The “Internal Brain” and
“External Brain” of Humanity

The human brain (the “internal brain”) is a high-energy, imperfect information storage system equipped with active mechanisms for deletion and distortion. Precisely because of this imperfection, humans began externalizing critical survival information early in their evolutionary history—from cave paintings to writing to computers to AI—constructing a continuously evolving “external brain” system.

This paper proposes a unified interdisciplinary framework that reconceptualizes the evolution and degradation of human cognition as a dynamic interaction between the internal brain and the external brain, driven jointly by environmental constraints (temperature, humidity, oxygen concentration), external threat levels, group size, and tool autonomy. By integrating cross-domain evidence from gas physics, neuroimaging (fMRI/fNIRS), evolutionary biology, paleoclimatology, paleoanthropology, memory psychology, and cognitive science, this paper argues five core propositions:

• As the most metabolically expensive biological organ, the internal brain’s size is subject to strict energy cost–benefit constraints
• The internal brain’s mechanisms of memory deletion and distortion render it incapable of reliably preserving precise information over the long term—this is the fundamental driver behind the emergence of the external brain
• Environmental variables directly affect internal brain cognitive performance through three pathways: oxygen supply, attentional resources, and oxidative stress
• The external brain has evolved from a passive storage tool into an autonomous AI, fundamentally altering the master–servant relationship between the internal and external brains
• Cognitive degradation operates simultaneously across four temporal scales—seconds, years, generations, and evolutionary epochs—and is masked by compensatory mechanisms at each level

This paper designates this framework the “Environmental Cognitive Pressure Hypothesis” (ECPH) and introduces the “Tipping Point Theory” of external brain autonomy evolution—when the external brain transitions from passive storage to active generation, the evolutionary equation governing human cognition undergoes a fundamental transformation.

1.1 The Body Temperature Paradox

Human body temperature is maintained at a constant 37°C, yet the optimal ambient temperature for human comfort lies between 18°C and 25°C. This approximately 12°C differential is not a design flaw but rather an exquisite product of evolution. Research by Casadevall and Bergman at the Albert Einstein College of Medicine demonstrated that a body temperature of 37°C achieves a precise optimal balance between the benefits of resisting fungal infection and the costs of sustaining metabolism—their mathematical model calculated an optimal temperature of 36.7°C, an almost perfect match with the actual human body temperature. For every 1°C increase in body temperature, the number of fungal species capable of infecting the host decreases by approximately 6%, which is why tens of thousands of fungal species can infect cold-blooded animals while only a few hundred can threaten mammals.

An ambient temperature below body temperature is a necessary condition for heat dissipation. More than 50% of the chemical energy released by human metabolism is converted to thermal energy for maintaining body temperature, and this heat must be dissipated through the skin into the surrounding environment. At an ambient temperature of 25°C, relative humidity of 50%, and airflow of 0.15 m/s, the human body achieves an ideal state of thermal equilibrium. This design reveals a deep thermodynamic coupling between the human body and its environment.

1.2 The Central Question

If ambient temperature exerts such profound effects on human physiology—determining heat dissipation efficiency, metabolic balance, and immune defense—has it also shaped the evolutionary trajectory of human cognition? The starting point of this paper is that the environment is not merely the backdrop of survival but the foundational infrastructure of intelligence evolution. Temperature, humidity, and oxygen concentration do not merely affect physical comfort; they directly influence whether the brain can operate at peak efficiency, and over evolutionary timescales, they have shaped the brain’s very size and structure.

1.3 Statement of Research Paradigm

This paper is a macro-level systems analysis of human cognition, not a single-variable experimental report. It does not pursue “absolute statistical-level” explanatory power for any single factor, but instead constructs a unified framework of multi-factor interactions. All variables—climate, oxygen, tools, group size—are contributing factors within an environment–cognition interaction system; none is the sole determinant.

2.1 Memory Is Not Playback but Reconstruction

Each retrieval of human memory is a process of reconstruction, not precise playback. This process involves four categories of systematic bias: omission errors (forgetting details or entire events), commission errors (adding nonexistent information), distortion errors (altering details while preserving the general structure), and source monitoring errors (remembering information but misattributing its origin). The mechanism of memory reconsolidation renders memories unstable during each retrieval; if erroneous information is present at that moment, it becomes integrated into the memory as though it were part of the original experience.

2.2 The Active Deletion Mechanism of Memory

Adaptive forgetting is not a malfunction of the brain but a design feature. Forgetting frees cognitive resources, prevents information overload, and enables the brain to extract useful patterns and regularities. However, the cost of this mechanism is that critical survival information—such as the migration routes of specific prey or the identifying markers of poisonous plants—may also be erroneously deleted.

2.3 The Active Distortion Mechanism of Memory

False memories are enhanced during sleep. Social conformity effects can rewrite individual memories—research by Edelson and colleagues demonstrated that social pressure can reshape memory traces at the neural level. A deeper finding is that susceptibility to false memories is positively correlated with categorization ability. The very capacity that makes human intelligence powerful—abstract generalization—is the same capacity that makes memory unreliable. In the process of extracting the “gist,” the brain sacrifices precision.

2.4 The Adaptive Deficiency Hypothesis

A review published in Trends in Cognitive Sciences confirmed that memory distortions reflect adaptive cognitive processes—imagination inflation aids simulation of future events, gist-based memory aids conceptual learning, and post-event misinformation facilitates memory updating. These processes contribute to the efficient operation of memory, but at the cost of producing distortions. Evolution selected for “imperfect but energy-efficient memory with strong generalization capabilities,” because the energy cost of perfect memory (such as eidetic memory) would be prohibitively high and might compromise abstract reasoning ability.

2.5 Core Proposition

2.6 Cognitive Biases—Systematic Reasoning Defects of the Internal Brain

The imperfection of the internal brain extends beyond the storage layer (memory) to pervade the processing layer (reasoning). The seminal work of Tversky and Kahneman (1974) revealed that human judgment under uncertainty relies on a limited set of heuristic principles, producing severe and predictable systematic biases. Confirmation bias inclines individuals to accept information supporting existing beliefs while disregarding contradictory evidence. The anchoring effect causes judgments to be disproportionately influenced by initially received information. The availability heuristic leads to overestimation of the probability of easily recalled events. The framing effect causes identical information, when presented differently, to produce entirely different decisions. These biases originate from fast decision-making strategies that were advantageous for survival in ancestral environments—the heuristic of “approximately right is good enough” held greater survival value than “precise but slow” analysis in ancient settings.

Recognizing this, the definition of the external brain must be expanded from “storage patch” to “full-stack cognitive patch.” The scientific method is a fix for confirmation bias—double-blind experiments prevent researchers from unconsciously selecting data that support their hypotheses. Statistical tests remedy the availability heuristic—replacing intuitive judgment with mathematical probability. Peer review addresses individual blind spots. The adversarial trial system in law corrects for the anchoring and framing effects. Mathematics itself remedies defects in human logical reasoning—the human brain cannot reliably perform more than three or four steps of logical deduction, but algebraic formulas can. Thus, the evolutionary lineage of the external brain encompasses not only the storage track of “cave paintings → writing → computers → AI” but also the processing track of “arithmetic → geometry → algebra → statistics → the scientific method → institutional design → AI reasoning.” Both tracks converge at AI—which simultaneously substitutes for both the storage and processing functions of the internal brain.

3.1 Temperature and Cognitive Performance

Using data from 1.15 million mental arithmetic games played by 31,000 users on the Lumosity platform, MIT CEEPR found that high temperatures significantly reduce cognitive performance, with the negative effects intensifying when the average temperature exceeds 21°C over consecutive days. A heat wave experiment by the Harvard School of Public Health found that students in dormitories without air conditioning exhibited 13.4% slower reaction times and 13.3% lower scores on arithmetic tests. An analysis of 10 million PSAT examinees by the Harvard Kennedy School revealed that cumulative exposure to high temperatures suppresses cognitive skill development, with extreme heat proving especially destructive. A Chinese longitudinal survey found that when the temperature on exam day exceeded 32°C, mathematics scores declined by approximately 0.088 standard deviations, equivalent to 0.3 years of education. The optimal cognitive temperature range has been confirmed as 22–24°C.

3.2 Temperature’s Precision Strike on Attentional Networks

An fMRI brain imaging experiment found that passive hyperthermia (50°C, 40% humidity) impaired executive function—complex cognitive control requiring sustained attention—but had no significant effect on simpler alerting and orienting responses. High temperature does not make the brain uniformly duller; rather, it delivers a precision strike against higher-order cognitive functions that “require sustained focused attention.” Activation patterns of the dorsolateral prefrontal cortex (DLPFC) were significantly altered. The “maximal adaptability model” showed that attentional resources are progressively depleted by heat stress. More critically, the anticorrelated relationship between the default mode network and the dorsal attention network was disrupted under high temperature, meaning that the brain’s ability to switch between “focused” and “mind-wandering” states was impaired. Eye-tracking experiments directly demonstrated that gaze patterns became disordered under extreme heat and humidity conditions.

3.3 Sex Differences and Humidity Effects

The “battle for the thermostat” experiment (N=543) found that across a range of 16–33°C, women performed better on mathematical and verbal tasks in warmer environments, while men performed better at cooler temperatures; moreover, the magnitude of women’s performance gains from rising temperatures was significantly greater than the magnitude of men’s decline. Climate chamber experiments demonstrated that humidity exerts an independent effect: at equivalent high temperatures (39°C), reducing humidity from 70% to 50% improved cognitive performance.

4.1 The Physical Relationship Between Temperature and Oxygen

The proportion of oxygen in the atmosphere remains constant at 21%, but according to the ideal gas law (n/V = P/RT), air density decreases as temperature rises, meaning fewer oxygen molecules are inhaled with each breath. Quantitative calculations reveal: compared with 20°C, breathing at 35°C yields approximately 4.9% fewer oxygen molecules per breath; the difference between 0°C and 40°C reaches 12.8%. Humidity further compounds this effect—water vapor molecules (molecular weight 18) displace heavier nitrogen (28) and oxygen (32) molecules, reducing air density by an additional ~0.5% at 100% relative humidity and 20°C. The double overlay of high temperature and high humidity in tropical environments reduces the actual number of inhalable oxygen molecules even further.

4.2 High Temperature Strikes the Internal Brain via the “Oxygen Pathway”

A 2025 review published in Experimental Physiology revealed a multi-pathway mechanism: elevated skin temperature → blood is redistributed to the skin for heat dissipation → simultaneous hyperventilation reduces arterial CO₂ partial pressure → these two mechanisms jointly reduce cerebral blood flow → thereby limiting the brain’s oxygen supply → impairing cognitive function. The study’s core conclusion was that cognitive impairment is more closely associated with rising brain temperature and reduced cerebral oxygen availability than with cerebral blood flow per se.

4.3 The “Triple Strike” Model of Tropical Environments

5.1 The Nonlinear Relationship

Oxygen is the fundamental “nutrient” for oxidative phosphorylation—the cell’s primary net energy-producing process. Mathematical models combined with experimental data reveal an exquisite nonlinear relationship: moderate hypoxia (2–8% oxygen) and mild hypoxia (8–15%) enhance cell proliferation, while severe hypoxia (<2%) induces cellular quiescence. Physiological oxygen concentrations in human tissues range from only 2–9%, far below the atmospheric level of 21%. The dentate gyrus of the hippocampus normally maintains a low-oxygen microenvironment that promotes neurogenesis.

5.2 Resolving the Key Paradox

Mild hypoxia promotes stem cell proliferation, but more proliferation means more mutations and faster telomere shortening—thus it does not equate to greater longevity. The “tumor suppressor theory of aging” posits that obesity and caloric restriction accelerate and decelerate aging, respectively, precisely because of their effects on cell proliferation—most mutations arise during cell division. Reactive oxygen species (ROS) attack the DNA base guanine, directly damaging telomeres and accelerating their shortening. Neurons, as post-mitotic cells, are especially vulnerable—they accumulate oxidative damage without the ability to “dilute” it through division.

6.1 Climate and Longevity

A panel data study published in PLOS Climate (191 countries, 1940–2020) reached the quantitative conclusion that for every 1°C increase in mean annual temperature, life expectancy at birth decreases by 0.44 years. The world’s five widely recognized longevity “Blue Zones” share one key characteristic: year-round temperate climates, with temperatures ranging between 17.4°C and 23.5°C. This closely aligns with the most comfortable temperature range for the human body (18–25°C) and the optimal cognitive temperature range (22–24°C).

6.2 Crystallized Intelligence and the Grandmother Hypothesis

The Seattle Longitudinal Study (ongoing since 1956) demonstrates that fluid intelligence begins declining in early adulthood, but crystallized intelligence—vocabulary, knowledge, judgment—continues to increase into the sixties and then remains stable. Hawkes (1998) proposed the Grandmother Hypothesis, arguing that post-reproductive individuals enhance offspring survival rates by providing care and transmitting knowledge. Among all vertebrates, only humans, pilot whales, and orcas possess a significant post-reproductive lifespan—all highly social species dependent on intergenerational knowledge transfer.

6.3 The Positive Feedback Loop

7.1 Out of Africa and Civilization’s “Goldilocks Zone”

A study published in Nature Communications used 300,000 years of high-resolution paleoclimate reconstructions to find that human migration out of Africa was not a single event but a series of wave-like dispersals, each occurring during climatic windows when warming and increased moisture opened “green corridors.” All major ancient civilizations clustered within the temperate–subtropical band between 17°N and 42°N latitude; no civilization originated in extreme polar regions or equatorial tropical rainforests. Civilizational centers have exhibited a northward drift: Mesopotamia (33°N) → Greece (38°N) → Rome (42°N) → Northwestern Europe (48–55°N). The east–west orientation of Eurasia enabled rapid technological diffusion within the same climatic zone, whereas the north–south orientation of the Americas significantly slowed diffusion due to the crossing of climatic belts.

7.2 The Maya Civilization—Not a Counterexample but the Strongest Evidence

The Maya independently developed astronomy and mathematics in the tropics, demonstrating the independent creative capacity of tropical civilizations. Yet at the historical cross-section of 1519, a few hundred Spanish conquistadors armed with firearms and carrying pathogens annihilated the entire Aztec civilization—precisely demonstrating that the technological level of temperate civilizations overwhelmed that of tropical civilizations at the same point in time. All Stone Age civilizations surviving into the 21st century on islands like the Andaman Islands and North Sentinel Island are located in the tropics; no Stone Age civilization persists in cold or temperate regions.

7.3 Singapore—Not a Counterexample but the Strongest Supporting Evidence

8.1 The Fates of Each Species

The uniqueness of Homo sapiens lies not in “choosing the right climate zone” but in the ability to actively create optimal micro-environments through clothing, fire, and shelter. Cognitive flexibility—rather than any fixed environmental adaptation—is the fundamental distinction from other Homo species. A two-million-year climate simulation published in Nature in 2022 confirmed that the appearance and disappearance of Homo species correlate with long-term climatic anomalies.

9.1 Predation Pressure and the Cognitive Arms Race

Artificial selection experiments in guppies found that large-brained females had a 13.5% higher survival rate than small-brained females in semi-natural streams with predators. The “Predator Intelligence Hypothesis” (PIH), formally proposed in a 2026 Nature Reviews publication, posits that cognitive challenges arising from predator–prey interactions drive a co-evolutionary cognitive arms race. Analysis of 623 predator–prey pairings revealed that prey species possess relatively larger brains than non-prey species, and that predator and prey brain sizes are mutually correlated. Reverse evidence is equally compelling: a cross-species study of 102 Chinese frog species found that low predation risk led to brain shrinkage accompanied by enhanced protective coloration. Domesticated animals universally undergo brain reduction, yet mink that return to the wild exhibit brain re-enlargement—demonstrating that cognitive degradation is reversible.

9.2 The Core Principle

10.1 The Brain Is Under “Intermediate Optimum” Selection

The coefficient of additive genetic variation (CVA ≈ 7.8) for human brain size is lower than that of any other organ or life-history trait in the human body, indicating that the brain has been under intense stabilizing (“mean optimum”) selection. The latest research from July 2025 confirmed that approximately 300,000 years ago, the genus Homo transitioned from directional selection (bigger is better) to stabilizing selection (intermediate is best). Highly intelligent individuals face survival costs: in guppy experiments, large-brained females exhibited stronger cognition but produced fewer offspring and grew more slowly. Individuals of moderate intelligence prevail in long-term evolutionary competition—cognition that is “good enough” combined with tolerable metabolic costs and normal reproductive output. The paradox of civilization is this: a small number of highly intelligent individuals drive civilizational progress, but the majority of the gene pool is composed of large numbers of moderately intelligent individuals.

10.3 The Selective Degradation Hypothesis—Cold Cognition Declines While Hot Cognition Advances

The human self-domestication hypothesis proposed by Wrangham and Hare provides a complementary explanation for brain shrinkage: natural selection targeted primarily not intelligence itself but reactive aggression. In late Homo evolution, selection for within-group prosociality—reduced reactive aggression, enhanced social tolerance, and cooperative ability—may have facilitated the emergence of complex behavior through cultural evolution. Cieri and colleagues (2014) found that craniofacial feminization in the Upper Paleolithic coincided with an explosion of cultural transformation. Self-domestication is thought to have promoted cooperation, language, knowledge sharing, and technological progress.

This raises a critical refinement of “cognitive degradation”: the decline of the internal brain may not be uniform but selective. Cold cognition—independent computation, spatial navigation, individual survival reasoning—capacities that were vital for small-group independent survival—is degrading. Hot cognition—cooperative negotiation, emotional regulation, empathic understanding, verbal communication—capacities crucial for large-group collaborative survival—is advancing. This corresponds precisely to the evolutionary transition from an individual self-sufficiency survival mode to a group division-of-labor survival mode. When group size expanded from 30 individuals to thousands and then millions, individuals no longer needed all-purpose cold cognition to independently cope with every survival challenge, but required stronger hot cognition to maintain cooperative relationships within increasingly vast social networks. The process of brain shrinkage may simultaneously encompass the reduction of cold cognition regions and the reorganization and optimization of hot cognition circuits.

11.1 Cave Paintings—The First External Brain

A breakthrough study in 2023 decoded the markings on cave paintings for the first time, demonstrating that at least 20,000 years ago, humans across Europe were recording temporal information about wild animals and their breeding cycles—constituting a lunar calendar system. The paintings depicted only important prey species, and prehistoric artists possessed a detailed understanding of animal anatomy and hunting techniques. “Human memory is fragile, but stone walls are enduring”—cave paintings served as a tribe’s shared memory repository, ensuring that vital knowledge would not be lost with the death of a single hunter.

11.2 The External Brain Is Not an “Extension” but a “Patch”

Clark and Chalmers’s Extended Mind thesis (1998) defined external tools as cognitive “extensions.” This paper redefines them: external records are “patches” for the information fidelity deficit of the internal brain. “Extension” implies an optional enhancement; “patch” implies a necessary repair. Merlin Donald’s theory of distributed cognition is consistent with this view: symbolic artifacts function as external memory storage systems, enabling complex cultural continuity to transcend the limitations of individual minds. In the Pleistocene, absent any technology, climate was the sole independent variable—the causal direction is unambiguous: climate → cognitive constraints → drive external brain creation → which in turn controls climate.

12.1 From Amplification to Substitution

From 2.6 million to 50,000 years ago (Phase One), tools and brains co-evolved: more complex tools → larger brains needed to design them → larger brains capable of crafting better tools. During the Upper Paleolithic explosion of material culture (approximately 50,000 years ago), each individual still needed to master all skills comprehensively, corresponding to the peak brain volume of 1,660 cc. From 10,000 years ago to the present (Phase Two), agriculture introduced social division of labor and writing enabled external information storage—individuals no longer needed to be self-sufficient, and the brain began to decline.

12.2 Scientific Confirmation of Cognitive Offloading and Precise Temporal Alignment

Cognitive offloading has been defined as “the use of physical actions to alter the information-processing demands of a task in order to reduce cognitive load.” Experiments have confirmed that reducing offloading diminishes immediate performance but improves subsequent memory; that using generative AI induces “metacognitive laziness”; and that performance after losing access to AI is worse than that of individuals who never used it. The reverse Flynn effect (IQ scores beginning to decline in developed nations from the mid-1990s onward) aligns perfectly in time with the rise of the internet and smartphones. The emergence of writing (approximately 5,000 years ago) falls within the confidence interval of brain volume reduction. Large-scale bureaucratic systems (approximately 3,000 years ago) precisely correspond to the acceleration point confirmed by change-point analysis. European female endocranial volume has decreased by approximately 240 mL over the past 10,000 years—a rate 36 times faster than the growth rate over the preceding 800,000 years.

12.3 The External Brain’s Reverse Creation of the Internal Brain—Neuronal Recycling

The external brain does not merely substitute for old functions of the internal brain; it also creates entirely new functions within it. Dehaene’s “neuronal recycling hypothesis” reveals this bidirectional sculpting process: cultural inventions can reorganize the brain by “recycling” existing neural circuits—co-opting cortical areas originally dedicated to other functions. There is no innate reading area in the human brain. The “visual word form area” (VWFA) in the left occipitotemporal region was originally dedicated to face and object recognition; over thousands of years of cultural transmission, the tool of writing “commandeered” this region into a module specialized for letter and word recognition. Writing systems evolved to fit the pre-existing constraints of the brain, rather than the brain evolving to accommodate writing—this explains why globally diverse writing systems share certain visual features.

Each generation of the external brain has carved new cognitive capacities into the internal brain: writing created “reading,” mathematical notation created “abstract numerical operations,” programming languages created “algorithmic thinking,” and AI may currently be creating “the skill of asking questions”—knowing how to pose good questions to AI is itself an entirely new cognitive skill. But this creation comes at a cost: neuronal recycling is a form of “destructive competition”—the acquisition of reading ability partially impairs face recognition ability, because both compete for the neural resources of the same region. Learning to read requires “unlearning” mirror invariance (distinguishing b from d), yet mirror invariance is originally a fundamental function of object recognition. Thus, the relationship between the external brain and the internal brain is not a unidirectional process of “substitution → degradation” but a bidirectional one of “substituting old functions + creating new functions”—though creating new functions may simultaneously damage certain pre-existing foundational abilities.

13.1 The Individual Information Load Formula

The social brain hypothesis discovered that brain size imposes a numerical constraint on social group size (approximately 150 individuals—Dunbar’s number). In Ice Age hunter-gatherer bands of 25–50 individuals, each person had to be a “complete database”—navigation, tracking, toolmaking, medicinal plants, and weather prediction were all stored in the individual brain. After the Agricultural Revolution, groups expanded from dozens to millions of members, distributing information across many more brains and dramatically reducing the individual information load. Evolutionary robotics models confirm this: the neural complexity of small-brained agents evolved under social interaction is comparable to that of large-brained agents evolved in solitary conditions. Ant models likewise confirm that optimizing total colony brain mass is more energetically advantageous than optimizing individual brain mass.

The “information fidelity transmission rate” in the formula is a critical variable. In the era of purely oral tradition (without an external brain), the internal brain’s memory deletion and distortion mechanisms caused information to decay progressively across generations—like a game of telephone, each generation loses or warps a portion of the information. The core contribution of the external brain is not only increasing the denominator’s “number of individuals available for sharing” but, more crucially, elevating the information fidelity transmission rate from the internal brain’s unstable value to a near-1.0 permanent storage value. Cave paintings first achieved high spatial fidelity (stone walls do not “forget”); writing achieved high temporal fidelity (texts can remain unchanged across millennia); digital storage pushed the fidelity rate toward 1.0. AI introduces a new complexity: its information fidelity transmission rate is extremely high, but its active generation function may introduce a new type of “distortion”—AI hallucination. This means that the external brain, while repairing the old defects of the internal brain (forgetting and memory distortion), may simultaneously be creating new forms of information corruption.

14.1 The Five-Generation Evolutionary Lineage of the External Brain

14.2 The Tipping Point Theory

The first four generations of external brains share a common characteristic: they were “dead”—cave paintings did not paint themselves, writing did not write itself, Word did not generate its own content. The internal brain was always the sole source and processor of information. AI is the first “living external brain”—it not only stores information but generates it; it is not merely retrieved but actively reasons; it is not just used by humans but surpasses the human internal brain in specific tasks.

When the external brain transitions from “passive storage” to “active generation,” the master–servant relationship between the internal and external brains begins to invert. Before the tipping point: the external brain is the internal brain’s tool—humans decide what to store and how. After the tipping point: the external brain becomes the internal brain’s environment—AI determines what information to present, in what manner, and at what moment. The four stages of cognitive decline: “I know” → “I know where to look it up” → “I know how to ask AI” → “AI knows what I need; I don’t need to ask.”

15.1 The Data Cannot Be Ignored

Over the past 35,000 years: 1,660 cc → 1,336 cc, a 20% reduction. Approximately 5 million neurons are lost per generation. The rate of reduction is roughly 25–36 times faster than the rate of growth over the preceding 3 million years. The brain took 3 million years to grow by 374% but only 35,000 years to shrink by 20%—the rate of regression far exceeds the rate of evolution. A 2023 study in Brain, Behavior and Evolution confirmed that brains during warm periods are significantly smaller than those from cold periods. The latest research published in Nature Communications in July 2026 revealed that brain growth was far less driven by directional natural selection than previously believed. Genetic evidence suggests that brain size, general cognitive ability, and educational attainment are currently being selected against in modern populations.

15.2 The Reverse Flynn Effect—Real-Time Validation

During the 20th century, IQ rose by approximately 3 points per decade (the Flynn effect), but since the mid-1990s, IQ has begun to decline in the most developed nations, including Norway, Denmark, France, and the United Kingdom. A 2018 Norwegian study of 730,000 men ruled out genetic factors, confirming that the cause is environmental. The timing of the reversal aligns perfectly with the rise of the internet and smartphones. Young people (ages 17–25) exhibit the highest dependence on AI and the lowest critical thinking scores (Gerlich 2025: correlation between cognitive offloading and AI use r = 0.72; correlation between cognitive offloading and critical thinking r = −0.75). It is worth noting that the tests measuring the Flynn effect have never controlled for environmental variables such as temperature, humidity, and oxygen.

15.3 The Limits of Brain Volume as a Proxy Variable

This paper uses brain volume (cc) as a proxy indicator for the direction of cognitive selection pressure on evolutionary timescales. It must be stated clearly: brain volume does not equal intelligence. The correlation between brain volume and IQ among modern human individuals is only approximately r = 0.3. What truly determines cognitive ability may be synaptic density, the organizational efficiency of white matter fiber tracts, and the strength of functional connectivity between the prefrontal cortex and other regions—not simple volume. The brain can become smaller while simultaneously becoming more efficient—much as chips progressed from vacuum tubes to transistors to integrated circuits, shrinking in size while increasing in computing power. The Flynn effect itself demonstrates that IQ can rise during the same period that brain volume declines.

Therefore, brain volume reduction does not mean every individual becomes less intelligent. What it reflects is that natural selection no longer requires maintaining maximum brain size. This in itself is an important signal—it means that the environment and tools are systematically reducing the survival necessity of individual cognition, rendering the metabolically expensive brain an insufficient return on investment. A shift in the direction of evolutionary pressure, even if it has not yet caused an absolute decline in intelligence, portends a long-term trend: if the environment continues to reduce the demands placed on individual cognition while the energy cost of the internal brain remains high, some form of cognitive simplification will ultimately be unavoidable.

16.1 The Four-Timescale Model of Cognitive Degradation

The degradation at each level is silent, gradual, and masked by compensatory mechanisms at the current level. Individuals compensate for attentional deficits with effort. Cultures compensate for developmental damage with education. Technology compensates for knowledge discontinuity with external storage. Civilization compensates for declining individual brain capacity with collective intelligence.

16.2 The Dual Drivers of Cognitive Degradation and the Core Principle

Driver One (Natural Environment): Climate deviates from the optimal zone → reduced oxygen + attentional depletion + oxidative stress → rising cost of brain maintenance → brain shrinkage. Driver Two (Tool Environment): Tool specialization → cognitive offloading → disappearance of individual self-sufficiency → disappearance of selection pressure on the brain → brain shrinkage.

16.3 The Ultimate Paradox

The human capacity to invent the external brain originated from the internal brain at its maximum size (the 1,660 cc Ice Age hunter), yet every upgrade of the external brain has eroded the rationale for the very internal brain that created it. When the external brain acquired autonomy (AI), this process underwent a qualitative shift—it no longer merely substituted for storage functions but began substituting for the function of thought itself. We may be undergoing cognitive degradation, and precisely because we are degrading, we cannot perceive that we are degrading.

The environmental control systems of the International Space Station (ISS) precisely maintain 25°C, 40–70% humidity, 21% oxygen, and 101.3 kPa atmospheric pressure—NASA has invested billions of dollars to sustain a “cognitive optimal zone,” which is itself the highest-level engineering confirmation that “environmental variables affect cognition.” Even under these optimal parameters, astronauts still exhibit deceleration in processing speed, working memory, and sustained attention. NASA’s Twin Study revealed increased risk-taking propensity and reduced cognitive speed. Ventricular system expansion and white matter reduction were observed, requiring years for recovery. In the microgravity environment, CO₂ accumulates around astronauts’ heads, causing localized oxygen deprivation—hypoxia symptoms are mistaken for fatigue, precisely demonstrating the imperceptibility of cognitive degradation.

• Climate Prediction: Global warming will accelerate evolutionary pressure on human brain size
• AI Prediction: The reverse Flynn effect will be most pronounced in countries and age groups with the highest AI adoption rates; groups highly dependent on AI will exhibit worse cognitive performance after losing access to AI than groups that maintained independent thinking
• Space Prediction: Deviations in life support systems during Mars missions will produce measurable progressive cognitive degradation, which astronauts may be unable to self-detect
• “Cognitive Offloading Red Line”: A tipping point exists—beyond which offloading shifts from “augmentation” to “substitution,” from “tool” to “prosthesis”
• Critical Experiment Design: Under fully controlled conditions of temperature, humidity, and oxygen, compare the cognitive performance of populations who have long resided in tropical versus temperate regions. If no difference is found, it demonstrates that climate effects can be entirely eliminated by technology (the Singapore model); if a difference exists, it demonstrates the presence of long-term evolutionary effects that transcend immediate environmental control

19.1 The Fundamental Distinction from Lynn/Rushton

Lynn and Rushton claimed that cold climates selected for more intelligent races (racial determinism). This paper argues that temperate climates reduced brain maintenance costs, thereby permitting larger brains (environmental energy economics). The key rebuttal: the earliest civilizations arose in warm river valleys rather than the cold north—a direct failure of the cold-climate theory’s predictions.

19.2 Deepening Clark-Chalmers’s Extended Mind Thesis

The Extended Mind thesis describes what the external brain “is.” This paper answers why the external brain “had to be born” (the information fidelity deficit of the internal brain) and “where it is headed” (the autonomy tipping point).

19.3 Integration with DeSilva’s Ant Model and Stibel’s Climate–Brain Size Research

DeSilva explains “when” the brain shrank and the collective intelligence hypothesis. Stibel demonstrates the correlation between climate change and brain size. Building upon both, this paper adds entirely new dimensions: the oxygen physics mediating mechanism, attentional neuroscience, the crystallized intelligence–longevity cycle, comparisons of extinct Homo species, the individual information load formula, and the autonomy evolution of the external brain.

20.1 Core Contributions of This Paper

• Proposes a unified analytical framework for the internal brain–external brain interaction system
• Identifies internal brain memory deficits as the fundamental driver of external brain emergence
• Identifies the tipping point in the autonomy evolution of the external brain
• Constructs a four-timescale model of cognitive degradation
• Proposes the individual information load formula
• Examines AI within the context of 300,000 years of evolutionary history

20.2 The Ultimate Questions

Can humanity consciously reverse a trend of cognitive degradation driven by natural selection? Can the external brain continue to be upgraded without damaging the internal brain? If external systems collapse, the neurons that have been “optimized away” will not return—does this constitute a civilization-level systemic risk? When the external brain acquires autonomy, are humans still the agents of thought, or have they already become accessory organs of the cognitive systems they created?

20.3 Methodological Note

This paper is based on a multi-hour human–machine collaborative thought experiment conducted on July 15, 2026. The human researcher contributed all core hypotheses and logical leaps; the AI assisted by conducting comprehensive evidence retrieval and alignment verification across the web. The human internal brain accomplished cross-domain intuitive inferential leaps that AI cannot yet make autonomously. The AI external brain accomplished instantaneous large-scale literature retrieval and data alignment that the human internal brain cannot perform. This process itself is a living demonstration of the “internal brain–external brain” collaborative model—and a microcosm of the tipping point this paper warns about.

20.4 The Epistemological Paradox

If the internal brain is indeed degrading, how can the reliability of conclusions drawn by a degrading internal brain analyzing its own degradation be guaranteed? This is not a rhetorical question—it is the projection of Gödel’s incompleteness theorems into cognitive science: a system cannot fully prove its own consistency. Can a brain that is losing neurons reliably diagnose that it is losing neurons?

The writing process of this paper offers a partial answer. The human internal brain generated all key hypotheses—cross-domain intuitive inferential leaps that current AI cannot autonomously produce. The AI external brain provided verification—instantaneous retrieval and alignment of hundreds of papers that the human internal brain cannot accomplish. The method for detecting cognitive degradation is precisely the collaboration of internal and external brains—using the objectivity and permanence of the external brain to compensate for the potential decay and biases of the internal brain. But this raises one final question: if detecting cognitive degradation itself depends on the external brain, and the external brain is one of the causes of cognitive degradation, then is the diagnostic process itself accelerating the very problem it seeks to diagnose? This paper chooses not to answer this question, instead leaving it as a fundamental inquiry into the entire enterprise of human cognitive research—for the reader.

Environment–Cognition

Casadevall, A. & Bergman, A. (2010). 98.6°F ideal temperature for keeping fungi away. Albert Einstein College of Medicine.

Krebs, B. (2022). Temperature and Cognitive Performance. MIT CEEPR Working Paper 2022-019.

Cedeño-Laurent, J.G. et al. (2018). Reduced cognitive function during a heat wave. PLOS Medicine.

Park, R.J. et al. (2020). Heat and learning. American Economic Journal: Economic Policy, 12(2).

Liu, K. et al. (2013). The impact of passive hyperthermia on human attention networks: An fMRI study. Behavioural Brain Research, 243.

Chang, T.Y. & Kajackaite, A. (2019). Battle for the thermostat. PLOS ONE, 14(5).

Nakata, H. et al. (2025). Modulation of cerebral blood flow and cognition by hyperthermia and hypoxia. Experimental Physiology.

Memory Deficits and the External Brain

Schacter, D.L. et al. (2011). Memory distortion: an adaptive perspective. Trends in Cognitive Sciences, 15(10).

Donald, M. (1991). Origins of the Modern Mind. Harvard University Press.

Clark, A. & Chalmers, D. (1998). The Extended Mind. Analysis, 58(1).

Bacon, B. et al. (2023). An Upper Palaeolithic Proto-writing System. Cambridge Archaeological Journal.

Climate–Brain Evolution

Stibel, J.M. (2023). Climate Change Influences Brain Size in Humans. Brain, Behavior and Evolution, 98(2).

Stibel, J.M. (2025). Did increasing brain size place early humans at risk of extinction? Brain and Cognition.

DeSilva, J.M. et al. (2021). When and Why Did Human Brains Decrease in Size? Frontiers in Ecology and Evolution.

Hubbe, M. et al. (2026). Evolutionary drivers of encephalization and facial reduction in Homo. Nature Communications.

Timmermann, A. et al. (2022). Climate effects on archaic human habitats and species successions. Nature, 604.

Raia, P. et al. (2020). Past Extinctions of Homo Species Coincided with Increased Vulnerability to Climatic Change. One Earth.

Cognitive Offloading and AI-Induced Degradation

Risko, E.F. & Gilbert, S.J. (2016). Cognitive Offloading. Trends in Cognitive Sciences, 20(9).

Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading. Societies, 15(1).

Dergaa, I. et al. (2024). AI-chatbots-induced cognitive atrophy (AICICA). Heliyon.

AI-overdependence and human cognitive decline (2026). ScienceDirect comprehensive review.

Brainrot: Deskilling and Addiction are Overlooked AI Risks (2026). arXiv:2605.03512.

Ferdman, A. (2025). AI deskilling is a structural problem. AI & Society, Springer.

Evolutionary Selection

Kotrschal, A. et al. (2015). Brain size affects female but not male survival under predation threat. Ecology Letters, 18.

Wooster, E. et al. (2026). Predator-prey interactions as drivers of cognitive evolution. Nature Reviews Biodiversity.

Hawkes, K. (1998/2024). Grandmothers and the Evolution of Human Longevity. American Journal of Human Biology.

Kotrschal, A. et al. (2013). Artificial Selection on Relative Brain Size in the Guppy. Current Biology, 23.

Cognitive Biases and Neuronal Recycling

Tversky, A. & Kahneman, D. (1974). Judgment under Uncertainty: Heuristics and Biases. Science, 185(4157).

Dehaene, S. (2009). Reading in the Brain: The New Science of How We Read. Viking.

Dehaene, S. & Cohen, L. (2007). Cultural recycling of cortical maps. Neuron, 56(2).

Dehaene, S. et al. (2010). How learning to read changes the cortical networks for vision and language. Science, 330(6009).

Self-Domestication and Selective Degradation

Hare, B. (2017). Survival of the Friendliest: Homo sapiens Evolved via Selection for Prosociality. Annual Review of Psychology, 68.

Wrangham, R. (2018). Two types of aggression in human evolution. PNAS, 115(2).

Cieri, R.L. et al. (2014). Craniofacial feminization, social tolerance, and the origins of behavioral modernity. Current Anthropology, 55(4).

Thomas, J. & Kirby, S. (2018). Self-domestication and the evolution of language. Biology & Philosophy, 33.