24 History, not wiring: a speculative sketch of experience
(ns meme-constructor.speculative.post-s1-monads
(:require [scicloj.kindly.v4.kind :as kind]
[meme-constructor.figures :as fig]))Part VI, post S1. Status: September 2026.
This post is speculative. Parts I to III state claims precisely and test them. This post does not. It collects ideas that come out of the framework but do not follow from it, and that cannot be formalised today. Where an idea could be tested, it says how. Where it only renames a problem, it says that too. Nothing in Parts I to III depends on it.
We conjecture that how unified a mind’s experience is depends on its history, on which concepts it formed together, and not only on its present wiring. This predicts two things that the leading theory of integration does not.
- a link between two brains makes one mind gradually and irreversibly, not suddenly and reversibly
- people with the same wiring but different histories differ in how unified they are
The rest of the post is the picture behind that conjecture. A concept fits many situations and matures under criticism. Remembering is copying a culture meme from an earlier self to a later one, and the memory trace is the channel, not the meme. The self is a concept that is always active. A quale is a pattern that a brain keeps re-instantiating. None of this explains why there is something it is like to be such a pattern. That problem is still there at the end.
24.1 1. A link between brains is first a channel
Integrated Information Theory (IIT) says that consciousness is maximally integrated information, and that only the system with maximal integration is conscious (Tononi 2004). Christof Koch describes a thought experiment (Koch 2019).
Our framework says something else. A meme is not moved from one brain to another. The receiver rebuilds it from evidence, in its own concepts (post 1). Two adult brains code their concepts each in their own way. So a signal from brain A is, for brain B, only evidence.
Conjecture S1.1 (coupling). A physical link between two brains is at first a channel, whatever its bandwidth. A shared strand of experience forms only through shared concept formation: both sides develop their coding together, gradually, as a child learns a language. Once formed, it is not undone by cutting the link.
- what decides: integrated information now
- course: sudden, at a threshold
- two adult brains, wide link: can merge
- cut the link after a merge: reversible, the two persons return
- what decides: concepts formed together
- course: gradual, like learning a language
- two adult brains, wide link: a channel; communication can err
- cut the link after a merge: irreversible, one history in two halves, like a split brain
This is how the two hemispheres of one brain normally develop: together, over the corpus callosum, from the start. The brain bridge itself is a thought experiment today. The next test is not.
24.2 2. History at equal wiring
Take three groups of people whose hemispheres are similarly connected today: people born without a corpus callosum (callosal agenesis), people whose callosum was cut in early childhood, and people whose callosum was cut as adults. IIT predicts that equal wiring means equal unity. The coupling conjecture predicts that unity depends on history. People born without a callosum formed their concepts together over the remaining connections, such as the anterior commissure, which in agenesis is often enlarged.
This can be done with existing patient groups. It is the strongest empirical anchor of this post.
What we know so far fits. The first finding is Pinto et al. (2017).
- perception is split, but patients do not report two selves
- the halves formed their concepts together before the operation
- brains joined at the thalamus; reportedly shared sensations, separate identities
- the link is at the level of the senses, not where concepts form
24.4 4. What a strand is made of: concepts
Why should concepts, rather than wiring, be what unifies? Because a concept fits many situations.
Take the concept “door”.
Hofstadter calls each such step an analogy and holds that analogy is the core of thinking (Hofstadter 2001; Hofstadter & Sander 2013).
Definition S1.3 (extent of a concept). Let \(\Sigma\) be a set of situations. The extent of a concept \(c\) is the set of situations it fits:
\[\mathrm{Ext}(c) = \{\, \sigma \in \Sigma : c \text{ fits } \sigma \,\}\]
The word is taken from formal concept analysis (Ganter & Wille 1999). We avoid Deutsch’s word reach here. His reach is the ability of an explanation to solve problems it was not made for (Deutsch 2011, ch. 1). Extent is only a set of situations. A concept with a wide extent need not have reach in his sense.
Concepts absorb small differences. Here the picture uses the Everett multiverse. Knowledge-bearing matter is the same across neighbouring branches, like a crystal, while a random structure is not (Deutsch 1997, ch. 8). Take a system \(S\) in branch \(b\), and let \(N(b)\) be the neighbouring branches, which differ in small ways: a photon went left instead of right, a leaf fell a centimetre further. Write \(W_{b'}(S)\) for the set of concepts active in \(S\) in branch \(b'\).
One might call \(S\) fully defined when every neighbour has the same set \(W_{b'}(S)\). That does not work.
- a dust grain has ∅ in every branch and would be fully defined
- a few coarse concepts would beat a person
Definition S1.4 (definite content, spread).
\[\begin{aligned} D_b(S) &= \bigcap_{b' \in N(b)} W_{b'}(S) && \text{(definite content: active in every neighbour)} \\ V_b(S) &= \bigcup_{b' \in N(b)} W_{b'}(S) \setminus D_b(S) && \text{(spread: active in some neighbours only)} \end{aligned}\]
A dust grain has no definite content. A system is well defined when its definite content is large and its spread is small.
Proposition S1.5 (wide concepts add no spread; sketch). Suppose the neighbours differ only in the situation \(S\) faces, with spread of situations \(U_b\), and a concept is active exactly when it fits the situation. A concept whose extent contains all of \(U_b\) adds to the definite content. A concept whose extent misses all of \(U_b\) adds nothing. Only a concept whose extent cuts through \(U_b\) adds to the spread.
Sketch. Read the second column of the table. ∎
A concept with a wide extent has its boundary far from ordinary branch differences. An episodic detail, “the leaf landed exactly here”, has its boundary everywhere. A concept is error correction at the level of meaning, as an attractor is error correction at the level of neurons.
Conjecture S1.6 (knowledge unifies). The more of its experience a system organises in wide concepts, the larger its definite content and the smaller its spread. On this view knowledge does not create experience. It unifies it. This turns the combination problem of panpsychism around.
This grades systems; it does not draw a line. Pigeons learn to sort photographs by whether they show people (Herrnstein & Loveland 1964), and honeybees learn “same” and “different” (Giurfa et al. 2001). Animals have concepts.
24.5 5. Concepts mature
Every use of a concept meets a new situation, and the situation can show that the concept is wrong.
Conjecture S1.7 (maturation). A successor to a concept is a new conjecture, not a summary of instances. Each activation of a concept is a chance to criticise it. When some aspect fails to fit, the host guesses a successor, and further situations test it.
Successors that survive often have a wider extent, but criticism can also split a concept, as when whales leave “fish”, and children narrow words they had stretched too far.
One might read maturation as abstraction, as formal concept analysis says that fewer shared aspects means more situations (Ganter & Wille 1999).
Maturation cannot simply widen the extent. A concept that fits everything tells no two situations apart and carries no information. A good concept is wide and still excludes something.
How to test it. Transfer studies measure whether what was learned in one setting is used in new ones. The conjecture predicts that a concept which has been criticised and revised more often transfers further than one that has only been used more often. Revision frequency has to be measured separately from use frequency.
Concepts about concepts. A brain’s content at a moment, its active concepts, is itself a situation. So a concept can fit a pattern of other active concepts: “cause”, “number”, “analogy” itself. This is what combination and abstraction need. One might think that concepts nested without a top level are what makes a person, and that animals lack them. That does not hold: no condition on what a system can reach separates a person from an enumerator (post I.2, Proposition I.2.14), and an unbounded space of concepts is such a condition. How deep an animal’s concepts nest is an empirical question.
A possible neural basis. Buzsáki argues that the brain generates its own patterns first, including preconfigured sequences in the hippocampus, which get meaning only through action and experience, and that the hippocampus acts as a pointer into neocortical content (Buzsáki 2019; compare Teyler & DiScenna 1986).
Conjecture S1.8 (a maturing index). A hippocampal pattern serves as a stable index for a concept, while the neocortical associations it points to are updated at each activation. The concept matures while its index stays the same, as a word keeps its spelling while its meaning deepens.
This is our reading, not Buzsáki’s claim. How to test it. The hippocampal pattern for a concept should stay stable across learning while the neocortical pattern it co-activates changes. Damage to the hippocampus after a concept is learned should stop the concept from maturing more than it stops it from being used.
24.6 6. The engram is a channel, not the meme
Remembering is rebuilding, not reading out (Bartlett 1932). In the terms of post 1, for a culture meme it is a copy task: a meme is copied from the earlier self to the later self, by the same creativity that copies memes between people.
Conjecture S1.9 (the engram is a channel). The engram, the physical memory trace, is not the meme. It turns a cue into a pattern of evidence, from which the later self rebuilds the meme. It is the inner counterpart of the channel between two people.
Three known findings fit.
A recalled memory becomes labile again and has to be stored anew (Nader, Schafe & LeDoux 2000). A false memory errs exactly as a meme copy can, by loss and by gain. In mice, reactivating a few hippocampal cells triggers a fear memory (Liu et al. 2012), and a false fear memory of a place can be written (Ramirez et al. 2013). The decoder dichotomy (F7) of post 2 leads one to expect this: a closed program such as fear conditioning can have a decoder, even an encoder. The prediction for humans follows.
The engram alone means nothing. Like a sentence, it is evidence only for a rebuilder with the right concepts. This may be part of why early childhood is forgotten: the early engrams may still exist (lost infant memories in mice can be restored by reactivating the tagged cells, Guskjolen et al. 2018), but the later self lacks the concepts to rebuild from them. Richard Semon, who coined the word “engram”, counted memory and heredity as one phenomenon (Semon 1904). The difference here is that the engram is the channel, not a recipe. The function of a culture meme has no recipe. It is nowhere in particular: it is the function, the abstract equivalence class, that is rebuilt anew between engram and rebuilding.
The receiver’s concepts are therefore the meme’s niche. A meme can be rebuilt only where there are concepts to rebuild it from, and it comes back only in situations those concepts fit.
Records and causes. A memory is a record because it was caused by what it records. A random fluctuation that happens to match a brain is a Boltzmann brain.
So such a pattern is not a remembering mind, whatever it is like to be it. Dust tells this as a story.
24.7 7. The self is a concept that is always on
“The self” is a concept that fits every situation a person lives through. A concept that is always active tells no two situations apart, so it carries no information: an attribute carries information only if the medium could have been otherwise (Deutsch & Marletto 2015). But it connects. Because it is active in every situation, every episode attaches to it. Two aspects have to be kept apart.
- constant
- without information
- connecting: every episode attaches to it
- my properties, history and aims
- variable
- informative
Conjecture S1.10 (when the strand begins). Autobiographical memory, and with it a strand of experience, begins when the self as index exists and there are enough concepts to attach experience to. Before that there is a bundle: several weakly defined worlds at once, and no strand that later memory can reach.
Two findings fit. Howe & Courage (1993, 1997) hold that the cognitive self, measured by the mirror test at the end of the second year, sets the lower limit for autobiographical memory. They call it necessary but not sufficient, and infantile amnesia ends only at three or four. In between lies language: children do not later describe events from before they could speak with words they learned afterwards (Simcock & Hayne 2002).
Against this, rodents have infantile amnesia too, and there high neurogenesis in the hippocampus causes it (Akers et al. 2014). So at least two mechanisms act, and only one of them can be conceptual.
Dissociative identity disorder fits as several self-indices on one base of concepts. Patients report amnesia between identities, yet objective tests show information passing between them (Huntjens et al. 2003). What is separated is only “this happened to me”.
How to test it: lucid dreams. In most dreams there is a point of view but little reflective self (Hobson 2009). The conjecture predicts that lucid dreams, in which the reflective self is active, are remembered better than ordinary ones. Higher-order theories of consciousness predict that too. They differ on ordinary dreams.
Dream diaries and awakening studies can measure this. Bundle tells the start of a strand from the inside.
24.8 8. A quale is a recurring pattern
A state of experience, say “this red, now”, is a pattern that a brain instantiates in some branches and moments and not in others. Groups of neurons that fire together wire together into cell assemblies (Hebb 1949), and such networks behave like attractors (Hopfield 1982).
So the red you see today is not merely similar to yesterday’s. Up to the error the attractor corrects, it is the same pattern, in the sense of the equivalence class of post 1.
Conjecture S1.11 (a quale is a recurring pattern). What is called a quale is, roughly, a pattern that a brain keeps re-instantiating by attractor dynamics and error correction. Its sameness over time is identity of pattern, not resemblance.
Functionally such a pattern is an index, a symbol that other processes address, as in Dennett’s picture of consciousness as a virtual machine running on the brain (Dennett 1991). On how it works, Dennett is right. The only addition is that the symbol is a real pattern whose recurrence is a fact, not a useful fiction.
How to test it. Only weakly. If error correction in a sensory attractor is disrupted, reports that “it looks the same as before” should fail before reports that “I see something”. A functionalist account without this picture can predict the same, so the test would not decide much. Same Red tells this as a story.
24.9 9. The frame, and what it leaves open
The picture can be put in Leibniz’s terms. His monads have no windows: none acts on another (Leibniz 1714, §7), yet all fit together in a pre-established harmony (§78). Read every real pattern in the multiverse as a monad, and the harmony becomes what we call the laws of physics: which patterns can follow or contain which. Unlike Egan’s dust (Egan 1994), where every pattern a reader can find in the noise is a world, only the patterns that the laws actually instantiate count. The frame orders the ideas above and adds no prediction of its own.
Which patterns fit is found out, not derived. One might derive the laws from a simplicity prior: short programs are more common, so simple laws dominate (Schmidhuber 1997). Program length depends on the universal machine that runs the programs. The invariance theorem says that it does so only up to an additive constant (Li & Vitányi 2019; post I.4, Proposition I.4.5).
“Why these laws and not those” is a finite ranking of a few candidates. For it, the prior presupposes a machine, and which machine is physically real is a question of physics. Deutsch calls the opposite assumption the mathematicians’ misconception (Deutsch 1997, ch. 10). So the argument holds for finite rankings and only there. Harmony tells this as a story.
The hard problem is still there. Nothing here explains why it feels like anything to be such a pattern (Chalmers 1995). Leibniz saw it: walk into a thinking machine as into a mill, and you find parts pushing each other, never a perception (§17). Walk into the attractor that keeps producing red, and you find the pattern, not the red.
A first idea was that experience is existence seen from inside. It brings nothing, because one could equate existence with mass or number just as well, and it tells no experienced state from an unexperienced one.
One way the whole picture fails. It rests on the Everett multiverse. If an objective collapse of the wave function were found, the footing would be gone. The simplest version of the Diósi-Penrose collapse model has been ruled out (Donadi et al. 2021); others are still being tested.
24.10 Short stories
The stories are fiction. They take the conjectures as true inside the story, to see what follows.
- Same Red. An implant weakens the error correction in a painter’s colour attractors. Red is never quite the same twice.
- The Bridge. Two adults are joined by a high-bandwidth link. They expected to merge. They have to learn a language instead.
- Bundle. The last months before a child’s first memory, told from inside the bundle.
- Dust. A noise source produces, by chance, the exact state of a small learning creature. Is it the creature?
- Harmony. Physicists inside a simulation try to guess the laws of the machine they run on, and find out that their idea of “simple” was their own machine’s.
24.11 Summary
- The unity of experience depends on history, on concepts formed together, not only on present wiring. Against IIT: a link between brains merges them gradually and irreversibly, not suddenly and reversibly.
- The feasible test compares callosal agenesis with early and late callosotomy at similar wiring. A strand is shared when a concept taught to one part is used by the other at once and with the same error profile.
- Concepts fit many situations. Wide concepts add definite content without spread across branches, so knowledge unifies experience rather than creating it.
- Concepts mature by conjecture and criticism. Revision frequency, measured apart from use frequency, should predict transfer; hippocampal damage should stop maturation more than use.
- The engram is a channel, not the meme: every recall of a culture meme is a copy that can lose and gain, and engram manipulation should work for fear and places, not for explanations.
- The self is a concept that is always on: no information, but everything attaches to it. The strand begins when it exists and enough concepts are there. A quale is a recurring pattern; functionally an index, as Dennett says. The hard problem remains.
24.12 Open problems
The dispute over the limbic route is in Pinto et al. (2017).
24.13 References
- Akers, K. G. et al. (2014). Hippocampal neurogenesis regulates forgetting during adulthood and infancy. Science 344(6184), 598–602. (not re-checked)
- Bartlett, F. C. (1932). Remembering. Cambridge University Press. (not re-checked)
- Buzsáki, G. (2019). The Brain from Inside Out. Oxford University Press. (not re-checked)
- Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies 2(3), 200–219. (not re-checked)
- Dennett, D. C. (1991). Consciousness Explained. Little, Brown. (not re-checked)
- Deutsch, D. (1997). The Fabric of Reality. Allen Lane. Ch. 8 (multiversal crystal), ch. 10 (mathematics and physics). (not re-checked)
- Deutsch, D. (2011). The Beginning of Infinity. Allen Lane. Ch. 1 (reach). (not re-checked)
- Deutsch, D. & Marletto, C. (2015). Constructor theory of information. Proc. R. Soc. A 471(2174): 20140540. (not re-checked)
- Donadi, S. et al. (2021). Underground test of gravity-related wave function collapse. Nature Physics 17, 74–78. (not re-checked)
- Egan, G. (1994). Permutation City. Millennium. (not re-checked)
- Ganter, B. & Wille, R. (1999). Formal Concept Analysis: Mathematical Foundations. Springer. (not re-checked)
- Giurfa, M. et al. (2001). The concepts of ‘sameness’ and ‘difference’ in an insect. Nature 410, 930–933. (not re-checked)
- Guskjolen, A. et al. (2018). Recovery of “lost” infant memories in mice. Current Biology 28(14), 2283–2290. (not re-checked)
- Hebb, D. O. (1949). The Organization of Behavior. Wiley. (not re-checked)
- Herrnstein, R. J. & Loveland, D. H. (1964). Complex visual concept in the pigeon. Science 146, 549–551. (not re-checked)
- Hobson, J. A. (2009). REM sleep and dreaming: towards a theory of protoconsciousness. Nature Reviews Neuroscience 10, 803–813. (not re-checked)
- Hofstadter, D. R. (2001). Epilogue: Analogy as the core of cognition. In D. Gentner, K. J. Holyoak & B. N. Kokinov (eds.), The Analogical Mind, MIT Press. (not re-checked)
- Hofstadter, D. R. & Sander, E. (2013). Surfaces and Essences. Basic Books. (not re-checked)
- Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective computational abilities. PNAS 79(8), 2554–2558. (not re-checked)
- Howe, M. L. & Courage, M. L. (1993). On resolving the enigma of infantile amnesia. Psychological Bulletin 113(2), 305–326. (not re-checked)
- Howe, M. L. & Courage, M. L. (1997). The emergence and early development of autobiographical memory. Psychological Review 104(3), 499–523. (not re-checked)
- Huntjens, R. J. C. et al. (2003). Interidentity amnesia for neutral, episodic information in dissociative identity disorder. Journal of Abnormal Psychology 112(2), 290–297. (not re-checked)
- Koch, C. (2019). The Feeling of Life Itself. MIT Press. (not re-checked)
- Leibniz, G. W. (1714). Monadology. Translated by Jonathan Bennett, earlymoderntexts.com. §7, §17, §78. (not re-checked)
- Li, M. & Vitányi, P. (2019). An Introduction to Kolmogorov Complexity and Its Applications. 4th ed. Springer. (not re-checked)
- Liu, X. et al. (2012). Optogenetic stimulation of a hippocampal engram activates fear memory recall. Nature 484, 381–385. (not re-checked)
- Nader, K., Schafe, G. E. & LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature 406, 722–726. (not re-checked)
- Pinto, Y. et al. (2017). Split brain: divided perception but undivided consciousness. Brain 140(5), 1231–1237. (not re-checked)
- Ramirez, S. et al. (2013). Creating a false memory in the hippocampus. Science 341(6144), 387–391. (not re-checked)
- Schmidhuber, J. (1997). A computer scientist’s view of life, the universe, and everything. In C. Freksa et al. (eds.), Foundations of Computer Science, LNCS 1337, 201–208. Springer. (not re-checked)
- Semon, R. (1904). Die Mneme als erhaltendes Prinzip im Wechsel des organischen Geschehens. Engelmann. (not re-checked)
- Simcock, G. & Hayne, H. (2002). Breaking the barrier? Children fail to translate their preverbal memories into language. Psychological Science 13(3), 225–231. (not re-checked)
- Teyler, T. J. & DiScenna, P. (1986). The hippocampal memory indexing theory. Behavioral Neuroscience 100(2), 147–154. (not re-checked)
- Tononi, G. (2004). An information integration theory of consciousness. BMC Neuroscience 5, 42. (not re-checked)
source: book2/notebooks/meme_constructor/speculative/post_s1_monads.clj