Memes as Abstract Constructors

(ns index
  (:require [scicloj.kindly.v4.kind :as kind]
            [meme-constructor.figures :as fig]))

1 Memes as Abstract Constructors

An introduction to the book. Status: September 2026.

1.1 Outcome

The book states memes, adaptation and software as tasks and constructors. Its main results in short:

The main finding: two levels of copying.

  • The closed level is the channel: sounds, words, writing, a fixed repertoire of behaviours. It is a closed alphabet that is fixed in the receiver at the time of copying, whether innate, learned in early childhood or invented; the fixed alphabet is the central property of this level. The receiver parses what it observes into the alphabet, with errors. Because the alphabet is fixed, errors below the distance between two symbols are absorbed, and the copy becomes arbitrarily accurate. Byrne’s behaviour parsing (Byrne 1999; Byrne & Russon 1998) is the mechanism we build on, and the digital alphabet that makes the copy arbitrarily accurate is our reading of it. We conjecture that the symbols of the alphabet are attractors in Sperber’s (1996) sense: every rendition is pulled to one of them. Its limits are that it is expensive to pass on and can only express what the alphabet composes (Deutsch 2011, ch. 16). The fixed alphabet and redundancy give this level high fidelity by another route than Marletto’s separation of recipe and vehicle, short of it for an error that every rendition shares and, at finite redundancy, for a small error that enters the string each generation: the string is mostly read back from a performance, and a recipe is copied from a recipe only where a text is copied from the text (post 1). The phoneme and the word are examples: they are fixed in the child by the time it hears the word, and a heard word is copied digitally.
  • The open level is the function. The receiver rebuilds it from the evidence in the open mode, from reasons and with its own creativity. Two heads share the signal alphabet but not the concepts: each builds its own. An utterance or a text is a recipe-like signal that the receiver reads as evidence and does not execute. No exact copy of the function can be guaranteed or confirmed (post 1).
  • An ape meme is the case where a string on the closed level fixes the whole repertoire. A culture meme is carried by a closed channel, but its function lives on the open level. The two are not a partition of memes. They are two levels of one copy task, and the word dog has both: the phonemes are copied digitally, what the word means is rebuilt by each child.
  • So the answer to Sperber is not that all copying is approximately accurate. Copying on the closed level is digital and arbitrarily accurate. Copying the function on the open level is reconstruction and never confirmed exact.

Definitions.

  • A meme is an abstract constructor: a repertoire \(T_m\) of tasks, carried by every brain state that can perform them all (post 1). A program is the same thing on a programmable constructor: the class of all programs with one behaviour (post I.1).
  • Copying is a task: an original is rebuilt from finitely many of its answers. It is synthesis from examples with a weaker specification than the one to be met. Its fidelity is the share of the original’s tasks that the copy keeps (post I.3).
  • Design is defined without a process. A piece of knowledge is designed to the degree that its one-step variants perform the task much worse (post I.5).
  • An adaptive process has two conditions: variation that does not read the task, and selection that does. Competition is not one of them (post I.5).
  • A universal explainer reaches every computable behaviour by its own steps, a universal computer only by programming from outside. This separates a person from an idle laptop, but not from a laptop running a program enumerator (post I.2). What does separate them is the problem of creativity, the central question of the book (post I.6, Section 7).

Findings.

  • Over an open language, verification is undecidable, and only behaviour witnessed by finitely many cases can be confirmed. So one has to run things, and a host whose programs form an acceptable language has no full decoder (posts I.3 and 2).
  • Copying on the closed level is arbitrarily accurate, because the receiver parses the signal into a fixed digital alphabet. The function has no alphabet of concepts to be parsed into: each head builds its own, so its copy is never confirmed exact (post 1).
  • No exact copy of an open idea can be guaranteed or confirmed. The evidence rejects reconstructions but does not produce them, so copying needs the receiver’s creativity (post 1).
  • Under an internal anchor, content is blind to reality. Under an exact external test, isolated populations converge on the same hidden points. Under internal anchors they agree on levers, never on the hidden points (posts 4 and 5).
  • Fidelity and growth are separate axes: an anti-rational population copies as faithfully as a rational one and does not grow (post 4).
  • Design needs variation and selection, not competition. In the experiments, pools from which nothing was removed found and kept the design when parents were drawn by the task (posts I.5, 7 and III.6).
  • A criterion that is itself varied keeps content only with an external anchor. Without one it becomes unmoored and the content is lost (post I.6).
  • Drift as the null hypothesis gives a measure of design, the constraint \(1 - k/\mu\). On real proteins it separates designed from weakly designed parts of one gene: insulin’s A and B chains change at a third of the rate of its C-peptide (post III.2).

Conjectures. Adaptationism is the best theory of design (post I.5). Creativity is persistence plus anchored criticism (post I.6). The brain runs an adaptive process with an externally anchored criterion (posts 7 and III.6).

What this book claims. I consider these the obvious steps, the low-hanging fruit, in applying constructor theory to memetics, adaptation and software. The book continues the abstract theory of replicators, which Dawkins (1976) began and Deutsch (2011, chs. 15 and 16) applied to memes, with constructor theory. That this step would be taken was clear, and Deutsch would have taken it himself. I consider the book a loose strand tied up, not a new theory. That even these surface steps reach so far, from memes to proteins to software, shows how much of constructor theory is still there to be worked through. Everything that goes beyond the obvious is a conjecture for now, and the posts mark it as one.

1.2 What this series is

This series asks two questions about ideas that spread from mind to mind. What is such an idea physically? And which ways of passing it on are possible at all? It answers both in the language of constructor theory, a way of stating physics that David Deutsch and Chiara Marletto have developed since 2012.

The series is for readers who are curious about cultural evolution, the philosophy of Karl Popper, or the physics of information. You do not need to know any of them in advance. This page introduces the three traditions the series draws on: memetics, Popper’s and Deutsch’s theory of knowledge, and constructor theory. Then it says why they belong together and in which order to read the posts.

The four sources of the book
the question
Memetics
  • Dawkins 1976, Blackmore 1999
  • ideas as replicators, copied brain to brain
how knowledge grows
Popper
  • conjecture and refutation
  • the searchlight, not the bucket
explanations and memes
Deutsch
  • good explanations are hard to vary
  • creativity was selected to copy memes
the language
Constructor theory
  • tasks, and which are possible
  • abstract constructors
Memetics asks the question; constructor theory gives the language to state it; Popper and Deutsch say who performs the copy.

The posts themselves are more technical. They contain definitions, a few proofs, and small simulations written in Clojure.

1.3 The parts

The book has parts I, II, III, V and VI. Part V is planned and has no posts yet. Part II, “Memes as Abstract Constructors”, is the series this page mainly introduces: posts 0 to 7 on what a meme is and how it is copied.

Part I, “A Constructor Theory of Computing and Design”, is the general half. It asks what computing is in constructor-theoretic terms, abstract enough to cover computers, brains, language models and biological evolution alike. It defines a computing system, two conditions that a person meets, synthesis (the task whose product is a program), design, without reference to any process, the adaptive process, of which evolution and the growth of knowledge are both cases, and fallibilism, a process of criticism that never makes anything final. Its first four posts run on Push3, a stack language by Lee Spector in which programs are data, so a program that builds programs is an ordinary program. The experiments on design and on criticism use a simpler bit-string model, in which the quantities they need can be computed exactly. Its main source besides constructor theory is Gregory Chaitin’s metabiology, which treats organisms as programs and measures mutations by their description length rather than by how many symbols they change. Part I reproduces part of Chaitin’s model on Push3 and adds what it lacks: populations, a criterion that can itself vary, and a designer that has to pay for its choices. The Part I posts give the references for Push3 and for Chaitin’s papers.

Part III, “A Constructor Theory of Adaptation”, takes design from Part I and asks how evolutionary biology produces it. It shows that the classic critiques of adaptationism explain non-designs, not a rival way to make design, and treats drift as the null hypothesis: how far a site falls below the rate at which neutral mutations are fixed measures how designed it is. It tests this measure on real protein data, insulin, haemoglobin and histone H4, and asks how strong selection must be, relative to the effective population size \(N_e\), for a population to hold a design. It then asks which programs can carry design at all, in a small RNA folding model, what sex and symbiogenesis change about variation, and what a brain needs in order to produce good ideas.

Part V, on gene-meme coevolution and evolutionary psychology, is planned. Part VI, “Concepts, Memory and Experience”, is speculative and marked as such on every page.

The parts divide the work. Part I owns the general definitions, including design and the adaptive process. Part II applies them to ideas in brains, and Part III to living things.

The map of the book
Part VI: Concepts, Memory and Experience
the stories go with it
speculative
Part V: gene-meme coevolution
no posts yet
planned
Part III: A Constructor Theory of Adaptation
design in living things
builds on I and II
Part II: Memes as Abstract Constructors
ideas in brains
builds on I
Part I: A Constructor Theory of Computing and Design
the general definitions
the foundation
Parts I to III, bottom up, in their order of dependence. Nothing in them depends on Part V or VI.

Every definition is stated in one place only, and the other parts link to it. The order of dependence is Part I, then Part II, then Part III. Part II builds on Part I: where a Part II post uses a notion such as self-reachability, synthesis, design or the adaptive process, it takes the definition and result as given and links to the post that proves it. Part III builds on Parts I and II. A “See also” box may point forward to a later part, but nothing depends on it. Read Part I first, or start with Part II and read posts I.1 to I.3 before or next to post 1, and post I.5 before post 7.

1.4 Memetics

The word meme comes from the last chapter of Richard Dawkins’s The Selfish Gene (Dawkins 1976, ch. 11, “Memes: the new replicators”). A meme is a unit of cultural transmission: a replicator, something that gets copied and whose copies get copied in turn, like a gene, but passed from brain to brain. Susan Blackmore took the coevolution of genes and memes seriously and called the human a meme machine (Blackmore 1999). David Deutsch treated memes better, in two chapters of The Beginning of Infinity (Deutsch 2011, chs. 15–16). His account rests on Popper’s theory of knowledge. It carries none of the baggage of cultural-evolution models, nor any commitment about how a meme is implemented in a brain.

This book builds on Deutsch. We do not retell the history of memetics here. The section on Deutsch below says what we take from him and where we differ.

Memetics has met three objections, and the series answers each. The second is Dan Sperber’s (Sperber 1996).

Three objections to memetics
objection
Ontology
A meme might be a pattern of neurons, a behaviour, a text or an idea. Memetics has no concept that says what these have in common.
objection
Fidelity
Ideas are not copied but reconstructed. What stays stable is explained by attractors that pull reconstructions, not by copying, so there is no replicator to select.
objection
Dynamics instead of possibility
Memetic models describe how often an idea occurs over time, not which transmissions are possible at all.

1.5 Popper: how knowledge grows

Karl Popper’s theory of knowledge starts from a negative claim. Knowledge does not come from induction, that is, from generalising observations until a law emerges. No number of white swans proves that all swans are white, and in any case observations do not arrive with their interpretation attached. Knowledge grows instead by conjecture and refutation (Popper 1963). We guess, and then we try to find out what is wrong with the guess.

Popper contrasted two pictures of the mind (Popper 1972, appendix 1, “The Bucket and the Searchlight: Two Theories of Knowledge”): the bucket theory and the searchlight theory. He rejected the bucket and defended the searchlight.

Two pictures of the mind
rejected
The bucket
  • the mind is a container
  • experience pours in through the senses
  • knowledge is what collects at the bottom
defended
The searchlight
  • the mind is active
  • expectations and theories come first
  • they decide what it looks at and what counts as an observation
Popper (1972, appendix 1).

This will matter for memes: if the bucket theory is false, an idea cannot simply be poured from one head into another. The receiver has to rebuild it, and what it is shown can only criticise its guesses.

A third idea of Popper’s is World 3, the world of the objective contents of thought: theories, problems and arguments, next to the physical World 1 and the mental World 2 (Popper 1972, ch. 3, first a 1967 lecture, “Epistemology without a knowing subject”). Its point for this book is simple. A theory that is written down is an object that others can criticise, instead of only a disposition of whoever holds it. That is why science works: our theories can die in our stead, as Popper put it. The book takes this as a conjecture, that criticism needs a representation (post I.6, Conjecture I.6.15).

Popper also called his theory evolutionary. He summarised the growth of knowledge in a four-step schema, often called the tetrad (Popper 1972, ch. 7):

\[P_1 \to TT \to EE \to P_2\]

Popper's tetrad
P₁
a problem
leads to→
TT
a tentative theory
subjected to→
EE
error elimination, by criticism or by testing
survivors raise→
P₂
usually a new problem
Popper (1972, ch. 7).

Popper applied the same schema to animals as well as to scientists. The difference is where the error elimination happens. An organism whose built-in expectation is wrong may die of it. A person can criticise the expectation and discard it instead. Popper put this as letting our theories die in our stead (the wording varies across his writings, so we paraphrase it here).

1.6 Deutsch: explanations and memes

We take the following from The Beginning of Infinity (Deutsch 2011) as given. A good explanation, an account of what is there and how it works, is hard to vary: almost every change to its details would make it stop explaining what it explains (ch. 1). People are universal explainers: anything that can be understood at all can in principle be understood by a person (chs. 3 and 6). Rational memes are created by criticism and depend on it to be copied faithfully. Anti-rational memes, in Deutsch’s words, “survive by disabling their holders’ critical faculties” and dominate static societies (ch. 15). And creativity was first selected to copy memes, not to invent them (ch. 16). Apes copy as well, without creativity: Byrne (1999) calls this behaviour parsing: the observer picks out recurring patterns in the stream of behaviour and copies its organisation (Byrne & Russon 1998; Deutsch 2011, ch. 16 discusses it). We read it as parsing into elements of a fixed repertoire and copying the string of elements. Such memes are expensive to pass on and can only express what the repertoire composes. Human culture memes are not limited that way, and the reason is that an idea inside another brain cannot be observed.

Why copying a culture meme needs creativity
idea in A's brain
cannot be observed
shows as→
behaviour
all the learner sees
learner guesses→
an idea behind it
corrects the guess→
a copy
rebuilt, not received

So copying has two levels, and the answer to Sperber’s objection that the series develops is different for each. The channel is a closed alphabet (sounds, words, writing) that the receiver has fixed by the time of copying, and copying on it is digital. The function is rebuilt on the open level. It is not that all copying is approximately accurate.

Two levels of copying
sounds, words, writing; the whole of an ape meme
Closed level: the channel
  • a closed alphabet, fixed in the receiver at the time of copying (innate, learned early or invented)
  • the receiver parses what it observes into the alphabet, with errors
  • the alphabet is digital, so errors below symbol distance are absorbed and the copy is arbitrarily accurate
  • fidelity from the fixed alphabet and redundancy, not from a separate recipe
what makes a culture meme
Open level: the function
  • the function is not a string: two heads share the signal alphabet but not the concepts
  • the receiver rebuilds it from the evidence, from reasons, with its own creativity
  • the utterance or text is a recipe-like signal that is read as evidence, not executed
  • no copy can be guaranteed or confirmed exact
Copying on the closed level is digital and arbitrarily accurate. Copying the function on the open level is reconstruction, and never confirmed exact. Sperber's attractors explain fidelity on both levels.

Faithful copying of a culture meme therefore requires creativity rather than being threatened by it.

Where we differ from Deutsch:

  • Copying on a closed channel is digital, because its alphabet is fixed. Deutsch gives the limits, cost and expressiveness of ape memes. We add that the fixed alphabet makes closed-level copying arbitrarily accurate, that this level carries every meme (the words of a culture meme as well), and that its symbols are Sperber’s attractors on this level. The fixed alphabet, not a separation of recipe and vehicle, is the source of its fidelity. The function of a culture meme is rebuilt from reasons and is not a recipe (post 1).
  • The meme gets a constructor-theoretic definition. Deutsch treats memes in prose. We define a meme as an abstract constructor: a class of brain states, fixed by the tasks a brain in such a state can perform (post 1).
  • The copy error is made precise. We split it into what the copy loses and what it gains. The gain holds errors and improvements alike, and only a criterion tells which (post 1).
  • Anti-rational memes also use creativity. If copying needs the receiver’s creativity, “disabling critical faculties” cannot be literally right. We propose that both kinds of meme use creativity and correct errors, and that they differ in the anchor of the criterion that decides which variants are kept: reality outside the hosts, or the hosts themselves (post 4).
  • Memes cannot be read off brains. A decoder that reads a meme off a brain state would be a verifier of the brain’s own programs, and for a brain that can take up any program from what it is told no such verifier exists. The general way to find out what someone thinks is to ask them. That asking is also the most efficient way is a conjecture (post 2).

1.7 Constructor theory

Constructor theory has been introduced many times: in Deutsch (2013), Deutsch & Marletto (2015), Marletto (2015), in §§2–3 of Deutsch & Marletto (2025), and without formalism in Marletto (2021).

Its central move is to take abstract machines seriously. A constructor is anything that causes a transformation and keeps being able to cause it again, and the theory asks only which transformations such machines can and cannot perform. Taking these machines as the basic entities makes a handful of concepts precise. The ribosome example is Deutsch’s (2013, §2.1), and the reading of software is argued in post I.1.

What constructor theory makes precise
Task
what a task is, and what makes it possible or impossible
Information
what can be copied and permuted, whatever it is made of
Programmability
a constructor whose task is set by a program it is given, as a ribosome's product is set by the RNA it reads
Abstract constructor
a constructor that is itself information, and does its work whatever it is instantiated in
the book's reading
Software
an abstract constructor on a programmable constructor, down a stack of interpreters to a circuit

These are the things we need to know about for computing, for biology, for psychology and for memetics. We do not repeat the theory’s introduction here. This section only fixes the names and the notation the book uses.

A law of constructor theory says which transformations are possible, which are impossible, and why. A substrate \(\mathcal{P}\) is a physical system that gets transformed, and an attribute is a set of its states. A task takes an input attribute \(x\) to an output attribute \(y\), written as in Deutsch & Marletto (2025, eq. 1):

\[x \to y \ \text{on}\ \mathcal{P}\]

In general a task is a set of such pairs: for each legitimate input, the outputs it allows (Deutsch 2013, §1.1). A constructor is something that performs a task and keeps the ability to do it again.

A task and its constructor
input attribute x
a set of states of the substrate 𝒫
constructor C→
output attribute y
on the same substrate
→
C unchanged
able to do it again

No real device does either perfectly, and the two ways it falls short have names, accuracy and reliability (Deutsch & Marletto 2025, §2, following Violaris & Marletto 2022).

Two ways a real device falls short
run once
Accuracy
“the distance between the desired output attribute of the task, and the actual output of the system being used to perform it, when run once”
run again and again
Reliability
“the degree to which such a system retains its ability to deliver the task to that accuracy, when re-used repeatedly”
Deutsch & Marletto (2025, §2).

Which distance to use is left to the particular theory. Real devices are approximate constructors. A constructor proper is their idealisation to perfect accuracy and reliability, a “fictitious ideal device” (ibid.).

A task \(X\) is possible, written \(X^{\checkmark}\), if there is no bound short of perfection on the accuracy and the reliability with which approximate constructors can perform it. Otherwise it is impossible, written \(X^{\text{✘}}\) (Deutsch & Marletto 2025, §3, eqs. 2–3). Write \(\alpha(K, X)\) for the inaccuracy and \(\rho(K, X)\) for the unreliability of an approximate constructor \(K\) on \(X\). Then, in our paraphrase of the verbal definition:

\[\begin{aligned} X^{\checkmark} &\iff \forall \varepsilon > 0\ \exists K_\varepsilon:\ \alpha(K_\varepsilon, X) \le \varepsilon \ \text{and}\ \rho(K_\varepsilon, X) \le \varepsilon \\ X^{\text{✘}} &\iff \neg\,\big(X^{\checkmark}\big) \end{aligned}\]

The \(\neg\) negates the statement “\(X\) is possible”, not the task.

Possible and impossible
X✓ possible
for every ε > 0 some approximate constructor does X with inaccuracy and unreliability at most ε
X✘ impossible
there is a bound, short of perfection, on how well any constructor can do X
There is no third option.

“Possible” does not mean “can be done perfectly”, and it does not mean “happens”. This carries a lot of weight in the book, because it builds room for copy errors into the notion of possibility.

Tasks combine. Tasks are built from tasks, and three operations are all the book needs (Deutsch 2013, §§1.1–1.2): the transpose \(X^{\sim}\); the serial composition \(AB\), where every output of \(B\) is an input of \(A\); and the parallel composition \(A \otimes B\), whose two substrates together form the composite system \(\mathcal{P}_1 \oplus \mathcal{P}_2\).

The task algebra
operation
written
what it does
example
transpose
X~
swaps inputs and outputs
{p → q}~ = {q → p}
serial composition
AB
B, then A, on the same substrate
a run of a program: its steps in series
parallel composition
A ⊗ B
A on 𝒫₁ and B on 𝒫₂
copying a meme, on evidence ⊕ brain

The copy task of a meme is a task on evidence and brain together, on \(E \oplus B\) (post 1). A possible task can have an impossible transpose (Deutsch 2013). That is how the theory states irreversibility without a time parameter.

A possible task with an impossible transpose
possible
work → heat
converting work entirely into heat
impossible, by the second law
heat → work
turning heat entirely into work

The first law Deutsch proposes is the composition principle (Deutsch 2013, §1.2).

The composition principle, and its converse
a law
Possible parts → possible whole
every network of possible tasks, the outputs of each fitting the inputs of the next, is possible
does not hold
Possible whole → possible parts
two impossible tasks that break a conservation law by equal and opposite amounts make a possible one

The book uses a few further notions from these papers, restated with section references in post 1. An abstract constructor is information that performs tasks whatever it is instantiated in, like the ideas that keep the Ship of Theseus in repair; knowledge is information that tends to keep itself instantiated (Deutsch 2013, §2.14–2.15). An information variable is a set of attributes that can be copied and permuted (Deutsch & Marletto 2015). Accurate self-reproduction under laws without design needs a digital recipe that is copied and a vehicle rebuilt from it each generation (Marletto 2015).

Recipe and vehicle
recipe
digital; copied
is copied, and read to build→
vehicle
rebuilt from the recipe
carries the copied recipe into→
the next generation
↺each generation
Marletto (2015). In an organism the genome is the recipe. For memes we conjecture that it applies to the closed level of copying, the channel of every meme.

In an organism the genome is the recipe and the rest is the vehicle. A timer is a constructor that says whether a given period has elapsed, which lets a theory without a time parameter speak of duration (Deutsch & Marletto 2025).

1.8 Why put these together

Memetics has a good question and no framework to answer it with. It lacks an account of what a meme is, and it has no theory of which copies are possible. Constructor theory supplies both. It lets us define a meme as an abstract constructor, identified by what a brain in a certain state can do rather than by the neurons that encode it. And its notion of “possible” already includes imperfect copying, so Sperber’s point that no copy is exact stops being fatal: copies on the closed level are arbitrarily accurate, and the copy of the function on the open level is never confirmed exact. What constructor theory does not say is who performs the copy task for a meme. Popper and Deutsch fill that gap: the copier is the receiver’s own creativity, guessing the idea behind the evidence and correcting the guess.

Conjecture (the series’ thesis, stated informally).

conjecture
The thesis
  • a meme is an abstract constructor
  • closed level (the channel, a fixed alphabet): copies are arbitrarily accurate, because the alphabet is fixed
  • open level (the function of a culture meme): the receiver's creativity is the constructor that copies it
  • a copy of the function can be exact, but never guaranteed or confirmed to be, so it is never final

A meme is an abstract constructor in Marletto’s sense, and, in the open mode, the constructor that copies it is the creativity of the receiver. A copy can be exact, but it can never be guaranteed or confirmed to be, so no copy is ever treated as final: it stays open to criticism even when it is exact. This holds for the open level, the function of a culture meme. On the closed level (the words and writing that carry it, and the whole of an ape meme, a recitation or a checklist) a fixed digital alphabet makes copies arbitrarily accurate. The programme post places the project among existing work and lists the individual theses with the post that treats each.

1.9 Reading guide

Part I: A Constructor Theory of Computing and Design.

Part II: Memes as Abstract Constructors.

Part III: A Constructor Theory of Adaptation.

Part VI (speculative): Concepts, Memory and Experience.

  • Post S1: History, not wiring: a speculative sketch of experience. Speculative. The 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, with the picture of concepts, memory and the self behind it. It is followed by five short stories that explore the same ideas in fiction. Nothing in Parts I to III depends on it.

Lexicon.

  • Lexicon. Every short code of the series (F1, E10, D2, …) with its name, a paragraph on what it is and how it turned out, and links to where it is introduced and used.

Glossary.

Every post is an executable notebook. The prose sits next to the Clojure code that computes its examples and simulations, and every random run names its seed. Several claims we started from did not survive being made precise. Where that happened, the idea was replaced by what survived, and the post says in one sentence, at the place where the old claim would be tempting, why it does not hold.

1.10 Further reading

  • Marletto (2021), The Science of Can and Can’t. The gentlest entry point to constructor theory.
  • Deutsch (2011), The Beginning of Infinity. Chapter 1 for good explanations, chapters 15 and 16 for memes and creativity.
  • Dawkins (1976), chapter 11. Short, and the origin of the whole debate.
  • Sperber (1996). The strongest criticism of meme-style copying, and an alternative account of cultural transmission.
  • Popper (1963), the introduction and chapter 1, for conjecture and refutation. Popper (1972), chapter 7 and appendix 1, for the evolutionary view and the bucket and the searchlight.
  • Deutsch (2013). The founding paper of constructor theory. §1 and §2.14–2.15 are the parts this series uses most.
  • Marletto (2015). The paper whose structure this series tries to carry over from biology to culture.

1.11 References

  • Blackmore, S. (1999). The Meme Machine. Oxford University Press. (verified)
  • Byrne, R. W. (1999). Imitation without intentionality. Using string parsing to copy the organization of behaviour. Animal Cognition 2(2), 63–72. doi:10.1007/s100710050025. (verified)
  • Byrne, R. W. & Russon, A. E. (1998). Learning by imitation: a hierarchical approach. Behavioral and Brain Sciences 21(5), 667–721. (verified)
  • Dawkins, R. (1976). The Selfish Gene. Oxford University Press. Ch. 11, “Memes: the new replicators”. (verified)
  • Deutsch, D. (2011). The Beginning of Infinity: Explanations that Transform the World. Allen Lane. Chs. 1, 3, 6, 15 (“The Evolution of Culture”), 16 (“The Evolution of Creativity”). (verified; the ch. 15 definitions and the ch. 16 argument checked against the text)
  • Deutsch, D. (2013). Constructor theory. Synthese 190(18), 4331–4359. doi:10.1007/s11229-013-0279-z. arXiv:1210.7439. §1.1, §1.2, §2.14, §2.15. (verified; quotations checked against the arXiv full text, §§1.1–1.2 and the transpose of work into heat re-checked on 2026-09-30)
  • Deutsch, D. & Marletto, C. (2015). Constructor theory of information. Proc. R. Soc. A 471(2174): 20140540. doi:10.1098/rspa.2014.0540. arXiv:1405.5563. (verified)
  • Deutsch, D. & Marletto, C. (2025). Constructor theory of time. arXiv:2505.08692 (v3, June 2026). (verified)
  • Marletto, C. (2015). Constructor theory of life. J. R. Soc. Interface 12(104): 20141226. doi:10.1098/rsif.2014.1226. arXiv:1407.0681. (verified)
  • Marletto, C. (2021). The Science of Can and Can’t: A Physicist’s Journey Through the Land of Counterfactuals. Allen Lane (UK), Viking (US). (verified)
  • Violaris, M. & Marletto, C. (2022). Irreversibility beyond Landauer’s principle: is it possible to perform erasure using the quantum homogenizer? New J. Phys. 24: 113030. (cited via Deutsch & Marletto 2025, §2; not checked separately)
  • Popper, K. R. (1963). Conjectures and Refutations: The Growth of Scientific Knowledge. Routledge & Kegan Paul. (verified)
  • Popper, K. R. (1972). Objective Knowledge: An Evolutionary Approach. Clarendon Press. Ch. 3 (“Epistemology without a knowing subject”, first 1967), ch. 7 (“Evolution and the Tree of Knowledge”), appendix 1 (“The Bucket and the Searchlight: Two Theories of Knowledge”). (verified: publisher, chapter and appendix titles, and the tetrad; the “die in our stead” line is paraphrased because its exact wording could not be checked against the book)
  • Sperber, D. (1996). Explaining Culture: A Naturalistic Approach. Blackwell. (verified)
source: book2/notebooks/index.clj