I haven't done the experiment yet (I'd like to), but I have a strong prediction of what the result will be. My prediction is that not all of the twenty teams will succeed in passing the skill intact down the line to their tenth members, but that a significant number of them will. In some of the teams there will be mistakes: perhaps a weak link in the chain will forget some vital step in the procedure, and everyone downstream of the mistake will then obviously fail. Perhaps team 4 gets as far as the 'catamaran' but falters thereafter. Perhaps the eighth member of team 13 produces a 'mutant' somewhere between the 'box with two lids' and the 'picture frame' and the ninth and tenth members of his team then copy the mutated version.

Now, of those teams in which the skill is transferred successfully to the tenth generation, I make a further prediction. If you rank the junks in order of 'generation' you will not see a systematic deterioration of quality with generation number. If, on the other hand, you were to run an experiment identical in all respects except that the skill transferred was not origami but copying a drawing of a junk, there would definitely be a systematic deterioration in the accuracy with which the generation 1 pattern 'survived' to generation 10.

In the drawing version of the experiment, all the generation 10 drawings would bear some slight resemblance to the generation 1 drawing. And within each team, the resemblance would more or less steadily deteriorate as you proceed down the generations. In the origami version of the experiment, by contrast, the mistakes would be all-or-none: they'd be 'digital' mutations. Either a team would make no mistakes and the generation 10 junk would be no worse, and no better, on average than that produced by generation 5 or generation 1; or there would be a 'mutation' in some particular generation and all downstream efforts would be complete failures, often faithfully reproducing the mutation.

What is the crucial difference between the two skills? It is that the origami skill consists of a series of discrete actions, none of which is difficult to perform in itself. Mostly the operations are things like 'Fold both sides into the middle.' A particular team member may execute the step ineptly, but it will be clear to the next team member down the line what he is trying to do. The origami steps are 'self-normalizing'. It is this that makes them 'digital'. It is like my master carpenter, whose intention to flatten the nail head in the wood is obvious to his apprentice, regardless of the details of the hammer blows. Either you get a given step of the origami recipe right or you don't. The drawing skill, by contrast, is an analogue skill. Everybody can have a go, but some people copy a drawing more accurately than others, and nobody copies it perfectly. The accuracy of the copy depends, too, on the amount of time and care devoted to it, and these are continuously variable quantities. Some team members, moreover, will embellish and 'improve', rather than strictly copy, the preceding model.

Words — at least when they are understood — are self-normalizing in the same kind of way as origami operations. In the original game of Chinese Whispers (Telephone) the first child is told a story, or a sentence, and is asked to pass it on to the next child, and so on. If the sentence is less than about seven words, in the native language of all the children, there is a good chance that it will survive, un-mutated, down ten generations. If it is in an unknown foreign language, so that the children are forced to imitate phonetically rather than word by word, the message does not survive. The pattern of decay down the generations is then the same as for a drawing, and it will become garbled. When the message makes sense in the children's own language, and doesn't contain any unfamiliar words like 'phenotype' or 'allele', it survives. Instead of mimicking the sounds phonetically, each child recognizes each word as a member of a finite vocabulary and selects the same word, although very probably pronounced in a different accent, when passing it on to the next child. Written language is also self-normalizing because the squiggles on paper, no matter how much they may differ in detail, are all drawn from a finite alphabet of (say) twenty-six letters.

The fact that memes can sometimes display very high fidelity, due to self-normalizing processes of this kind, is enough to answer some of the commonest objections that are raised to the meme/gene analogy. In any case, the main purpose of meme theory, at this early stage of its development, is not to supply a comprehensive theory of culture, on a par with Watson-Crick genetics. My original purpose in advocating memes, indeed, was to counter the impression that the gene was the only Darwinian game in town — an impression that The Selfish Gene was otherwise at risk of conveying. Peter Richerson and Robert Boyd emphasize the point in the title of their valuable and thoughtful book Not by Genes Alone, although they give reasons for not adopting the word 'meme' itself, preferring 'cultural variants'. Stephen Shennan's Genes, Memes and Human History was partly inspired by an earlier excellent book by Boyd and Richerson, Culture and the Evolutionary Process. Other book-length treatments of memes include Robert Aunger's The Electric Meme, Kate Distin's The Selfish Meme, and Virus of the Mind: The New Science of the Meme by Richard Brodie.

But it is Susan Blackmore, in The Meme Machine, who has pushed memetic theory further than anyone. She repeatedly visualizes a world full of brains (or other receptacles or conduits, such as computers or radio frequency bands) and memes jostling to occupy them. As with genes in a gene pool, the memes that prevail will be the ones that are good at getting themselves copied. This may be because they have direct appeal, as, presumably, the immortality meme has for some people. Or it may be because they flourish in the presence of other memes that have already become numerous in the meme pool. This gives rise to meme complexes or 'memeplexes'. As usual with memes, we gain understanding by going back to the genetic origin of the analogy.

For didactic purposes, I treated genes as though they were isolated units, acting independently. But of course they are not independent of one another, and this fact shows itself in two ways. First, genes are linearly strung along chromosomes, and so tend to travel through generations in the company of particular other genes that occupy neighbouring chromosomal loci. We doctors call that kind of linkage linkage, and I shall say no more about it because memes don't have chromosomes, alleles or sexual recombination. The other respect in which genes are not independent is very different from genetic linkage, and here there is a good memetic analogy. It concerns embryology which — the fact is often misunderstood — is completely distinct from genetics. Bodies are not jigsawed together as mosaics of phenotypic pieces, each one contributed by a different gene. There is no one-to-one mapping between genes and units of anatomy or behaviour. Genes 'collaborate' with hundreds of other genes in programming the developmental processes that culminate in a body, in the same kind of way as the words of a recipe collaborate in a cookery process that culminates in a dish. It is not the case that each word of the recipe corresponds to a different morsel of the dish.

Genes, then, co-operate in cartels to build bodies, and that is one of the important principles of embryology. It is tempting to say that natural selection favours cartels of genes in a kind of group selection between alternative cartels. That is confusion. What really happens is that the other genes of the gene pool constitute a major part of the environment in which each gene is selected versus its alleles. Because each is selected to be successful in the presence of the others — which are also being selected in a similar way — cartels of co-operating genes emerge. We have here something more like a free market than a planned economy. There is a butcher and a baker, but perhaps a gap in the market for a candlestick maker. The invisible hand of natural selection fills the gap. That is different from having a central planner who favours the troika of butcher + baker + candlestick maker. The idea of co-operating cartels assembled by the invisible hand will turn out to be central to our understanding of religious memes and how they work.