Richard Dawkins spent most of The Selfish Gene making the case that biology's real protagonist is the replicator rather than the organism carrying it about. Then, in the closing pages of that 1976 book, he asked a mischievous follow-up question. If copying with variation is the engine, is there anything else on this planet that replicates? His answer was culture, and culture needed a unit. He borrowed the Greek mimeme, meaning a thing imitated, and clipped it to one syllable that would sit comfortably beside gene. The examples he reached for were deliberately unglamorous: tunes, catchphrases, fashions in clothing, the knack of building an arch. A meme was any scrap of culture that travelled between minds by being copied.
The word leaves the biology department
For roughly two decades the idea led an academic life. Philosophers and psychologists tried to build a discipline around it, and for a while there was even a journal devoted to the project. The objections were serious and never fully answered. Genes are discrete, physically located and copied with high fidelity; a joke is none of those things. Where exactly does one idea stop and the next begin? Culture mutates every time it is retold, which makes the analogy with faithful replication awkward. By the turn of the millennium, memetics as a formal research programme had largely stalled, and Britannica's entry on the term records both the coinage and the difficulty. The word survived anyway, because a much larger population had quietly taken it out of the seminar room.
What the internet meant by it instead
Message-board users in the 1990s and 2000s began applying the word to something far more specific than Dawkins had in mind: a picture with text on it, a stock phrase with a fixed rhythm, a video that everyone was making their own version of. The narrowing looks like sloppiness. It was actually a useful clarification, and the media scholar Limor Shifman made the case for it in Memes in Digital Culture, published by MIT Press. Her definition treats an internet meme not as a single item but as a group of items sharing content, form or stance, created in awareness of one another, and circulated and transformed by many users.
The most valuable thing in that definition is the line it draws between a meme and a viral. A viral is one object that a great many people pass along unchanged: a video, a photograph, a news item. A meme is a family. The originating object matters less than the template it leaves behind, and the template stays alive only as long as people keep producing new members of the family. This is why going viral and becoming a meme describe genuinely different events, even though the two are constantly muddled together.
The four rooms a format passes through
Watch enough of them and a rough itinerary emerges. A format usually begins inside a community small enough to have its own dialect: a subreddit, a group chat, a fandom, some corner of an art platform. At this stage the joke is partly about membership. Recognising the reference is most of the point, and outsiders looking at the same image see very little. Then a bridge appears. An aggregator account, a screenshot posted somewhere much larger, a journalist hunting for a column, some connector carries the format across into a wider population that adopts it without the original context attached.
The third room is saturation. The format is now legible to everyone, including brands, local councils and people who have to explain it to their relatives. Volume peaks, and so does the ratio of weak versions to good ones, because the crowd producing them no longer shares the taste that shaped the first hundred. The fourth room is exhaustion, which is rarely a clean death. Some formats vanish outright. Others are demoted into ordinary vocabulary. Reaction images in particular tend to slide from joke to punctuation, used thousands of times a day by people who could not name their origin and would not care if you told them.
Copying is participation, not transmission
The reason this itinerary looks nothing like an epidemic curve is that memes do not really spread by being forwarded. They spread by being remade. Henry Jenkins has spent a career on that distinction, first in his research on television fandom and later in Convergence Culture, which argued that audiences had stopped being an endpoint and turned into a production line. Fans had always rewritten what they loved. What changed was that distributing the rewrite stopped costing anything.
Applied to memes, the insight has a practical edge. Formats that thrive are the ones with a gap in them: a blank caption, an unfilled slot, a rhythm you can drop your own grievance into. A finished, perfect artefact invites admiration and nothing else. A slightly broken one invites a reply. Shifman makes a related observation about texts whose very flaws are what make them productive, and anyone who has watched an amateur video accumulate ten thousand imitations will recognise the pattern. The barrier to making a version is far lower than the barrier to making an original, and that gap in difficulty is the entire mechanism.
Architecture decides which formats can live
Every platform is a set of constraints dressed up as software, and those constraints select ruthlessly for particular shapes. Anonymous image boards, where nobody accumulated reputation, produced the captioned template, because the joke had to carry itself with no author attached. Character-limited text feeds rewarded compressed sentence frames that could be typed from memory. Blogging platforms with reblog buttons favoured long chains of accumulated commentary, so the unit of humour became the thread rather than the picture.
Short-video apps changed the terms again by making the audio track a first-class object. Once a sound has its own page and its own button, the sound becomes the template and the visuals become the variable, which is why so much of the past decade's meme culture is built on gestures, dances and reused fragments of speech rather than captioned stills. Add recommendation feeds that surface material from strangers rather than friends, and adoption stops depending on who you happen to follow. A format can now reach a million people without passing through anyone either of you knows.
Why nobody manufactures one on demand
All of which makes memes easy to explain backwards and nearly impossible to plan forwards. The traits we credit for success, such as simplicity, an obvious slot, emotional charge and a whiff of transgression, are equally visible in thousands of formats that went nowhere at all. We only ever inspect the survivors. Marketing departments discover this expensively and repeatedly, because a campaign built to become a meme almost never does. It arrives finished, polished and legally approved, with nothing left over for anyone to add.
Fifty years on, the two senses of the word have reached an uneasy peace. The Dawkins version remains a provocative analogy about culture at large. The internet version is narrower, more concrete and, for once, easier to study, because every copy leaves a timestamp behind it. What both share is the underlying claim that a piece of culture succeeds according to how readily it can be reproduced rather than how good it is. Anyone who has watched a genuinely funny joke die quietly while a mediocre template runs for three years already understands the argument.
