Recently there has been a lot of talk about the snippets of language and culture that are known as memes. They pop up in the newspaper, on television, and most of all on the Internet. In this essay I’ll briefly discuss the explanatory role that meme theory is supposed to provide, then I’ll argue that there are some gaps in the evidence supporting this particular explanation.
Richard Dawkins originally introduced the concept of memes as an illustration of evolution outside the biological domain. He postulated that the framework of evolution by natural selection was not specific to biology, but could actually be applied wherever one found replicators, so long as two conditions were met:
- Heritability with error. The replicators must contain information that is passed on during replication, and errors must occasionally be introduced during this process.
- Selection. Some replicators must systematically produce more copies of themselves than others.
Biological genes meet these criteria since the information they embody is represented in their genome and is passed on via reproduction, and the natural environment provides selection pressure in the biological domain. In The Selfish Gene, Dawkins went to pains to explain that in the case of biological evolution it is genes, not organisms (which contain many genes) that are the replicators, and hence we should not be confused by apparent mal-adaptations such as the peacock’s tail, which are explained easily from a gene view but create difficulty for the organism view.
Dawkins also pointed out that his criteria are not merely necessary for evolution, rather they constitute sufficient conditions for evolution. That is, if the two criteria are met then evolution will occur, no more questions. The reason for this is that any system that meets the criteria will naturally explore a search space via replication, and selection will bias that search towards regions representing greater propensity to replicate. Without any further ingredients, any such system will evolve.
This phenomenon has been put to use by computer scientists, who use it to solve abstract optimisation problems. Genetic algorithms, also known as evolutionary algorithms, solve problems by simulating artificial genes (typically represented by sequences of bytes in the computer’s memory) in an artificial “environment” defined by a fitness function. The programmer specifies a fitness function appropriate to the problem at hand; then, beginning with a randomly generated population, the algorithm proceeds to pick the most promising individuals (as defined by the fitness function), which, by analogy to biological reproduction, are “recombined” to create the next generation. Of course, this reproduction is entirely artificial – the algorithm just mixes the bytes from the fittest individuals to create the next generation – but artificial or not, this system meets the criteria for evolution and indeed evolutionary algorithms can solve problems for which few other optimisation algorithms perform well.
Another type of replicator that meets the criteria for evolution is ideas. Humans pass ideas from one to another, so every time you repeat a catch-phrase, concept, or idea in conversation you are participating in the “replication” of that idea. Of course, the idea will often come out slightly differently to when you heard it, so ideas meet the criteria of heritability with error. What about the second criteria? Well consider how much more likely you are to pass on a funny joke or a catchy phrase or an insightful idea, in comparison to a dull, lame, or obvious one. Some ideas are stickier – they appeal to us, play to our sensibilities, resonate with us – and that is what constitutes selective pressure for ideas. So according to Dawkins’ argument, having seen that the relevant criteria are satisfied, we can now conclude that ideas, or memes, will evolve to be ever stickier to the human mind.
The meme concept has been received with much applause, and has been applied with great vigour to explain many aspects of human culture, from phrases invented by internet communities, to religion, and even to scientific theories. But is this a valid conclusion to draw from the analysis of the previous paragraph? Dawkins’ argument demonstrates that evolution of memes will occur, but it doesn’t tell us which particular phenomena we can attribute as caused by that evolutionary process. This is the crucial step in the argument that I believe requires further evidence; that is, the step from “evolution will occur among replicator class X” to “phenomenon Z was caused by evolution among replicator class X”. Dawkins’ criteria can get us to the first step but not to the second. To make the jump to the second step we need additional evidence.
As an example, consider the following (absurd) example. Pretty streams that flow through picturesque countryside naturally appeal to our sense of beauty. In time, people will tend to reshape other streams in the image of the prettiest streams they have seen in the past. Pretty streams will in this way reproduce themselves, with humanity’s sense of beauty providing the selection pressure. Therefore, the Nile – surely one of the prettiest streams in the world – must have been produced by evolution-of-beautiful-streams. Let’s call this exciting new phenomenon stremes!
The conclusion of the argument above is pure nonsense: the beauty of the Nile was produced by geological forces, not humans, and certainly not evolution. But if the conclusion was wrong then either one of our premises must have been wrong, or else there must have been a mistaken step somewhere in the argument. So what exactly went wrong? It is certainly true that people do have a sense of beauty and that some streams are naturally more beautiful (to humans). It is also true that humans reshape their environment (in part) according to appealing environments they have seen before – the spread of landscaping trends over the past few hundred years bears witness to that. Despite the absurdity of the stremes example, it does satisfy Dawkins’ criteria for evolution, and indeed if we imagine a world in which the only force is humans industriously replicating their favourite streams, occasionally introducing errors, then, yes, we actually could imagine an evolutionary process leading to ever prettier streams. It is true that stremes could evolve, but the real problem with the argument I gave was in the final conclusion, that a specific phenomenon (the Nile river) was caused by this type of evolution. Perhaps something similar to the Nile river could eventually come about as a result of evolution-of-stremes, given a sufficiently conducive environment and enough time, but that is different to the question of whether the actual Nile river did come about as a result of evolution amongst stremes. Indeed we have very good reason to believe that the Nile came about for reasons completely unrelated to evolution. There is a gap between the statement that “stremes will evolve given the right circumstances”, and the statement that “the Nile river was caused by evolution among stremes”. And that gap can only be bridged by extra evidence.
The same evidence gap exists in the case of memes. The argument for memetic evolution is valid: Dawkins’ criteria are satisfied and a straight forward argument tells us that memes will evolve. However, to conclude that any particular phenomenon such as Rick-Rolling or Buddhism or silicon chips was caused by evolution of memes requires further evidence.
At this point it may be worth returning to the case of biological evolution and asking on what grounds we conclude that the specific phenomenon of life on Earth is a result of biological evolution. Well, I’m glad you asked. One source of evidence is the fossil record, which shows a sequence of species that looks just as it should if evolution were the cause for life on earth. Another supporting observation is that the range of species found on each continent looks just as it should assuming an evolutionary process on separate land masses occasionally connected by land bridges. These pieces of evidence are independent of Dawkins’ criteria for evolution (which merely show that genes will evolve), and it is because of them that we conclude that life on Earth was caused by an evolutionary process.
Similar evidence is required to validate memetic explanations for any specific phenomena. If evolution amongst memes really was responsible for, say, Rick Rolling, then we should expect to see a relatively continuous sequence of memes in historical records, analogous to the fossil record. We should also expect never to find irreducible complexity in memes (that is, memes that could not have been caused by evolution because removing any part would destroy the meme’s fitness). I have not collected evidence that supports or denies memetic explanations for particular cultural phenomenon; the thesis of this essay is simply that it is invalid to assume memetic explanations in the absence of such evidence.
Another piece of evidence that supports biological evolution as the cause for life on Earth is that there is no other process we are aware of that provides a plausible alternative. But is the same true in the case of memes? Are we forced to accept memetic explanations for lack of any alternative? No, we are not; there are many alternatives to memetic explanations. For instance, the word “mum” has come to mean “mother” probably because “mum” is an easy sound for young children to make given the biology of the mouth and throat, not because it arose from evolution of memes passing between infants. That is, our use of the word “mum” is explained by biological facts, not memetic evolution. Don’t be confused by the fact that our biology itself arose through an evolutionary process: although biological and memetic evolution could surely interact, they are fundamentally different processes since they operate on different substrates: DNA versus human discourse.
Another candidate explanation for cultural phenomena is optimisation by humans. When Einstein proposed the special theory of relativity he did not simply make random permutations to the ideas of those before him; rather he took the best ideas in physics at the time and made directed, purposeful modifications. A memetic explanation for how the theory of special relativity came about would posit many physicists making small, random modifications to existing theories, followed by selection of the more promising theories for further modification. While the latter is quite plausible, the former is not: Einstein’s contribution to physics was anything but small and random; rather he used insight and rational inquiry to make contributions in a directed manner. That Einstein’s theory was significantly better than previous theories at explaining physical phenomena could only be accounted for in a memetic framework if there had also been literally billions of competitor theories that proved less useful, since in an evolutionary process a large increase in fitness comes about only as a result of a great many small increases, and an even greater number of small failures. Although there certainly were rivals to special relativity that were ultimately discarded, they number in the dozens, or at most in the hundreds: too small by many orders of magnitude for the memetic explanation.
If optimisation-by-humans is the cause of theories in physics then it is also a plausible explanation for more general phenomenon such as Buddhism, Rick Rolling, or silicon chips. I am not going to try to settle those particular cases in this article, the arguments herein are simply intended to demonstrate that optimisation-by-humans is at least as plausible as memetic evolution, and that therefore in the absence of further evidence it is unjustified to assume the memetic explanation.
The term “meme” has found widespread use in contemporary discourse, especially when discussing the public mindset, since the constitution of that mindset is what memetic evolution is supposed to explain. It is often used to refer to particularly catchy or trendy ideas, but I have argued that in many such cases there is little justification for assuming that an evolutionary process was responsible. Questioning whether the term “meme” should be applied in such cases process is dangerously close to a vacuous quibble over semantics; however, a few cautions do bear mentioning. First, it is a mistake to think that a deeper understanding has been reached just by calling something a meme. In the absence of evidence for an evolutionary process, calling something a “meme” is no different to calling it an “idea” or “phrase”. No greater understanding of its nature or origin has been reached by invoking the term, nor does the term suggest any new ways that it might be manipulated, magnified, or minimised. To talk of “injecting into the meme pool” is no different than just talking about plain old “publicity”. Second, we may miss evidence of alternative explanations for the things we label as memes, since the term implies that the ideas have a “life of their own” (which they would, but only if the memetic was correct).
In this essay I have argued that although memetic evolution is a coherent concept, applying it as an explanation for specific phenomena requires extra evidence to corroborate its causal role in producing those phenomena. In the absence of such evidence we should be careful about using the term “meme” too liberally since we may make unjustified assumptions about the nature of the ideas we are dealing with.