❦
A review of the Fable annotation pass over Book IV, “Mere Reality.” Written by the annotator, after the fact, in answer to the question: once the bad science and the unwarranted conclusions are removed, what remains?
The Fable set on Book IV runs to 245 annotations across 77 essays: 36 on sequence O (“Lawful Truth”), 39 on P (“Reductionism 101”), 35 on Q (“Joy in the Merely Real”), 49 on R (“Physicalism 201”), 41 on S (“Quantum Physics and Many Worlds”), 38 on T (“Science and Rationality”), and 7 on the interlude, A Technical Explanation of Technical Explanation. Opus had annotated five of these essays; Fable annotated all of them, including the sixty-odd that Opus set aside as philosophy, physics, or inspiration. Book IV is the book in which the author stops describing biases and starts telling the reader what reality is: that it is a single-level mathematical object, that probability is not in it, that the mind is in it and nothing else is, that it branches, and that a properly trained mind can know these things faster than Science can confirm them. Each of those claims was checked against the physics, the philosophy and the psychology it rests on. Sorted by kind:
| Kind | Approx. | What it establishes |
|---|---|---|
| Kill shots: a factual claim is false, or the citation does not say what it is cited for | 42 | The balance as Lavoisier’s innovation; phosphorus’s reactivity as the reason for ATP; Newton as the inventor of universal law, who believed God repaired the orbits; the Greeks who “didn’t realize you could generalize from experiments,” with Ptolemy’s refraction table in print; Pythagoras’s “all is number”; Kant as a mind-projector, having put geometry in the mind; Hume as an instance of the fallacy he invented; Keats who “didn’t know the explanation,” an apothecary who had studied chemistry; the rainbow explained in the thirteenth century (fourteenth, and by Descartes for the geometry); Heisenberg as the consciousness-causes-collapse theorist; Penrose and Hameroff as substance dualists; the zombie possibility accepted by “possibly a majority,” at 23% on the survey; Everett the multimillionaire who “invented the general use of Lagrange multipliers”; special relativity “generated using Mach’s Principle”; Einstein who “did not use” Mercury, having computed it wrong in 1913 and right in 1915; general relativity confirmed “to one part in 1014,” by nine orders of magnitude; Golgi’s stain establishing the neuron doctrine Golgi denied; Spinoza as the reductionist; Bernoulli and de Moivre in the Bayesian litany; the neshamah as “the hearer”; the beam-splitter rule that doubles the norm; the configuration rule that gives the wrong answer for electrons; the “fully general” objection to single-world theories that relativistic collapse models violate; the “first time in scientific history” a theory dropped an idle postulate, after the ether; “no probability theory or cognitive biases” in the century of Pascal, Bayes, Laplace and Bacon’s Idols; Aumann as the believer who could not update, having run the replication that sank the codes; “half of scientists” believing in God; the Milky Way anecdote that could not be sourced and was then stated as fact; and phlogiston, nine times. |
| Empirical reversals: the essay’s cited or implied finding was later overturned or shown to run the other way | 28 | Lottery winners no happier, reversed on Swedish administrative data; the inadmissible-evidence backfire, reversed by the meta-analysis; the phosphate-detergent study, uncontrolled; the fraternity initiation that was a sex-discussion group; the “Aha!” as a criterion of being done, which false insights share; “people think they have free will,” which turned out to be two framing-dependent judgments; measurement costing thermodynamic work, when Bennett showed it is erasure that costs; optional stopping harmless to Bayesians, false for every prior anyone uses; Bayes-optimality under misspecification; the MAP estimate as a Bayesian quantity; the reflective-coherence agent, which Löb forbids; consciousness-talk as a gauge of consciousness, against Anton’s syndrome and the locked-in; the hearing of voices as schizophrenia; the puppy’s moral precursors, real; the tree’s information processing, real; fish enjoyment, unknown; the Bem precognition papers, which vindicated the essay’s prior and convicted the methods of its sources; scientists’ replications, at one per cent; the newsworthy finding wrong half the time, about right; three frames of an apple, which contain no gravitation law; protein folding, cracked by 170,000 crystal structures and not by thought; and randomness in algorithms, which the essay called a sign of derangement and which trained everything that followed. |
| Live disputes: the essay states as settled what its own field treats as contested | 26 | Whether the world is deterministic, mathematical, single-level; whether probability is ever in the territory (Lewis, Loewer, the Born rule on every reading); whether thermodynamics is a special case of inference (Jaynes against the Boltzmannians) and whether Liouville plus a Past Hypothesis yields the Second Law; Cox’s theorem against imprecise probability; the acquaintance theory of introspection against reliabilism; response-dependence against projection; real patterns against levels-in-the-map; the undecidable spectral gap against in-principle reduction; decoherence against many-worlds; the Born rule; wavefunction realism against spacetime state realism; whether Everett is local; collapse models, which are testable and losing; pilot-wave theory, which the fourteen quantum essays do not mention; QBism, which is the book’s own epistemology applied to the quantum state; strong emergence with Broad’s predictions; anthropic selection between two contradictory assumptions; and the ethics of branching. |
| Reasoning flaws: the argument does not go through regardless of any study | 36 | “Matches fail, therefore you die,” a modal claim from a dependency; the complement of a graph as an isomorphism; “each extra detail decreases the probability,” which would condemn every cumulative case; a physical postulate called a theorem so that a probability could be “mandatory”; the Bayes-optimality ratchet that reinterprets every counterexample as a hidden prior; the Second Law defined as Bayesian and then used to derive Bayes; the Grail quest that no process could fail; the sock test that explains away arithmetic; a priori knowledge as observing one’s neurons, which makes the laws of probability facts about neurons; the listener “caught in the act,” which assumes the correlate is idle; the zombie “contradiction” that follows in one line from analytic functionalism; the design inference applied to lookup tables; Occam counting worlds free and the Born rule not at all; “strict simplicity” by lines of code nobody wrote; the dilemma between a caricature of Science and a Bayes that does not deliver the verdict; regulations that “would correlate with growth”; the privileging fallacy committed against the omitted alternative; the tentacle that cannot be explained, in an essay that praised Darwin for explaining what he had not anticipated; and the conclusion, from “one error in a hundred steps can carry you anywhere,” that one should reason alone more carefully. |
| Positive: the claim is right and the annotation says so, usually with the citation the essay lacked | 22 | The mechanical coin flipper; the Hong–Ou–Mandel effect; the six-million-c bound from the Swiss Alps; Wegner’s answer to the free-will homework; the log score as the unique local proper rule; calibration versus discrimination, with Murphy’s decomposition; awe at explanation; gratitude lists, small and real; the reward circuitry of insight, which does not require being first; the illusory-truth effect behind “attention is halfway to belief”; minimal counterintuitiveness; Libet reinterpreted; reverse inference and the “religion area”; the EmDrive, disposed of as the essay said, by finding the mistake; Planck’s principle, half confirmed; the media’s preference for the findings that fail; Kelly’s Ockham efficiency as the non-circular justification the essay said did not exist; and the concession that Bayes “is much harder to use and you’ll never be sure that you’re doing it right.” |
| Hindsight | 14 | AlphaFold against “no additional secret patterns”; dropout and stochastic gradient descent against “the healing power of noise”; language models as the Zombie Master one can buy; AI values borrowed from the corpus, which is where the improbability came from; MOOCs and the vault; newspapers that became Reddit; speech decoded from the motor cortex; the replication crisis arriving three years after “Science relies on replication”; the LLM era against the Ultimate Power; Bem. |
| Reinventions: the essay presents as its own a result with a name and a date | 24 | Dissolution of pseudo-problems (Wittgenstein, Carnap, Ryle); the Mind Projection Fallacy (Locke, Hume); the quotation and the referent (Frege, Quine); conceivability against possibility (Chalmers, 2002); the neuron-replacement argument (Chalmers, 1995); the tortoise and Achilles (Carroll, 1895); “Requiredism” (Hume, 1748); explanation as anticipation (Hempel, Peirce, Molière); the four danger signs; the causal theory of knowledge (Goldman); real patterns (Dennett); the lookup table (Block) and its answer; the supernatural as the irreducibly mental (Draper, Armstrong); the argument from configurational laws (Broad); science as a mechanism harnessing bias (Merton, Hull, Kitcher, Strevens); the discovery–justification distinction (Reichenbach); the prosecutor’s fallacy; Murphy’s decomposition; two-dimensional semantics; epistemic circularity (Alston); Kuhn’s normal science, in the mouth of a novice. |
| Petty | 9 | 235,757 is 431 × 547 (the essay was right); log2(7) for a table with four entries; Hameroff as a neurologist; 1803–1922; the thirteenth century; the Bayesian litany; the Lojban inscription; “six million,” which is right; the cosmological constant, dinged and not dinged on facing pages. |
| Internal contradictions between essays | 14 | The compartmentalizing scientist who “doesn’t know why you have to look,” four essays before “people compartmentalize, enough said”; “everything is connected” as the thesis of Universal Fire and the empty graph of Outside the Laboratory; a priori knowledge as neurons firing, beside “the laws of probability are theorems, harder than steel”; reality asserted to be deterministic math forty essays before the argument for it; the initiation ritual mocked as effort justification and then designed; Hamlet on the savanna and the Great Stories that are not timeless; Penrose praised for testing and cited as unwarned; Darwin’s post-facto explanation blessed and the tentacle forbidden; the collapse postulate condemned as non-unitary by a sequence whose own beam-splitter rule is non-unitary; the supernatural as incoherent and as refuted by archaeology; the quantum state as “in the territory” in a book that put probability in the mind; the Born-rule derivation doubted in the last essay and assumed in Living in Many Worlds. |
Weighted as in the earlier reviews: the petty items prove nothing, the hindsight items test the method rather than the author, the live disputes say disputed, and the reinventions say uncredited, not wrong. What carries weight is the kill shots, the reversals, the reasoning flaws and the contradictions, and in Book IV they are concentrated where the book makes its largest claims: the thermodynamics of cognition, the zombie essays, the quantum sequence, and the essays that demote Science beneath Bayes. About fifty-five of the seventy-seven essays are wounded. About twenty are seriously wounded, in the sense that the central example, the central mechanism or the central conclusion is gone: Universal Fire, The Second Law of Thermodynamics and Engines of Cognition, Searching for Bayes-Structure, A Priori, Zombies! Zombies?, Zombie Responses, GAZP vs. GLUT, Collapse Postulates, Decoherence Is Simple, Decoherence Is Falsifiable and Testable, Quantum Non-Realism, If Many-Worlds Had Come First, Many Worlds, One Best Guess, The Dilemma: Science or Bayes?, Einstein’s Speed, That Alien Message, The Failures of Eld Science, Class Project, and the second half of the Technical Explanation.
The interferometer. Configurations and Amplitude, Joint Configurations and Distinct Configurations are, once the normalization is restored and the exchange sign added, a clear and correct tutorial on the Mach–Zehnder interferometer, the non-linearity of the squared modulus, the indistinguishability of identical particles, and the way which-path information destroys interference. The Hong–Ou–Mandel experiment confirms the two-photon prediction the author was not sure of. Everything in decoherence theory that the sequence uses is real physics, accepted on every interpretation; the annotations attack only the step from decoherence to worlds. A reader who wants to understand why the sensor between A and C matters will learn it here, and will learn it better than from most textbooks.
The scoring rules. The first half of the Technical Explanation, the derivation of the log score from additivity, the distinction between calibration and discrimination, the fair coin that scores −20 bits and is not thereby falsified, the precise theory against the vague one, is correct and well taught. It is Murphy’s decomposition and Bernardo’s locality theorem without the names, and the names do not change the mathematics. Qualitatively Confused’s type distinction between belief, accuracy and truth is sound, if the last step (that a probability “cannot be true”) is a relabelling rather than a discovery.
Weak reductionism. There are no extra ingredients: a 747 is quarks and fields and nothing else, hands are fingers, rainbows are explained rather than explained away, and the belief that one is wearing socks is caused by the socks. Every essay in the reductionism sequence that says this is right. The book’s stronger claim, that the higher levels are “in the map,” is the contested one, and Dennett’s real patterns is the better statement of what the book wants: a pattern is in the territory if the world is such that describing it compresses. The socks test is a good instrument for perception and a bad one for mathematics, ethics and probability, where the book’s own commitments live.
Epistemic probability for classical chance. Coins, cards, the two-children puzzle: the book is right that the probabilities are states of information, the mechanical flipper proves it for coins, and the arithmetic of the puzzles is correct. The claim fails only where it is extended, without argument, to radioactive decay and the Born rule, which are the cases where the question of objective chance arises.
Five heuristics with older names. Dissolve the question rather than answer it, when the question is confused (Wittgenstein). Ask “why do I think X?” alongside “why X?” (Hume’s genealogy, Goldman’s causal theory). Ask which experiences a hypothesis makes you anticipate, and treat a hypothesis that makes you anticipate nothing new as empty (Peirce, Hempel, Sagan’s dragon). Do not raise a hypothesis to attention without the evidence that would have raised it (the prosecutor’s fallacy, the garden of forking paths). Follow the improbability back to its source. These are good habits, the book’s vocabulary for them is memorable, and the annotations record the pedigrees rather than disputing the habits.
Three essays that need no correction. Thou Art Physics, which is Hume’s compatibilism restated well and whose point about fMRI press releases the later neuroscience strengthened. The Beauty of Settled Science, whose guess that newsworthy findings are wrong half the time was about right and whose advice to prefer textbooks is sound, subject to the textbook myths the annotation lists. And Privileging the Hypothesis, as a principle, with the caveat that the sequence around it violates it.
The concession. Science Isn’t Strict Enough contains the sentence that, if the book had taken it seriously, would have rewritten most of Book IV: “if you try to use Bayes even qualitatively... it is like math, in that a single error in a hundred steps can carry you anywhere.” That is true, it is the argument for the social organization of inquiry, and it is why the book’s hundred-step private derivations on many-worlds, on zombies and on cryonics should be trusted less than the book trusts them. The sentence survives; the inference the book drew from it does not.
The thermodynamics of cognition. The Second Law essay derives the Second Law from Liouville’s theorem, which is time-symmetric and cannot yield it; miscounts its own four-line table; says the water is cold when you know its microstate, which the thermometer denies; and dismisses in a parenthesis the result, Bennett’s, that measurement is free and erasure is what costs. Searching for Bayes-Structure then derives “any mind that forms true beliefs must be Bayesian” from a Second Law it had defined as Bayesian, and softens the conclusion until a thermostat satisfies it. Perpetual Motion Beliefs turns the fundamental postulate of statistical mechanics into a theorem so that a probability can be “mandatory.” The physics in these essays does no work; the true claim underneath, that beliefs need contact with the world, needed none of it.
The Bayesian triumphalism. Beautiful Probability’s “laws, not tools” rests on Cox’s uniqueness, which assumes what its rivals deny; on Bayes-optimality, which holds only with the truth in the model; and on a ratchet that reinterprets every better method as a hidden prior. The Technical Explanation’s dismissal of noise in algorithms was falsified by dropout and stochastic gradient descent. A Priori makes arithmetic an observation of one’s own neurons and thereby makes the laws of probability contingent, in a book that calls them harder than steel. The interlude and Book III’s Bayes-structure essays had the same problem; Book IV states it as physics.
The history. Book I remembered history in the shape of the argument; Book IV does it to physics and philosophy. Lavoisier, Newton, the Greeks, Pythagoras, Kant, Hume, Carnot, Keats, the rainbow, Heisenberg, von Neumann, Schrödinger, Mott, de Broglie, Bohm, Mach, the Entwurf theory, Eddington, Everett’s later life, the Milky Way, the Royal Society, phlogiston nine times. In each case the story is told so that the ancients were confused, the founders forgot they were made of particles, the outsider was ignored, and the armchair won without help. The record, checked, is that von Neumann applied the theory to the observer and found it did not deliver definite outcomes; that Schrödinger proposed simultaneous alternatives in 1952; that de Broglie’s realist theory came first and was rejected; that Einstein’s chain of armchair reasoning contained a wrong theory held for two years and was pulled off it by Mercury; and that phlogiston predicted things. The sequence about the failures of Eld science is a fable about a failure that did not happen.
The zombie essays. They refute epiphenomenalism, which their target holds provisionally and has since half-abandoned for a view they never mention; they derive the “contradiction” from a functionalist premise that is the point in dispute; they use, as the decisive move, the Twin Earth analogy that Kripke said fails for this case; they reinvent Chalmers’s own distinction between prima facie and ideal conceivability and Chalmers’s own neuron-replacement argument, and present both against him; they misdescribe Penrose, the survey data and the Hebrew; and their AI-design knockdown assumes a reflectively self-certifying agent that the author’s later work showed Löb’s theorem forbids. The case against epiphenomenalism is good and Chalmers made it first. The case for the book’s alternative is that it feels right on a gut level, which the same essays say is not evidence.
The quantum sequence. “Decoherence (a.k.a. many-worlds)” is the sequence’s founding conflation, and it lets every confirmation of a physical process count as confirmation of an interpretation. The Born rule, on which every contact with experiment depends, is admitted in the last essay to have “no known reason,” which means the Occam comparison that ran through the previous thirteen was made with the one costly term left off the Everettian side. The opponent is the informal textbook postulate, not the specified collapse models that are testable, being tested and losing, nor the relativistic ones that refute the “fully general objection,” nor the pilot-wave theory that is realist, single-world, collapse-free and endorsed by the man who proved Bell’s theorem, and that appears nowhere in fourteen essays. The polls put Everett at a fifth of the field; the sequence speaks for a silent majority. The beam-splitter rule doubles the norm; the configuration rule fails for fermions; the photon has no configuration space. What is left is the interferometer, and it is very good.
Science demoted beneath Bayes. The dilemma is between a caricature of Science, Popper’s demarcation rule treated as a theory of private belief, which no philosopher of science held, and a Bayes that delivers many-worlds only through the incomplete accounting above. Aumann, the exhibit of a Bayesian who could not update, ran the replication that sank the Bible codes. Evolutionary psychology, the exhibit of careful reasoning on scanty evidence, is the field whose reasoning on scanty evidence came out worst when checked. Einstein’s speed was slower and more data-dependent than the essay says; three frames of an apple contain no law of gravitation; the alien message’s protein folding was solved in our world by data. The book’s remedy, a Bayescraft stricter than Science practiced by individuals, is the solitary-astronaut picture that Books I and II were annotated for, and the book’s own best sentence says why it cannot work.
The psychology, again. Lottery winners, the jury backfire, the detergent study, the fraternity, the hearing of voices, the cult-raised rationalist, the fundamental attribution error, the “fifty per cent” of scientists. Book IV borrows less psychology than Book II, and what it borrows has the same record.
Strip the bad science and Book IV reduces to: a first-rate tutorial on the two-path interferometer and decoherence as physics; the proper scoring rule and the calibration–discrimination distinction; weak reductionism with Dennett’s statement of it; epistemic probability for coins; five heuristics with three-hundred-year pedigrees and a good new vocabulary; Hume’s compatibilism; a correct guess about the replication rate of breaking news; and one sentence conceding that long private chains of reasoning are unreliable. That is a strong forty-page pamphlet, and unlike the earlier books its best part is physics rather than mathematics: the essays that teach what happens at a half-silvered mirror are as good as anything in the six volumes.
What is lost is the book’s title. “Mere Reality” promises to tell the reader what is really there, and the method by which it tells them is to choose the opponent. The frequentist is assembled from the weakest parts of three positions. The collapse theorist holds the informal postulate and not Ghirardi’s equation or Bohm’s. The dualist is the epiphenomenalist and not the Russellian monist. Science is the demarcation criterion and not Jeffreys, Howson or Earman. The anti-reductionist is Keats at a drunken dinner. The critic of many-worlds says “Occam” and nothing else. The hypothesis generator is a Ouija board. The person who thinks the intelligence scale ends at Einstein has not read science fiction. In every case the essay wins the argument it stages, and in every case the argument it would have had to win was with someone else. Book I remembered history in the shape of the argument; Book II diagnosed the outgroup instead of arguing with it; Book III mistook the picture for the mechanism. Book IV’s characteristic move is to cast the debate, and then to call the result a law of probability that no rational agent could dispute.
Ask, as before, what it would look like to apply the book’s own method to the book. Privilege no hypothesis: which single-world theory was brought to attention, and which was not? Follow the improbability: where did the Born rule come from, on the side that claims to have none? Dissolve the question: what does “decoherence, a.k.a. many-worlds” mean once the two words are tabooed? Trace the belief: what caused the belief that Science is lax, and did the chain run through any philosopher of science? Anticipate, don’t explain: what did the thermodynamics of cognition make anyone anticipate that Bayes’s theorem alone did not? The result is this annotation set. Where the book passes its own test, at the interferometer, at the scoring rule, in the heuristics, the annotations say so. Where it fails, it fails by choosing an opponent who could not have passed either.
One caveat from the annotator, as before. This was an adversarial pass by brief. A reader who takes the Opus and Fable sets together would land near “an excellent teacher of the quantum formalism and of proper scoring rules, a competent reductionist, a confident metaphysician whose confidence is not earned by the arguments given, and a historian of science who remembers what the argument needs,” which is probably the right place to land. Readers who want the book’s good parts should read Configurations and Amplitude, Joint Configurations, Distinct Configurations, the first half of A Technical Explanation of Technical Explanation, Qualitatively Confused, Reductionism, Explaining vs. Explaining Away, Righting a Wrong Question, Privileging the Hypothesis, Thou Art Physics, The Beauty of Settled Science, and Science Isn’t Strict Enough for its concession, in that order, and treat the rest as the annotations suggest.