Memetics as evolutionary science: a mechanism-first research briefing

The core claim of memetics is that Darwinian natural selection operates on cultural information independently of — and sometimes in opposition to — biological genes. This briefing synthesizes primary-source research on the theory’s intellectual foundations, the meme concept at every level of analysis, the Darwinian fitness properties that drive memetic evolution, the unit problem that haunts the discipline, and the memeplex structures that explain why belief systems resist rational persuasion. It is organized to support writing approximately 5,000 words of academic primer text.

---

1. Orientation: what memetics is, what gap it fills, and why it provokes

The discipline’s intellectual identity

Memetics is the application of Darwinian evolutionary theory to cultural information — ideas, behaviors, artifacts, skills, and symbolic structures that propagate between minds through imitation and communication. Its central premise is that culture contains entities (memes) that function as replicators in the technical sense Dawkins defined: self-copying informational units subject to variation, differential selection, and heredity. This makes cultural evolution not merely analogous to biological evolution but a genuine second instantiation of the same abstract algorithm.

The theoretical gap memetics fills sits at the intersection of three fields that, before 1976, lacked a unifying framework. Evolutionary biology could explain the human capacity for culture (large brains, social learning) but not the specific trajectory of cultural change — why celibacy spreads in religious orders, why fashions cycle, why scientific paradigms resist displacement then collapse suddenly. Cognitive science could explain individual learning and memory but lacked a population-level theory of how mental contents propagate and compete across minds. And traditional social science studied cultural transmission descriptively (diffusion of innovations, sociology of knowledge) without a selectionist mechanism that could explain why some cultural variants persist while others vanish.

Why the evolutionary lens was both necessary and controversial

The necessity argument rests on what Dennett called the “universal acid” property of Darwinian logic. If culture exhibits variation (ideas differ), differential fitness (some ideas spread more than others), and heredity (ideas are transmitted from person to person), then culture satisfies Lewontin’s three conditions for evolution by natural selection. Denying that Darwinian logic applies requires showing that one of these three conditions fails — and empirically, all three clearly hold. The printing press, the internet, religious conversion, language change, and scientific revolutions all involve information varying, competing, and being inherited.

The controversy derives from three sources. First, autonomy: memetics claims memes are replicators in their own right, not merely tools of genetic fitness. This breaks with sociobiology and evolutionary psychology, which treat culture as ultimately serving genetic interests. Second, agency: the “selfish meme” hypothesis implies that our beliefs may serve the meme’s replicative interests rather than our own welfare — a deeply unsettling claim that challenges commonsense notions of rational agency and free will. Third, scientificity: critics (Sperber, Boyd & Richerson, Edmonds) argue that memetics lacks the empirical precision of competing frameworks, that memes are too vaguely defined to count or measure, and that the gene-meme analogy is taken too literally.

Positioning relative to neighbors

Memetics is best understood as one branch within the broader field of cultural evolutionary theory, which also includes Boyd and Richerson’s dual inheritance theory (DIT) and Sperber’s cultural attraction theory (CAT). DIT uses population-genetics mathematics to model cultural transmission through biased social learning — content bias, prestige bias, conformist bias — without requiring discrete replicating units. It is considered the academic mainstream. CAT argues that cultural stability arises not from high-fidelity replication but from cognitive attractors: points in representational space toward which mental reconstructions converge because of shared cognitive architecture. Memetics occupies the most radical position: it insists on treating cultural information as a genuine second replicator with its own evolutionary dynamics, potentially in conflict with genes.

---

2. Intellectual genealogy: the logic of each step

Gabriel Tarde (1890) — the idea’s-eye view before Darwin was applied to culture

Tarde’s Les Lois de l’Imitation proposed that the fundamental unit of social life is not the individual or the group but imitation — the process by which ideas, desires, and practices propagate person-to-person. He identified three processes: invention (generation of novel variants by roughly 1% of the population), imitation (geometric propagation through populations following S-curve dynamics), and opposition (conflict when mutually exclusive imitative waves collide). His critical structural contribution was adopting the perspective of the propagating idea itself rather than the human carrier. As Marion Blute observed, Tarde was the “Mendel of memetics” — a precursor whose contribution went unrecognized until the framework he anticipated was independently reinvented. He even noted that imitation is never perfectly mechanical, producing variation that anticipates memetic mutation.

Frederic Bartlett (1932) — the empirical foundation for memetic mutation

Bartlett’s serial reproduction experiments (most famously using the Native American folk tale “The War of the Ghosts”) demonstrated that cultural information mutates systematically during transmission. Stories became shorter, culturally unfamiliar elements were rationalized into familiar forms (conventionalization), unusual details were smoothed out (leveling), and certain features were exaggerated (sharpening). His core finding: memory is reconstructive, not reproductive. People do not retrieve stored traces but actively rebuild information using pre-existing schemas. This has a dual significance for memetics. It provides the empirical basis for memetic variation (each retelling introduces mutations). But it simultaneously challenges the fidelity requirement — if cultural copying is this error-prone, can memes sustain the lineages that cumulative evolution requires? This tension between Bartlett’s findings and the fidelity threshold for evolution remains central to the debate between memetics and Sperber’s attraction model.

Donald Campbell (1960) — the universal selection framework

Campbell’s “blind variation and selective retention” (BVSR) framework, developed across papers from 1960 to 1974, established that all knowledge-gaining processes — genetic evolution, individual learning, scientific discovery, cultural change — operate through the same algorithm: unjustified generation of variants, selection by environmental criteria, and retention of selected variants. He constructed a hierarchical model of increasingly sophisticated “vicarious selectors” from genetic mutation up through cultural institutions. His key structural contribution was establishing that Darwinian selection is substrate-neutral — not confined to biology. This was the formal backbone that made Dawkins’s meme concept conceptually possible: if BVSR operates on culture, something must be varied, selected, and retained. Campbell himself, however, was skeptical of discrete cultural replicators, emphasizing that learned behavior must be independently re-learned, not passively copied.

Claude Shannon (1948) — information as discrete, transmittable, substrate-independent

Shannon’s mathematical theory of communication decomposed information into discrete, measurable units (bits), showed that messages can be transmitted reliably through noisy channels via appropriate encoding, and — crucially — declared that the semantic aspects of communication are irrelevant to the engineering problem. This provided the conceptual infrastructure for treating ideas as discrete, quantifiable, substrate-independent units. Before Shannon, “information” was vague; after Shannon, it could be mathematized. The implications for memetics: (a) cultural content can be decomposed into discrete propagating units; (b) transmission involves noise (mutation); (c) channels shape what can be transmitted; (d) the same information can exist in radically different physical media. This last point is essential for Universal Darwinism.

Karl Popper (1963–1972) — knowledge as evolutionary and autonomous

Popper’s evolutionary epistemology framed scientific progress as structurally identical to biological evolution: bold conjectures (analogous to mutations) are subjected to refutation (analogous to natural selection). His formula P1→TT→EE→P2 (Problem→Tentative Theory→Error Elimination→New Problem) describes a non-teleological generate-and-test cycle. His World 3 concept — the realm of objective thought contents (theories, arguments, problems) that exist independently of any knowing subject and interact causally with the physical world — directly prefigures the idea that memes have autonomous existence. If knowledge lives in an objective realm and undergoes selection, treating ideas as evolutionary entities becomes philosophically coherent.

Marshall McLuhan (1964) — transmission channels as selection environments

McLuhan’s “the medium is the message” established that information vehicles shape the information they carry. Media are not neutral pipes but active shapers of cultural selection. Each new medium (print, broadcast, internet) creates a radically new selection environment for memes: a tune meme propagates differently via oral tradition than via recording; a religious meme propagates differently via manuscript than via printing press. McLuhan anticipated the memetic insight that the ecology of available transmission channels determines which memes can survive and how they mutate.

Richard Dawkins (1976) — naming the replicator and the Universal Darwinism argument

In Chapter 11 of The Selfish Gene (“Memes: The New Replicators”), Dawkins synthesized all preceding threads. After ten chapters arguing that genes are selfish replicators using organisms as survival machines, he deliberately broadened: “I am an enthusiastic Darwinian, but I think Darwinism is too big a theory to be confined to the narrow context of the gene.” His logic: the fundamental requirement for Darwinian evolution is not DNA but replication with variation and selection. “What, after all, is so special about genes? The answer is that they are replicators.” If another entity could meet those criteria, it too would evolve by natural selection, regardless of substrate.

He coined the term meme (abbreviated from the Greek mimeme, to sound like “gene”) for “a unit of cultural transmission, or a unit of imitation.” Examples: tunes, catch-phrases, fashions, ways of making pots or building arches. “Just as genes propagate themselves in the gene pool by leaping from body to body via sperms or eggs, so memes propagate themselves in the meme pool by leaping from brain to brain via a process which, in the broad sense, can be called imitation.” He identified the three Darwinian fitness properties — fidelity, fecundity, longevity — and noted that meme mutation is continuous and quasi-Lamarckian (each retelling introduces directed changes), unlike the discrete, random mutation of DNA.

Dawkins also introduced the concept of coadapted meme complexes (organized religion was his example) and ended with a deliberately paradoxical claim: “We, alone on earth, can rebel against the tyranny of the selfish replicators.” This framing — that we are built as gene machines and cultured as meme machines but can rebel — set up the tension that subsequent theorists would develop.

Douglas Hofstadter (1979–1985) — self-reference and the coining of “memetics”

Hofstadter’s Gödel, Escher, Bach explored strange loops — self-referential systems that can represent and modify themselves. This provided the conceptual vocabulary for understanding how ideas can encode instructions for their own replication. His 1983 Scientific American column explicitly explored “virus-like sentences and self-replicating structures” — chain letters, self-referential commands, informational parasites. His Metamagical Themas collection (1985) is credited with coining the term “memetics” for the study of memes. Structural importance: he showed that self-reference is a fundamental property of complex information systems, crucial for understanding how memes can be “selfish” — a meme encoding instructions for its own propagation is a strange loop in the cultural domain.

Daniel Dennett (1991, 1995) — philosophical legitimation and the meme-consciousness link

Dennett made two landmark contributions. In Darwin’s Dangerous Idea (1995), he argued that natural selection is a universal acid — an algorithmic process so powerful it dissolves all prior explanatory frameworks. His metaphor: cranes versus skyhooks. Skyhooks are miraculous forces invoked to explain complexity (divine design, irreducible consciousness); cranes are natural mechanisms that amplify the design process. Memes are cranes — they provide a non-miraculous explanation for cultural complexity. In Consciousness Explained (1991), he made the more radical claim: human consciousness is substantially constituted by memes. The mind is not a Cartesian Theater but a “multiple drafts” process where memic contents compete for dominance. This means “it cannot be ‘memes versus us,’ because earlier infestations of memes have already played a major role in determining who or what we are.” This philosophically legitimized memetics and connected it to consciousness studies.

Susan Blackmore (1999) — radical meme theory and the selfplex

Blackmore radicalized Dawkins’s meme concept in three ways. First, her memetic drive hypothesis: once early hominids began imitating, the most successful memes created selection pressure for brains better at imitation, creating a gene-meme coevolutionary spiral that explains the anomalously large human brain — it evolved to serve memes, not just genes. Second, she proposed that language evolved primarily to increase memetic fidelity and fecundity, not for direct genetic advantage. Third, and most radical, the selfplex: the human self is itself a memeplex, a vast complex of mutually reinforcing memes. There is no inner “I” controlling the machine — the self is what the memes construct. She also introduced the important distinction between copy-the-product (low fidelity, like recreating a soup from taste) and copy-the-instructions (high fidelity, like copying a recipe), paralleling the genotype/phenotype distinction.

Aaron Lynch (1996) — epidemiological mechanisms

Lynch specified seven concrete modes of memetic transmission: quantity of parenthood (ideas promoting reproduction), efficiency of parenthood, proselytic mode, preservational mode (discouraging abandonment), adversative mode (attacking competitors), cognitive advantage (ease of processing), and motivational advantage (emotional satisfaction). He developed epidemiological mathematics to model idea transmission, shifting memetics from general claims about “ideas evolving” toward quantitative specification of how specific idea-types gain competitive advantage.

Robert Aunger (2002) — the neuroscience grounding attempt

Aunger argued that memes can only exist as patterns of electrochemical activity in the brain — neuromemes. Critically, he argued memes don’t literally jump between brains but use “instigator signals” (behaviors, artifacts, speech) that trigger recreation of a similar neural pattern in another brain. He examined prions and computer viruses as proof that non-DNA replication exists but is constrained. His key contribution was insisting memetics take the physical reality of its replicator seriously, though the specific neural theory remained speculative and was criticized for the multiple realizability problem — the same idea can be implemented in different neural configurations.

---

3. The meme concept — core theory at maximum resolution

Four levels of analysis

The meme can be defined at four levels, each with distinct strengths and problems:

Neurological (Aunger): the meme as a specific configuration of neural activity capable of self-replication within and between neural nodes. Problems: resembles the elusive “engram”; the multiple realizability objection (same idea, different neural implementations); empirically unverifiable with current neuroscience.

Behavioral (Blackmore): the meme as any behavior or skill copied through true imitation (not mere social learning or stimulus enhancement). This is the most operationally tractable definition — you can observe what people imitate. The connection to mirror neurons provides a candidate neural mechanism, though Blackmore noted mirror neurons are necessary but not sufficient (many species have them without true imitation).

Artifactual (Dawkins A / externalists like Gatherer): the meme as an external artifact — a book, tool, building, or digital object. This treats artifacts as analogous to Dawkins’s “extended phenotype.” Advantage: artifacts are directly observable and quantifiable. Problem: artifacts don’t self-replicate; they need minds to copy them.

Informational (Dennett, Heylighen): the meme as an abstract information pattern instantiatable across multiple substrates. This is the most general definition and the one most consistent with Universal Darwinism, since it makes the meme substrate-independent. The same meme can exist as neural patterns, ink on paper, or digital bits. Heylighen formalized this as the distinction between memotype (information in memory) and mediotype (meme expressed in external medium). Critics (Sperber) argue this risks treating information as disembodied “form” divorced from material substrate.

The three Darwinian properties in depth

Fidelity (copying accuracy): This is the most contested property. Dawkins acknowledged the difficulty: memes seem to lack “the particulate, all-or-none quality of gene transmission.” DNA polymerase achieves error rates of approximately 1 in 10⁹ per base pair; human imitation is vastly less precise. The critical question is whether memetic fidelity exceeds the error catastrophe threshold (from Eigen and Schuster’s work on RNA viruses) — the minimum copying accuracy below which evolutionary “good tricks” cannot accumulate. Blackmore’s argument: the existence of cumulative culture empirically demonstrates sufficient fidelity, whatever the mechanism. Key fidelity-maintaining mechanisms include language (discrete combinatorial units constraining variation), writing and print (dramatic fidelity increase; Plato’s ideas survive 2,400 years without unbroken chains of memory), digital storage (near-perfect copying), and institutions (schools, churches, legal systems function as error-correction systems for memetic transmission). Blackmore’s copy-the-product versus copy-the-instructions distinction is critical here: copying instructions (a recipe, a score, a blueprint) is high-fidelity and parallels the genotype; copying products (recreating a soup from taste) is low-fidelity and parallels the phenotype.

Fecundity (copying rate): What determines spread rate includes ease of imitation (simpler memes spread faster), emotional resonance (memes exploiting fear, humor, or outrage spread more readily), access to transmission channels (modern technology massively amplifies fecundity), and compatibility with existing mental structures. Heylighen and Chielens formalized this: memetic fitness F = A × R × E × T, where A = proportion assimilated, R = proportion retained, E = number of times expressed, T = number of potential hosts reached per expression. F is zero if any component is zero. The progression from oral tradition through writing, print, broadcast, and internet represents exponential increases in T and E. Memes have vastly higher fecundity than genes: horizontal transmission across unrelated individuals, instantaneous global reach, no generational bottleneck.

Longevity (persistence): A meme must persist long enough to be copied at least once, but longer persistence multiplies replication opportunities. External storage media revolutionized memetic longevity: Aunger noted that memes can store themselves in artifacts that await a signal to replicate — “a signal that may not come until the meme’s living hosts all are dead.” The progression from ephemeral speech to durable text to indefinite digital storage represents a qualitative transformation of the memetic fitness landscape.

Mapping to fitness: Memetic fitness = the product of fidelity × fecundity × longevity, operating within a fitness landscape where memes compete for finite cognitive resources (limited attention, memory capacity). The fitness landscape is shaped by memorability, emotional resonance, ease of transmission, compatibility with cognitive biases, and access to transmission infrastructure — not by truth or utility to the host.

The unit problem — why it’s hard and why it matters

The unit problem asks: what counts as “one meme”? Is Beethoven’s Fifth Symphony a single meme, or is the four-note motto (da-da-da-DUM) a meme by itself? The problem directly parallels the gene boundary problem in biology — even now, “gene” has multiple definitions (Mendelian, molecular, evolutionary). George Williams (1966) defined the evolutionary gene functionally: “any hereditary information for which there is a favorable or unfavorable selection bias equal to several or many times its rate of endogenous change.” The meme’s boundaries may similarly depend on function, not structure.

Proposed solutions: (i) The meme as the smallest element that replicates with reliability and fecundity (Dennett). Problem: cultural transmission doesn’t decompose neatly into minimal units. As Balkin objected, a three-note phrase can be a meme in one context (the opening of “Three Blind Mice”) but not another. (ii) The meme as whatever unit selection acts on — “the largest units of socially transmitted information that reliably and repeatedly withstand transmission” (Pocklington and Best, 1997). This functional definition makes the unit problem empirical rather than definitional. (iii) The meme as information pattern at whatever level of granularity is analytically useful — accepting that memes exist in nested hierarchies with context-dependent boundaries.

Why it matters: Without discrete units, there can be no population-level counting, no measurement of frequency changes, and no rigorous science of memetics. As one critic put it: “denying memes are a unit denies the cultural analogy that inspired Dawkins to define them.” This unresolved problem is arguably why the Journal of Memetics ceased publication in 2005 and why the academic mainstream migrated to Boyd and Richerson’s DIT, which sidesteps the unit problem by modeling “cultural variants” operationally for specific research questions without requiring a universal cultural atom.

The replicator concept and Universal Darwinism

A replicator in Dawkins’s technical sense is any entity that makes copies of itself with fidelity, fecundity, and longevity. The key insight: replicators are defined by function (what they do), not form (what they’re made of). This makes the concept substrate-neutral. Dawkins distinguished replicators from vehicles — entities that carry replicators and interact with the environment (organisms carry genes; brains/cultures carry memes). David Hull refined this to replicator/interactor/lineage: interactors are causally active (not passive “vehicles”), and lineages are the historical sequences of replicator copies.

Why a second replicator was needed: Genes alone cannot explain celibacy (genetically maladaptive but culturally successful), martyrdom (self-sacrifice for ideas contradicts genetic self-interest), birth control and adoption (genetic “mysteries” easily explained by memetic advantage), immense cultural diversity not traceable to genetic variation, and the speed of cultural change (orders of magnitude faster than genetic change).

Universal Darwinism is the claim that Darwinian evolution is an abstract algorithm running on any substrate exhibiting variation, selection, and heredity. Key formulators: Campbell (1960, BVSR as universal principle), Dawkins (1983, coined the term), Dennett (1995, called it a “universal acid” — “a scheme for creating Design out of Chaos without the aid of Mind”). The critical implication: if culture contains information copied with variation and selection, cultural evolution is not metaphorical — it is a genuine instance of the Darwinian algorithm.

The selfish meme hypothesis

A meme is “selfish” in Dawkins’s technical sense: it spreads because doing so benefits the replicator itself, not necessarily its host. “Selfish” is a metaphor for the outcome of differential replication, not a claim about meme consciousness.

Key examples: Chain letters explicitly instruct recipients to make copies — they do not benefit hosts but are highly successful at replication. Celibacy memes in Catholicism are genetically disastrous (carriers don’t reproduce) but memetically brilliant — celibate priests devote all energy to propagating religious memes. Martyrdom/suicide memes: the host’s death can actually increase the meme’s spread (martyrdom inspires others) — completely inexplicable from the gene’s perspective but perfectly explicable from the meme’s. Boudry and Hofhuis (2018) analyzed medieval witch persecution as “socio-cultural design without designers” — the belief complex (Devil’s pact, flying brooms, Sabbaths) harmed everyone involved but was remarkably effective at self-replication.

Philosophical implications for agency: Dennett argued that if human consciousness is itself “a huge complex of memes,” then “the ‘independent’ mind struggling to protect itself from alien and dangerous memes is a myth.” The “agent” who might resist selfish memes is itself a memetic construct. Blackmore pushed to the conclusion: “We have no free will, and our consciousness is not the driving force of our behaviour.” Dawkins himself was more moderate, arguing we “can rebel against the tyranny of the selfish replicators” — but as Blackmore noted, this claim is difficult to sustain within the framework Dawkins himself constructed.

Strongest counterarguments: (1) The agency objection — people actively interpret, modify, accept, or reject cultural information; cultural evolution is arguably Lamarckian, giving humans creative agency. (2) Explanatory emptiness — if selfish memes explain everything, they explain nothing (Boudry and Hofhuis). (3) Sperber’s reconstruction critique — ideas are reconstructed, not replicated, so there is no true replicator to be “selfish.” (4) No memetic code — Benitez-Bribiesca points to the absence of a code analogous to DNA, making “replication” chaotically imprecise.

---

4. Memeplexes: co-adapted complexes and immunizing strategies

What memeplexes are and how they form

A memeplex (meme-complex) is a collection of mutually supporting memes that have co-evolved into a symbiotic relationship, tending to replicate together as a unit. Dawkins introduced the concept as “coadapted meme complexes” in The Selfish Gene; Blackmore popularized the abbreviated term in The Meme Machine. The mechanism of co-adaptation: memes that happen to co-occur in the same host mind and enhance each other’s fidelity, fecundity, or longevity will tend to be transmitted together. Over time, selection favors bundles that propagate more effectively as a group than any individual meme could alone. This parallels coadapted gene complexes in biology (a cheetah’s sharp teeth, specialist digestive system, and speed are individually less viable but collectively powerful).

The linkage mechanism differs from genetics: genes are physically linked on chromosomes, but memes are linked through cognitive coherence — memes that fit together logically, emotionally, and socially tend to be adopted and transmitted as a package. This creates path dependence: once a person adopts some elements of a memeplex, the remaining elements become easier to accept.

Religion as the paradigmatic memeplex

Component memes and their mutual reinforcement: God exists (core ontological claim enabling all others) → God is all-knowing/judging (makes moral codes enforceable without human policing) → Afterlife with heaven/hell (ultimate enforcement mechanism) → Sin/guilt (creates unworthiness, making salvation necessary) → Salvation (makes the system seem worthwhile) → Faith as virtue (immunizing meme protecting against evidence-based critique) → Proselytizing duty (maximizes fecundity) → Punishment for apostasy (raises cost of defection) → Sacred text (fixed “recipe” for copying, increasing fidelity) → Clergy/hierarchy (professional transmitters and enforcers). Each component increases the replicative fitness of the others. As Dawkins argued: “Roman Catholicism and Islam were not necessarily designed by individual people, but evolved separately as alternative collections of memes that flourish in the presence of other members of the same memeplex.”

Nationalism as memeplex

Component memes: shared mythic history (founding narratives), sacred symbols (flag, anthem), sacrifice norms (military service), in-group/out-group distinction (citizenship, borders), territorial sovereignty, cultural superiority claims, reverence for national language, threat narratives activating solidarity. These co-adapt because each reinforces the others: mythic history justifies territorial claims; sacred symbols emotionalize abstract loyalty; threat narratives make sacrifice seem necessary.

The scientific method as a structurally different memeplex

Components: empiricism, falsifiability, peer review, replication, parsimony, provisional knowledge, mathematical formalization, open publication, skepticism as virtue. The crucial structural difference from religious memeplexes: science contains built-in self-correction mechanisms rather than immunizing strategies. Where religion says “doubt is sin,” science says “doubt is methodology.” Where religion has faith (an immunizing meme protecting core beliefs from evidence), science has falsifiability (a vulnerability meme that deliberately exposes core beliefs to evidence). However, at the institutional level, science can develop immunizing tendencies — as Kuhn’s analysis of “normal science” shows, paradigms resist displacement through auxiliary hypotheses and sociological pressure.

Immunizing memes — the memeplex’s immune system

Immunizing memes are component memes that protect the complex from competing memes, disconfirming evidence, or rational critique. Heylighen identified four meme-centered selection criteria that function as immunizing and propagating strategies:

Self-justification: Components mutually justify each other in circular fashion. “God exists because the Bible says so” + “You should believe the Bible because it is the word of God.” Impervious to external critique because no independent support is needed.

Self-reinforcement: The meme stimulates its host to rehearse it — through prayer, chanting, meditation, ritual repetition. This deepens neural encoding, making the meme harder to dislodge.

Intolerance: The meme induces the host to reject competing memes a priori. This creates what Glenn Grant called “meme allergies” — hostile reactions to threatening ideas that bypass rational evaluation.

Proselytism: The meme urges maximum spread to other hosts, increasing fecundity and creating social reinforcement networks.

Specific immunizing examples in religious memeplexes: “Faith” as virtue — reframes absence of evidence as a test of character rather than a reason for doubt. “The devil plants doubts” — attributes skeptical thought to an enemy agent. “Persecution proves we’re right” — transforms external criticism into confirmation, making the memeplex antifragile. “You’ll understand when you’re saved” — dismisses outsider critique by denying the critic’s epistemic standing. Dennett’s concept of “belief in belief” — even non-religious people often believe that belief itself is virtuous, creating a meta-level immunizing meme that protects religion from external critique.

Why immunizing memes explain resistance to rational persuasion

Duch’s neural modeling of conspiracy theories provides a mechanistic account: memeplexes create basins of attraction in neural networks. New information entering a mind with an established memeplex is distorted by the existing attractor, assimilated into the complex, and further strengthens it. “Once a set of distorted memory states is entrenched it becomes a powerful force, attracting and distorting all information that has some associations with these states.” Contradicting arguments arouse only “transient weak activations” and are ignored or actively erased through retrieval-induced forgetting. This is a physical neural process, not merely a psychological bias, which explains why simple persuasion fails against well-established memeplexes.

Conspiracy theories are the paradigmatic self-sealing memeplex: every piece of counter-evidence becomes evidence of deeper conspiracy; all critics are suspect as dupes or conspirators; internal coherence substitutes for external evidence; emotional arousal deepens encoding.

Co-adaptation mechanisms and inter-memeplex competition

Memes co-adapt through mutual reinforcement of the three fitness parameters: sacred texts increase copying fidelity; proselytizing memes increase fecundity; institutional structures (churches, universities, political parties) increase longevity. Multiple levels of selection operate: individual memes compete within a memeplex (specific dietary rules may be eliminated while the core persists), entire memeplexes compete against other memeplexes (Christianity vs. Islam vs. Buddhism), and memeplexes compete for finite cognitive resources — human attention, memory space, and commitment.

A crucial distinction from gene complexes: memeplexes face less pressure to benefit their hosts because memes propagate horizontally (any person to any person) rather than primarily vertically (parent to child). Vertically transmitted cultural items face selection for host benefit (because the host must survive and reproduce to pass them on); horizontally transmitted items face selection only for transmissibility. This structural difference explains why memeplexes can be parasitic in ways gene complexes typically cannot.

Connection to Kuhn’s paradigm shifts

The relationship is direct and illuminating. Kuhn’s paradigms are essentially scientific memeplexes. Normal science = the period of memeplex stability, where internal coherence creates a strong attractor basin. Anomalies are handled through auxiliary hypotheses (adding components to the memeplex). Crisis = accumulating anomalies degrade coherence. Paradigm shift = the old memeplex is replaced wholesale by a new one. This explains why paradigm shifts are rare and revolutionary — they require overcoming the entire self-reinforcing structure of a memeplex, not just refuting individual claims.

---

5. Critical context: the mainstream alternative and the key critique

Boyd and Richerson’s dual inheritance theory

DIT models cultural transmission using population-genetics mathematics and transmission biases rather than discrete replicating units. Their core biases: content bias (some variants are intrinsically more attractive), prestige bias (copying high-status individuals, which can drive runaway cultural selection analogous to sexual selection), conformist bias (disproportionately adopting the most common variant, which homogenizes within-group behavior and strengthens cultural group selection), success bias (copying demonstrably successful individuals). DIT became the academic mainstream because it builds on rigorous mathematical foundations, makes testable population-level predictions, generated substantial empirical research, and avoids contentious ontological claims about cultural replicators. It successfully explained lactase persistence (selection coefficient 0.09–0.19 in Scandinavians from dairy-culture-driven gene-culture coevolution), the demographic transition, and large-scale cooperation.

Sperber’s epidemiology of representations

Sperber’s central critique: “neither communication nor imitation are replication mechanisms.” Cultural stability arises from cultural attractors — points in representational space toward which mental reconstructions converge because of shared cognitive architecture. Even with imperfect transmission, if transformations are biased toward attractors, cultural items stably concentrate around a few configurations — “mimicking replication without actual replication.” This is directly supported by Bartlett’s findings. The attractor framework is mathematically formalizable and empirically grounded in cognitive science. However, as some reviewers note, it may be possible to recover a Darwinian theory by treating all representations within the same basin of attraction as instances of the “same” meme — making attractors and replicators complementary rather than competing frameworks.

Heylighen’s four-stage selection framework

Heylighen formalized meme replication as a four-stage process: assimilation (the meme enters a new host’s memory — requires being noticed, understood, and accepted), retention (it remains in memory long enough to spread), expression (it emerges from memory into physical form perceivable by others), and transmission (the expression reaches another individual via a stable medium). At each stage, different selection criteria apply — objective (distinctiveness, invariance), subjective (novelty, simplicity, coherence, utility), intersubjective (authority, conformity, group utility), and meme-centered (self-justification, self-reinforcement, intolerance, proselytism). This framework operationalizes memetic fitness and was tested empirically using virus hoaxes as paradigmatic self-replicating cultural items.

Why memetics failed institutionally but persists conceptually

The Journal of Memetics ceased publication in 2005. Bruce Edmonds’s final-issue assessment: “the closer work has been to the core of memetics, the less successful it has been.” Key failure reasons: definitional gridlock (endless debates about what constitutes a meme), lack of empirical traction (theories too vague for testing), the gene analogy pushed too far (futile pursuit of a cultural atom), and competing frameworks proving more productive. Yet the concept persists: Dawkins and Dennett continued advocating it; the term “meme” entered everyday language; researchers continue applying memetic frameworks; and the useful insight — that cultural information has its own evolutionary dynamics, that the “information’s-eye view” is illuminating, and that Darwinian logic applies to culture — has been largely absorbed into broader cultural evolutionary theory. As Wilkins concluded: “It is not that there are no replicators in culture, but rather that there is no single kind of replicator, and moreover that not all cultural evolution requires replication in order to proceed.”

Transmission, models, media, and critique

Part IV: Transmission mechanisms in depth — cognitive architecture as meme filter

The central claim of memetic theory is that cultural items replicate differentially. But differential replication requires a mechanism, and that mechanism is the human mind. The brain is not a passive recording device; it is a battery of evolved inference systems, each with its own biases, thresholds, and sensitivities. These systems act as filters through which candidate memes must pass. Understanding which memes succeed therefore demands understanding the cognitive architecture that selects among them. The relevant biases fall into three analytically distinct categories — content biases, context biases, and evolved psychological hooks — which operate at different levels and can reinforce, compete with, or override one another.

Content biases: what the meme itself brings to the table

Content biases arise from intrinsic properties of the information interacting with stable features of human cognition. The most fundamental is cognitive ease. Daniel Kahneman’s dual-process framework distinguishes System 1 (fast, automatic, associative) from System 2 (slow, effortful, rule-governed). System 1 processes information with minimal cognitive load, and items that pass through System 1 processing without friction acquire what Kahneman calls “cognitive fluency” — they feel true, familiar, and safe. Memes that are simple, rhythmic, symmetrical, or easily visualized exploit this processing advantage. A slogan that rhymes (“if it doesn’t fit, you must acquit”) is judged more truthful than its unrhymed paraphrase, not because rhyming correlates with truth but because phonological fluency is misattributed as evidential weight. The memetic consequence is stark: in any competition between a cognitively fluent meme and a more accurate but effortful rival, the fluent variant enjoys a systematic transmission advantage that compounds across retelling chains.

Emotional salience constitutes the second major content bias, and here a pronounced asymmetry governs the landscape. Negative high-arousal emotions — fear, disgust, and outrage — amplify transmission far more powerfully than positive emotions of equivalent intensity. Baumeister and colleagues established in their landmark 2001 review that “bad is stronger than good” across virtually every psychological domain, from impression formation to memory encoding. Heath, Bell, and Sternberg demonstrated the mechanism specifically for cultural transmission in their study of urban legends: controlling for informational content, people were significantly more willing to pass along narratives that evoked stronger disgust, and legends containing more disgust motifs enjoyed wider distribution on legend-aggregation websites. The evolutionary logic is straightforward and follows from error management theory. The fitness cost of ignoring a genuine threat vastly exceeds the cost of a false alarm; natural selection therefore calibrated threat-detection systems toward oversensitivity. Fessler, Pisor, and Navarrete formalized this as negatively-biased credulity: people find hazard-related claims more believable than equivalent benefit-related claims because the asymmetric cost structure of errors made skepticism about danger more costly than gullibility. Crucially, the relevant variable is arousal, not mere valence. Berger and Milkman showed that high-arousal positive emotions like awe also drive sharing, while low-arousal negative states like sadness actually suppress it. But the asymmetry persists because the space of high-arousal negative emotions (anger, fear, disgust, outrage, anxiety) is larger and more readily activated than its positive counterpart.

Narrative structure operates as a third content bias — a transmission scaffold that organizes information into a form the mind preferentially processes. Humans are, as Jonathan Gottschall argued, storytelling animals. Information packaged in story form — with characters, causation, temporal sequence, and resolution — enjoys enormous advantages in encoding, retention, and retelling over equivalent information presented as data, lists, or abstract propositions. The mechanism is partly mnemonic (narrative provides retrieval cues through causal and temporal chaining) and partly motivational (stories generate suspense, empathy, and curiosity that sustain attention). Green and Brock’s research on narrative transportation demonstrates that absorbed readers lower their critical defenses, making story-embedded claims more persuasive than identical claims presented argumentatively.

A fourth content bias operates through moral coding. Jonathan Haidt’s Moral Foundations Theory identifies six evolved psychological systems — Care/Harm, Fairness/Cheating, Loyalty/Betrayal, Authority/Subversion, Sanctity/Degradation, and Liberty/Oppression — each shaped by recurrent adaptive challenges. These foundations function as pre-wired receptor sites for memes. Content that triggers a moral foundation acquires automatic emotional salience and bypasses deliberative evaluation. Brady, Wills, Jost, Tucker, and Van Bavel found that on Twitter, each additional moral-emotional word in a tweet increased its retweet rate by approximately 20%. The Sanctity/Degradation foundation is particularly potent as a transmission amplifier because it is grounded in disgust psychology and the behavioral immune system. Wheatley and Haidt demonstrated that hypnotically induced disgust intensifies moral judgments even for morally neutral scenarios, and Dehghani and colleagues found that purity-based moral language creates tighter social clustering than any other foundation — meaning purity-triggering memes generate stronger echo chambers.

Context biases: who says it matters as much as what is said

Context biases operate not on the content of the meme but on properties of the source, the audience, or the social environment. Prestige bias — preferentially copying the cultural variants of admired, high-status individuals — is the most consequential. Henrich and Gil-White’s influential 2001 model explains the mechanism with considerable precision. Natural selection favored social learners who could identify and copy the most skilled models. To improve copying fidelity, learners evolved dispositions to ingratiate themselves with chosen models — conferring deference, gifts, and attention to gain prolonged proximity. Once widespread, these deference patterns became publicly visible, creating a second-order cue: the size and enthusiasm of a model’s “clientele” serves as a reliable proxy for that model’s information quality. New entrants exploit this shortcut, copying the most-deferred-to individual without independently evaluating their skills.

The evolutionary dynamic here parallels Fisher’s runaway sexual selection. Boyd and Richerson showed formally that when the criterion used to evaluate models is itself culturally transmitted alongside the target trait, a positive feedback loop emerges. The preference for prestigious individuals and the traits those individuals carry are transmitted together, creating co-evolutionary escalation. Just as peahens’ preference for elaborate tails drives ever-more-elaborate tails, copiers’ deference drives ever-greater prestige markers — and, critically, can propagate maladaptive traits when prestigious individuals are copied beyond their domain of expertise.

Conformist bias operates through a different mechanism: frequency-dependent copying where the probability of adopting a variant disproportionately exceeds its population frequency. If 60% of observed models use behavior A, conformist bias means more than 60% of learners adopt it. Boyd and Richerson demonstrated that conformist transmission evolves when environments change at moderate rates — too slowly for genetic tracking but not so rapidly that individual learning dominates. The stabilizing effect is powerful: conformist bias maintains between-group cultural variation while making it difficult for novel variants to invade established populations. It functions as a cultural conservatism mechanism, preserving locally adaptive norms against drift and individual-level experimentation.

Success bias (directly observing and copying the behavior of demonstrably successful individuals) and kin bias (preferential learning from family members) complete the context-bias repertoire. The critical analytical point is that content and context biases can conflict. A cognitively fluent, emotionally salient meme (strong content biases) transmitted by a low-prestige source (weak context bias) may lose to a less intrinsically compelling meme endorsed by a prestigious figure. The outcome depends on the relative weighting of each bias, which itself varies across domains and individuals.

Evolved psychological hooks: exploiting deep cognitive architecture

Beyond content and context biases lie deeper features of evolved cognitive architecture that specific meme types exploit with particular effectiveness. Justin Barrett’s Hyperactive Agency Detection Device (HADD) names the evolved tendency to over-attribute intentional agency to ambiguous stimuli. Grounded in signal detection theory and error management, HADD reflects the asymmetric cost structure of agency-detection errors: failing to detect a predator is potentially fatal, while falsely detecting one costs only a momentary startle. The memetic consequence is that agent-invoking concepts — stories featuring ghosts, gods, conspirators, hidden manipulators — enjoy a systematic processing advantage over non-agent explanations for the same phenomena. Pascal Boyer extended this analysis by noting that agentive concepts are cognitively richer: our Theory of Mind module generates extensive predictions about agents’ goals, beliefs, and desires, making agent concepts more inferentially productive and therefore more interesting, attention-grabbing, and memorable than mechanistic explanations.

Boyer’s minimal counterintuitiveness (MCI) theory identifies a separate cognitive hook. Boyer proposed that concepts can be classified by how many violations they commit against the default ontological expectations associated with their category (person, animal, artifact, natural object). A person who walks through walls violates one expectation of the person category (solidity). A statue that listens to prayers transfers a psychological property from persons to artifacts. Boyer and Ramble’s cross-cultural experiments — conducted in France, Gabon, and Nepal — demonstrated that concepts with one or two such violations are recalled significantly better than either fully intuitive concepts (unremarkable, quickly forgotten) or maximally counterintuitive ones (too many violations overwhelm working memory and prevent meaningful inference). The cognitive optimum sits at one to two ontological violations: enough to capture attention through surprise, few enough to remain inferentially tractable. Barrett and Nyhof confirmed that this recall advantage persists across three retellings and survives a three-month delay. The successful supernatural concepts populating the world’s religions cluster overwhelmingly at this optimum — virgin births, burning bushes that speak, prophets who ascend to heaven — while maximally bizarre theological abstractions remain the province of specialists.

Coalitional instincts provide yet another hook. Humans evolved in contexts where inter-group competition was intense, generating psychological adaptations for identifying in-group members, detecting defectors, and coordinating collective action against rivals. Memes that activate in-group/out-group psychology — tribal markers, loyalty tests, narratives of external threat — exploit these adaptations to secure rapid adoption and vigorous transmission. The moral foundations of Loyalty/Betrayal and Sanctity/Degradation are particularly effective here: loyalty-invoking memes demand sharing as proof of group commitment, while purity-invoking memes mark out-group members as contaminated or polluted, triggering disgust-based avoidance.

The Heylighen four-stage selection model

Francis Heylighen synthesized these disparate mechanisms into a unified framework by modeling meme fitness as the product of success at four sequential stages. The assimilation stage determines whether a meme is noticed and cognitively processed by a potential host. Selection criteria here include distinctiveness (standing out from background noise), novelty (differing enough from existing mental contents to attract attention), simplicity (being processable without excessive cognitive effort), coherence (fitting with the host’s existing belief network), and source authority (coming from a prestigious or trusted transmitter). The tension between novelty and coherence is critical: the optimally assimilable meme is novel enough to capture attention but coherent enough with existing schemas to be comprehensible.

The retention stage determines whether an assimilated meme persists in memory. Selection criteria include utility (memes perceived as useful are rehearsed and retained), invariance (memes that prove consistent across observations gain credibility), emotional intensity (strongly valenced memes receive deeper encoding), and self-reinforcement (memes that prompt their host to rehearse them — through prayer, recitation, or rumination — strengthen their own memory traces). The expression stage determines how frequently and vigorously a retained meme is externalized by its host. Here, utility again matters (useful memes are applied and discussed), but proselytism (memes that contain an explicit or implicit instruction to spread them) and emotional arousal (which activates the sympathetic nervous system and primes behavioral output) are the dominant drivers. The transmission stage determines how many potential new hosts receive each expression. Publicity (breadth of the medium), formality (clarity and standardization of expression), and network position of the host all matter here.

Heylighen formalized the overall fitness of a meme as F = A × R × E × T, where each factor ranges from zero to one (for A and R) or zero to unbounded (for E and T). The multiplicative structure carries a crucial implication: fitness is zero if any single stage fails completely. A perfectly memorable meme that no one is motivated to express has zero fitness. A highly expressed meme that cannot be assimilated by recipients likewise fails. This framework explains why the most successful memes are not necessarily those that maximize any single property but those that achieve adequate performance across all four stages — a constraint that favors robust generalists over narrow specialists.

---

Part V: Formal models — mathematical structure of cultural dynamics

Intuitive accounts of meme transmission can only carry analysis so far. Formal models impose the discipline of logical consistency, reveal non-obvious implications, and generate quantitative predictions that can be tested against empirical data. The mathematical toolbox available for modeling cultural dynamics draws from epidemiology, evolutionary game theory, Bayesian inference, agent-based simulation, and computational data science.

Epidemiological models: the SIR framework and its extensions

The most natural formal analogy for meme spread is epidemic contagion. The SIR model, formulated by Kermack and McKendrick in 1927, partitions a population into three compartments — Susceptible (S), Infected (I), and Recovered (R) — governed by coupled differential equations. In memetic translation, Susceptible individuals have not yet encountered the meme; Infected individuals have adopted it and actively transmit it; Recovered individuals have encountered the meme but lost interest or become resistant. The transmission rate β captures the meme’s intrinsic contagiousness — its shareability, emotional charge, and cognitive fluency — while the recovery rate γ captures the rate at which carriers lose interest and stop transmitting.

The fundamental quantity is the basic reproduction number R₀ = β/γ, representing the average number of new carriers generated by one active carrier in a fully susceptible population. When R₀ exceeds one, the meme spreads epidemically; below one, it dies out. R₀ is not a fixed property of the meme alone but an emergent quantity depending on meme properties (β), host psychology (γ), and crucially, network structure. In a well-mixed population, the epidemic threshold is clear-cut. But real social networks are not well-mixed.

Network topology fundamentally transforms epidemic dynamics. On Erdős-Rényi random networks, where connections are randomly distributed, a well-defined percolation threshold exists: below a critical transmission probability proportional to the inverse of the mean degree, large-scale cascades cannot occur. On Barabási-Albert scale-free networks, however, where degree distribution follows a power law and a few hub nodes maintain vastly more connections than typical nodes, Pastor-Satorras and Vespignani showed that the epidemic threshold effectively vanishes. Even weakly contagious memes can spread epidemically because superspreader hubs amplify transmission disproportionately. This finding has direct relevance for social media, where influencer accounts function as hub nodes capable of seeding cascades that would fail on more homogeneous networks.

The SEIR extension adds an Exposed compartment between Susceptible and Infected, capturing a latency period during which individuals have encountered the meme but are not yet transmitting it. In memetic terms, this latency period encompasses the time required for cognitive processing, deliberation about whether to share, waiting for the right social context, and the gradual internalization that transforms passive awareness into active advocacy. The latency period varies enormously across meme types: a funny image may have near-zero latency (immediate sharing), while a political ideology may incubate for months before expression.

The SIR framework, despite its elegance, fails for memes in at least three critical respects. First, recovery is not permanent: people can re-adopt memes they previously abandoned, making SIS (Susceptible-Infected-Susceptible) or SIRS models more appropriate for many cultural dynamics. Second, multiple meme “strains” coexist and compete for the same attentional resources — there is no memetic equivalent of the assumption that infection with one pathogen confers immunity to others. Third, memes mutate during transmission far more rapidly than most pathogens, with each retelling potentially altering content in ways that change transmissibility. These disanalogies do not invalidate the epidemiological approach but require extensions that push beyond the classical framework.

Replicator dynamics: the algebra of cultural competition

Where epidemiological models track absolute numbers of carriers, replicator dynamics track the relative frequencies of competing cultural variants. The replicator equation, formalized by Taylor and Jonker in 1978, states: ẋᵢ = xᵢ(fᵢ − f̄), where xᵢ is the frequency of variant i, fᵢ is its fitness, and f̄ is the population’s average fitness. The equation’s mechanism is straightforward but powerful. The rate of change of a variant’s frequency is proportional to two factors simultaneously: its current frequency xᵢ (rare variants change slowly because they have few carriers) and its fitness advantage over the average (fᵢ − f̄). Variants fitter than average grow; those below average shrink. Fitness is always relative — what matters is not absolute performance but performance compared to the current competition.

The equation generates several important dynamics in cultural contexts. Fixed points occur when all variants have equal fitness or when one variant has driven all others to extinction. An Evolutionarily Stable Strategy (ESS) is a cultural variant that, once established at high frequency, cannot be invaded by any small group of individuals using an alternative — the cultural equivalent of a deeply entrenched norm. But not all systems converge to stable equilibria. When fitness is frequency-dependent — when the success of a strategy depends on what others are doing — replicator dynamics can produce sustained cycling. The classic case is Rock-Paper-Scissors dynamics: each strategy beats one rival and loses to another, generating perpetual oscillation rather than fixation. Cultural analogs include fashion cycles, where any dominant style creates conditions favoring its antithesis, and cyclical political ideologies, where the failures of the currently dominant approach generate demand for its opposite.

The replicator equation’s limitations are equally important to understand. It assumes a well-mixed population (no network structure), does not introduce new variants (all strategies must pre-exist), and treats fitness as externally given rather than endogenously determined by cultural context. For understanding the full dynamics of cultural evolution, a more general tool is needed.

The Price Equation: separating selection from transmission bias

George Price derived in 1970 what may be the most general equation in all of evolutionary theory. The Price Equation partitions evolutionary change into two terms: w̄Δz̄ = Cov(w, z) + E(wΔz), where z̄ is the population mean of some trait, w is fitness, and Δz represents parent-offspring change in trait value.

The first term, Cov(w, z), is the selection term. It captures the covariance between fitness and trait value: if individuals with higher trait values also have higher fitness (more cultural “offspring” — more people copying them), the population mean shifts upward. This is cultural selection in its purest form. The second term, E(wΔz), is the transmission bias term. It captures systematic changes in trait value during the transmission process itself, weighted by fitness. In genetic evolution, this term is typically negligible because DNA replication is highly faithful. In cultural evolution, it is often the dominant force. When learners systematically simplify, embellish, emotionally intensify, or otherwise transform cultural content during acquisition and retransmission, this term drives evolutionary change independently of selection.

Daniel Nettle emphasized in his 2020 analysis that this distinction is invisible to replicator dynamics. The Price Equation reveals that cultural change can be driven by how people transform ideas during learning, not merely by which ideas differentially attract learners. The design-like properties of cultural traditions — the apparent “adaptation” of myths, norms, and technologies to their cognitive and social environments — may owe as much to convergent transmission biases (everyone’s memory systems distort in the same direction) as to competitive selection among variants. The Price Equation also enables decomposition of the selection term into within-group and between-group components, providing the formal foundation for cultural group selection models. El Mouden, André, Morin, and Nettle used this decomposition to demonstrate that cultural fitness and genetic fitness are distinct quantities, and that antagonistic coevolution between genes and culture occurs whenever they diverge.

Agent-based models: from equations to simulations

Equation-based models purchase analytical tractability at the cost of simplifying assumptions — homogeneous agents, well-mixed populations, deterministic dynamics. Agent-based models (ABMs) relax all three assumptions simultaneously. In an ABM, each agent is a distinct computational entity with individual properties (susceptibility thresholds, network position, belief states, decision rules) that interacts locally with connected agents according to specified protocols. Population-level outcomes emerge from the aggregation of millions of individual interactions, often in ways that no equation could predict.

The key contribution of ABMs to cultural evolution research has been the discovery of threshold effects and cascade dynamics. Building on Granovetter’s classic threshold model, Watts demonstrated that global cascades — system-wide adoption events — emerge only within a narrow window of parameter space. Outside this window, perturbations remain local. Inside it, a single adopter can trigger a domino chain that transforms the entire population. The size and location of this cascade window depends on the distribution of individual thresholds and on network topology in ways that defy analytical prediction.

Centola’s work on complex contagion produced the field’s most counterintuitive finding. For simple contagions (where a single exposure suffices for transmission), Granovetter’s weak ties accelerate spread by bridging otherwise disconnected clusters. But for complex contagions — where adoption requires social reinforcement from multiple independent sources — weak ties actually impede diffusion because they provide only single-source exposure without reinforcement. What complex contagions require are “wide bridges”: multiple shared connections between clusters that can deliver the redundant social proof needed to cross resistance thresholds. Centola confirmed this experimentally in 2010, showing that health behaviors spread farther and faster on clustered-lattice networks than on random networks — the exact opposite of simple-contagion predictions. His subsequent work with Baronchelli identified the 25% tipping point: when a committed minority reaches approximately one-quarter of the population, they can overturn established social conventions with remarkable speed.

Axelrod’s culture model demonstrated another emergent phenomenon: despite a mechanism of local convergence (interacting agents become more similar), populations can freeze into stable multicultural configurations with sharp cultural boundaries. Global homogenization is not the inevitable outcome of cultural mixing; instead, the system exhibits a phase transition between a homogeneous regime and a polarized one, governed by the number of available cultural traits.

Bayesian models of cultural learning

Bayesian models reconceptualize cultural learning as rational inference under uncertainty. The framework is elegant: an agent’s existing beliefs constitute the prior distribution P(state); new evidence — whether from personal experience or social observation — generates a likelihood P(data|state); the posterior belief P(state|data) is proportional to prior times likelihood. Perreault, Moya, and Boyd’s foundational 2012 model showed that individual learning and social learning need not be treated as separate strategies but emerge as endpoints of a single inferential continuum. The optimal weighting depends on environmental stability (social information is more reliable when environments change slowly), the noisiness of personal experience (costly or unreliable individual learning pushes agents toward heavier social weighting), and the number of available cultural models (more models make social information more reliable through averaging).

The Bayesian framework naturally explains two apparently contradictory phenomena. Cultural conservatism — resistance to new ideas — arises when agents have strong priors: well-established beliefs require overwhelming evidence to overturn. Cultural susceptibility — rapid adoption of novel memes — arises when agents have flat priors: in domains of genuine uncertainty, even weak social signals can shift beliefs dramatically. Conformist bias emerges naturally from Bayesian inference: if each model’s behavior is a noisy signal about the optimal behavior, majority-following is simply optimal signal aggregation. Muthukrishna, Morgan, and Henrich confirmed experimentally that both social learning and conformist bias increase with the number of cultural traits in the environment and with the adaptive relevance of the domain.

Computational memetics: what large-scale data reveals

The digital revolution has transformed memetics from an armchair discipline into one capable of large-scale empirical investigation. Phylogenetic lineage tracking — applying the computational methods developed for reconstructing biological evolutionary trees to cultural data — enables researchers to trace meme genealogies, measure mutation rates, and map fitness landscapes with unprecedented precision. Studies of internet meme ecology on platforms like Reddit have tracked meme “species” dynamics including competition, speciation, and extinction, finding patterns consistent with ecological theory.

The most consequential empirical finding in computational memetics is almost certainly Vosoughi, Roy, and Aral’s 2018 Science paper analyzing the spread of true and false news on Twitter. Examining approximately 126,000 rumor cascades tweeted by roughly 3 million people more than 4.5 million times between 2006 and 2017, they found that falsehoods were 70% more likely to be retweeted than true stories, that truth took approximately six times as long as falsehood to reach 1,500 people, and that the top 1% of false cascades reached between 1,000 and 100,000 people while truth rarely exceeded 1,000. False political news traveled deepest and broadest of any category. Critically, when all bot accounts were removed using sophisticated detection algorithms, the differential persisted: the spread advantage of false news is driven by human behavior, not automated amplification. Analysis of emotional profiles revealed that false news inspired greater surprise, fear, and disgust, while true news inspired sadness, anticipation, joy, and trust. The novelty hypothesis provides the likely mechanism: false news was measurably more novel (quantified using information-theoretic divergence measures), and novelty drives sharing because it captures attention, aids decision-making, and confers social status.

---

Part VI: The meme-media nexus — ecology, attention, and algorithmic selection

If cognitive architecture constitutes the internal selection environment for memes, media constitute the external one. Each communication technology creates a distinct ecology with its own fitness criteria, and transitions between dominant media technologies function as mass extinction events that destroy previously adapted meme lineages while opening niches for new ones.

Media as meme ecology: from oral culture to algorithmic feeds

Marshall McLuhan’s foundational insight — “the medium is the message” — translates into memetic terms as the claim that the transmission channel imposes selection pressures independent of content. The content of any medium, McLuhan argued, is “a juicy piece of meat carried by the burglar to distract the watchdog of the mind” — the real action lies in how the medium restructures cognition and social organization.

In oral culture, memes must survive in biological memory and be reconstructed through performance. The selection environment ruthlessly favors memorability: rhythm, rhyme, narrative structure, emotional intensity, proverbial compression, and formulaic phrasing. The Homeric epithets (“rosy-fingered dawn,” “wine-dark sea”) are not poetic decoration but mnemonic technology — chunking devices that reduce cognitive load during oral composition and reception. Complex abstract arguments, quantitative data, and ideas that resist narrative packaging are strongly selected against. The fitness landscape rewards memes that are, in Walter Ong’s terms, aggregative rather than analytic, redundant rather than economical, and agonistically toned rather than neutrally presented.

The invention of writing created the first major media transition. Externalized storage liberated memes from the constraints of biological memory, enabling unprecedented length, complexity, and precision. But writing also shattered the oral memeplex: communal performance was replaced by individual reading, formulaic composition gave way to original authorship, and memes that thrived through redundancy and participatory retelling lost their adaptive advantage. The meme equivalent of a mass extinction event eliminated oral-tradition specialists — bards, griots, memorizers — as dominant cultural transmitters.

Print amplified writing’s effects by orders of magnitude. McLuhan’s Gutenberg Galaxy traces how the printing press created mass-reproducible identical copies, enabling standardization of language, law, and doctrine while selecting for memes that exploit visual-linear processing: logical argument, systematic classification, sequential exposition. Print culture rewards extended reasoning, evidential citation, and authorial credibility. Neil Postman’s Amusing Ourselves to Death documented the print-era memetic ecosystem of American public discourse, where the Lincoln-Douglas debates featured seven-hour exchanges of sustained argument — a meme format that would be instantly fatal in later media environments.

Broadcast media — radio and television — imposed a new selection regime favoring emotional immediacy, visual impact, and entertainment value over logical complexity. Television in particular, as Postman argued, restructured public discourse around entertainment: news became narrative, politics became personality, and argument became spectacle. The memetic mass extinction was severe. Print-adapted memes requiring sustained attention and sequential reasoning found their ecological niche collapsing. The meme types that flourished — sound bites, vivid images, emotional narratives, charismatic personalities — bore the same relationship to print-era memes that mammals bore to dinosaurs after the Cretaceous impact: radically different body plans adapted to radically different selection pressures.

The internet created the most dramatic media transition in human history by combining characteristics of all previous media while adding participatory production, algorithmic curation, and network effects. The resulting selection environment favors memes that are shareable (optimized for forwarding, not just reception), remix-friendly (modular enough to be recombined), emotionally arousing (capable of interrupting the scroll), visually compressed (communicating in the fraction of a second before attention moves on), and platform-adapted (conforming to the specific affordances of each distribution channel). Memes adapted to the broadcast era — one-to-many, professionally produced, requiring passive consumption — face their own mass extinction as the attention economy restructures around participatory, algorithmically curated, peer-to-peer transmission.

The attention economy as selective pressure

Herbert Simon identified the foundational constraint in 1971: “a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information sources that might consume it.” Simon recognized that in an information-rich world, the scarce resource is not information but the cognitive capacity to process it. Tim Wu’s The Attention Merchants traced the institutional consequences: since Benjamin Day’s penny press in the 1830s, a business model has persisted in which free content is exchanged for audience attention, which is then sold to advertisers. Each new medium — newspapers, radio, television, social media — has intensified the competition for attention while expanding the supply of information competing for it.

The memetic consequences of attention scarcity are specific and predictable. Finite working memory (Miller’s 7±2 items) creates a hard bandwidth constraint on meme assimilation. Dual-process cognition means that under attentional pressure, System 1 dominates, favoring cognitively fluent and emotionally salient content over analytically demanding material. Interference competition between memes — where processing one meme displaces another from working memory — selects against complexity and nuance. The resulting selection gradient systematically favors memes that are emotionally arousing (to capture attention from competitors), identity-relevant (because identity-related processing is prioritized by evolved threat-detection systems), simple (to minimize processing demands), and novel (to overcome habituation). Note what this gradient selects against: nuance, qualification, statistical reasoning, multi-causal explanation, and acknowledgment of uncertainty. The attention economy does not merely favor engaging content; it creates a systematic bias against epistemic virtues.

Virality: mechanism, not metaphor

The phenomenon of “going viral” — explosive, cascading transmission through social networks — requires mechanistic explanation, not just empirical correlation. Three mechanisms interact to produce viral dynamics. First, social signaling theory explains why sharing happens at all. Sharing content is a form of identity performance: what you share signals your values, knowledge, taste, and group affiliations to your social network. Research grounded in Tajfel and Turner’s social identity theory demonstrates that meme-sharing functions as social grooming, signaling subcultural literacy and in-group membership. The selection implication is that memes optimized for identity signaling — those that make the sharer look smart, moral, funny, or loyal — enjoy sharing advantages independent of their truth value or informational content.

Second, Elaine Hatfield’s emotional contagion theory explains the affective mechanism. Hatfield, Cacioppo, and Rapson documented that humans unconsciously and nearly instantaneously mimic others’ emotional expressions, with mimicked expressions feeding back to generate corresponding felt emotions. In digital contexts, emotional contagion operates through textual cues, images, video, and emoji. The 2014 Facebook emotional contagion experiment confirmed that manipulating News Feed emotional valence significantly altered users’ own posting behavior. Content that evokes strong emotions triggers mimicry-mediated emotional convergence in the receiver, which in turn motivates further sharing — creating a self-amplifying emotional cascade.

Third, network cascade mechanics determine how local sharing decisions aggregate to global spread. Jonah Berger’s STEPPS framework identifies six mechanistic drivers of sharing: Social Currency (sharing remarkable content enhances the sharer’s status), Triggers (environmental cues that keep memes cognitively accessible), Emotion (physiological arousal from high-arousal emotions primes action including sharing), Public (observable behavior generates social proof that drives imitation), Practical Value (sharing useful information signals the sharer’s value as a social partner), and Stories (narrative packaging provides retrieval structure and lowers critical defenses). Each component operates through a distinct psychological mechanism, and the most viral content typically activates multiple STEPPS simultaneously.

The critical analytical distinction, contributed by Centola’s research, separates simple contagion from complex contagion. Simple contagion — where a single exposure suffices for transmission — spreads fastest through networks rich in weak ties and random long-range connections. Jokes, news items, and simple meme formats spread this way. Complex contagion — where adoption requires social reinforcement from multiple independent sources — spreads through dense clustered networks with “wide bridges” connecting communities. Behavioral commitments, identity shifts, and risky norm violations require complex contagion. The profound implication is that information about a behavior can go viral (simple contagion through weak ties) while the behavior itself fails to spread (complex contagion requiring strong ties). Surface-level meme formats propagate as simple contagions; the deeper ideological commitments they encode propagate as complex contagions through fundamentally different network pathways.

Algorithmic amplification as artificial selection

Recommendation algorithms on major platforms constitute perhaps the most powerful selection mechanism memes have ever encountered. These algorithms function as artificial selection — analogous to the selective breeding of domesticated species, but operating at computational speed and planetary scale. The mechanism is a feedback loop: algorithms measure engagement signals (clicks, likes, shares, comments, watch time); content generating higher engagement is amplified to larger audiences; amplification generates more engagement data; the algorithm refines its amplification criteria based on this data. Content creators, observing which formats and topics are algorithmically rewarded, increasingly produce engagement-optimized content. The result is a directed evolutionary process where memetic fitness becomes decoupled from truth value, host benefit, or social welfare and is instead optimized for engagement maximization.

The evolutionary consequences are profound. Memes evolve toward properties that maximize engagement metrics — emotional arousal, outrage, novelty, identity threat, controversy — regardless of accuracy. A 2022 pre-registered algorithmic audit of Twitter found that engagement-based ranking amplified emotionally charged, out-group-hostile political content, and that users did not prefer the political tweets selected by the algorithm over reverse-chronological presentation. Internal Facebook research reportedly demonstrated that algorithmic targeting was responsible for 64% of extremist group joins, establishing a causal link between algorithmic amplification and radicalization.

Shoshana Zuboff’s analysis of surveillance capitalism provides the institutional framework for understanding this dynamic. Zuboff argues that platforms extract behavioral data as “free raw material,” transform it into “prediction products” sold in “behavioral futures markets,” and have progressed from merely monitoring behavior to actively shaping it through “economies of action.” The algorithmic selection mechanism is characterized by what Zuboff calls “radical indifference” — the system ceaselessly escalates engagement without regard for what generates it. The memetic implication is that platform infrastructure has become a selection environment that systematically rewards engagement-maximizing meme variants and penalizes those optimized for accuracy, nuance, or social benefit.

Filter bubbles, meme speciation, and platform-specific evolution

Algorithmic personalization creates what can be analyzed through the lens of allopatric speciation. In evolutionary biology, allopatric speciation occurs when geographic barriers isolate populations, subjecting them to divergent selection pressures until they can no longer interbreed. Eli Pariser’s filter bubble concept identifies the memetic analog: algorithmic clustering creates divergent informational environments that subject different user populations to different selection pressures. Memes within each bubble evolve toward within-bubble fitness — triggering the specific moral foundations, identity commitments, and emotional patterns of that bubble’s population — becoming increasingly maladapted for cross-bubble transmission. The endpoint is not merely disagreement but memetic incompatibility: groups develop such different information diets, inferential frameworks, and vocabularies that their memes become mutually unintelligible.

Axel Bruns’ critique provides important empirical qualification. Bruns argues that filter bubbles are real but “less total than claimed” — search and social media users generally encounter a more centrist media diet than non-users, and cross-platform exposure persists even when single-platform patterns appear highly homophilous. Gentzkow and Shapiro found that online ideological segregation was lower than offline segregation. Christopher Bail’s experimental work, however, revealed a more troubling dynamic: exposure to opposing views on Twitter actually increased polarization among conservatives, suggesting that cross-bubble contact may intensify rather than reduce memetic divergence. The mechanism may not be isolation but hostile contact — encountering out-group memes that are optimized for out-group offense triggers identity-protective responses that deepen in-group commitment.

Each major platform functions as a distinct ecological niche with specific fitness criteria. Twitter/X selects for brevity, topicality, wit, and outrage — the character limit imposes formal constraints analogous to a sonnet’s meter, and the real-time feed rewards speed over deliberation. Reddit’s upvote-downvote system creates community-enforced selection where memes must satisfy subcultural norms to survive; the subreddit structure produces isolated ecosystems where specialist meme formats evolve. TikTok’s audio-first, short-form video format selects for rhythm, visual impact, participatory template structure, and sub-30-second narrative completeness — the algorithm’s For You Page rewards immediate engagement, creating intense selection pressure at the assimilation stage. YouTube rewards watch time, selecting for parasocial relationship dynamics and long-form authority formats that sustain viewer attention across minutes or hours. The migration of memes across platforms — originating in niche communities, spreading through Twitter, reaching mainstream audiences via Facebook, and declining into “normie” status — recapitulates the ecological dynamics of inter-niche invasion, with each migration requiring adaptive modification to the new platform’s selection regime.

---

Part VII: The complete critique — testing memetics against its strongest objections

Any theoretical framework must be evaluated not by its elegance but by its capacity to withstand its strongest objections. Memetics has attracted vigorous criticism from multiple directions, and intellectual honesty requires engaging these critiques at their full force before assessing what, if anything, the memetic framework uniquely contributes.

The unit problem revisited: is discreteness necessary?

We must press deeper into the unit problem logical consequences. The objection, stated precisely, is that memes lack the discrete, well-bounded character of genes, and that without discrete units, evolutionary dynamics cannot operate. Mary Midgley insisted that “thought is not granular”; Jablonka and Lamb argued that cultural evolution features “no discrete unchanging units with unchanging boundaries that can be followed from one generation to the next”; Boyd and Richerson declared flatly that “cultural variants are not replicators.”

The logical consequences of discreteness failure are serious. Without countable units, population-genetic-style frequency tracking becomes imprecise. Without clear boundaries, fitness measurement becomes ambiguous — is the fitness of a religious meme the fitness of the entire doctrinal system, a single commandment, or a particular phrasing? Lineage tracking requires some criterion for identity across transmission events, and without discreteness, identity claims become judgment calls rather than empirical observations.

However, the comparison to biology’s own “species problem” is instructive and cuts against treating discreteness failure as fatal. Biology has operated productively for over a century with at least twenty competing species concepts (biological, morphological, phylogenetic, ecological), persistent boundary cases (ring species, hybridizing populations, horizontal gene transfer), and ongoing debates about whether species are “real” natural kinds or convenient abstractions. The gene concept itself was operationally useful for decades before molecular biology clarified its physical substrate, and even now “gene” carries multiple meanings depending on context. If evolutionary biology can function with fuzzy units, the demand that memetics produce crisper units than biology itself possesses appears disproportionate. The unit problem is genuinely awkward but not logically fatal — it parallels rather than exceeds the conceptual difficulties that biology has managed to work around through pragmatic operationalization rather than definitive resolution.

The fidelity problem and Eigen’s error catastrophe

The fidelity objection is perhaps the most technically precise critique of memetics. Manfred Eigen established in 1971 that for evolution by natural selection to maintain genetic information against entropic degradation, copying fidelity must exceed a critical threshold. The key relationship is Q = (1−μ)^L, where Q is the probability of error-free replication, μ is the per-element error rate, and L is the genome length (number of independently mutable elements). Evolution requires Q to exceed 1/a, where a is the selective advantage of the master sequence over random competitors.

Applied to memes, the arithmetic is alarming. If per-element copying accuracy for cultural transmission is approximately 70% — a generous estimate for behavioral imitation, given the substantial transformations observed in serial reproduction experiments — and a memeplex contains 10 semi-independent components, then the probability of faithful transmission is 0.7^10 ≈ 2.8%. This falls catastrophically below any plausible error threshold. At such infidelity, meme lineages should dissolve into noise within a handful of transmission events, precluding the cumulative cultural evolution that the memetic framework purports to explain.

Dan Sperber built his entire alternative framework — the Cultural Attraction Theory — on this problem. His solution reconceives transmission as fundamentally transformative rather than preservative. Cultural items “remain self-similar not because they are replicated again and again but because variations that occur at almost every turn in their repeated transmission tend to gravitate around cultural attractors.” Cognitive biases, ecological constraints, and social interaction patterns create basins of attraction in the space of possible cultural variants. A myth persists not because each retelling faithfully copies the previous version but because each teller’s cognitive biases push their version toward the same attractor point. Cinderella’s wicked stepmother recurs across retellings not through copying fidelity but because kin-selection psychology makes stepparent conflict a cognitive attractor. Norenzayan and colleagues found that the culturally successful Grimm’s fairy tales clustered at Boyer’s minimal counterintuitiveness optimum — again, not through faithful copying but through convergent cognitive processing.

The attractor solution is elegant but does not entirely resolve the tension. Gil-White countered that near-perfect copying fidelity is important in genetic systems but “is not the requirement for any and all Darwinian systems.” Henrich and Boyd showed formally that when multiple attractors exert competing influence, the dynamics involve contests between discrete alternatives, and “something close to particulate selection still happens.” Simulation work by Miśta demonstrated that stronger attractors can compensate for lower fidelity, and vice versa — the two forces are partially substitutable.

The most decisive response to the fidelity problem, however, comes from digital storage. Written text, printing, and digital files replicate with near-perfect fidelity (Q ≈ 1.0 for verbatim digital copying). The invention of writing did not merely extend memory; it fundamentally transformed the error regime of cultural transmission. As legal scholar Jack Balkin noted, “the next great advance in memetic fecundity, transmissibility, and longevity was the invention of external forms of information storage.” This creates a dual regime: high-fidelity digital transmission for externally stored memes (texts, images, code, recorded performances) alongside low-fidelity analogic transmission for behavioral and tacit-knowledge memes. The error catastrophe applies primarily to the oral-behavioral channel; digital storage effectively eliminates it for the textual-symbolic channel, and most of the memes that memetics claims to explain — religious doctrines, political ideologies, scientific theories, internet memes — are predominantly transmitted through the high-fidelity channel.

The Lamarckian problem: can directed evolution be Darwinian?

Cultural evolution is structurally Lamarckian: acquired characteristics are directly inherited. A scientist who improves an experimental protocol through deliberate reasoning transmits the improved version to students, who in turn improve it further. There is no Weismann barrier in culture — no separation between the soma that acquires modifications and the germ line that transmits them. Everything learned, invented, or refined can be directly passed on.

Biological Lamarckism failed empirically because molecular mechanisms prevent acquired somatic changes from being encoded in heritable DNA. Cultural Lamarckism succeeds precisely because the “acquired character” — a refined skill, an improved design — is what is transmitted. Kronfeldner distinguished carefully among three Lamarckian aspects: inheritance of acquired characteristics (present in culture), transformational evolution (partially present), and directed variation (strongly present). She concluded that culture exhibits both Darwinian and Lamarckian features, making it a Darwinian variational system with Lamarckian components, not a purely transformational one.

Boyd and Richerson incorporated Lamarckian dynamics through the concept of “guided variation”: individuals transform cultural variants in non-random (often adaptive) directions through learning and reasoning, then transmit these modified variants. Their formal models showed that guided variation is tractable within an evolutionary framework and that “Lamarckian effects are added easily to models, and the models remain evolutionary so long as rationality remains bounded.” The key insight is that intentional design functions as a variant-generation mechanism — analogous to mutation but strongly biased toward improvement rather than randomly distributed. The overall process remains Darwinian because intentionally generated variants still face selection (not all improvements succeed, not all intentions produce improvements), and cumulative change still depends on transmission and retention.

The deeper challenge is whether substantially directed cultural evolution needs the Darwinian framework at all. If most cultural change results from intelligent agents deliberately solving problems, then variation is not blind, selection is not environmental but intentional, and inheritance is not passive copying but active teaching. Dennett countered by noting that artificial selection in biology — where breeders intentionally direct variation — is universally treated as a special case of Darwinian selection, not as a refutation of it. The Darwinian framework does not require blind variation; it requires heritable variation subject to selection, and intentional cultural change satisfies this criterion. The Price Equation makes the relationship precise: the transmission bias term E(wΔz) captures directed cultural change, and the equation shows exactly how this term interacts with the selection term to produce overall evolutionary dynamics. Cultural evolution may be less “purely Darwinian” than genetic evolution — Nettle’s point — but it remains analyzable within the Darwinian framework as long as both terms are tracked.

The testability problem: retrodiction versus prediction

A framework that can explain anything predicts nothing. Bruce Edmonds charged that memetics is “merely a way of thinking rather than a scientific discipline,” and the editors of the Journal of Memetics acknowledged upon its closure in 2005 that the field had failed to produce sufficient empirical research. The circularity objection — that “fit” memes are defined post hoc as those that survived — parallels the “survival of the fittest” tautology critique in biology.

The comparison to evolutionary biology is clarifying. Biology also struggles with macroevolutionary prediction — no one predicted the Cambrian explosion, the K-Pg extinction, or the specific trajectory of hominin evolution. Much of paleontology is retrodictive, and “just-so stories” were a recognized problem long before memetics existed (Gould and Lewontin’s “Spandrels of San Marco”). Biology overcame these limitations through micro-level experimental testability (population genetics, laboratory selection experiments), quantitative predictions about allele frequency changes, and phylogenetic hypothesis testing. The question is whether memetics can achieve comparable empirical grounding.

Several areas demonstrate genuine progress. Epidemiological models of meme spread have been successfully fitted to internet propagation data, achieving excellent quantitative fits for content exhibiting initial spike-and-decay dynamics. Large-scale empirical studies of meme ecology on Reddit have tracked species dynamics — competition, speciation, extinction — finding patterns consistent with ecological theory. Inoculation theory, which directly operationalizes the epidemiological metaphor by exposing people to weakened forms of misinformation to build resistance, has produced robust and replicable experimental results. Viral content properties identified by Berger and others generate testable predictions about what content will spread. Hofhuis and Boudry made a specifically memetic prediction about witch-hunt dynamics — that witch hunts propagated because the witch-hunt meme was self-reinforcing rather than because it served host interests — and found supporting historical evidence. Meme lineage tracking using phylogenetic computational methods is now empirically feasible at scale.

The honest assessment is that memetics currently excels at retrodiction and generates relatively few novel predictions, but the predictive deficit is narrowing as computational and experimental methods improve. The field’s testability is roughly comparable to early evolutionary ecology — limited but growing — and the framework generates research programs that progressively improve empirical grounding rather than remaining permanently unfalsifiable.

The reductionism problem: can meme-level explanation reach structural causation?

Historical materialism challenges memetics from the left: Marx held that “it is not the consciousness of human beings that determines their existence, but their social existence that determines their consciousness.” Ideas, in this framework, are part of the ideological superstructure arising from and ultimately corresponding to the economic base. Memetics inverts this causal arrow, treating ideas as autonomous replicators driving cultural change. The Marxist rejoinder is that memetics attributes independent causal power to ideas that are actually determined by material conditions of production.

John Searle’s institutional critique challenges memetics from a different direction. Searle argued that institutions are constituted by collective intentionality — shared mental states assigning status functions through the formula “X counts as Y in context C.” Money exists as money because of collective recognition, not because a “money meme” outcompeted rival memes. Institutions create deontic powers (rights, duties, obligations) that cannot emerge from mere copying dynamics. The institutional fabric of social reality requires “we-intentionality” that is biologically primitive and irreducible to competition among replicators.

The emergence problem generalizes this concern. Do memeplexes have causal properties irreducible to individual meme dynamics? The grammatical system of a language cannot be decomposed into competing word-memes; it has emergent systemic properties that constrain individual memes and are not predictable from their aggregation. Legal systems, economies, and scientific paradigms exhibit similar emergent behavior. If memeplexes possess downward causation — constraining which individual memes can survive within them — then meme-level explanation is incomplete without memeplex-level analysis, and the reductionist program of explaining culture through individual meme dynamics fails.

Dennett distinguished “good reductionism” (acknowledging material causation at lower levels) from “greedy reductionism” (explaining away higher-level phenomena). Memetics, properly practiced, need not be greedy: it can acknowledge that memes operate within material constraints, institutional contexts, and emergent cultural structures while insisting that the information’s-eye view adds explanatory value that materialist and institutionalist frameworks miss. The persistence of demonstrably maladaptive cultural practices — witch hunts, conspiracy theories, costly rituals that impoverish their practitioners — is difficult to explain from either a materialist (how does this serve material interests?) or institutionalist (how does this constitute functional social reality?) perspective, but follows naturally from a meme’s-eye view where cultural items persist because they are good at propagating, not because they benefit their hosts.

The agency problem: selfish memes and moral responsibility

The deepest philosophical tension in memetics concerns agency itself. If memes are selfish replicators, and the self is, as Dennett put it, a “coalition of memes,” then moral responsibility appears undermined. If our beliefs, values, and choices are determined by which memes happen to have colonized our brains, the concept of autonomous agency dissolves into an illusion generated by the memeplex that constitutes the self.

The practical tension is acute. Counter-disinformation work — prebunking, media literacy, fact-checking — presupposes that agents can evaluate memes critically and resist manipulation. But if the strong memetic framework is correct, resistance is just another meme competing with the meme being resisted, and “critical thinking” is itself a memeplex that either wins or loses the competition without any genuine agent directing the process.

Dennett’s compatibilist response threads this needle with considerable sophistication. He argues that “it cannot be ‘memes versus us,’ because earlier infestations of memes have already played a major role in determining who or what we are.” The self is a coalition of memes; therefore, when memes direct behavior, we direct behavior — there is no pre-memetic self being puppeteered. Meaningful free will, for Dennett, consists not in supernatural escape from causal chains but in possessing the cognitive equipment to reason about alternatives, evaluate evidence, and respond to reasons. These capacities are themselves memetic accomplishments — products of cultural evolution that confer genuine (if not metaphysically ultimate) control. Keith Stanovich’s “Robot’s Rebellion” framework makes the practical implication explicit: understanding memes as potentially parasitic should motivate reflective evaluation of one’s own belief system, with the caveat that this evaluation must proceed from within the system being evaluated — a “Neurathian bootstrap” of rebuilding the ship while sailing it.

Boudry and Hofhuis navigate the tension by drawing an analogy with artificial selection in biology. Just as plant and animal breeders exercise genuine agency in directing selection while genes remain the units of selection, humans exercise genuine agency in cultural choice while memes remain the units being transmitted. The meme’s-eye view and the agent’s-eye view are complementary perspectives on the same process, not competitors for explanatory supremacy. Dismissing “panmemetics” — the worldview treating all culture as collections of selfish memes — they argue that the meme concept is most productive when applied selectively to cases where the replicator’s-eye view adds genuine explanatory value, rather than as a universal acid dissolving all cultural phenomena into replicator competition.

What memetics uniquely provides

After engaging the full battery of critiques, what remains? The unique and irreplaceable contribution of memetics is the meme’s-eye view — the perspective that cultural items can succeed because they are good at propagating, not because they benefit their human hosts. This perspective is not merely a restatement of cultural evolution; it is a specific claim that the interests of cultural replicators can diverge from the interests of their carriers, and that this divergence explains cultural phenomena that no host-benefit framework can account for.

Stewart-Williams estimated that perhaps 95% of cultural evolutionary science can proceed without invoking the meme’s-eye view. But for the remaining 5% — witch hunts, chain letters, viral conspiracy theories, pyramid schemes, self-destructive ideological commitments, religions that demand costly sacrifices — the meme’s-eye view provides explanations that neither rational-actor models, nor functionalist cultural theories, nor historical materialism can match. The concept has also generated practical interventions: the epidemiological metaphor directly inspired inoculation-based approaches to misinformation resistance that have produced replicable experimental successes. The meme concept survives the closure of the Journal of Memetics and the absence of dedicated academic departments because it captures something real — a level of explanation that exists whether or not the institutional infrastructure to study it has been adequately built. The information’s-eye view, stripped of overreach and properly constrained by its strongest critiques, remains an indispensable tool in the cultural evolutionist’s analytical repertoire.

---

SECTION 1: APPLIED MEMETICS

1a. Political memetics and information warfare

The memetic warfare toolbox — Firehose of falsehood. Christopher Paul and Miriam Matthews at RAND Corporation (2016, RAND Perspectives PE-198) codified the Russian propaganda model with four distinguishing features: (1) high volume and multichannel delivery (RT’s ~$300M annual budget, IRA troll farms producing up to 135 comments per shift per troll, bot networks); (2) rapid, continuous, and repetitive output exploiting first-mover advantage — first narratives anchor all subsequent processing; (3) no commitment to objective reality — freely mixing partial truths, complete fabrications, and staged events; (4) no commitment to consistency — spokespersons contradict each other without concern for “information fratricide.” The psychological mechanisms making the firehose effective map cleanly onto known cognitive biases: the illusory truth effect (repetition breeds perceived validity), source confusion (multiple channels create an illusion of independent corroboration), social proof exploitation (false impression of majority consensus), and cognitive overload (volume pushes audiences from System 2 analytical processing toward System 1 heuristic processing). Paul and Matthews concluded: “Don’t expect to counter the firehose of falsehood with the squirt gun of truth.”

IRA specifics (2016 campaign). The Oxford Internet Institute’s Computational Propaganda Research Project (Howard et al., 2018, commissioned for the Senate Select Committee on Intelligence) and New Knowledge’s analysis (DiResta et al., 2018) revealed the core mechanism: the IRA operated accounts on both sides of divisive issues simultaneously — organizing pro-police and anti-police-brutality events in the same city on the same day. The primary target was African Americans (~60% of racial messaging), using themes of police brutality and voting futility. Nelson (2025, Journal of Community & Applied Social Psychology) identified three argument classes micro-targeted at three distinct audiences: racial injustice narratives for Black Americans, liberal inequity narratives for white progressives, and white grievance narratives for conservatives. Scale: between 2013–2018, IRA campaigns reached tens of millions of US users; over 30 million shared IRA content between 2015–2017. The key performance metric was converting online engagement to offline action.

Astroturfing and conformist bias. Kovic et al. (2018, Studies in Communication Sciences) define astroturfing as “manufactured, deceptive and strategic top-down activity on the Internet initiated by political actors that mimics bottom-up activity by autonomous individuals.” The mechanism exploits Asch’s (1951) conformity finding: 75% of participants went along with obviously false consensus at least once. Chan (2024, Philosophy & Public Affairs) formalized the connection — astroturfing affects decision-making “just like any crowd action would” by triggering information cascades (Sunstein, 2019) and pluralistic ignorance. Documented operations: China’s “50 Cent Army” (paid commentators seeding pro-regime narratives as spontaneous sentiment); US Air Force 2010 “persona management” contract ($2.6M to Ntrepid) enabling single operators to manage multiple online personas.

Cambridge Analytica and psychographic targeting. The mechanism relied on Kosinski, Stillwell, and Graepel’s research (2013, PNAS) showing ~100 Facebook likes can estimate personality traits on the OCEAN/Big Five model. Data acquisition: Aleksandr Kogan’s app “thisisyourdigitallife” — 270,000 direct users; Facebook’s Friends API yielded profiles on up to 87 million. Cambridge Analytica developed dozens of ad variations on immigration, economy, and gun rights, each tailored to different personality profiles. Military psy-ops origins through parent company SCL Group — whistleblower Christopher Wylie called it “a psychological warfare mindfuck tool.” Validation: Matz et al. (2017, PNAS) demonstrated psychologically tailored advertising can be effective. Critical caveat: Nature (2018) noted “the scant science behind Cambridge Analytica’s controversial marketing techniques” — correlation between likes and personality traits was modest, and some researchers (Eitan Hersh) argue techniques were not fundamentally different from prior campaigns.

Dog whistles as dual-audience memetic structures. Henderson and McCready (2018, Springer; expanded into Signaling without Saying, Oxford UP) provide the formal framework: dog whistles “send one message to an outgroup while simultaneously sending a second (often taboo) message to an ingroup.” They identify two types — Type 1 operates through speaker persona transmission only (e.g., “cosmopolitan” signaling anti-Semitism); Type 2 carries truth-conditional impact, adding an “addendum” for a sub-audience (e.g., “inner city” → African American neighborhoods). The recovery mechanism: listeners recover a particular persona for the speaker — what kind of person would use this expression this way? Plausible deniability is the key feature: the same expression is deployable in two practices. Jennifer Saul (2018, Oxford UP) adds a crucial distinction between overt dog whistles (in-group recognizes as such) and covert dog whistles (raising attitudes to salience without audience awareness — functioning as perlocutionary speech acts succeeding only when the audience is unaware). Albertson (2015, Political Behavior) demonstrated experimentally that “multivocal appeals” resonate as religious with select audiences but have no religious content for others.

Wedge memes and coalition fractures. The IRA specifically operated both Black Lives Matter-adjacent pages AND Blue Lives Matter pages simultaneously. The goal was not to win for either side but to deepen the fracture itself. DiResta et al. (2018): the campaign was “designed to exploit societal fractures, blur the lines between reality and fiction, and erode trust in media entities and the information environment.” Nelson (2025) confirmed messages reinforced pre-existing beliefs and strengthened in-group solidarity on both sides through micro-targeted Facebook advertising.

NATO cognitive warfare doctrine. François du Cluzel (retired Lt.-Colonel, French Army; Head of Innovative Projects, NATO ACT Innovation Hub, Norfolk, Virginia) authored the primary document: “Cognitive Warfare” (November 2020, 45 pages, ACT-sponsored study). Core framing: “The brain will be the battlefield of the 21st century.” “Humans are the contested domain.” CW is proposed as the sixth operational domain alongside land, sea, air, space, and cyber. Du Cluzel defines cognitive warfare as “the art of using technologies to alter the cognition of human targets, most often without their knowledge and consent.” Critical distinction from information warfare: “Information warfare aims at controlling the flow of information… Cognitive warfare degrades the capacity to know, produce or thwart knowledge.” The target is trust itself: “the individual becomes the weapon, while the goal is not to attack what individuals think but rather the way they think.” The report explicitly uses the term “participatory propaganda.” It analyzes Russian CW under “Reflexive Control Doctrine” and China’s “Military Brain Science.” Follow-up: Claverie and du Cluzel (2022) published the formal concept definition. The first NATO Scientific Meeting on Cognitive Warfare occurred June 21, 2021, with proceedings published as Cognitive Warfare: The Future of Cognitive Dominance (2022, NATO CSO, ISBN: 978-92-837-2392-9).

QAnon as a participatory memeplex. Q “drops” were cryptic, fragmentary posts on 4chan (October 2017) and later 8chan/8kun — deliberately ambiguous, low-fidelity meme seeds requiring active reconstruction. Marwick and Partin (2024, New Media & Society) term this “populist expertise” — the systematic construction of alternative facts through collective research programs. Followers (“Bakers”) decoded drops through what Francesca Tripodi (2018) calls “scriptural inference” — close-reading techniques learned from Bible study remediated through social media. Ethan Zuckerman (2019, MIT Journal of Design and Science) characterized QAnon as a “big tent conspiracy theory” — narratives from flat earth, anti-vax, sovereign citizen, and wellness communities could all be reinterpreted to fit the QAnon grand narrative.

The mechanism by which participatory decoding increased rather than decreased engagement operates through multiple reinforcing pathways. The IKEA effect (Norton, Mochon, Ariely, 2012): when people invest cognitive labor in constructing meaning, they value resulting beliefs more highly than passively received content. ARG designer Reed Berkowitz (2020): “Finding something out, doing that independent research will give you a dopamine hit.” Lantian et al. (2017, Social Psychology) found “need for uniqueness” drives conspiracy belief — QAnon satisfies this by making followers feel like elite insiders. The collaborative process creates what Lave and Wenger (1991) would call a “community of practice.”

The gamification mechanic: Davies (2022, Acta Ludologica) provides a comprehensive ludological analysis finding QAnon shares core mechanics with ARGs — puzzle-solving structure (“bread crumbs”), rabbit-hole entry points, and the “This Is Not A Game” (TINAG) principle taken literally. Adrian Hon (ARG designer, author of You’ve Been Played) recognized QAnon as behaving precisely like an alternate reality game. De Zeeuw and Gekker (2023, Social Media + Society) introduce “conspiracy fictioning” — the recursive process by which conspiracies bootstrap their own alternate realities. The “Great Awakening” at the center of the QMap provides a messianic/eschatological framework giving participants a proselytizing directive (Conner, 2023, Frontiers in Sociology).

Counter-memetic operations and why they fail. The original Nyhan and Reifler (2010, Political Behavior) “backfire effect” — corrections strengthening false beliefs — became a widely cited rationale for why fact-checking fails. However, this effect has not survived replication. Wood and Porter (2019, Political Behavior) tested 52 issues across 5 experiments with N > 10,100 and found “no corrections capable of triggering backfire.” Nyhan himself (2021, PNAS) wrote that interpretations “quickly outstripped the findings” of his 2010 paper. The four researchers (Nyhan, Porter, Reifler, Wood) jointly concluded the backfire effect does not explain the durability of misperceptions. The “familiarity backfire” (repeating a myth to debunk it increases familiarity and perceived truth) has also been challenged — Swire-Thompson, DeGutis, and Lazer (2020) conclude it has “little to no empirical support.”

What is robust: Lewandowsky’s continued influence effect (CIE) — corrected misinformation continues influencing judgments even after the correction is acknowledged and accepted. Classic paradigm (Johnson and Seifert, 1994; Wilkes and Leatherbarrow, 1988): participants read unfolding reports of a warehouse fire, initial reports cite flammable chemicals, a retraction states the closet was empty — yet participants still make inferences consistent with the discredited chemicals claim even when they remember and accept the retraction. Mechanisms: the retraction leaves a causal gap in the mental model; the misinformation remains automatically activated from memory; the “negation tag” may fail to be retrieved under cognitive load. Walter and Tukachinsky’s (2019) meta-analysis confirmed corrections reduce but do not fully eliminate belief in misinformation.

What actually works: inoculation (detailed below), identity-preserving correction (Kahan’s “identity-protective cognition” — framing corrections to be consonant with target group values; Feinberg and Willer 2015 on moral reframing), providing alternative causal explanations (Lewandowsky et al. 2012 — “X is false, and actually Y” far more effective than “X is false”), and personalized AI dialogue. Costello, Pennycook, and Rand (2024, Science) achieved ~20% reduction in conspiracy belief strength using GPT-4 Turbo (“DebunkBot”) engaged in personalized, evidence-based dialogue with N = 2,190 participants — effects persisted at 2-month follow-up, worked even for identity-central beliefs, and shifted behavioral intentions. This suggests previous failures to correct conspiracy beliefs stemmed from counterevidence being insufficiently compelling and personalized, not from any inherent imperviousness.

1b. Public health memetics

Vaccine hesitancy as a memeplex with immunizing architecture. The vaccine hesitancy memeplex contains specific memes that immunize against medical authority: the “natural is better” meme (exploiting a naturalistic bias deeply rooted in evolved food-safety heuristics), the “pharmaceutical profit motive” meme (grafting distrust of corporate motives onto the medical establishment), and the “herd immunity is enough” meme (a free-rider strategy — “enough other people are vaccinating, so I don’t need to”). These memes form a co-adapted complex: the naturalistic meme provides the emotional core, the profit-motive meme provides the conspiracy narrative, and the herd-immunity meme provides the rational-sounding excuse. Each meme reinforces the others, and the memeplex includes anti-defection mechanisms: anyone who vaccinate despite belonging to the hesitant community faces social sanction framed as betrayal.

The WHO’s infodemic concept. The WHO’s framework distinguishes an infodemic from ordinary misinformation by scale and speed: an infodemic is “too much information including false or misleading information in digital and physical environments during a disease outbreak,” overwhelming information-processing capacity and creating confusion that undermines health responses. What distinguishes it: volume exceeds processing capacity at a population level, false information actively competes with health guidance in the same channels, and the speed of spread outpaces official verification timelines.

COVID-19 mask meme competition. “Masks don’t work” vs. “masks protect others” competed through asymmetric fitness properties. “Masks don’t work” was simpler (lower cognitive load), aligned with personal convenience (lower behavioral cost to the host), and exploited the naturalistic bias (breathing through fabric feels unnatural). “Masks protect others” required altruistic framing (higher cognitive complexity), demanded behavioral change (higher cost), but gained fitness through institutional backing and prosocial identity signaling. The competition illustrates how meme fitness is not intrinsic but context-dependent — during high-fear phases, protective memes gained fitness through fear-reduction; during fatigue phases, anti-mask memes gained fitness through offering behavioral relief.

Inoculation theory: from McGuire to prebunking at YouTube scale. William McGuire (1961) drew a direct analogy from medical vaccination: attitudes can be made resistant to persuasion through preexposure to weakened counterarguments. Motivated by Korean War defections — nine US POWs elected to remain with captors, suggesting soldiers had no “mental defenses” against ideological attack. McGuire and Papageorgis (1961, Journal of Abnormal and Social Psychology) tested cultural truisms (e.g., value of brushing teeth): participants pre-exposed to weakened counterarguments plus refutations showed 20–30% more resistance than controls. Two core components: threat (recognition that one’s position is vulnerable) and refutational preemption (exposure to weakened counterarguments + their refutation, triggering counterarguing). Banas and Rains (2010) meta-analysis: effect size of approximately g ≈ 0.41 for attitude resistance across dozens of studies.

The paradigm shift came from Roozenbeek and van der Linden: shifting from “refutational-same” (inoculating against the specific counterargument) to “refutational-different” — exposing people to the techniques underlying misinformation production rather than specific false claims. This “broad-spectrum” inoculation sidesteps the arbiter-of-truth problem, works across topics and political ideologies, and doesn’t require predicting which misinformation will emerge. The DROG “Bad News” game (English version launched February 2018, developed with Dutch media collective DROG): a ~15-minute browser game where players take the role of a fake news producer, earning badges for six manipulation techniques — impersonation, emotional language, polarization, conspiracy theories, discrediting opponents, trolling (based on academic literature plus NATO StratCom’s “Digital Hydra” report). Initial study (Roozenbeek and van der Linden, 2019, Humanities and Social Sciences Communications): N ≈ 15,000, significant reductions in perceived reliability of manipulative posts. Rigorous replication (Basol et al., 2020, Journal of Cognition): N = 196, reduced perceived reliability by 21% vs. Tetris control.

The landmark scaling study: Roozenbeek, van der Linden, Goldberg, Rathje, and Lewandowsky (2022, “Psychological inoculation improves resilience against misinformation on social media,” Science Advances, 8(34), eabo6254). Phase 1: six preregistered RCTs, total N = 6,464, testing 90-second animated inoculation videos targeting five techniques (emotionally manipulative language, incoherence, false dichotomies, scapegoating, ad hominem). Phase 2: partnership with Google’s Jigsaw unit — ~5.4 million US YouTubers exposed to inoculation videos in pre-roll ad slots; almost 1 million watched for 30+ seconds; 22,632 completed a voluntary test question within 24 hours. Result: 5–10% average boost in correct manipulation-technique recognition vs. control. Average test participation ~18 hours post-viewing, suggesting durability. Cost: ~US$0.05 per view. The videos make no claims about what is true or false — they are “effective for anyone who does not appreciate being manipulated.” Google subsequently rolled out prebunking to hundreds of millions of people, with campaigns in Poland, Slovakia, and Czech Republic (September 2022) targeting anti-refugee disinformation related to the Ukraine war. John Cook (creator of the Cranky Uncle game) is a contributor to the broader “inoculation science” research program at van der Linden’s Cambridge Social Decision-Making Lab (SDML).

1c. Religion as memeplex — the academic analysis

Harvey Whitehouse’s two modes of religiosity. Whitehouse developed the Divergent Modes of Religiosity (DMR) theory from ethnographic fieldwork among the Pomio Kivung movement in Papua New Guinea, proposing that religions coalesce around two cognitive-mnemonic “attractor positions” — not two types of religion but two organizing principles that can coexist within a single tradition.

The doctrinal mode: high-frequency transmission (daily prayers, weekly services, regular recitations), low arousal (routinized, generating little affect), exploiting semantic memory (repetitive experiences stored as general schemas, like learning multiplication tables), transmitted by centralized priestly authority, creating large-scale categorical bonds with anonymous co-religionists who share identity markers. Orthodoxy is maintained through repetition; innovation must originate from authoritative sources. Danger: the “tedium effect” — frequent repetition leads to boredom and mindless repetition. Examples: Holy Communion, Islamic call to prayer, mainstream institutional Christianity and Islam. Historical role: enabled the transition from small-scale to large-scale societies.

The imagistic mode: low-frequency transmission (once-in-a-lifetime initiations, periodic rites of passage), high arousal (painful initiations, traumatic experiences, ecstatic states), exploiting episodic memory (vivid, autobiographical flashbulb-like memories — emotional intensity “illuminates the scene and preserves it forever”), generating Spontaneous Exegetical Reflection (SER) where participants ruminate extensively on meaning, producing a sense of multivalence. Creates small, exclusive groups with intense relational bonds and identity fusion — an extreme form of cohesion where personal and group identity merge, motivating extreme altruism and self-sacrifice when the group is threatened. Examples: Melanesian initiation rites, hazing rituals, ecstatic ceremonies.

These map to two distinct memetic transmission strategies: doctrinal = fidelity through repetition (like a photocopier running many copies — each impression weak but cumulative exposure reliable); imagistic = commitment through emotional intensity (like a single photograph with flash — one exposure burns permanently). Whitehouse explicitly frames this as a selectionist model: traditions with frequently repeated rituals see doctrinal-mode features enjoy selective advantage and coalesce; traditions with rare high-arousal rituals see imagistic features coalesce. Key publications: Inside the Cult (1995), Arguments and Icons (2000, Oxford UP), Modes of Religiosity (2004, AltaMira). Empirically tested against 645 rituals from the HRAF (Human Relations Area Files) global ethnographic database (Atkinson and Whitehouse, 2011), confirming the anticipated clustering. Extended beyond religion to football fans and armed militias; Whitehouse and Lanman (2014, Current Anthropology) connected the theory to identity fusion.

Pascal Boyer’s minimally counterintuitive agents. Boyer (Religion Explained, 2001) argues religious concepts are not arbitrary cultural inventions but predictable outputs of ordinary cognitive processes. His framework relies on intuitive ontological categories — mental templates for reasoning about persons, animals, plants, natural objects, and artifacts. A minimally counterintuitive (MCI) concept violates exactly one or two intuitive expectations while preserving most others: an invisible person violates physics but preserves psychology and biology; an omniscient being preserves agency but violates normal limits on knowledge; a talking tree violates the expectation that plants lack psychology. MCI concepts occupy a cognitive optimum: surprising enough to be memorable (the violation flags them as noteworthy, triggering attention and deeper processing) but not so bizarre as to be unprocessable (preserving most expectations means they remain inferentially rich — you can reason about what an invisible person wants, thinks, and feels). Barrett and Nyhof (2001) demonstrated MCI concepts are better remembered and more faithfully transmitted in serial recall tasks; Boyer and Ramble (2001) provided cross-cultural evidence.

Justin Barrett (Why Would Anyone Believe in God?, 2004) contributed the Hyperactive Agency Detection Device (HADD): an evolved cognitive module that systematically errs toward false positives (detecting agents that aren’t there) because the ancestral cost of failing to detect a predator (death) far exceeded the cost of false detection (momentary fear). HADD is triggered by non-inertial movement, unexpected sounds, and patterns suggesting design. When HADD fires and no visible agent is found, the mind generates the concept of an invisible agent — a spirit, ghost, or god. Cultural frameworks then provide names and stories. Building on Stewart Guthrie’s Faces in the Clouds (1993). Note: empirical evidence for HADD is limited — Van Leeuwen and Van Elk (2019) found no compelling evidence for a positive bias in agency detection; recent work (2025, Religion, Brain & Behavior) proposes replacing HADD with a motivation-based theory. The connection: HADD generates the raw intuition; MCI theory explains why resulting concepts have the specific form they do and why they persist in transmission. Gods, spirits, and ancestors exploit pre-existing cognitive architecture, giving religious memes a built-in transmission advantage.

Ara Norenzayan’s “Big Gods” hypothesis. Big Gods: How Religion Transformed Cooperation and Conflict (2013, Princeton UP). Core hypothesis: prosocial “Big Gods” — omniscient, morally concerned, punishing deities — solved the free-rider problem at scale. In small bands (~150 people), face-to-face monitoring sufficed; as societies grew larger and more anonymous, free-riders could exploit cooperation undetected. Belief in an all-seeing supernatural monitor who punishes transgressions enabled trust among strangers at previously impossible scales. Key principles: “Watched people are nice people” (supernatural monitoring increases prosocial behavior); “Hell is stronger than heaven” (Shariff and Norenzayan, 2011: mean-God belief reduces cheating, nice-God belief does not); Big Gods for Big Groups (Roes and Raymond, 2003: correlation between society size and belief in moralizing gods across 186 cultures). Evidence: God-priming studies (Shariff and Norenzayan, 2007) showing priming God concepts increases prosocial behavior in anonymous economic games even among nonbelievers; Henrich et al. (2010): 15-society cross-cultural study showing world religion participation predicts prosocial behavior.

Criticisms: Whitehouse et al. (2019, Nature) used the Seshat Global History Databank to claim complex societies precede Big Gods — but the paper was retracted in 2021 due to problems with missing-data treatment. Whitehouse et al. (2022, Religion, Brain & Behavior) published a corrected analysis maintaining the same conclusion — warfare intensity and agricultural productivity, not Big Gods, were the primary drivers. A special issue of Religion (Stausberg, 2014) criticized the hypothesis for neglecting that early complex civilizations (Mesopotamia, Egypt, China, Mesoamerica) had powerful gods who were not consistently morally concerned. Peter Turchin supports a feedback loop: a strong causal arrow from social scale to Big Gods, with weaker feedback from Big Gods to scale, but only once societies exceed a complexity threshold.

Why secular counter-memeplexes struggle. Religious memeplexes possess interlocking, mutually-reinforcing features creating extraordinary transmission advantages. Specific fitness gaps: (1) the afterlife meme offers literal personal immortality — “asserting there is nothing beyond the grave has never been a great selling point”; (2) the divine purpose meme provides cosmic meaning, while secular alternatives require self-constructed meaning, which is cognitively demanding; (3) supernatural monitoring cannot be matched by secular institutions (police, courts) because they lack perceived omniscience; (4) ritual reinforcement structures time through liturgical calendars — secular equivalents lack compulsory rhythm; (5) anti-defection mechanisms (heresy, apostasy, framing doubt as moral failure) protect the memeplex — secular movements, valuing autonomy and critical thinking, actively undermine their own protective mechanisms; (6) exploitation of cognitive biases — religious memes exploit HADD and MCI, while secular memes often work against these biases.

The Sunday Assembly movement (founded January 2013, London, by Sanderson Jones and Pippa Evans — “all the best bits of church, but with no religion”) peaked at ~70 congregations worldwide, declined to ~22 by 2023 — predating COVID-19. Overwhelmingly middle-class, urban, white demographic. Earlier precedents: Auguste Comte’s Church of Positivism (19th century) — only one active congregation remains (Brazil). Alain de Botton (Religion for Atheists, 2012) argued religions are “giant machines for making ideas vivid and real in people’s lives” and that secular society has foolishly discarded practical wisdom about community, ritual, and consolation along with theology.

1d. Conspiracy theories as memetic ecosystems

Epistemic closure as self-reinforcing system. Conspiracy theories are unfalsifiable not by accident but because unfalsifiability dramatically increases memetic fitness — no evidence can exit the system. Disconfirming evidence is reinterpreted as further confirmation (it was planted, it’s what “they” want you to think). Zuckerman (2019) noted QAnon is “self-sealing” — any disproof becomes just another puzzle to solve, generating more engagement. This is not a bug but an evolved feature: memeplexes with this architecture systematically outcompete falsifiable alternatives because they cannot be dislodged by counter-evidence.

Cognitive attractors. Proportionality bias: large events must have large causes (a president’s assassination cannot be the work of a lone gunman — the effect is too large for such a small cause). HADD applied to complex social events: agency detection overreads intentional action behind structural or stochastic phenomena (“someone must be behind this”). Apophenia: pattern recognition in noise — danah boyd (2019) identified this as central to QAnon’s spread: “the social mechanisms of conspiratorial thinking are rooted in reality. It’s the pattern that’s non-existent.”

QAnon’s specific innovation. Crowdsourced sense-making — the “Great Awakening” memeplex structure where initiates construct the conspiracy themselves. Baća (2024, Current Sociology) terms these “epistemic communities of the unreal” — bottom-up, horizontal, collective meaning-making. De Zeeuw and Gekker (2023) analyze QAnon as “conspiracy fictioning” — recursive self-bootstrapping of alternate realities, reviving the concept of “hyperstition” (fictions that make themselves true). Conner (2023, Frontiers in Sociology): 1,000+ hours of ethnographic observation confirmed QAnon absorbed flat earth, Alex Jones, anti-vax, sovereign citizens, New Age wellness, and neo-shamanistic content — three unexpected cultural entry points were yoga/wellness groups, neo-shamanistic circles, and psychics.

Why fact-checking fails. Lewandowsky’s continued influence effect (Lewandowsky et al., 2012, Psychological Science in the Public Interest, 13(3), 106–131): corrected misinformation continues influencing judgments even after correction is acknowledged. Multiple mechanisms: retracting misinformation leaves a causal gap in the mental model; without an alternative explanation, people fall back on discredited information to maintain coherent narratives; the “negation tag” may fail to be retrieved, especially under cognitive load; misinformation remains fluently processed, which is misattributed to validity. Fact-checking also faces the reach problem — misinformation goes viral; corrections rarely do (Vosoughi et al., 2018, Science). And fact-checking is always reactive — running behind the curve.

What actually works. Inoculation (technique-based, as above). Identity-preserving correction (Feinberg and Willer, 2015: framing environmental protection in purity/sanctity terms for conservatives). Providing alternative causal accounts (Lewandowsky et al., 2012: “X is false and actually Y” is far more effective than “X is false”). The Costello et al. (2024, Science) AI dialogue study — personalized, empathic, evidence-based conversation, sharing key principles with motivational interviewing (non-judgmental engagement, exploring the person’s own reasons, tailored counterarguments, rapport-building before challenging). Effect sizes: rational counterarguments alone produce d = 0.13–0.29 (small); inoculation meta-analytic effect g ≈ 0.41; AI personalized debunking ~20% belief reduction durable at 2 months.

---

SECTION 2: MEMETICS AND THE SUBCONSCIOUS

2a. Below the threshold of awareness

Bourdieu’s habitus. Bourdieu (2000, pp. 86–87) defines habitus as “a system of lasting, transposable dispositions which, integrating past experiences, functions, at every moment as a matrix of perceptions, appreciations, and actions.” Formation occurs through early childhood socialization — repeated performance of cognitive, affective, and bodily repertoires — before critical faculty is operational. “The principles embodied in this way are placed beyond the grasp of consciousness, and hence cannot be touched by voluntary, deliberate transformation, cannot even be made explicit” (Outline of a Theory of Practice, 1977). Social conditions are inscribed in the body through what Bourdieu calls “the hidden persuasion of an implicit pedagogy, capable of instilling a whole cosmology, an ethic, a metaphysic, a political philosophy, through injunctions as insignificant as ‘stand up straight’ or ‘don’t hold your knife in your left hand.’”

Habitus operates as pre-reflexive “practical sense” — “a durably installed generative principle of regulated improvisations” (1978, p. 78). It generates dispositions that feel natural and chosen but are socially structured. It produces doxa: situations in which “the natural and social world appears as self-evident.” Crucially, habitus is “purposeful without being questionable; it is transmitted but not actively taught.”

Bodily hexis: “political mythology realised, embodied, turned into a permanent disposition, a durable manner of standing, speaking and thereby of feeling and thinking” (1977). Drawing on Marcel Mauss’s “techniques du corps” and Merleau-Ponty’s phenomenology, Bourdieu shows social structures encoded in posture, gait, deportment, accent — forms of bodily movement “ingrained, difficult to consciously alter, that literally reveal our background.” This challenges simple memetic agency models: if memes are transmitted through habituated body-schemas formed before critical faculty, the “host” has no choice about reception — there is no moment of evaluation or selection. The habitus “operates beneath the level of rational ideology.” Key works: Outline of a Theory of Practice (1977), Distinction (1979/1984), The Logic of Practice (1990).

Embodied cognition and tacit knowledge memes. Memes encoded in gesture, posture, rhythm, spatial organization, and kinesthetic practice constitute “tacit knowledge” memes (Polanyi) that resist verbalization and explicit transmission. Bourdieu’s “practical taxonomies” that structure perception “are rooted in and only make sense from the point of view of the body.” These memes are transmitted not through linguistic instruction but through bodily co-presence, imitation, and repeated practice — through what Lave and Wenger call “legitimate peripheral participation.” They represent a category of meme transmission entirely invisible to the standard memetic model of explicit, propositional cultural transmission.

Priming and conceptual activation. Bargh, Chen, and Burrows (1996) reported that students primed with elderly-related words (Florida, forgetful, bald, gray, wrinkle) in a sentence-unscrambling task walked more slowly down a hallway afterward. Kahneman featured this as the “Florida effect” in Thinking, Fast and Slow (2011). However, Doyen et al. (2012, PLoS ONE) failed to replicate using objective laser measurement — no walking-speed difference — and demonstrated the effect was likely an experimenter expectancy effect (experimenters told to expect slower walking got slower walking). Kahneman effectively retracted his endorsement by 2017; Bargh never conducted new replications. Crucial distinction: cognitive/semantic priming (recognizing “doctor” faster after “nurse” — short-duration, attention-level effects) is well-established and replicable. Social/behavioral priming (dramatic unconscious behavioral effects from subtle conceptual manipulations) has largely failed replication. Z-curve analysis of Bargh’s most-cited articles suggests extensive reliance on questionable research practices. For the primer: the general mechanism — low-level meme exposure altering subsequent cognition without awareness — is real and has robust neural underpinnings (semantic priming is solid), but the dramatic behavioral effects dominating popular accounts were overstated.

2b. Archetypal memes and depth psychology

Jung’s archetypes through a memetic lens. The reinterpretation: archetypes as ancient meme lineages sufficiently old that gene-culture coevolution has partially neurologically instantiated them, creating the “felt” interior quality of archetypal experience. The Hero, Shadow, Anima/Animus as cross-cultural convergent forms — not mystical entities but meme lineages so ancient and so consistently selected for that they have become partially hardwired, creating a cognitive landscape with built-in attractor basins. This would explain why archetypal experiences feel qualitatively different from merely encountering a new idea — they activate pre-structured neural pathways shaped by millennia of gene-culture coevolution.

Freudian concepts reinterpreted memetically. Repression as meme suppression: culturally dissonant memes (sexual, aggressive, socially unacceptable) are driven below conscious articulation by memetic immune mechanisms of the dominant cultural memeplex. Defense mechanisms as memetic immunity strategies: denial, projection, rationalization function to protect the core memeplex (selfplex) from threatening memes by distorting or deflecting them. Transference as activation of archaic memeplexes: in the therapeutic relationship, early-acquired relational memes (from childhood attachment figures) are activated and projected onto the therapist — the analyst as screen for archaic meme-complexes.

The non-Jungian alternative. Sperber’s cognitive attractor explanation: universal myths emerge not through transmitted lineages but through convergent reconstruction by minds with similar cognitive architecture. Just as Boyer’s MCI theory predicts similar religious concepts will independently emerge across cultures (because similar cognitive biases produce similar outputs), the hero myth, the trickster, the flood narrative may reflect convergent attractors in cognitive space rather than vertically transmitted meme lineages. This is the key distinction: lineage versus convergence. Jung says the archetype IS the lineage; Sperber says the lineage is illusory — it’s the same cognitive architecture independently generating similar outputs.

2c. Advertising and applied meme engineering

Edward Bernays. Freud’s double nephew (mother was Freud’s sister; father was brother of Freud’s wife), who explicitly applied psychoanalytic meme logic to mass persuasion. Published Crystallizing Public Opinion (1923) and Propaganda (1928), arguing “the conscious and intelligent manipulation of the organized habits and opinions of the masses is an important element in democratic society.” Drew on Gustave Le Bon’s mass psychology, Wilfred Trotter’s herd instinct, and Freud’s theories of unconscious drives. Called his technique the “engineering of consent.”

The Torches of Freedom campaign (1929): client George W. Hill (American Tobacco Company) wanted to break the taboo on women smoking publicly. Bernays consulted psychoanalyst A.A. Brill, who stated: “Cigarettes, which are equated with men, become torches of freedom.” On March 31, 1929 (Easter Sunday), Bernays had secretary Bertha Hunt pose as a women’s rights advocate, sending telegrams to debutantes: “In the interests of equality of the sexes… I and other young women will light another torch of freedom by smoking cigarettes while strolling on Fifth Avenue.” 6–12 women walked Fifth Avenue smoking; press photographers were pre-tipped; neither Bernays nor American Tobacco were mentioned. The New York Times front page April 1: “Group of Girls Puff at Cigarettes as a Gesture of ‘Freedom.’” Coverage went nationwide.

Memetic engineering analysis: precise meme grafting — attaching a new meme (smoking = freedom) to an existing high-fitness memeplex (women’s liberation). By associating cigarettes with a powerful identity narrative, Bernays bypassed rational product evaluation. The mechanisms: appealing to unconscious desires, activating herd instinct via fashionable leader-figures, creating a pseudo-event generating free media amplification, and framing commercial interest as political protest. Critical note: Murphree (2015, American Journalism) provides evidence the campaign’s impact was partly a Bernays-driven myth — women had been smoking publicly for years; the broader market trend (5% of cigarettes purchased by women in 1923, 12% in 1929, 18.1% in 1935) cannot be attributed solely to one campaign.

Sonic branding and earworms. Beaman and Williams (2010, British Journal of Psychology) published the foundational study on involuntary musical imagery (INMI) — “earworms” as a form of intrusive thought reflecting “a systematic failure in mental control.” Over 90% of people experience INMI at least weekly (Liikkanen, 2012, N = 11,910). Mean episode duration: ~8 minutes; episodes most frequent during low cognitive-demand states. Jakubowski et al. (2017, Psychology of Aesthetics, Creativity, and the Arts) found INMI tunes share specific melodic features: common global contour patterns, smaller pitch intervals, and unusual interval patterns. Why music is a high-fidelity, low-effort meme vehicle: melodic contour is easily encoded, rhythm provides temporal scaffolding, repetition is built into musical structure (verse-chorus form), lyrics provide semantic hooks. The Zeigarnik effect applied to musical incompletion (unfinished tasks remembered better than completed ones) is theoretically plausible but has received mixed empirical support — McCullough Campbell and Margulis (2015) and Killingly et al. (2021) found no significant truncation effect in controlled experiments.

Mere exposure effect. Zajonc (1968, Journal of Personality and Social Psychology): repeated exposure increases preference even without conscious recognition — demonstrated across nonsense words, Chinese-style characters, geometric shapes, and photographs, including subliminal presentations. Neural basis: fMRI shows reduced amygdala activation for familiarized vs. novel stimuli (Zebrowitz and Zhang, 2012, Social Neuroscience) — familiarity reduces threat assessment, which increases preference. Also increased reward-circuit activity (ventral striatum, VTA). Mechanism: perceptual fluency — familiar stimuli are easier to process, and processing ease is misattributed to liking (Reber, Winkielman, and Schwarz, 1998).

IKEA effect. Norton, Mochon, and Ariely (2012, Journal of Consumer Psychology): labor increases valuation of self-made products — participants willing to pay 63% more for self-assembled furniture than equivalent pre-assembled items. Critically, labor leads to love “only when labor results in successful completion.” Connection to meme engineering: participatory construction increases memetic commitment — the same mechanism explains QAnon’s crowdsourced decoding (active cognitive labor in interpreting Q drops increases perceived value of resulting beliefs), Wikipedia’s editor commitment, and open-source community loyalty.

---

SECTION 3: MEMETICS AND SEMIOTICS

Saussure: signifier/signified. The Saussurean sign (signifier + signified) maps onto memetic transmission: the meme is the transmission unit of the sign — what actually moves between minds. But Saussure emphasized that signs derive meaning from systemic relations to other signs (structural difference), not intrinsic properties — predicting that meme fitness depends not just on internal structure but on relational position within a sign system. The arbitrary but conventionalized signifier-signified bond explains why memes can be recoded (same concept, different form) and why form matters for fitness (some signifiers are more transmissible than others for the same signified).

Peirce’s icon/index/symbol taxonomy and cross-cultural meme fitness. Icons (resembling their referent — photographs, mimicked gestures, recognizable cartoons) should have the highest cross-cultural fitness since they require no prior cultural knowledge. Indexes (associated through causal/real connection — smoke/fire, pointing finger) travel moderately well where shared causal knowledge exists but are context-dependent, with lower decontextualized fitness. Symbols (related by convention — words, traffic signals) have the highest within-community fitness but lowest cross-community fitness, since they require shared cultural codes. Example: a thumbs-up gesture is symbolic with culturally variable meaning (positive in the West, offensive in parts of the Middle East and West Africa), while a smile icon has near-universal fitness (iconic, exploiting universal facial-expression recognition).

Barthes’ mythology as memeplex naturalization. Roland Barthes (Mythologies, 1957): myth operates as a second-order semiological system — “that which is a sign in the first system becomes a mere signifier in the second.” The function: myth takes historical contingency and presents it as natural necessity — “making contingency appear eternal” and “giving a historical intention a natural justification.” Myth “empties reality and presents itself as depoliticised speech.” This is the most fitness-enhancing meta-meme because a meme that appears “natural” is virtually immune to critical examination — it has been removed from the competition space by making alternatives unthinkable. “This is just how things are” naturalization prevents the meme from even being recognized as a meme.

Eco’s unlimited semiosis as memetic mutation mechanism. Eco (A Theory of Semiotics, 1976), building on Peirce: the interpretation of one sign leads to more signs in an endless chain. Each interpretant itself functions as a further sign, generating another interpretant ad infinitum. For memetics: each interpretive act potentially generates a variant — aberrant decodings with higher fitness than “intended” readings are the seeds of meme speciation. A joke misunderstood becomes a new joke; a political slogan reinterpreted in an unintended context gains new life. Eco later constrained this (1990), distinguishing Peirce’s growth-of-signs from “anything goes” overinterpretation — paralleling the idea that not all memetic mutations survive selection.

Intertextuality as meme recombination. Kristeva (late 1960s), building on Bakhtin’s dialogism: “any text is a mosaic of quotations; any text is the absorption and transformation of another.” Bakhtin: language is inherently “double-voiced,” participating in ongoing dialogue across texts. Kristeva adds “transposition” — employing pre-existent signifying practices for different purposes. Allusion, parody, homage, remix function as the sexual-reproduction analog in culture — meme recombination where fragments of existing cultural products are combined and transformed, generating novel variants with potentially higher fitness than any parent form. This is the mechanism behind remix culture, where each iteration carries traces of all previous versions while introducing variation.

---

SECTION 4: FRONTIER RESEARCH

4a. Computational memetics

Leskovec, Backstrom, and Kleinberg’s MemeTracker (2009). “Meme-tracking and the Dynamics of the News Cycle,” KDD 2009 (Cornell/Stanford). Dataset: ~900,000 news stories and blog posts per day from ~1 million online sources during a three-month period covering the 2008 US presidential election, tracking over 17 million different phrases. Methodology: scalable algorithms clustering textual variants of short quoted phrases using directed phrase graphs based on approximate inclusion and overlap, weighted by edit distance. Key finding: a typical lag of 2.5 hours between peak attention to a phrase in mainstream news and subsequent peak in blogs. The framework enabled tracking how phrases mutate as they propagate — an observable analogue to genetic mutation. The MemeTracker DAG approach using edit distance between phrase variants is essentially the same mathematical operation as sequence alignment in genomics — making “meme phylogenetics” a natural extension. The analogy to purifying selection (high-fitness genetic sequences experiencing stronger selection pressure against mutation) is theoretically compelling and enabled by the framework, though the primary published findings focused on temporal dynamics.

Weng, Flammini, Vespignani, and Menczer (2012). “Competition among memes in a world with limited attention,” Scientific Reports, 2, 335 (Indiana University/Northeastern/Harvard). Agent-based model where agents share messages on a social network but can only attend to a portion of information received. Key finding: massive heterogeneity in meme popularity — a few go viral while most don’t — can be explained by competition for limited attention combined with social network structure, without assuming different intrinsic values among ideas. Popularity is largely a stochastic product of attentional competition and network topology. Follow-up (Weng et al., 2013, Scientific Reports): most memes spread as complex contagions (requiring social reinforcement, trapped by community boundaries), but rare viral memes spread across communities as simple contagions. Early community concentration predicts future virality.

Ferrara et al. (2013, IEEE/ACM ASONAM): algorithms for identifying, clustering, and tracking memes in Twitter streams, distinguishing genuine diffusion from bot-driven manipulation. Part of NSF-funded project “Meme Diffusion Through Mass Social Media.” Also work on sentiment’s effect on information diffusion (PeerJ Computer Science, 2015) and computational fact-checking.

Machine learning predictive features. Network structure features (community concentration, seed node centrality, early cascade breadth), temporal features (early growth rate, time-to-peak), content features (emotional valence, sentiment, self-relevance, social utility), and social features (influencer involvement, cross-community penetration). Goel, Anderson, Hofman, and Watts (2016, Management Science) analyzed a billion diffusion events on Twitter: popularity is largely driven by broadcast size, not viral chain length — “structural virality” is typically low.

Meme phylogenetics. Biological phylogenetic methods have been productively applied to cultural data: Tehrani and Collard (2007, Journal of Anthropological Archaeology) traced evolution of Turkmen textiles; Gray, Drummond, and Greenhill (2009) applied Bayesian phylogenetic methods to Austronesian language evolution; Greenhill (2015, PNAS) applied weighted string alignment from genomics to track phonetic and lexical change across ~1,000 Eurasian languages. Gabora (1995, 2013) developed the “Meme and Variations” computational model inspired by genetic algorithms.

4b. Neuroscience of meme transmission

Mirror neurons. Rizzolatti and Craighero (2004, Annual Review of Neuroscience): neurons in macaque premotor cortex area F5 fire both when the monkey performs a goal-directed action AND when observing the same action (originally Gallese et al., 1996; di Pellegrino et al., 1992). The resonance model: observing an action activates the same motor programs as performing it, creating automatic, pre-reflective motor resonance — a direct mapping from observed actions to the observer’s own motor repertoire. Human mirror system: a fronto-parietal circuit in inferior parietal lobule (IPL) and ventral premotor cortex plus posterior inferior frontal gyrus (IFG), overlapping with Broca’s area — suggesting a connection between the mirror system and language evolution. Rizzolatti and Arbib (1998) proposed the mirror system provided the evolutionary foundation for language by solving the “parity problem” (creating shared semantic framework between sender and receiver). Memetic relevance: mirror neurons provide a candidate neural substrate for imitation-based meme transmission — the core mechanism of memetic replication.

Falk and Lieberman fMRI virality studies. The key study: Falk, Morelli, Welborn, Dambacher, and Lieberman (2013, Psychological Science, 24(7), 1234–1242). Paradigm: N = 19 undergraduate “interns” scanned via fMRI while reading descriptions of TV pilot ideas, then pitched favorites to “producers.” Ideas that successfully spread were associated with greater mentalizing system and reward system activation during initial encoding — before communicators knew they would recommend them. Individual differences in TPJ activation predicted persuasiveness as an “idea salesperson.” The finding: successful communicators spontaneously engage mentalizing — simulating how others will receive the message — during initial exposure. Neural activity predicted idea propagation beyond self-reported intentions.

Major follow-up: Scholz, Baek, O’Donnell, Kim, Cappella, and Falk (2017, PNAS, 114, 2881–2886). Two studies (N = 41 + 39), 80 real NYT health articles. Brain activity in self-related processing regions (MPFC, PCC), mentalizing regions (TPJ, DMPFC, precuneus), and valuation regions (ventral striatum, VMPFC) predicted both individual sharing decisions AND population-level virality — how many times each article was actually shared by the entire NYT readership. Brain signals from 80 individuals predicted article popularity among millions. Chan et al. (2023, PNAS) replicated cross-culturally: Dutch participants’ neural signals predicted US article virality, suggesting brain-based predictions capture more universal mechanisms than self-report.

Default Mode Network. Raichle et al. (2001) identified the DMN; Menon (2023, Neuron — “20 years of the default mode network”) reviews comprehensively. Key regions: mPFC, PCC, precuneus, angular gyrus, middle temporal gyrus. Functions: self-referential processing, social cognition/mentalizing (extensive overlap with the “social brain”), mental simulation, autobiographical memory, future planning, and internal narrative construction. Menon argues the DMN “integrates memory, language, and semantic representations to create a coherent internal narrative reflecting individual experiences, central to construction of a sense of self.” Connection to memetic “selfplex”: the DMN provides the neural infrastructure for maintaining and updating the selfplex — the memetic construct of personal identity that determines which new memes are accepted or rejected. Molnar-Szakacs and Uddin (2013, Frontiers in Human Neuroscience): embodied simulation (mirror system) and mentalizing (DMN) provide complementary routes to understanding others, both mediated through self-referential processing.

Amygdala-hippocampal interaction. McGaugh (2004, Annual Review of Neuroscience): the basolateral amygdala (BLA) modulates hippocampal memory consolidation. Emotional arousal triggers adrenal stress hormones activating β-noradrenergic receptors in BLA; BLA then modulates synaptic plasticity in hippocampus, enhancing long-term potentiation. The locus coeruleus-norepinephrine (LC-NE) system is central. Dual enhancement: immediate affect-biased attention enhances encoding; delayed amygdala-hippocampal coupling during consolidation preferentially stabilizes emotional memories. Qasim et al. (2022, Nature Human Behaviour): using intracranial EEG in 148 participants, demonstrated increased high-frequency activity in both hippocampus and amygdala during successful encoding of emotional stimuli; electrical stimulation of hippocampus selectively diminished emotional memory, confirming a causal role. Memetic implication: emotional memes (fear, outrage, humor, awe) enjoy systematic transmission advantage through amygdala-hippocampal encoding — not because of truth value but because of ability to activate this circuit.

4c. Temes — the third replicator

Blackmore’s concept. Introduced at TED 2008 (“Memes and Temes”), elaborated in New Scientist (2009, “Evolution’s Third Replicator”), and ongoing (2023, “The New Evolution,” New Scientist). Later shifted terminology from “teme” to “treme” (tertiary meme/third replicator). Definition: digital information stored, copied, varied, and selected by machines — the third replicator after genes and memes. The critical threshold: when all three processes of the evolutionary algorithm (copying, variation, selection) are carried out by machines without human involvement. Blackmore: “We humans like to think we are the designers, creators and controllers of this newly emerging world but really we are stepping stones from one replicator to the next.” Replicator hierarchy: just as memes piggyback on genes, temes piggyback on both. Each replicator transition is dangerous for existing biota.

Current evidence. Recommendation algorithms perform variation (A/B testing), selection (promoting content maximizing engagement), and copying (distributing successful formats) — largely without human editorial involvement. LLMs generate cultural artifacts at scale, creating content that enters the cultural ecosystem and competes with human-created information. The selection environment for temes: engagement metrics, click-through rates, ad conversion — purely metric-based selection pressures, not cognitive. The optimization target disconnect: selection pressures on temes (maximize engagement) can diverge radically from human welfare (truth, wellbeing, cohesion).

Co-evolutionary implication. Teme evolution increasingly constrains the meme landscape available to humans. If algorithmic curation controls information access, human meme evolution occurs within an environment shaped by teme evolution. Feedback loop: humans provide behavioral data → algorithms optimize for engagement → humans’ cultural environment is reshaped → behavior changes → new data feeds algorithms. Critiques: the neuroanthropology critique (Downey and Lende, 2010) argued the framework is overly reductive; general critiques of memetics (Boyd, Richerson, Pigliucci) regarding the replicator analogy apply equally to temes. Despite critiques, the teme concept has gained informal traction as a framework for thinking about AI-driven cultural evolution post-2022.

4d. Meme-gene coevolution

Lactase persistence. The canonical case: the dairying meme (cultural practice of herding and consuming milk) spread through northern European and East African populations, creating selection pressure for lactase persistence alleles. Gene-culture coevolutionary model well-established (Richerson, Boyd, Henrich). Allele frequency change measurable and documented — a cultural practice (meme) literally shaped the human genome.

Language and brain evolution. Pinker and Bloom’s linguistic adaptationist argument: language evolved through natural selection for communication. Blackmore’s memetic drive hypothesis: language evolved for memetic fidelity and fecundity rather than direct genetic advantage. If true, memes drove brain evolution — individuals better at imitating and transmitting memes had higher genetic fitness because meme-rich environments selected for meme-processing brains. The Rizzolatti-Arbib hypothesis connecting mirror neurons to language evolution supports this: the neural infrastructure for imitation (and thus meme transmission) is architecturally linked to the neural infrastructure for language.

Cooking memes and morphological change. Wrangham’s cooking hypothesis (Catching Fire, 2009): the cultural practice of cooking (a meme) reduced selection pressure for large digestive organs and tough enamel, freed metabolic energy for brain growth, and thus shaped human morphological evolution. Cooking memes literally transformed the human body.

---

SECTION 5: FIELD SYNTHESIS

5a. Key axes of disagreement

Strong vs. weak memetics. Strong memetics (Blackmore, early Dawkins): memes are genuine replicators with their own interests; the self is a memeplex (“selfplex”); meme-level selection is the primary explanatory framework for culture. Weak memetics (Dennett, later practitioners): the meme’s-eye view is a useful heuristic — a way of asking productive questions about cultural persistence and transmission — but not a literal claim about discrete replicating entities. Most working researchers occupy the weak position.

Internal vs. external meme location. Internalists (Dawkins’ original formulation): memes are neural patterns — information stored in brains. Externalists (Dennett, later Blackmore): memes are behavioral artifacts, instructions, or publicly observable cultural items. The distinction matters because it determines what counts as meme replication (neural copying vs. behavioral copying) and what evidence is relevant. Neither position has won decisively.

Memetics vs. DIT vs. CAT. What each framework is actually good for: Memetics excels at explaining maladaptive cultural persistence (why harmful practices persist despite reducing host fitness), counter-cultural memes (memes that spread against host interests), and information warfare (deliberate memetic engineering). DIT (Dual Inheritance Theory, Boyd and Richerson) excels at population-level cultural dynamics — gene-culture coevolution, mathematical modeling, quantitative predictions about cultural change rates and equilibria. CAT (Cultural Attraction Theory, Sperber) excels at explaining content stability mechanisms — why certain cultural forms recur cross-culturally (cognitive universals, attractor landscapes) without requiring high-fidelity copying. The competent practitioner uses all three frameworks selectively, matching tool to problem.

5b. Tacit vocabulary and standard moves

The meme’s-eye view move. Asking “what does it mean for this to be a good meme, not just a good idea?” — reframing analysis from the perspective of the information pattern’s interests rather than the host’s. This is the most productive single move in applied memetics. It explains why some ideas persist not because they’re true or useful but because they’re structurally optimized for transmission.

Phenotype vs. genotype for memes. The behavior (ritual performance, spoken catchphrase, shared post) is the phenotype — the observable expression. The information encoding it (in brain, text, or artifact) is the genotype — the replicating pattern. Same memetic genotype can produce different phenotypes in different host environments (like the same gene producing different traits in different organisms). Mutations can occur at either level.

Vertical vs. horizontal transmission. Vertical (parent-to-child) transmission predicts strong host-benefit alignment — memes transmitted vertically that harm hosts will be selected against because they reduce the host’s (and therefore the meme’s) reproductive success. Horizontal (peer-to-peer) transmission predicts weaker alignment — horizontally transmitted memes can harm hosts because their fitness depends on transmission rate, not host welfare. Oblique transmission (teacher-to-many, media-to-audience) allows even greater divergence between meme and host interests. This predicts: religions transmitted primarily vertically (family traditions) will be more host-beneficial than religions transmitted primarily horizontally (proselytizing, conversion-focused).

How to distinguish a good meme story from a just-so story. Four requirements: (1) Is there a specified selection pressure? (2) Is there a described transmission mechanism? (3) Is there a fitness differential that precedes and explains the spread? (4) Is the explanation generating testable predictions, or merely providing post-hoc narration? A meme story that cannot specify these is unfalsifiable and therefore vacuous.

The “who benefits?” heuristic. Genes? Host organism? Meme itself? These come apart in the most analytically interesting cases. Celibacy memes benefit neither genes nor host reproductive fitness — but benefit the memeplex (Catholic Church) by channeling resources toward institutional propagation. Suicide bombing memes obviously harm the host but benefit the memeplex by demonstrating commitment credibility (costly signaling) and generating fear-based attention.

5c. Common novice errors

Conflating internet memes with memes. Internet memes (image macros, viral videos) are one empirically accessible instance of the general category — not the general category itself. Technical memes include languages, cooking techniques, mathematical proofs, religious rituals, fashion trends, architectural styles, and institutional norms. The internet meme is to the meme as the laboratory fruit fly is to the insect — a tractable model organism, not the whole domain.

Treating memetics as a theory of intentional persuasion. Memetics is a theory of differential replication — intentional meme engineering (advertising, propaganda) is a special case. Most meme evolution is non-intentional, driven by cognitive biases, structural features of transmission networks, and competition for limited attention. The vast majority of cultural change is not “designed” by anyone.

The “it’s all memes” error. Treating every cultural item as equally analyzable through the memetic lens. The 95/5 heuristic: most cultural evolution doesn’t need the selfish-meme framing — standard explanations (people adopt useful practices, institutions enforce norms, rational agents respond to incentives) suffice. The selfish-meme framework adds genuine explanatory value in roughly 5% of cases — precisely those where cultural persistence cannot be explained by host benefit.

Confusing fitness with truth. A meme’s memetic success tells you about its transmissibility, not its truth value. Highly fit memes may be true (germ theory), false (conspiracy theories), or truth-irrelevant (fashion trends). The fitness-truth conflation leads to the naturalistic fallacy applied to culture: “it spread because it’s true” or “it’s true because it spread.”

The “greedy memetics” error. Reducing all culture to meme competition, missing structural causation (economic systems, material conditions, institutional power), emergent phenomena (market dynamics, demographic transitions), and historical contingency. Memetics is one analytical lens among several; it illuminates certain phenomena brilliantly and obscures others. Greedy memetics (by analogy with Dennett’s “greedy reductionism”) tries to explain everything at the meme level, ignoring the causal power of structures, materials, and institutions.

5d. The live frontier

Computational cultural evolution. The productive core of memetics is being absorbed into DIT-style formal frameworks powered by big data. The future is not “memetics vs. DIT” but computational cultural evolution using population-level models, phylogenetic methods, machine learning, and massive digital datasets — retaining the useful insights of the meme’s-eye view within mathematically rigorous frameworks.

The AI/teme question. As AI systems become dominant cultural producers, the selection environment for human memes is increasingly shaped by algorithmic systems evolving according to engagement metrics. Practical consequences: for democracy (citizens’ information environment curated by engagement-optimizing temes, not editorial judgment), for epistemics (truth value decoupled from visibility), and for culture (homogenization pressures as algorithms converge on attention-maximizing content forms).

Mental immune theory. Andy Norman (Mental Immunity, 2021) treats epistemic health as analogous to physical immune health. Just as physical immune systems distinguish self from non-self and pathogen from benign, mental immune systems distinguish credible from incredible claims. “Cognitive immunology” develops inoculation-based interventions for memetic manipulation — treating susceptibility to misinformation as an immune deficiency rather than a character flaw.

Prebunking at scale. The empirical evidence from Roozenbeek, van der Linden, and Cook showing inoculation works at YouTube scale (Jigsaw partnership, 2022, Science Advances), at Instagram scale (van der Linden et al., 2026, Harvard Kennedy School Misinformation Review, N = 375,597), and cross-culturally (campaigns in Poland, Slovakia, Czech Republic). The mechanism: exposing manipulation techniques, not specific content. Why it works better than debunking: it is proactive rather than reactive, scales efficiently, works across political ideologies, doesn’t require adjudicating truth claims, and targets the production method rather than the product. The most promising empirical program in applied memetics — demonstrating that memetic defense can be engineered and deployed at population scale.