Memes and virality as copying: why viral is only a past tense

A joke that spreads does what a joke told at a table cannot: it gets copied by people who never met its author, altered a little each time, and judged by strangers who see only the latest version.

The word for that process predates the internet. The internet is the first environment where it can be watched happening.

This page is about the mechanics: what makes something copyable, what copying costs, why the copies change, and why the story of where any of it started is nearly always wrong.

What to take away

The word and what it carries

The term meme was coined as an analogy by Richard Dawkins in The Selfish Gene, the popular science book Oxford University Press first published in 1976: a unit of culture that copies itself from mind to mind the way a gene copies itself from body to body. Dawkins wanted a monosyllable that sounds a bit like "gene", so he shortened the Greek-rooted "mimeme" to "meme". Whether the analogy holds is argued about; what it usefully carries is the idea that the unit is selected by how well it copies, not by how good it is.

That single idea explains most of what follows. A format that is funnier but harder to reproduce loses to a format that is less funny and easier. Quality is not the selection pressure. Reproducibility is.

The three costs

Every act of participation has a price, paid in time and in the risk of getting it wrong. Formats that spread are formats that have driven all three prices close to nothing.

The image macro, a picture with a fixed layout and replaceable text, is the clearest example of a format shaped by selection to score well on all three. It is not the funniest form.

Anyone with no skill and no time can produce a passable instance in under a minute. That is why it dominated so long: image macros were one of the most common forms of Internet meme in the 2000s. Lolcats, pictures of expressive cats with text, are considered the first notable occurrence, and the Advice Animal stock-character macros, one fixed character per set of jokes, are the clearest case of a fixed part and a slot. Even the name has a documented source: the term "image macro" originated on the Something Awful forums, where a short bit of text expanded into the code for a pre-defined image.

Why the copies change

Faithful copies do not travel. A version identical to what an audience has already seen is not worth passing on; the value in forwarding something is in having something new to forward. So the versions that spread are the ones that changed, and the format drifts with every hop.

Drift has a direction. Each community that receives a format bends it toward its own references, its own tone and its own targets. The version that crosses from one community to the next is the one that was legible to both, which usually means the most stripped-down version, with the community-specific layers removed.

Formats therefore get simpler as they travel and more elaborate as they settle. A format that has crossed many boundaries is a skeleton of what it was anywhere in particular.

This is also why irony layers accumulate. Once a format is familiar enough to be dull, the next move is to use it wrongly on purpose, and then to use the wrong use wrongly. A format's late life is spent parodying itself, and a reader who arrives at that stage cannot reconstruct what the original meant.

Viral is a past tense

Nothing is viral while it is happening. Something is being copied at some rate, and nearly everything being copied stops being copied soon. The label is attached afterwards, to the small fraction that did not stop, and then the story is rewritten so that the outcome looks like it was visible from the start.

It was not. The signals available at the time, a rising count here, a share by someone with an audience there, are present for enormous numbers of things that then vanished. Anyone claiming to have predicted which would spread is describing the survivor and forgetting the field.

This has a practical consequence for anyone writing about the subject: the reasons given for why something spread are constructed after the fact, from the fact that it did. They are plausible because they were chosen to be. The same reasons would have been available for a thousand things that did not spread, and nobody wrote those up.

The origin problem

Ask where a format came from and you will get a confident answer: usually a site and a year. Treat that as a claim about the earliest copy the answerer's sources happened to index, not the origin.

The gap can be years. It is invisible from the inside, because unindexed copies leave no trace of their absence.

The definitional problems that make chronology hard for early social networks apply here with more force. A format has no operator, no announcement and no launch.

Its record is whatever was copied, and copying favored what was already popular, which is circular. Dating a claim properly is set out under platform timelines: "earliest located" is a finding, "first" is a guess.

One format traced: lolcats

Lolcats show every part of this page in one record, and they involve no real person.

The same three questions, what made it cheap, where the earliest located copy sits, and when the label was applied, work on any format.

When the material is a person

Sometimes the fixed part is a photo of a real person, a clip, or a sentence they said. The process treats it like a drawing: copy, vary, strip context, layer irony.

The person is not consulted and cannot withdraw. They watch their face or worst moment become a slot for other people's jokes.

The pattern is consistent enough to describe without an example. A private moment becomes public through someone else's decision. The first wave is amused. The second wave is larger, has never seen the context, and does not know there is a person.

The person's name becomes searchable alongside the format permanently. Every attempt to explain, object, or ask it to stop becomes new material. Employers, family, and strangers find the format before they find the person.

Communities differ in how they handle this, and that difference is a good test of what a community is. Those described under internet communities as having real norms tend to develop a rule about it, while mere audiences do not.

This site does not name the people, describe the incidents, or reproduce the formats. Where human cost matters to the argument, the pattern is enough.

Common questions

Is there a way to make something spread on purpose?

Everything that can be done is done constantly by people with budgets, and nearly all of it fails. Removing the three costs raises the odds; it does not produce the result. Anyone selling a method is selling survivorship.

Why do formats come back after they seem dead?

Because a new audience arrives that has not seen them, and because a format familiar enough to be parodied has a second life as a parody. Nothing online is finished; it is only quiet.

Does a format mean the same thing everywhere it is used?

No. It is bent by every community it passes through, and a version that has crossed many boundaries carries almost none of the meaning it had in any one of them. Reading a format from one community with the assumptions of another is the most common way to misunderstand it.

Why does this site not give examples?

An example carries an origin claim, a popularity claim and often a person. All three are things this site declines to assert without evidence that, for this subject, mostly does not exist. The lolcats record above is one of the exceptions where the evidence is published, and even there it shows three competing origins rather than one.