Even though conventional Darwinian selection of genes might have favoured psychological predispositions that produce religion as a by-product, it is unlikely to have shaped the details. I have already hinted that, if we are going to apply some form of selection theory to those details, we should look not to genes but to their cultural equivalents. Are religions such stuff as memes are made on?

This chapter began with the observation that, because Darwinian natural selection abhors waste, any ubiquitous feature of a species — such as religion — must have conferred some advantage or it wouldn't have survived. But I hinted that the advantage doesn't have to redound to the survival or reproductive success of the individual. As we saw, advantage to the genes of the cold virus sufficiently explains the ubiquity of that miserable complaint among our species.*) And it doesn't even have to be genes that benefit. Any replicator will do. Genes are only the most obvious examples of replicators. Other candidates are computer viruses, and memes — units of cultural inheritance and the topic of this section. If we are to understand memes, we have first to look a little more carefully at exactly how natural selection works.

In its most general form, natural selection must choose between alternative replicators. A replicator is a piece of coded information that makes exact copies of itself, along with occasional inexact copies or 'mutations'. The point about this is the Darwinian one. Those varieties of replicator that happen to be good at getting copied become more numerous at the expense of alternative replicators that are bad at getting copied. That, at its most rudimentary, is natural selection. The archetypal replicator is a gene, a stretch of DNA that is duplicated, nearly always with extreme accuracy, through an indefinite number of generations. The central question for meme theory is whether there are units of cultural imitation which behave as true replicators, like genes. I am not saying that memes necessarily are close analogues of genes, only that the more like genes they are, the better will meme theory work; and the purpose of this section is to ask whether meme theory might work for the special case of religion.

In the world of genes, the occasional flaws in replication (mutations) see to it that the gene pool contains alternative variants of any given gene — 'alleles' — which may therefore be seen as competing with each other. Competing for what? For the particular chromosomal slot or 'locus' that belongs to that set of alleles. And how do they compete? Not by direct molecule-to-molecule combat but by proxy. The proxies are their 'phenotypic traits' — things like leg length or fur colour: manifestations of genes fleshed out as anatomy, physiology, biochemistry or behaviour. A gene's fate is normally bound up with the bodies in which it successively sits. To the extent that it influences those bodies, it affects its own chances of surviving in the gene pool. As the generations go by, genes increase or decrease in frequency in the gene pool by virtue of their phenotypic proxies.

Might the same be true of memes? One respect in which they are not like genes is that there is nothing obviously corresponding to chromosomes or loci or alleles or sexual recombination. The meme pool is less structured and less organized than the gene pool. Nevertheless, it is not obviously silly to speak of a meme pool, in which particular memes might have a 'frequency' which can change as a consequence of competitive interactions with alternative memes.

Some people have objected to memetic explanations, on various grounds that usually stem from the fact that memes are not entirely like genes. The exact physical nature of a gene is now known (it is a sequence of DNA) whereas that of memes is not, and different memeticists confuse one another by switching from one physical medium to another. Do memes exist only in brains? Or is every paper copy and electronic copy of, say, a particular limerick also entitled to be called a meme? Then again, genes replicate with very high fidelity, whereas, if memes replicate at all, don't they do so with low accuracy?

These alleged problems of memes are exaggerated. The most important objection is the allegation that memes are copied with insufficiently high fidelity to function as Darwinian replicators. The suspicion is that if the 'mutation rate' in every generation is high, the meme will mutate itself out of existence before Darwinian selection can have an impact on its frequency in the meme pool. But the problem is illusory. Think of a master carpenter, or a prehistoric flint-knapper, demonstrating a particular skill to a young apprentice. If the apprentice faithfully reproduced every hand movement of the master, you would indeed expect to see the meme mutate out of all recognition in a few 'generations' of master/apprentice transmission. But of course the apprentice does not faithfully reproduce every hand movement. It would be ridiculous to do so. Instead, he notes the goal that the master is trying to achieve, and imitates that. Drive in the nail until the head is flush, using as many hammer blows as it takes, which may not be the same number as the master used. It is such rules that can pass unmutated down an indefinite number of imitation 'generations'; no matter that the details of their execution may vary from individual to individual, and from case to case. Stitches in knitting, knots in ropes or fishing nets, origami folding patterns, useful tricks in carpentry or pottery: all can be reduced to discrete elements that really do have the opportunity to pass down an indefinite number of imitation generations without alteration. The details may wander idiosyncratically, but the essence passes down unmutated, and that is all that is needed for the analogy of memes with genes to work.

In my foreword to Susan Blackmore's The Meme Machine I developed the example of an origami procedure for making a model Chinese junk. It is quite a complicated recipe, involving thirty-two folding (or similar) operations. The end result (the Chinese junk itself) is a pleasing object, as are at least three intermediate stages in the 'embryology', namely the 'catamaran', the 'box with two lids' and the 'picture frame'. The whole performance does indeed remind me of the foldings and invaginations that the membranes of an embryo undergo as it morphs itself from blastula to gastrula to neurula. I learned to make the Chinese junk as a boy from my father who, at about the same age, had acquired the skill at his boarding school. A craze for making Chinese junks, initiated by the school matron, had spread through the school in his time like a measles epidemic, then died away, also like a measles epidemic. Twenty-six years later, when that matron was long gone, I went to the same school. I reintroduced the craze and it again spread, like another measles epidemic, and then again died away. The fact that such a teachable skill can spread like an epidemic tells us something important about the high fidelity of memetic transmission. We may be sure that the junks made by my father's generation of schoolboys in the 1920s were in no general respect different from those made by my generation in the 1950s.

We could investigate the phenomenon more systematically by the following experiment: a variant of the childhood game of Chinese Whispers (American children call it Telephone). Take two hundred people who have never made a Chinese junk before, and line them up in twenty teams of ten people each. Gather the heads of the twenty teams around a table and teach them, by demonstration, how to make a Chinese junk. Now send each one off to find the second person in his own team, and teach that person alone, again by demonstration, to make a Chinese junk. Each second 'generation' person then teaches the third person in her own team, and so on until the tenth member of every team has been reached. Keep all the junks made along the way, and label them by their team and 'generation' number for subsequent inspection.