, and the sense of a personal “I” are not separate topics; they’re different cross-sections through the same evolutionary event.

If that picture is roughly right, you’d expect three kinds of biological evidence:

1. Recent evolutionary change in brain and cognitive traits – especially after the human–Neanderthal split, intensified in the late Pleistocene and Holocene.
2. Tight entanglement of language, social cognition, and psychosis risk – the same regions and regulatory programs doing double duty.
3. Signs of ongoing selection on psychiatric risk – schizophrenia and bipolar disorder as costs of a system pushed up against a cliff edge, not free-floating “malfunctions.”

The genome won’t give us a movie of Eve waking up in her own head, but it should show that cognition, language, and psychosis are (a) unusually human, (b) unusually recent, and (c) under strong selection. The ten papers below are not written to support EToC, but if you read them with that hypothesis in mind, they fit together disturbingly well.

Ten Papers at a Glance#

Call this list a small fossil record of the Snake Cult in our DNA.

1. Human Evolution Markers Are Enriched in Schizophrenia Risk#

Srinivasan et al. (2016) asked a deceptively simple question: are schizophrenia SNPs randomly scattered across the genome, or do they pile up in regions that look “recently human”? They intersected GWAS hits with two kinds of evolutionary annotations: Neanderthal selective sweep markers and other human-specific evolutionary regions. They found that schizophrenia risk variants are significantly enriched in these evolution-tuned regions compared to expectation.

Banerjee’s later work on human-specific methylation echoed the same theme: recently evolved human-specific differentially methylated regions (DMRs) show a clear enrichment for schizophrenia association signals, stronger than for many other traits. ²

From an EToC lens:

• Psychotic risk isn’t some ancient mammalian thing; it’s disproportionately tied to human-specific genomic modifications.
• The relevant regions are regulatory – enhancers and methylation patterns tuning gene expression in cortex – exactly where you’d expect selection to carve out new modes of self-modeling.

If schizophrenia were just the broken version of a very general “brain disease,” you wouldn’t expect this kind of targeted enrichment in recently evolved regulatory real estate.

2. Genome Dating: When Did the Cognitive Variants Show Up?#

Libedinsky et al. (2025) used a “genome-dating” approach to estimate when variants contributing to trait polygenic scores (PGS) first arose in the human lineage, aggregating across many GWAS. They show that variants affecting cortical surface area, subcortical volume, general intelligence, and multiple psychiatric risks cluster at distinct evolutionary epochs.

The key points for our purposes:

• A nontrivial share of alleles shaping brain structure and cognition emerged after the human–Neanderthal split, often in the last few hundred thousand years.
• Some psychiatric-linked variants show dates compatible with very recent selection pressures, overlapping the late Pleistocene and Holocene.

On its own, this doesn’t prove EToC, but it nails one major background claim: the cognitive and psychiatric architecture of Homo sapiens is not frozen at 200kya. It has continued to mutate and be winnowed through the window when EToC wants consciousness to “bite its own tail.”

Eve’s awakening is not trying to ride on a static genome. The underlying genotype for self-reflection and its failure modes is still being pushed around.

3. Eastern Eurasian Ancient DNA: Directional Selection on Mind#

Piffer (2025) turns this general dating picture into something sharper: using Eastern Eurasian ancient genomes, he reconstructs polygenic scores through time for traits like educational attainment, intelligence, and autism risk. He reports evidence for directional selection on these PGS, especially in the Late Pleistocene and Holocene.

Two pieces line up with EToC:

• Cognition-linked PGS tend to increase through time in these lineages, consistent with selection for more sophisticated planning, cooperation, and abstract reasoning.
• Autism-related PGS show evidence of selection and shifting frequencies, hinting that “edge” phenotypes around social cognition and pattern detection are also under active evolutionary tuning, not static side-noise.

In EToC language, Piffer’s data look like the genomic shadow of a population climbing out of Golden-Age automaticity into a world where metacognitive sophistication and its costs are under continuous selection.

4. A Human-Accelerated Neuron Type and Autism#

Starr & Fraser (2025) identify a specific cortical neuron type with human-accelerated regulatory evolution and a strong link to autism-associated genes.

The key ingredients:

• Using single-cell transcriptomics and evolutionary annotations, they show that this neuron type sits at the intersection of human-specific regulatory change and ASD risk.
• Genes expressed in this neuron are enriched both for human-accelerated regions and autism GWAS hits, implying that the same molecular tweaks that made this cell type special also made it vulnerable.

From an EToC angle, this is exactly what you’d expect if the human nervous system has developed dedicated machinery for high-level social modeling and self-representation:

• Build a new kind of cortical cell;
• Wire it into language and social brain networks;
• Accept that a subset of minds will be pushed into autistic or psychotic territory when the system goes out of range.

This is not “brain in general” being fragile. It’s newly evolved human gear behaving like a high-performance race car: fast, but prone to spinning out.

5. Pervasive Directional Selection: Less Psychosis, More g#

Akbari et al. (2024) is the Reich-lab paper that finally does what people wanted from ancient DNA ten years ago: a joint analysis of hundreds of loci under selection in West Eurasians over the last ~10,000 years. They detect pervasive directional selection on traits including pigmentation, metabolism, immunity—and, crucially, cognition and psychiatric risk.

Results that matter for EToC:

• Polygenic scores for educational attainment and intelligence appear to have been pushed upward in multiple lineages.
• Polygenic scores for schizophrenia and bipolar disorder show evidence of selection against risk alleles in the Holocene West Eurasian record.
• This pattern persists even after controlling for demography and background selection.

Read in EToC mode, Akbari et al. show a post-Neolithic world where:

• Populations are selecting for more flexible, future-oriented cognition;
• At the same time, they are pruning the tails of extreme psychosis and mood instability that come with it.

It’s literally the “cliff-edge” idea turned into a time-series plot: a trait distribution moved toward the cliff for a long time (late Pleistocene) and then trimmed back as complex societies made full psychotic breakdown more costly.

6. European Trait Trajectories: Intelligence and Education Move#

Kuijpers et al. (2022) reconstruct polygenic trajectories for 40 complex traits in European ancient genomes, from Mesolithic hunter-gatherers through Neolithic farmers and into the Bronze and Iron Ages.

For cognition they find:

• Post-Neolithic increases in PGS for educational attainment and related cognitive traits.
• Shifts that cannot easily be explained by mere population replacement; they require some mix of selection and non-neutral processes.

Taken together with Akbari’s more explicit selection tests, this supports the EToC intuition that:

• The Bronze Age mind is not the same beast as the Upper Paleolithic mind.
• Whatever happened around the time Eve first realized she was thinking about herself, evolution did not simply stop; states required more self-control, planning, and cooperation, and the genome obliged.

EToC is not about a single magical mutation; it’s a long, ugly polygenic story. Kuijpers et al. give you the macro-shape: mental PGS have moved, and recently.

7. HAQERs: Rapidly Evolved Regions and Language Ability#

Casten et al. (2025) introduce HAQERs – “Human Ancestor Quickly Evolved Regions” – sequences that rapidly accumulated substitutions after the human–chimp split. They then show that polygenic scores restricted to HAQERs significantly predict language abilities in present-day humans, but not nonverbal IQ. ³

Highlights:

• HAQER-based PGS correlate with a core language factor derived from detailed longitudinal assessments in children.
• These regions alter binding motifs for Forkhead transcription factors, including FOXP2, the canonical “language gene.”
• The language-associated HAQER variants show a complex pattern of balancing selection and pleiotropic trade-offs (including links to birth complications).

For EToC, this is nearly on-the-nose:

• There really are fast-evolving regulatory elements that specifically target language circuitry.
• Their evolutionary story involves trade-offs and constraint, not a simple monotone improvement—mirror image of EToC’s claim that language-powered selfhood is both a superpower and a liability.

If language is the medium through which the self narrates itself into existence, HAQERs are the script editors: a thin layer of human-specific code that changed how voices sound inside the skull.

8. Recent Natural Selection for Schizophrenia Protection#

González-Peñas et al. (2023) use GWAS and evolutionary metrics to show that alleles reducing schizophrenia risk have been subject to recent natural selection via non-antagonistic pleiotropy. The Evolution & Medicine Review summary puts it crisply: recent selection seems to favor derived alleles that protect against schizophrenia, consistent with earlier hints by Liu et al. that ancestral alleles often increase risk. ⁴

The EToC-compatible pieces:

• Risk alleles for schizophrenia are now being pushed down by selection, but they haven’t been fully purged—which matches a cliff-edge model where most carriers benefit, a minority fall off.
• The selection is “non-antagonistic pleiotropy”: the protective alleles are not obviously bad for other traits, suggesting that the compromise is not a simple creativity-vs-sanity tradeoff, but something more structural in neurodevelopment.

In other words, schizophrenia is neither a random curse nor a cute “price of genius” story. It’s what you get when the species spends a long time ratcheting cognition up toward a cliff and then, in the Holocene, begins shoring up the guardrails without rebuilding the road.

9. Human-Specific Methylation and Schizophrenia Signals#

Banerjee et al. (2018) look at recently evolved human-specific differentially methylated regions and ask which complex traits’ GWAS signals are enriched there. Schizophrenia stands out as the only trait reaching clear enrichment that survives stringent permutation and bootstrap tests. ²

Key points:

• The enrichment is not seen for Neanderthal or Denisovan DMRs, only human-specific ones.
• The effect size is comparable to that seen in regions tagged by Neanderthal selective sweep markers and stronger than for classic Human Accelerated Regions (HARs).
• There is some enrichment for height, but psychiatric risk is the main standout.

This is essentially a molecular-epigenetic version of the Srinivasan result:

• Recently evolved regulatory modifications in the human cortex are disproportionately loaded with schizophrenia risk.
• Whatever we did to our cortical circuitry in the last slice of evolutionary time, it entangled us more deeply with psychotic failure modes.

From an EToC vantage, the picture is blunt: the move from bicameral, myth-saturated Golden Man to self-narrating Eve left fresh scars in methylation space. Those scars line up with the fault lines of schizophrenia.

10. Cliff-Edges and the Self: Aftab’s Evolutionary Genetics of Schizophrenia#

Awais Aftab’s “Evolutionary Genetics of Schizophrenia” is not a primary research paper, but it’s a rare synthesis that actually takes the modern genetics seriously. He engages with Nesse’s cliff-edge fitness model and the Mitteroecker & Merola formalization, arguing that schizophrenia is best understood as the tail of a polygenic trait that is beneficial across most of its range. ⁵

Aftab emphasizes several points that dovetail with EToC:

• High polygenicity and mechanistic heterogeneity: schizophrenia risk is smeared across thousands of alleles; there is no single “SZ gene” to purge. That’s exactly what you expect if what’s under selection is a mode of cognition, not a discrete organ.
• Evidence for recent negative selection: newer studies cast doubt on older claims of strong positive selection at SZ loci but still support the idea that risk variants are under ongoing purifying selection.
• Compatibility with a cliff-edge model: weak positive benefits across most of the distribution plus steep costs at the tail can coexist with present-day negative selection on high-risk alleles.

EToC’s Eve is precisely a mind “poised near the edge of a cliff”: the system that generates a continuous narrative self is adaptive in the middle ranges and catastrophic at the extremes. Aftab’s synthesis shows that this is not just poetic; the genetics, if anything, push us to think in exactly these terms.

How This All Looks from Eve’s Vantage Point#

Put all ten papers together and you get something like this:

1. The playground is human-specific.
   Schizophrenia risk variants and associated epigenetic marks sit disproportionately in human-specific, recently evolved regulatory regions. ²

2. Language is a specialized, recently tuned system.
   HAQERs and human-accelerated neuronal types shape language and social cognition, with trade-offs that include autism and, plausibly, psychotic vulnerability. ³

3. Cognitive architecture is still moving.
   Genome-dating and ancient DNA show that alleles affecting brain size, intelligence, educational attainment, and psychiatric risk have shifted throughout the Late Pleistocene and Holocene, with clear signs of directional selection.

4. Selection is now trimming, not building.
   Recent selection seems to favor reduced schizophrenia risk without obvious tradeoffs elsewhere—in line with a system that already overshot the “safe” zone and is being re-tuned. ⁴

In the language of Eve Theory of Consciousness:

• Golden Man was a control system: closed perception–action loops, low-dimensional self-model, gods as external voices.
• Over tens of thousands of years, selection loads the genome with regulatory tweaks that enhance language, social cognition, and long-range planning, using new neuron types and human-specific regulatory regions as the scaffolding.
• The cost of this climb is a tail of minds that fall off the cliff into schizophrenia, bipolar disorder, autism, and other extreme phenotypes—precisely along the axes of voice, agency, and self-world boundary that EToC treats as central.
• In the Holocene, as societies densify and the cost of extreme psychosis rises, selection starts pushing back on the worst alleles while still sustaining the general architecture that made Eve possible.

The point isn’t that any one paper “proves” EToC. It’s that, taken together, they make it increasingly hard to pretend that selfhood, language, and psychosis are independent problems. The same thin edge of cognitive steel that let Eve look back at herself in horror is visible in our methylomes, accelerated regions, and ancient genomes.

FAQ#

Q 1. Do these papers show that schizophrenia itself was positively selected? A. Not cleanly; newer work argues that most SZ risk alleles show signs of negative or background selection, but their enrichment in human-specific regions and the need for recent selection on protective alleles fit a cliff-edge “by-product of a beneficial architecture” story rather than simple maladaptation. ²

Q 2. How does language evolution connect to Eve Theory of Consciousness? A. HAQERs and human-accelerated neurons show that small, recent regulatory changes can strongly affect language ability and social cognition; EToC treats this language-saturated internal modeling as the medium in which a narrative self crystallizes, with psychosis as one failure mode. ³

Q 3. What role do ancient DNA polygenic scores actually play here? A. They don’t tell us what individuals “were like,” but their time-series shifts show that selection has moved the population distribution of cognitive and psychiatric traits during exactly the window when EToC posits dramatic changes in consciousness.

Q 4. Could all this just be demographic noise rather than selection? A. Some early signals probably were, but the newer methods explicitly model demography and still find directional selection for higher education/IQ and lower psychosis risk, which is hard to explain with drift alone.

Footnotes#

Sources#

1. Srinivasan, S. et al. “Genetic markers of human evolution are enriched in schizophrenia.” Biological Psychiatry 80 (2016): 284–292.
2. Banerjee, N. et al. “Recently evolved human-specific methylated regions are enriched in schizophrenia signals.” BMC Evolutionary Biology 18 (2018): 63. ²
3. Libedinsky, C. et al. “The emergence of genetic variants linked to brain and cognitive traits in human evolution.” Cerebral Cortex (2025).
4. Piffer, D. “Directional selection and evolution of polygenic traits in Eastern Eurasia: Insights from ancient DNA.” Human Biology (2025).
5. Starr, A. & Fraser, H. “A general principle of neuronal evolution reveals a human-accelerated neuron type potentially underlying the high prevalence of autism in humans.” Molecular Biology and Evolution (2025).
6. Akbari, A. et al. “Pervasive findings of directional selection realize the promise of ancient DNA to elucidate human adaptation.” preprint, 2024.
7. Kuijpers, C. C. H. J. et al. “Evolutionary trajectories of complex traits in European populations of modern humans.” Molecular Biology and Evolution (2022).
8. Casten, L. G. et al. “Rapidly evolved genomic regions shape individual language abilities in present-day humans.” bioRxiv preprint (2025). ³
9. González-Peñas, J. et al. “Recent natural selection conferred protection against schizophrenia by non-antagonistic pleiotropy.” Scientific Reports 13 (2023): 1–12. See also summary in “Schizophrenia: Recent selection for protective alleles,” Evolution & Medicine Review (2023). ⁴
10. Aftab, A. “The Evolutionary Genetics of Schizophrenia.” Psychiatry at the Margins (2025). ⁵