Memetics does provide a useful way
of understanding cultural evolution.

In Contemporary Debates in Philosophy of Biology, Ed. Francisco Ayala and Robert Arp, (2010) Chichester, Wiley-Blackwell, 255-72

(Version submitted January 2008. This may have been edited prior to publication)


Imagine a planet on which a self-replicating molecule appears, is copied with variation and selection, and evolves to build itself living creatures that protect and propagate it; those creatures eventually spread all over the planet, changing its atmosphere and environment as they do so. Now imagine that one of those creatures begins to copy behaviours from one individual to another, and this information gets copied with variation and selection. In other words, a second kind of replicator emerges. This transforms the original creatures to make them better at protecting and propagating it, and goes on to build cultural objects that eventually spread all over the planet, changing its atmosphere and environment all over again.

This is the vision that memetics provides – that human beings alone on earth are the product of two replicators, not one. In consequence they are meme machines as well as gene machines, and culture uses them as its replication machinery. Language, music and technology are all parasites, often turned symbiotic, that evolved for their own sake, not for the sake of individual humans or their genes.

This vision is different from most theories of cultural evolution or gene-culture co-evolution which generally treat genes as the final arbiter and culture as an adaptation or an aspect of the human phenotype. Whether this new vision of human and cultural evolution is valuable or not is the question for this debate.

Origins of the meme meme

The term “meme” was coined by Richard Dawkins in his 1976 book The Selfish Gene. His intention was not to create an analogy with genes but rather to illustrate the principle of universal Darwinism and his contention that “Darwinism is too big a theory to be confined to the narrow context of the gene.” (get pp from 1976 ed). Universal Darwinism is the general principle that whenever you have information that is copied with variation and selection, then you must get evolution. Darwin (1887) applied this process to languages as well as evolving organisms, and universal Darwinism has a long history (Plotkin 1993). The process has also been described as “blind variation with selective retention”, a phrase used by Campbell (1960) to describe the evolution of knowledge or “evolutionary epistemology” and now used mainly in cybernetics. The three steps, that is variation, selection and heredity, can be understood as a simple algorithmic process, the evolutionary algorithm; or “a scheme for creating Design out of Chaos without the aid of Mind.” (Dennett 1995, 50).

Dawkins called the information that is copied in such a process the “replicator” and contrasted this with “vehicles” (such as living organisms) that carry the replicators around, protecting and propagating them. In a slightly different scheme, Hull (1988) refers to replicators, interactors and lineages. Dawkins wanted “to claim almost limitless power for slightly inaccurate self-replicating entities, once they arise anywhere in the universe” (Dawkins 1989 p 322). His point was that genes are just one example of a replicator, and the general principle is more important than the specifics of the way genes have evolved on earth. This is why, at the end of the book, he asked his now famous question “do we have to go to distant worlds to find other kinds of replicator and other, consequent, kinds of evolution?” (1989 p 192). His answer was, of course, no. Staring us in the face, drifting about in its “primeval soup” of culture, is a new replicator: tunes, ideas, catch-phrases, fashions, and ways of making things. These show heredity (they are copied from person to person), variation (for example through errors in transmission or memory, and through recombination), and they are selected (people choose what to copy and from whom). So they must be replicators. He wanted a name that would sound a bit like gene while conveying the central meaning of “that which is imitated” so, from the Greek mimeme he took the abbreviation “meme”.

The term “meme” has been a successful meme, spreading widely, and being accepted into the Oxford English Dictionary in 1997. The science of memetics, however, has not grown correspondingly. The reasons for this are not clear. Memetics has some potential advantages over other theories, for example in how it deals with informational viruses such as email viruses, cults and religions; and in its novel approach to the evolution of language, the big brain, and the coevolution of genes and culture. Nevertheless, there is no thriving science of memetics, and no specialist journals or conferences. Perhaps memetics is just plain wrong; perhaps it is currently too difficult to test its predictions; or perhaps its implications are too scary or unappealing for people to accept. Certainly part of the problem is that memetics is frequently misunderstood. I shall therefore begin by discussing some of the most common confusions before turning to the value of memetics itself.

Do memes exist?

This question, although often asked, reveals a misunderstanding of the concept of memes. McGrath complains that “There is no direct evidence for the existence of “memes” themselves” (2005, p 121) and Aunger “their existence has yet to be proven.” (2000 p 7). Wimsatt claims they are “both misnamed and mischaracterized, and perhaps even that they do not exist.” (Wimsatt 1999, p 280). Yet, as Laland and Odling-Smee point out, “the pertinent question is not whether memes exist, as suggested by Aunger … but whether they are a useful theoretical expedient.” (2000 p 121).

To see why the existence question is misplaced it is only necessary to remember how the idea was first conceived, that is, to think of skills, habits, words or stories as replicators. There is nothing hypothetical about teenagers listening to i-PODs, wearing pre-torn jeans, or piercing their eyebrows after seeing others do the same. There is no question about the existence of money, railways, bicycles, telephones, furniture, skyscrapers, holiday brochures, football, or the days of the week. They are all information encoded in some kind of matter and energy, and they can all be copied or not. The core definition of memes is “that which is imitated” or that which is copied. It may be difficult to pin down just what has been copied, but whatever it is, that “something” is, by definition, a meme.

Accepting that memes exist does not, of course, mean that memetics is a good idea. To give a fanciful equivalent, imagine that I decided to take the class of all city streets that have plane trees in them and call them “plemes”. I could then go around counting the number of plemes in Bristol, or London, or the whole world. I could investigate whether there are more houses, or more cats, or more deaths in plemes as opposed to other streets. I doubt I would find out anything interesting, that is unless there is truly something peculiar about that beautiful bark, or a poison in the leaves. My guess is that the science of plemetics would be a complete waste of time – not because plemes don’t exist, they do, but because dividing the world up this way reveals no general principles, gives rise to no new insights, and is completely pointless.

Memetics might be like this, completely pointless, or it might not. So the existence question is a distraction. Memes exist. The interesting questions concern just what we should and should not count as memes, whether they really behave as replicators, whether they are inside or outside brains or both, how close the analogy with genes proves to be, and whether memetics gives rise to any useful theoretical or practical progress. Those are the questions I shall briefly consider here.

Trouble with analogies and units

It should be clear from the origin of the term ‘meme’ that it was not primarily conceived as a gene analog. Yet that is often how it is described. For example, Gould called memes “a useless analogy”, and Midgley “a meaningless metaphor” (Blackmore 1999). McGrath goes further and claims that “The case for the existence of the “meme” rests on the questionable assumption of a direct analogy with the gene, which proves incapable of bearing the theoretical weight that is placed upon it.” (McGrath 2005, p 121) and Richerson and Boyd argue against memetics on the grounds that “the best evidence suggests that cultural  variants are only loosely analogous to genes” (2005 p 60).

Such criticisms are common, the argument often going something like this: – memes were invented by analogy with genes, genes are real physical entities made of DNA, no equivalent physical entity can be found for memes, therefore the notion of memes is false (or is useless, or memes don’t exist). But this misses the point of Dawkins’s reason for inventing the term, which was to apply universal Darwinism and treat cultural information as a replicator. Once you take it this way it becomes clear that analogies between genes and memes are secondary, and must be treated with care.

So I prefer to see it this way: genes are replicators and we know a great deal about how they work. Memes are replicators and we know rather little about how they work. This means that some analogies will be fruitful because memes and genes are both replicators and will therefore show some profound similarities, but other analogies will be false because in many ways the two replicators are so different. For example they use entirely different copying mechanisms, their sources of variation are quite different, and they differ in fidelity, storage, and in many other ways. Also one replicator has been evolving along with its replication machinery for approximately four billion years and has settled down to a very high fidelity system with separated germ line and phenotypes; the other has been around for at most a few million years and is, as Dawkins put it “still in its infancy, still drifting clumsily about in its primeval soup” (1989 p 192).

I therefore agree with McGrath that “analogies can be dangerously misleading.” (2005 p 132) and with Wimsatt (1999) who points out many dissimilarities between genes and memes, and their relevance for attempts to model cultural evolution. These differences mean that we must treat gene-meme analogies with great care, using them where appropriate to build new hypotheses or explore possibilities, but not relying on them as a guide to how memes work.

Another source of confusion is the problem of units. Memes are often described as “units of culture” or “units of imitation” (although some definitions avoid the word by talking about “elements of culture”, “contagious ideas” or “cultural instructions”). Some critics have made the problem of units into a major objection to the whole of memetics. For example Midgley complains that culture cannot be neatly divided up into units, and nothing is to be gained by “atomising thought” for “Thought is not granular”  (Midgley 2000 p 67). Similarly Jablonka and Lamb argue that in cultural evolution “There are no discrete unchanging units with unchanging boundaries that can be followed from one generation to the next.” (Jablonka and Lamb 2005 p 212) as though that disqualified the concept of memes.

Even stronger are the objections made by Richerson and Boyd to the idea that “culture must be divisible into tiny, independent genelike bits that are faithfully replicated” or “tiny snippets of information” (2005 p 60), but in defining memes as “discrete, faithfully replicating, genelike bits of information” (Richerson and Boyd 2005 p 6) they are departing far from the original definition. Some memes are discrete and some are not; some replicate with very high fidelity (e.g. printed text) and some do not (e.g. dance steps); some memes are like genes in some ways and not in others. Going back to the original idea of memes as information that is copied, or “that which is imitated” helps to avoid these unrealistic claims.

To some extent it is true that proponents of memes describe them as units, but this is at least partly because it’s so much easier to talk about entities than to talk about “that which is imitated” in the abstract. Just try rewording some simple sentences about memes without inadvertently implying units: “Memes are more accurately copied by fax than by phone”,  “Soduko spread rapidly across Northern Europe” or “The meme for texting is doing better in Britain than in the USA”. All these make sense, and may seem to imply units, but in no case need there be precise units, let alone “discrete unchanging units with unchanging boundaries” or “independent genelike bits”.

Dennett (1995) tackles the problem by defining the units of memes as “the smallest elements that replicate themselves with reliability and fecundity” (p 344) or as a cultural item “with enough Design to be worth saving – or stealing or replicating.” (1995 p 143). In art, for example, a whole gallery of works is too large a unit for selection pressures to work on; but a blob of pink paint is too small to be selected, enjoyed, disliked, photographed or thrown away. The single painting is the natural unit, as when we remember Picasso’s Guernica or buy posters of Monet’s Water Lilies. Styles of painting, such as cubism or impressionism, can also be copied and can therefore count as memes, but they can hardly be divided up into units and would be much harder to track than an individual painting – but this does not mean the concept is useless. Similarly with written text a single word is too short to copyright and an entire library too long, but we can and do copyright 3 word advertising jingles and 300,000 word books. Either can count as memes because either can be copied with sufficient reliability and fecundity for selection to operate.

The same applies to music. Everyone’s favourite example is the first four notes of Beethoven’s Fifth symphony, which have spread as though they have an independent existence all across the globe in cultures where the whole symphony is scarcely heard. Wilson (1999) asks “What do we gain by thinking of the first four notes of Beethoven’s fifth symphony as a powerful meme? … The ability to define fitness independently of what evolves saves the concept of natural selection from being a tautology. For the meme concept to escape the same problem, we must define cultural fitness independently of what evolves. If the first four notes of Beethoven’s fifth is a powerful meme only because it is common, we have achieved no insight.” (1999, p 206).

I agree. So what is fitness for a short melody? It is the ability to survive and reproduce, which in terms of music means being copied, stored and reproduced more frequently than other melodies. This ability certainly ought to be measurable independently of how common the melody is. Inspired by Wilson’s criticism, Larry Bull, Nick Rose and I tried to find out. We chose the first four notes of three other well-known symphonies as comparison, and tested children aged between 4 and 8 hoping to catch them before they were familiar with this music. They could press four keys to play any of the tunes in any order and as often as they liked until they got bored, and we asked them which they liked best and which they had heard before. We predicted that they would play the Beethoven more often than any other (thus giving it more chances of being copied) even if they did not like it best. The experiment entirely failed because we could not find children old enough to do the task but naïve to the music. Indeed one five year old said he knew all of the tunes because he’d heard them in Peter and the Wolf.

Nevertheless, the principle remains that, as happens in biology, it should be possible to measure the ability of memes to survive and reproduce independently of how common they are, to understand their behaviour, and to predict how they will perform in new environments. Studies of this kind have been done, for example, in such widely different areas as the competition between different chemical terms in Western science and in China (Wright 2000), the survival of replicating text in the Internet (Pocklington and Best 1997) and the competition between different kinds of notation in music (Gersh 2007). Although not using the concept of memes, Wimsatt (?) has explored the evolution of Punnett squares as a technique for visualising problems in genetics. He treats the squares as model organisms for studying cultural evolution, and shows how the way they exploit human cognitive capacities helps them to attain high scientific fitness and rapid evolution. Such studies show that the problems of defining units and drawing analogies do not prevent effective research.

What is a meme?

The literature is full of arguments about what counts as a meme, as well as lists of examples, including ideas, behaviours, objects and brain states. Dawkins mentions habits, skills and stories, and Dennett lists “arch, wheel, wearing clothes, vendetta, right triangle, alphabet, calendar, the Odyssey, calculus, chess, perspective drawing, evolution by natural selection, impressionism, “Greensleeves”, and deconstructionism” (Dennett 1995 p 344).

Can a useful theory really be built out of such a curious assemblage? I agree with Dennett that it can. The way to see what unites them is to return to the origin of the term meme as “that which is imitated” or more generally “that which is copied”. An easy way to decide whether something is a meme or not is to ask yourself whether you copied it from or to someone else or invented it for yourself. If it was copied from person to person, or person to book or computer or artefact, then it’s a meme: otherwise not. In this light Dennett’s examples make sense. All of them (with the possible exception of wearing clothes) are words, concepts, technologies, skills or stories that are copied from person to person. They have all evolved over varying lengths of time and are survivors in the competition to be copied and stored.

I mention wearing clothes as a possible exception because different people might independently discover this useful ability rather than copying it from each other. In that case it would not count as a meme because no information has been copied from person to person. Alternatively they might pick up the general idea but then use totally different things as clothing. Culture is full of examples of this kind, such as the famous case of Sequoyah, the Cherokee Indian who reputedly invented the Cherokee syllabary after seeing Europeans reading and writing. In this case trying to pin down the meme (or what exactly was copied) is probably impossible, but the principle remains clear that Sequoyah created his writing system because he had seen a previous one, and so writing spread through many more people than it otherwise would, having effects on a whole society as it did so.

To give more modern examples, many learned skills entail a complex mixture of individual learning and memetic transmission. Riding a bike means watching someone else do it but then one has to learn the physical skills of balance, steering and braking for oneself through operant conditioning. The same can be said of driving a car, gardening, cooking, surfing, using a word processor or painting a portrait. None of these things can be done without the evolved scaffolding provided by the complex society we live in (Wimsatt 1999). The bike itself is a meme, copied by complex processes involving raw materials, technology, other people, and factory processes; riding a bike is a meme, copied by one person seeing another doing it and wanting to do the same, but the physical skill is acquired by individual practice and is not a meme.

Does this differ from other theories of cultural evolution, and is there any point in thinking memetically here? The difference is that memetics treats all of the copied information and its products as selfish replicators, spreading for their own sake. The bikes, cars, gardens, kitchens, surf boards, computers and paint brushes are all there because they have won in the competition to get humans to copy them. They, in turn, provide the selective environment in which more memes evolve.

This relates to an interesting debate in memetics over the extent to which memes are really copied as opposed to being reconstructed by the imitator. Jablonka (1999) points to the distinction between reproduction and replication of behaviours, and Sperber (2000 p 164) argues that a “fundamental objection to the meme model” is that most cultural items are “re-produced” in the sense of being produced many times, but are not “reproduced” in the sense of being copied from one another. He gives the example of two drawings, one a more or less random scribble, the other a slightly wonky five pointed star. What happens when people copy them? The first is hard to copy and results in a chain of copies in which each one differs slightly from the previous one. The second is easy to copy because people recognise it as a familiar shape and can re-produce (i.e. produce again) a shape they have seen and drawn many times before. As Dawkins (1999) points out, you could ask an independent observer to try to put the copies into the order in which they were made. In the first case they would be likely to succeed, in the second case they would not. The second case is “self-normalising”. Sperber sees this as a serious objection to memetics because the drawing is not actually copied but is “re-produced”, whereas Dawkins sees it as an example of the increasing fidelity of transmission as people learn to draw and to think about familiar shapes.

A new replicator or culture on a leash?

Questions about reproduction and copying are important because of the central claim of memetics that imitation creates a new replicator (i.e. information copied with variation and selection). It may help to think in terms of a continuum. At one extreme are events that are pure re-production. Sperber gives the example of laughter which is not a meme. Laughing when someone else laughs, like other examples of contagion (e.g. yawning or smiling), involves no copying. The second person’s laugh is triggered by hearing the first, and is their own laugh, not a copy of the way the first person laughed (that is, unless he is deliberately mimicking him for fun). Contagion occurs in other species too and can be useful, as when one individual picks up a warning cry from another, or flees because another does. This is entirely triggered re-production not meme replication. Then there are examples in which a combination of individual learning and social learning (but not true imitation) end up in cultural transmission that is not memetic. The classic example here is the pecking of milk bottle tops by tits in England. This novel behaviour was tracked from its beginnings, in several independent locations, in the 1940s (Fisher and Hinde 1949). Although the behaviour clearly spread from one bird to another, this is thought to have been by stimulus enhancement and not true imitation (Sherry and Galef 1984): that is, one bird discovers, by luck or accident, that there is cream underneath the foil bottle top. This bird then leaves a jagged, shiny foil top which is highly visible to other birds, so they come to investigate, thus learning that foil tops are associated with cream. They then land on unbroken foil tops and do what they always do – peck, and find more cream. And so the behaviour spreads. Is this a meme? No because there is no copying with variation and selection, and so no replicator and no cultural ratchet effect (Tomasello 1999). The tits do only what they always do, peck for food: all they have learned is to associate foil tops with food.

At the other extreme consider learning to dance. At first everything is unfamiliar; the steps, the rhythms, the precise sequence of movements all have to be watched carefully and then reproduced. Trial, error, self-assessment and hard work all result in the student being able to dance the same steps as the teacher. By the time the student is an accomplished dancer, new dances can be learned easily. This becomes more like the five-pointed star than the random scribble. The student has learned all sorts of routines that can be mixed and matched to make new dances. By this stage the process may look quite different, but if anything it more clearly reveals the principle of replicators at work.

Richerson and Boyd (2005) disagree, stating emphatically that “Cultural variants are not replicators” (p 82), thus effectively rejecting memetics. Part of their argument is that the peculiarities of biased transmission, behavioural attractors and error prone imitation are reasons to reject the idea of memes as replicators, but I suggest these only show how complex memetic transmission must be. Their other contention is that copying must be perfect for a replicator to count as such, and imitation does not produce perfect copies. But this makes no sense. If copying were perfect there would be no variation and no evolution. More interesting is to ask how high the fidelity has to be for an evolutionary process to get off the ground. If human imitation is good enough to support cumulative evolution then we should be justified in treating memes as replicators. And clearly, whatever its shortcomings, it is.

This question of whether memes are replicators or not is crucial. Although Richerson and Boyd’s theory is very close to memetics in some respects, for example they occasionally use the term “selfish meme”, and describe us and our culture as like obligate mutualists, they still maintain that “Culture is an adaptation” and that “Culture is on a leash, all right” even if the dog on the end is big and clever. This refers to Wilson’s famous claim that “genetic natural selection operates in such a way as to keep culture on a leash.”  (Lumsden and Wilson 1981 p 13). Here we see the fundamental difference between memetics and other theories of gene culture coevolution. For most theories, culture is an adaptation of benefit to human genes, and ultimately kept under control by them (i.e. on a leash). For memetics, memes are not (and never were) an adaptation; they are a new replicator that was accidentally let loose by the human capacity for imitation. In 1976 Dawkins complained of his colleagues that “In the last analysis they wish always to go back to ‘biological advantage’.” (Dawkins 1976, p 193). This is what Richerson and Boyd wish to do, and why their theory, although close, is fundamentally different from memetics.

Do memes have memotypes?

In asking “what counts as a meme” I listed many widely differing examples, and these raise another important question; are memes inside brains, outside of brains, or both? For example is the meme for “Cinderella” supposed to be the story itself, the spoken words, the written words, or their representation inside people’s heads. Should we take an analogy from genes and count the representation inside someone’s head as the actual meme and the written story as something like a meme phenotype, or vehicle or interactor? Wimsatt asks

“How would we know when a meme corresponded to a gene? a chromosome? a gamete? a genotype? or the gene pool or pools of a family? Or even of a whole ecosystem?” (Wimsatt, 1999 p 284). Or are these all false analogies?

Proponents and critics have argued each way on this one and no consensus has been reached. For example, in his original formulation Dawkins made no distinction between memes and their vehicles. Later, in The Extended Phenotype he revised this to make “the distinction between the meme itself, as replicator, on the one hand, and its ‘phenotypic effects’ or ‘meme products’ on the other. A meme should be regarded as a unit of information residing in a brain.” (1982, p 109). These two views are sometimes known as “Dawkins A” and “Dawkins B” (Gatherer 1998).

Among followers of “Dawkins B” are those who distinguish between the actual meme and its products or vehicle or, using Speel’s (1997) terminology, they distinguish between the memotype and the phemotype. The first to do this, even before Dawkins’ coining of the term ‘meme’, was Cloak (1975) who distinguished the i-culture (cultural instructions inside people’s heads) from the m-culture (the products of those instructions); the ultimate function of both being to maintain the i-culture. Delius (1989) and Aunger (2002) both propose that memes are some form of neural pattern inside the brain, and that their products or vehicles are outside of the brain. By contrast Benzon (1996) argues precisely the opposite, that the replicators are outside brains and their vehicles inside.

This contentious issue has been used by some critics to explore the difficulties and limitations of memetics (e.g. Wimsatt 1999) and by others to reject memetics altogether (e.g. Jablonka and Lamb 2005)|. Calling them “the dreaded memes”, Jablonka and Lamb argue that “The flaw stems from the distinction that is made between replicators (memes) and their vehicles (human brains, human artefacts, and humans themselves are all given this role).’ (2005, p 208). However, not everyone makes this distinction. Blackmore (1999) and Dennett (1995) do not. And Speel (1997) argues that “ideas, songs and norms reside in the brain somehow, as they are copied as units from human to human, but also from brains to books, or to physical air-waves… under the definition which makes physical memes replicators, brain-memes are also replicators.” (1997).

Closely related to this issue are two other common objections: that memetic inheritance is Lamarckian (depending on the inheritance of acquired characteristics) and that cultural variation is directed (Aunger 2000, Richerson and Boyd 2005). For some critics these two points seem to rule out memetics as a valid enterprise, while others are more concerned with whether culture can evolve if these two claims are true (Kronfeldner 2007). I think the concern over Lamarckism is misplaced because the whole idea of calling memetic evolution “Lamarckian” rests on drawing an inappropriate memotype/phemotype distinction.

Hull (2000) argues that since memes are replicators, not interactors, they are analogous to genes and not to phenotypic traits. So how you view such things as words, stories, or clothes fashions depends where you look from. From the perspective of genes, they are part of the phenotype, but from the memes’ eye view, they are memes. I think the whole problem stems from drawing too close an analogy between genes and memes. In fact these two replicators are different in important ways. Genes are neatly packaged inside bodies and passed on accurately down the germ line while their vehicles or interactors are not copied at all. For such a system Weisman’s law applies and there are good reasons why characteristics acquired by the phenotype are not passed on to the next generation (that is, leaving aside some forms of epigenetic effects). But most memes are not part of such a system and so the concept of Lamarckian inheritance simply does not apply (Blackmore 1999).

I think that, as with so much of memetics, it helps to take a step back and look at the processes in a very general way. Genes have ended up with a system of genotypes and phenotypes because it is more effective to copy the instructions for making something than it is to copy the product itself. This is not only because you can make multiple copies from the same instructions (so increasing fecundity) but because any accidents that befall the product, for example in its construction or during its lifetime, are not passed on (so increasing fidelity). Dennett (1995) describes this as a system for retaining any “good tricks” that the evolutionary process stumbles upon. By contrast most memes have not yet evolved such a system (but see below). Similarly genes have evolved a system of blind variation which is more efficient than the varied sources of variation occurring among memes.

Old genes: New memes

All of the objections and confusions I have discussed here can be clarified by remembering that genes are old and memes are relatively new. Genes are part of an ancient system that has evolved over about four billion years into the amazingly complex and high-fidelity system that we see today in which a phenomenal number of different kinds of creatures are all built from the recombination and mutation of information that is stored (almost entirely) in one single kind of molecule. Although we do not know precisely what were the predecessor chemicals, it is clear that DNA evolved from simpler systems into the modern system of genotypes and phenotypes, meiosis and mitosis, blind variation and natural selection (Maynard Smith and Szathmáry 1995). The whole process can be seen as a long and successful coevolution between a replicator and its copying machinery.

By contrast memes have been around for only a few million years. So perhaps it is not surprising that much of the copying is of low fidelity, much of it entails copying-the-product rather than copying-the-instructions (Blackmore 2001), generations are not clearly distinct, transmission is piecemeal, and new copying machinery is appearing all the time. Yet we might expect memetic evolution to be improving and indeed it appears to be doing so.

For example, memetic equivalents of a germ line and phenotypes have appeared in several domains. Take music as an example. Originally people would have listened to others singing or playing and then copied the sounds as best they could. Some people still do that, and in this case there is no replicator/vehicle distinction. But most music is now produced in recording studios and then copied in factories onto various media, ending up listened to in people’s homes or on mobile phones, MP3 players and other personal music systems. The information on a CD or music file is instructions for making sounds. When someone hears music they like, they don’t copy the sounds directly but go and buy or download the instructions from somewhere else. So this is all copy-the-instructions.

The same applies to cars, clothes, fridges, furniture and almost all the household goods we take for granted. Few people make their own; instead they see ones they like and then go and buy similar, or even identical, ones. If lots of people buy a particular model of car then the factory makes more from the production line it already had. It does not copy the cars already out on the streets.

Another change is the increase in fidelity and longevity created by the switch from analog to digital storage. DNA is a digital system of very high fidelity. Human brains use both digital and analog systems (e.g. information is carried by neurons either firing or not, but also as rates of firing, and the spatial summation of electrical potentials across membranes). Early memes were mostly analog but technological memes are far more often digitised, and increasingly so. Similarly we may guess that culture is evolving towards systems that depend on blind variation rather than directed variation. We can already see this happening in some computer software and on the Web, where software randomly recombines different memes to make new ones and then subjects them to selection. At the moment most selection is still done by human beings, although processes of data mining and search engines are rapidly taking over some of this process. Eventually we may suppose that all three processes, copying, variation and selection, will be taken over by machinery outside of human brains. At this point a third level replicator will have emerged from the first two. These “technological memes” and their machinery may even deserve names of their own, such as “temes” and “teme machines” (Blackmore in press).

Religions, cults and viral information

Memetics is best known in popular culture for its treatment of viral information. As Dawkins originally pointed out, chain letters, pyramid schemes and email viruses are all information that replicates for its own sake rather than for the benefit of the people who do the copying. The structure of all of these viruses is the same: a copy-me instruction backed up with threats and promises. This, he argued, is precisely the structure of the major religions as well (Dawkins 1993). Christianity and Islam, for example, have effective mechanisms for ensuring they keep infecting new hosts, many of which are familiar from biology (though again we must be careful with analogies). Many religions have mechanisms for ensuring that their central beliefs are passed on together in a package (e.g. confirmation classes or learning the catechism), and that the package is not picked apart. Threats and promises are used to ensure compliance (e.g. heaven, hell, or excommunication), and to prevent people from throwing off the infection (e.g. death sentences for apostates), and many other tricks are used such as piggy-backing beliefs on beautiful art and music, and persuading followers that they are superior, more altruistic or more spiritual (Blackmore 1999).

Cults, religions, and alternative belief systems may be thought of as lying on a continuum from the purely viral, such as the most dangerous cults, to the symbiotic, and even beneficial. Where any particular example lies must depend on the effects it has on the people who carry and propagate it, and on the groups who espouse it (Wilson 2002). There is now increasing evidence of the harm that religions do to both individuals and societies (Harris 2006, Paul 2005), although also of some benefits such as providing social cohesion and benefits to the group (Wilson 2002).

If memetics is to be useful in the analysis of belief systems it should be able to make predictions dependent on treating cults and religions as evolving systems parasitic on human carriers. For example, it should be able to predict which meme tricks religions need to use to survive in different circumstances, such as the difference between vertical transmission (as in Judaism and Catholicism for example) and horizontal transmission (as in evangelical sects). Dawkins (1976) discusses the power of celibacy in ensuring that priests spend their time and energy promoting their memes rather than their genes, and the effectiveness of martyrdom as a gene-destroying meme-spreader. Lynch (1996) considers rules against masturbation as a way of promoting more offspring in vertically transmitted religions. An effective memetics should be able to understand these processes and predict the kinds of culture in which different systems will thrive.

Human evolution

Memetics provides a completely different way of thinking about human evolution from other theories. The fundamental difference is that culture is seen, not as an adaptation of benefit to early hominids and their genes, but as a parasitic second-level replicator that appeared when our ancestors became capable of imitation. From then on the two replicators evolved together, the memes restructuring human bodies and brains to become better at propagating the very memes they copied. In this view the turning point in human evolution is not seen to be the advent of tool use, or language, or symbolism (see e.g. Deacon 1997, Dunbar 1996, Jablonka and Lamb 2005), but the advent of imitation of sufficiently high fidelity that it inadvertently let loose memes: intelligence and language are seen as consequences of this transition rather than its causes.

I have used the term ‘memetic drive’ to describe the way this might have operated (Blackmore 1999, 2001), and this process is different from those proposed in gene-culture co-evolution theories, in evolutionary psychology, or in theories based on sexual selection (Miller 2000). The idea is that once imitation became good enough, memes started spreading, possibly including ways of using fire, cooking food, wearing clothes or making artefacts. These memes competed to be copied, with the winners affecting the selective environment such that people incapable of copying them would be at a biological disadvantage. This alone provides pressure for imitation ability to increase and, assuming that imitation requires brain space, pressure for increasing brain size.

However, the difference from other theories is that in the memetic view the direction taken by the evolution of the memes affects biological evolution too, and this is true not only for useful memes such as lighting fires, but ones that are useless for survival such as wearing feathers in your hair. So, for example, if a fashion for performing ineffective rituals, wearing particular animal skins, or humming tunes took off then people who could not copy these skills would lose out. Many of the assumptions here are similar to those in other theories (e.g. Lumsden and Wilson 1981, Richerson and Boyd 2005) but the main difference is that memes are seen a replicators that evolve for their own sake and the direction they happen to take then drags genetic evolution along. A common example used in gene-culture coevolution is the effect of dairying practices on genes for lactose tolerance (Durham 1991) but this, like many such examples, is confined to a one-dimensional change. The equivalent for memetic drive would be if the evolution of cheeses and yogurts, and eventually TV dinners, microwaved chips, greasy hamburgers or hot curries, drove complex changes in human digestive systems such that those systems showed traces of the history of the memetic evolution that caused them.

This is what I suggest happened with language, as well as the evolution of music, religious behaviours, and many other aspects of human nature. That is, over the past two or three million years, memes evolved by competing with each other, and the winners sculpted the direction of biological evolution to create brains that were not only larger but were better adapted to copying the very memes which drove these changes. In the case of language, people began copying sounds, sounds of higher copying fidelity increased in the meme pool (for example by processes of digitising sound streams into discrete words, or linking sounds with seen objects) and then people who were good at copying those sounds thrived and passed on any genes conducive to that ability. In this way brains gradually became better and better at copying the particular kinds of sounds that won in the memetic competition. In this view language itself is seen as a parasite turned symbiont using human brains and vocal apparatus as its copying machinery. The same can be said of music, religion and art; that all these memetic systems have evolved in their own ways and so helped design human brains that are fond of music, ritual and art.

One implication of this approach is that any system capable of imitating sounds and following gaze should give rise to its own language (Blackmore 1999). Perhaps surprisingly this has turned out to be true for both simple imitating robots and autonomous agents (e.g. Steels 2000, 2006). That is, language and reference can evolve without its being of any benefit to its carrier. In other research, simulations and computer modelling have explored the proposed interactions between the two replicators and the possible effects on brain size (Bull, Holland and Blackmore 2000, Higgs 2000). Another prediction is that the parts of the human brain that maximally increased in size should also be those involved in imitation, and this has been confirmed by brain scanning studies (Iacoboni et al 1999).

The human brain was the first meme machine but now other, far more effective, meme machines are being created. Initial small steps towards increasing the longevity, fecundity and fidelity of memes included the invention of writing, printing, roads and railways, telephones and fax machines. All these developments can be seen as memes co-evolving with their copying machinery. Now the step to a third-level replicator can be discerned. We can look at it this way; the first-level replicator (genes) constructed vehicles (Dawkins 1976) or interactors (Hull 1988) that protected and propagated them. One of these vehicles became capable of imitation and so let loose a second-level replicator (memes). That is, the vehicle from the previous level became the copying machinery for the next. A similar step may be evident in the way that computer systems that began as vehicles for storing and manipulating memes are fast becoming capable of carrying out all three processes of copying, varying and selecting information. This would make them true teme machines, and we should not expect them to be subservient to the old meme machines (Blackmore in press).

Consciousness, creativity and the nature of self

A common objection to memetics is that it undermines human autonomy and the creative power of consciousness, and treats the human self as a complex of memes without free will. These ideas follow naturally from the universal Darwinism on which memetics is based. That is, the idea that all design in the universe comes about through the evolutionary algorithm and is driven by replicator power. This means that human creativity emerges from the human capacity to store, vary and select memes, rather than from some special creative spark, or power of consciousness (Blackmore 2007). The human self may also be a construct of memetic competition, surviving because it protects and propagates memes, including the many memes that make up a person (Dennett 1995). In this view the self is not a continuously existing entity with consciousness and free will but is a persistent illusion. This memetic view of human beings as the evolved creation of two replicators may be unsettling but it has the advantage of uniting biological and human creativity into one, and providing new ways of understanding human nature, self and consciousness.

Conclusion

These are just a few examples of a meme’s eye view on human and cultural evolution. I have outlined most of the major objections to memetics and shown why none of them effectively prevents memetics from being a viable scientific theory. In terms of the wider view of human nature and evolution, memetics clearly provides a novel way of looking at the world. But is it truer than existing models? The real test for memetics is whether its novel hypotheses and predictions can be tested and how they will fare in those tests. For that it is still too early to say.

References

Aunger, R.A. (Ed) (2000) Darwinizing Culture: The Status of Memetics as a Science, Oxford University Press

Aunger, R.A. (2000) Introduction to Aunger, R.A. (Ed) Darwinizing Culture: The Status of Memetics as a Science, Oxford University Press

Aunger, R.A. (2002) The Electric Meme: A New Theory of How We Think. New York, The Free Press

Benzon,W. (1996) Culture as an evolutionary arena.  Journal of Social and Evolutionary Systems,  19, 321-362

Blackmore,S.J. (1999) The Meme Machine, Oxford, Oxford University Press.

Blackmore,S. (2001) Evolution and memes: The human brain as a selective imitation device. Cybernetics and Systems, 32, 225-255.

Blackmore, S. (2007) Memes, minds and imagination. In Imaginative Minds (Proceedings of the British Academy). Ed. Ilona Roth, Oxford University Press, pp 61-78

Bull,L., Holland,O. and Blackmore,S. (2000) On meme-gene coevolution. Artificial Life, 6, 227-235

Darwin,C. (1871) The Descent of Man and Selection in Relation to Sex.London, John Murray

Dawkins,R. (1976) The Selfish Gene Oxford, Oxford University Press (new edition with additional material, 1989)

Dawkins,R. (1982) The Extended Phenotype. Oxford, Oxford University Press

Dawkins,R. (1993) Viruses of the mind. In B.Dahlbohm (ed) Dennett and his Critics: Demystifying Mind. Oxford, Blackwell

Dawkins, R. (1999) Foreword to The Meme Machine S.J. Blackmore, Oxford, Oxford University Press.

Deacon,T. (1997) The Symbolic Species: The Co-evolution of Language and the Human Brain. London, Penguin

Delius, J. (1989) Of mind memes and brain bugs, a natural history of culture. In W.A.Koch (Ed) The Nature of Culture. Bochum, Germany, Bochum Publications. 26-79

Dennett, D. (1995) Darwin’s Dangerous Idea, London, Penguin

Dunbar,R. (1996) Grooming, Gossip and the Evolution of Language. London, Faber & Faber

Durham,W.H. (1991) Coevolution: Genes, Culture and Human Diversity. Stanford, Ca., Stanford University Press.

Fisher,J. and Hinde,R.A. (1949) The opening of milk bottles by birds. British Birds, 42, 347-357

Gatherer, D. (1998). Why the Thought Contagion Metaphor is Retarding the Progress of Memetics.  Journal of Memetics – Evolutionary Models of Information Transmission, 2.

Gersh, J. (2007) Selection Pressure, Memetic Drift, and Progress in Musical Evolution, Paper presented at the conference “Music and Evolutionary Thought”, Durham, June 2007.

Harris, S. (2006) The end of faith: Religion, terror, and the future of reason. Free Press

Higgs, P.G. (2000) The mimetic transition: a simulation study of the evolution of learning by imitation. Proceedings of the Royal Society London B., 267, 1355-1361

Hull,D.L. (1988) Interactors versus vehicles. In H.C.Plotkin (Ed) The Role of Behaviour in Evolution. Cambridge, Mass, MIT Press

Hull,D.L. (2000) Taking memetics seriously: Memetics will be what we make it. R.A.Aunger (Ed) Darwinizing Culture: The Status of Memetics as a Science, Oxford University Press, 43-67

Jablonka,E. (1999) Between development and evolution: the generation and transmission of cultural variations. Paper presented at “The Evolution of Cultural Entities” The British Academy, April.

Iacoboni,M., Woods,R.P., Brass,M., Bekkering,H., Mazziotta,J.C. and Rizzolatti,G. (1999) Cortical mechanisms of human imitation, Science, 286, 2526-2528

Jablonka, E. and Lamb, M.J. (2005) Evolution in Four Dimensions: Genetic, Epigenetic, Behavioral and Symbolic Variation in the History of Life. Bradford Books

Kronfeldner, M.A. (2007) Is cultural evolution Lamarckian? Biology and Philosophy, 22, 493-512

Lumsden,C.J. and Wilson,E.O. (1981) Genes, Mind and Culture. Cambridge, Mass., Harvard University Press.

Lynch,A. (1991) Thought contagion as abstract evolution. Journal of Ideas, 2, 3-10

Maynard Smith,J. and Szathmáry,E (1995) The Major Transitions in Evolution. Oxford, Freeman

Midgley,M. (2000) Why memes? In. H. Rose and S. Rose (Eds) Alas, Poor Darwin, London, Cape, 67-84

Miller, G. (2000) The Mating Mind: How Sexual Choice Shaped the Evolution of Human Nature, London, Heinemann

Paul, G.S. (2005)   Cross-National Correlations of Quantifiable Societal Health with Popular Religiosity and Secularism in the Prosperous Democracies: A First Look. Journal of Religion and Society, 7, 1-17

Plotkin,H. (1993) Darwin Machines and the Nature of Knowledge. Cambridge, Mass., Harvard University Press

Pocklington,R and Best,M.L. (1997) Cultural evolution and units of selection in replicating text. Journal of Theoretical Biology, 188, 79-87

Richerson, P.J. and Boyd, R. (2005) Not by Genes Alone: How culture transformed human evolution,  Chicago, University of Chicago Press

Sherry,D.F. and Galef,B.G. (1984) Cultural transmission without imitation: milk bottle opening by birds. Animal Behavior, 32, 937-938

Speel, H.C.A.M. (1997) A short comment from a biologist on William Benzon’s essay ‘Culture as an Evolutionary Arena’. Journal of Social and Evolutionary Systems 20 (3): 309-322

Sperber, D. (2000) An objection to the memetic approach to culture. In Aunger, R.A. (Ed) (2000) Darwinizing Culture: The Status of Memetics as a Science, Oxford University Press, 163-173

Sperber, D. (2006) Why a deep understanding of cultural evolution is incompatible with shallow psychology. In Nick Enfield and Stephen Levinson (eds.) Roots of Human Sociality, 431-449

Steels,L (2000) Language as a complex adaptive system. Lecture Notes in Computer Science. Parallel Problem Solving from Nature – PPSN-VI. Volume Editor(s): Schoenauer et al, Springer-Verlag, Berlin.

Steels, L. (2006) How to do experiments in artificial language evolution and why. In Proceedings of the 6th International Conference on the Evolution of Language, pages 323–332.Tomasello, M (1999) The Cultural Origins of Human Cognition, Harvard University Press.

Wilson, D.S. (1999) Flying over uncharted territory. A review of The Meme Machine by S. Blackmore. Science, 285, 206

Wilson, D.S. (2002) Darwin’s Cathedral: Evolution, Religion and the Nature of Society, University of Chicago Press

Wimsatt, W. (1999) Genes, memes, and cultural heredity. Biology and Philosophy, 14, 279-310

Wimsatt, W. (?) The Analytic Geometry of Genetics: Part I:  The structure, function, and early evolution of Punnett squares

Wright, D. (2000) Translating Science: The Transmission of Western Chemistry into late Imperial China, 1840-1900 Brill, Leiden