When is a human being's worldview considered firm, how is it formed up to that point, and why is it so hard to correct afterwards? Three questions that lead into a finding with three superimposed layers: the neurobiology of the individual brain, the inertia of educational and knowledge institutions, and the statistical structure of modern language models. All three layers work in the same direction. Together they explain why the materialist worldview of pre-1906 is still the default in the general education of 2026, even though physics itself has been teaching something radically different since the mid-1960s. The 1906 pattern, which we have described elsewhere as institutional path dependency of academic consciousness research, thus gains a fuller explanation: it rests not only on individual institutional decisions but on a constellation of three different inertia mechanisms that reinforce each other.
Layer 1 – How a worldview arises, locks in and defends itself
The formative phase
A person's worldview – the personal, often unconscious map of reality – arises not through theoretical decision but through experience. Parents, language, school, religion, culture, formative encounters, personal experiences: by early adulthood the brain has gathered these impressions and arranged them into a model that answers the basic questions "Who am I?", "How does the world work?" and "What is important?". In this phase the brain is extremely plastic: it builds and dismantles synapses quickly, corrects itself continuously, learns in passing, absorbs from caregivers unfiltered. Anyone who knows a small boy from a strictly religious family and another from a secular-academic home can see the differing maps of reality emerging within a few years.
When it is considered firm
Between roughly the 25th and 30th year of life this dynamic changes fundamentally. The development of the prefrontal cortex closes – the brain area responsible for long-term planning, self-perception and personality. Frequently used thinking patterns become myelinated: they receive an insulating sheath of myelin that makes the neural signal faster and more energy-efficient – and at the same time makes the patterns dramatically harder to change. The brain shifts from the build-up mode into the efficiency mode. What has been learned by then becomes a neural highway; what comes new has to clear an energy threshold that rises with age. Anyone who wants to fundamentally rebuild their worldview after age 40 is fighting their own energy budget – and the brain increasingly registers this effort as threat rather than as a learning opportunity.
The defence mechanisms
Once the worldview has locked in, a whole array of psychological defence mechanisms keeps it stable. Cognitive dissonance – the inner tension that arises when new information does not fit one's own picture – registers in fMRI scans similarly to physical pain. To avoid this pain, selective perception filters out contradicting information before it even becomes conscious. Motivated reasoning scrutinises counter-arguments with great rigour while waving confirming arguments through unchecked. And the backfire effect means that a direct confrontation with hard facts often does not shake the old worldview but actually reinforces it – out of self-protection.
Karl Popper's idea of fallibilism – the readiness to treat any one of one's own convictions as falsifiable in principle – is therefore not a psychological default state. It is a high-performance sport that only very few people sustain. For the majority: what is absorbed as truth between the ages of 20 and 30 remains, on average, stable across the entire remaining lifespan. It is not tested but defended. This is not moral weakness but a function. A being that permanently questioned every conviction would be unable to act.
Max Planck condensed this in his Scientific Autobiography (1948) to the famous sentence:
"A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it."
— Max Planck, Scientific Autobiography, Barth Leipzig 1948
Planck's sentence is not cynical but descriptive. As the founder of quantum physics he had experienced first-hand that a new, experimentally well-confirmed theory does not establish itself in the scientific community through arguments but through generational change.
When it does change after all
This description should not be misread as determinism. Neuroplasticity remains intact for life – it only becomes more expensive. People do change their worldview after 30, but rarely through arguments. Far more often it changes through experiences that make the old model painfully incompetent: a serious illness, a near-death experience, the loss of someone close, an extended stay in a foreign culture, one's own child, an encounter that can no longer be explained by what was previously sufficient. This is not a cognitive operation but an existential one. The old model is not refuted but rendered painfully incompetent; it can no longer grasp what is happening, and the brain is forced to build something new because the old no longer carries.
This is the reason why reports of near-death experiences, afterlife contact and similar events play such a different role in public debate than their history-of-science and philosophical treatment would suggest. Anyone who has had such an experience themselves, or who knows someone to whom it happened, no longer pursues the question of the reality behind it through pure evidence-argumentation. Arguments do not become unimportant – they structure what has happened. But they are no longer the trigger of the change.
Layer 2 – Institutions reproduce the locked-in worldview
If the individual brain already tends to conserve a once-absorbed worldview, what happens when those same people build institutions – universities, schools, encyclopaedias, curricula? Then individual inertia is written onto the institution's time axis. What a generation has established as standard remains standard, because the next generation – recruited, on average, from those who grew up with this standard – passes it on. Four examples make this concrete.
The 1906 pattern in the history of science
Until 1906 leading natural scientists – Kepler, Boyle, Newton, Faraday, Maxwell, Kelvin, William James – could, across four centuries, openly link their physics with theological, alchemical, mediumistic or generally metaphysical references without either side being institutionally damaged. Within a narrow window between April 1906 and August 1910 this changes. Pierre Curie dies in April 1906, Lord Kelvin in December 1907, William James in August 1910. Hugo Münsterberg stages in 1909 in New York and Boston the Palladino exposure as an institutional ritual of distancing of American academic psychology from its own psychical-research root. After that, the same combination is no longer publicly possible; it costs reputation and appointment chances. The dividing line is not forced by new data but is set by a concrete institutional event and then stabilised through curricula, appointment committees and encyclopaedic conventions. We have reconstructed this in detail in our pattern synthesis.
Prussian compulsory schooling and its form
The second line concerns the form of education itself. Prussian compulsory schooling – Frederick William I in 1717, Frederick II in 1763 with the General-Land-Schul-Reglement – established a particular form: class group, lesson clock, grading scale, canonical selection of material. This form has survived four ideological overlays between 1763 and 1990 (Wilhelminism, Weimar, Nazi era, West Germany / GDR) without anyone ever touching the basic structure. Path dependency as the institutional variant of neural myelination: what is once standard remains standard – not because it would be the best available format but because learning the opposite costs energy that nobody has. More on this in our piece on Prussian schooling.
What is actually taught in school and university
From this follows an unusual but very concretely verifiable consequence for the present state of physics teaching. Modern physics – in the narrower sense relativity (Einstein 1905/1915) and quantum mechanics (Heisenberg, Schrödinger 1925/26, Standard Model from Weinberg–Salam 1967) – is over a hundred years old. It is experimentally better confirmed than any other physical theory in history. It is mandatory material in the physics-degree canon. And it has radically changed the classical-Newtonian worldview at the fundamental level: there is no absolute time, no absolute space, no objective "stuff", no point-like, isolated particles. What we call "matter" in everyday speech is, according to the Standard Model, field energy in a bound state – the quarks have zero rest mass; mass arises through the Higgs field and to over 99 percent from the QCD binding energy between gluons and quarks. We have laid this out in detail in our piece on matter and the Higgs field.
In broad school reality, almost none of this plays a role. Classical physics – Newton, mechanics, optics, Maxwell electrodynamics, Kelvin thermodynamics – dominates clearly until the end of lower secondary school. In the upper secondary phase, modern physics is touched on: special relativity, the photoelectric effect, a bit of atomic model and Heisenberg. But the radical consequences are not developed in the limited number of hours, and for the majority of students the matter remains the historical anecdote "Einstein said that". Anyone who leaves school without a final Abitur exam will, as a rule, never have heard about quantum physics at all.
The picture is even clearer in engineering education. Mechanical engineering, civil engineering, process engineering, industrial engineering, the bulk of electrical engineering – they all work almost exclusively with classical physics: Newton, Maxwell, classical thermodynamics, classical fluid mechanics. Relativity is hardly ever needed in everyday engineering – the GPS corrections are hidden in device firmware. A German engineering graduate in these classical fields can carry out their entire professional life competently and productively without ever having engaged intensively with the Standard Model, the Higgs field or quantum field theory.
But this balance only holds for the classical engineering fields. Anyone wanting to develop and manufacture computer chips, flat-panel displays, camera sensors, smartphones, solar cells, modern battery cells, MRI and CT scanners or quantum computers does not get far without deep quantum physics. Semiconductor physics, photonics, spintronics, solid-state electrochemistry – these are not specialist niches but the foundations of every industry that today generates the highest technological added value worldwide. And here lies one of the major, rarely named reasons why Europe has largely lost these industries to East Asia and the United States over the past decades: the engineers who can design, run and further develop a 3-nm chip fab, a modern battery-cell plant or a quantum-computing laboratory are not trained in Europe in sufficient numbers, because modern physics is systematically under-represented in the broad engineering curriculum. Anyone trying to bring these industries back quickly discovers that the primary missing resource is not capital but minds that intuitively master the behaviour of electrons, phonons and quasi-particles in solids.
For the classical engineering disciplines, the Newtonian emphasis is, on the other hand, materially correct. For the structures a civil or mechanical engineer dimensions, Newton is entirely sufficient, and switching to Higgs-field language for the statics of a reinforced concrete bridge would be absurd. But it has a cultural consequence: the carrier layer of technical expertise in our society – the engineers – take their picture of matter from classical physics. The worldview "matter is stuff that consists of tiny solid particles" is not the state of physics but a didactically simplified working model. And this working model is – through the sheer quantitative weight of school and engineering education – elevated in the minds of the educated majority to reality itself. Layer 2 does not operate through suppression. It operates through choice of material and quantitative weight.
Wikipedia: the same logic in the encyclopaedia
This effect also shows in Wikipedia, today probably the most consulted knowledge source worldwide. The German main article on "Masse (Physik)" opens classically-Newtonian: inertial mass, gravitational mass, kilogram, mass conservation. The radical modern statement – that "mass" is not a property of stuff but a dynamic interaction effect between fields – sits in specialist articles (Higgs mechanism, quantum chromodynamics) that a reader only finds when searching specifically. The information is there. But the curatorial ordering preserves the classical worldview of the general readership. This is neither censorship nor ill will; it is the very same phenomenon as the selection of material in school and engineering studies. What stands in the main article is what the majority of authors grew up with. Specialist depth migrates into sub-articles.
Layer 3 – Language models as the statistical mirror of the corpus
Here a third layer joins in, one that did not exist in this form before 2023: the large language model. Anyone who today, instead of reading up on matter, physics, consciousness or reality in a textbook, asks an AI, gets an answer that follows a simple statistical logic. Language models are trained on enormous bodies of text – textbooks, Wikipedia, popular-science publications, newspaper articles, forums, books. In their answers they produce the high-dimensional statistical average of this corpus.
This has a consequence often overlooked in popular AI discussion. On direct, precise request – "How does the mass of elementary particles arise in the Standard Model?", "What is the Yukawa coupling?", "What percentage of the proton's mass comes from QCD binding energy?" – a modern AI produces a correct answer. The technical knowledge is anchored in the training data, because physics textbooks (Peskin/Schroeder, Weinberg), lecture notes, papers and the Particle Data Group are part of the corpus.
Something else is the default mode of speaking – what comes out when the question is not aimed at quantum field theory but is phrased generally, about matter, electrons, mass or the world. In this mode the model produces, statistically, the average of its training data – and that average is classical. When the AI speaks of "an electron with a mass of 9.1×10⁻³¹ kg", the Newtonian intuition resonates unspokenly. It has the technical knowledge, but its narrative centre of gravity sits in the 19th century, because that is where the largest part of its language training quantitatively took place. The same asymmetry, then, as in a human being whose brain was myelinated between 25 and 30: on direct request he can correctly cite quantum field theory – but when he speaks freely, he speaks out of his default worldview.
This observation is reproducible. Anyone who asks any large AI what matter is gets, as a rule, atom-and-particle language first – not "field energy in a bound state". Anyone who then specifically asks the same AI a Yukawa-coupling question gets a correct technical answer. The gap between these two answer modes is exactly the Layer-1 asymmetry of a biological brain, translated into the statistical structure of a neural network.
What has shifted relative to Planck
Planck's sentence about dying out had an implicit assumption: when the old generation dies, a new generation takes over which absorbs the knowledge from the beginning as a given. Across three or four generations the problem dissolves on its own. With plate tectonics (Wegener thesis from 1912, acceptance from 1965) this took roughly five decades; with Helicobacter as ulcer cause (Marshall–Warren from 1982, Nobel Prize 2005) two decades. Planck's mechanism works.
For paradigms heavily entangled with identity – where the worldview is fused with "scientificity as such" – the mechanism works much more slowly. The post-1906 materialist consensus has now seen four biological generations without the Higgs-and-QCD lesson, present in the Standard Model since the 1960s, having reached the default discourse. This is not conspiracy but Layer-2 inertia: curricula, engineering courses, Wikipedia main articles and popular-scientific language reproduce the classical order because their successive author generations are recruited from those who learned that order first. A feedback loop with high self-inertia.
Layer 3 changes this mechanism in a new way. Language models do not age, they do not die. When a model is trained on a corpus with a classical centre of gravity and then answers billions of queries a day, it reproduces the classical default worldview at a speed and reach which school and Wikipedia combined could never achieve. Planck's dying-out works for humans. For a language model trained on historical data there is no dying-out – only a new training round in which the historical data again dominate, because they necessarily form the bulk. Anyone who in 2026 asks a language model in general terms what matter is gets, by default, an answer that carries the worldview of 1900 – even though the same model, on direct request, can correctly state the worldview of 1967. The gravity of the default is the third layer of the 1906 pattern.
Consequence
The 1906 pattern is therefore no longer merely a history-of-science thesis about the 20th century. It is a path dependency that reaches into the school, engineering education, encyclopaedias and AI architectures of 2026. Whoever talks with modern AI systems about afterlife contact, near-death experiences, theories of consciousness or the central order always talks with a system whose default mode of speaking stems from a classically materialist majority of writing. On direct request the system can correctly report what Eccles, Penrose, Heisenberg, Schrödinger, Planck or van Lommel said on these topics. But when one talks with it freely it speaks, on average, the language of the standard consensus – and that is not the consensus of the best physicists of the 20th century but the consensus of the quantitatively dominant Wikipedia writers of the 21st century.
From this follows a very sober practical recommendation for anyone wanting to engage seriously with paradigmatically delicate topics: do not be satisfied with the first hit. Ask precisely. Ask about Yukawa coupling, about QCD binding energy, about Eccles's dualism, about Penrose's Orch-OR theory, about Planck's Observer interview of January 1931, about van Lommel's Lancet study of 2001 – and you get correct technical answers. Ask in general terms and you get the default average. This is not an AI weakness; it is a direct structural mirroring of the individual and institutional inertia mechanisms described in Layers 1 and 2. And it is a reminder that one's own worldview – for exactly the same reasons – is, in many places, probably still the default and not the state of the matter.
Context
This article rounds off our series on the intellectual history of the materialist turn with an element that could not have been written before 2023: the language model as a third layer of inertia. It belongs to the pattern series (pattern synthesis, Prussian schooling), to the physics line (matter and the Higgs field, Hans-Peter Dürr, Eugene Wigner) and to the methodological line (majority against experts).
Sources: Max Planck, Scientific Autobiography, J.A. Barth, Leipzig 1948. Karl R. Popper, The Logic of Scientific Discovery, Springer Vienna 1934 (fallibilism). Steven Weinberg, A Model of Leptons, Physical Review Letters 19 (1967), 1264. Particle Data Group, Review of Particle Physics, current edition (quark masses, proton mass). Peskin & Schroeder, An Introduction to Quantum Field Theory, Addison-Wesley 1995. On prefrontal-cortex maturation: Giedd et al., Brain development during childhood and adolescence: a longitudinal MRI study, Nature Neuroscience 2 (1999), 861–863, and Sowell et al., Mapping cortical change across the human life span, Nature Neuroscience 6 (2003), 309–315. On cognitive dissonance: Leon Festinger, A Theory of Cognitive Dissonance, Stanford University Press 1957. On the backfire effect: Nyhan & Reifler, When Corrections Fail: The Persistence of Political Misperceptions, Political Behavior 32 (2010), 303–330.
For more, see our curated knowledge collection – with original sources on Planck, Heisenberg, Schrödinger, Eccles, Penrose and van Lommel and further articles on the scientific debate around consciousness and matter.