We all know memes. They’re the funny (and frequently not so funny) pictures of cats, socially awkward penguins, Wojaks and Bad Luck Brians that go viral on our message boards and social media feeds.
But, as many of you will know, the term “meme” extends beyond internet humor. It was originally coined by Richard Dawkins in his book The Selfish Gene (1976) to refer to “a unit of culture” or a “unit of imitation”. Memes in this sense are something we pick up by copying what we’ve seen other people say, do and create, such as words, fashions, jokes, ways of making clay pots and, of course, internet memes.
Dawkins’ aim in The Selfish Gene was to establish that evolution doesn’t primarily act on individuals, groups, or species, but rather on the gene level (in biology-speak, the claim is that the gene is the main “unit of selection”). This helped explain things that were otherwise mysterious, such as kin selection, where individuals may act in ways detrimental to themselves for the benefit of their kin. If evolution acts primarily on individuals, it’s hard to understand how sacrificing one’s own reproductive success in such a way could be selected for, but if one sees that genes for such behavior can benefit (since your kin who are genetically similar to you will have increased chances of survival) the mystery disappears, hence the “selfish” genes.
The meme is meant to be analogous to these selfish genes, in that they are the main unit of selection for cultural evolution. By applying Darwinian principles to these memes, we get an account of the spread and variation of cultural elements similar to how biological evolution explains the variation and origin of species. The field that studies this is called memetics.
For Dawkins, this whole meme business was something of an afterthought, introduced in the last chapter of his book mostly to illustrate the idea that Darwinian principles apply outside of biology as well (his doctrine of Universal Darwinism). But in the decades following The Selfish Gene, memetics caught on in a big way, as people realized the potential revolutionary ramifications of this seemingly simple idea. Richard Brodie’s book Virus of the Mind (1995) claimed that memetics would have at least as big an impact on our lives as the “emergence of atomic physics did in the Cold War”, and memes play a central role in Daniel Dennett’s account of consciousness in Consciousness Explained (1991). The zeitgeist reached its zenith in 1999 with the publication of Susan Blackmore’s The Meme Machine.
With this book, Blackmore presents a unified account of memetics, from the rise of the memes in the distant past through their impact on every aspect of our lives today, culminating in some rather gloomy speculation about what the future of memes will bring us. The Meme Machine is one of those books which wants to explain almost everything, including the origins of language, consciousness, the self, religion and technology, just to name a few. If her theories are right, it isn’t a stretch to say that they would have a huge impact on how we think about our minds, society and the future.
Yet, in the 25 years that have passed since its release, memetics has seen a steady decline. The online Journal of Memetics closed in 2005, and its last issue included an article claiming that memetics had failed completely to produce substantive results, and Wikipedia states that “many researchers of cultural evolution regard memetic theory of this time a failed paradigm”.
Some people however, including Dennett and Blackmore, still hold on to the revolutionary aims of memetics (since the time I wrote this, Dennett has sadly passed away). In this review of The Meme Machine, I will discuss Blackmore’s theories (and some of Dennett’s as well), and examine whether they live up to their promise. In particular, I will focus on three big questions, and show how memetics matter to them: the origins of language, consciousness and the evolution of institutions.
The review is divided into five sections:
- Meme basics: How Blackmore suggests memes arose, and her basic theory of memetics.
- The origins of language: The memetic theory of humanity’s ability to speak and understand language.
- Consciousness and the self: Exploring Dennett’s memetic approach to consciousness, and how it fuels Blackmore’s views about the non-existence of the self.
- Institutions: Adopting a memetic approach to explaining how institutions come to act for their own self-preservation, often at the cost of their stated purpose, and speculation about what this means for the future.
- Conclusion: Discussing if memetics is a worthwhile pursuit in the light of objections.
This is a lot of ground to cover, and the review is rather lengthy (roughly 14 000 words). I’d encourage you to read the first section, and then to jump into whatever interests you the most amongst the other topics, as they are written to be mostly independent. In the last part, I try to evaluate whether memetics is due for a revival in the light of the different memetics theories discussed. Be warned that the part about consciousness is slightly more technical than the others, so feel free to skip that if it isn’t your cup of tea.
Before delving in, I’ll also say a few words about the author. There are more than a few notable cases of brilliant scientists who after starting out doing good, serious research basically lose their minds to pseudoscience and woo. Rupert Sheldrake, Kary Mullis and John Lily all made significant contributions to science before turning their attention to “morphic resonances”, alien racoon abductions and Cosmic Coincidence Control Centers.
Susan Blackmore is sort of the opposite of this. When she was at Oxford as an undergraduate, she had a weed-induced out-of-body experience, in which she felt herself drifting away from the room she sat in (watching herself from above), out over the rooftops of her college (noting the design of the chimneys to be able to later confirm that what she saw was real), and eventually up into the cosmos, where she “explored strange scenes and entered mystical states beyond space and time”.
After the experience, she went to check if the chimneys matched what she’d seen, only to find that there weren’t any chimneys at all on her college rooftop. Despite this, the experience left her with an unshakeable belief that there are phenomena beyond what scientific materialism can account for. She went into parapsychology, trying to prove that things like precognition and ESP were real.
Few people who go so deep into the rabbit hole pseudoscience ever make it back to normality.
There’s simply too much sunk cost in terms of time and ego. But somehow, after many years of being unable to produce any evidence of the phenomena she was investigating, Blackmore changed her tune completely, denouncing parapsychology as pseudoscience and becoming a skeptic.
She instead turned to the question of trying to explain out-of-body experiences and paranormal beliefs from a scientific standpoint. She writes that “Instead of looking to the red herring of psi to answer my questions I tried to work out what had happened during [my out-of-body experience]”. She became famous for explaining experiences of alien abductions as instances of sleep paralysis, and for combating the idea that anything needs to physically leave the body in order for it to feel like something does.
To critics, Blackmore’s fondness of memetics amounts to a lapse back into the big but ultimately fruitless ideas of her parapsychology years. But I believe it’s actually her interest in combating pseudoscience, and in trying to explain why seemingly normal people (you don’t have to be out of your mind to feel like you’re out of your body) fall for bad explanations of what they experience, that led her to memetics, since it offers a simple account of how ideas which are incorrect and untestable may nevertheless be highly successful memes that spread not because they do anyone much good, but simply because they are good at getting into our heads, as we will shortly see.
1. Meme basics
The basic idea of evolution is well known. The survival of the fittest means that organisms best adapted to their environments are more likely to reproduce. This enables them to pass on the traits that made them successful to their offspring. Because of this, apparent design will appear over time, as a species increasingly adapts to its environment. Darwin famously marveled at the differences in beaks between different species of finches, which seemed specifically designed to obtain food from the food sources that were available in their respective habitats. Evolutionary theory allows us to explain such design without appealing to a conscious designer – all that’s required is that, over many generations, certain beak shapes that were more effective in accessing food led to greater reproductive success, gradually refining the beaks until they were optimally suited to their environment.
More generally, Blackmore holds that we have an evolutionary process whenever these three conditions are met:
- Variation - There is variation in which traits different elements have (i.e, different organisms have different genes)
- Selection - Depending on these traits, elements differ in how well they do in their environment, and not all get to survive/reproduce (i.e, some genes are better suited for the environment than others)
- Heredity/retention - Traits are passed on from one generation to the next (i.e genes are passed on from parents to their children)
As we see in the case above, variation, selection and heredity applies to genes. This makes genes a replicator, which just means that they are the type of thing that sustains an evolutionary process. Evolutionary processes are substrate neutral, meaning that as long as we have variation, selection and heredity, which medium the process takes place in doesn’t matter. For most of the history of life, genes have been the only real replicator. But according to Blackmore, there is now a second replicator, whose speed of evolution far outruns the genes. This second replicator is, of course, the meme.
How did this second replicator come into being? For Blackmore, the answer is imitation. According to her, imitation is a uniquely human ability, which only exists in very limited ways in the animal world. Being able to imitate each other, as even infants are hard-wired to do, is what distinguishes us from other species, more so than our intelligence, our consciousness or our language, (since Blackmore thinks all of these are the ultimate results of our ability to imitate). When we imitate each other, something is clearly passed on from person to person (a way of behaving, speaking, making things etc). For Blackmore, a meme is whateverit is that is passed on in imitation[64]. So, if I copy a gesture or a word or a way of weaving a basket by imitating someone, a meme has gotten replicated by getting itself into my head.
Clearly, there is variation in memes (there are innumerable different behaviors we could imitate, with large variation even in ways of weaving baskets). There is also selection – we cannot imitate everything, so memes have to compete to get into our heads, which some will do better at than others. Finally, there is heredity/retention, since when a meme is passed on (such as me teaching you to weave), something is retained (a way of weaving) or else we wouldn’t say that you had copied what I was doing. So, memes meet all the criteria of being a replicator. This means memes also sustain an evolutionary process, just as genes do, and so we should expect to see design without a designer when it comes to memes and culture as well.
One striking example of this is how technology can evolve to become better, simply due to the mechanism of natural selection on memes. Dennett likes to use the following quote to illustrate this:
Every boat is copied from another boat. Let’s reason as follows in the manner of Darwin. It is clear that a very badly made boat will end up at the bottom after one or two voyages, and thus never be copied. One could say then, with complete rigor, that it is the sea herself who fashions the boats, choosing those which function and destroying the others.
The point is that it is the boats that don’t sink that get copied by others. Therefore, it is these boat-memes that will flourish, meaning that boats will become better designed over time (less likely to sink). This is an example of technological improvement without anyone being the mastermind behind it, what Dennett calls “competence without comprehension”.
Another important aspect of memes is that they can group together in memeplexesthat give the constituent memes higher fitness than they otherwise would have. As Dawkins writes in the introduction to The Meme Machine:
Memes, like genes, are selected against the background of other memes in the meme pool. The result is that gangs of mutually compatible memes - [...] memeplexes - are found cohabiting in individual brains. This is not because selection has chosen them as a group, but because each separate member of the group tends to be favoured when its environment happens to be dominated by the others.
An example of a memeplex may be music genres – someone who already likes older rock music may more readily become host to a new Beatles song (compared to say a Skrillex song), and spreading that song may in turn result in a better environment for other memes of the same genre to spread. We shall see that memeplexes play a central role in our discussion of institutions later[65].
One of the crucial points of memetics is the relationship between memes and genes. Blackmore holds that, due to some selectional pressure or other, humans came to cross a certain threshold ability in imitation. Initially, this ability to imitate had to give us a genetic advantage in one way or another, or else evolution wouldn’t have selected for it. But once our general ability to imitate reached that threshold level, it gave birth to the second replicator, and what’s good for the memes doesn’t necessarily benefit our genes.
In fact, Blackmore claims that many of the ways we humans have evolved has been in order to become better spreaders of memes. She calls this memetic driving, which is when memes come to dictate the evolutionary direction taken by the genes. Her examples of this are, amongst other things, our big brains and the development of language.
Blackmore’s idea is basically this: Once we have some elements of culture, such as primitive tool making or ways of fishing, being able to imitate these things confers a big genetic advantage, and a similar disadvantage for those who just can’t learn the trick. Therefore, there is an increased pressure in favor of genes for good imitative abilities, which will increase towards fixation in the population (this is an instance of the Baldwin effect). This also means there’s a selectional pressure to mate with the best imitators, since their offspring are more likely to learn these survival-enhancing skills.
But the memetic twist on this is that memes evolve and compete as well, and not all memes are equally useful for the genes. If there’s twenty ways of fishing, they may range from useless (throwing sticks at the fish) to highly useful (nets), and so deciding who and what to imitate (and how to select the “best imitator” to mate with) will be vitally important.
According to Blackmore, these factors give rise to genetic filters that try to track which memes that are useful and which aren’t, based on principles like “imitate the successful”, “imitate those good at imitation”, “imitate things to do with food, sex and power” (since genes for selecting genetically useful memes will thrive). However, because there is competition amongst the memes as well, memes that aren’t useful to us may evolve to get past these filters.
So there is essentially an arms race between the memes and the genes, with the genes trying to track the useful memes and the memes evolving to get into our heads whether they are useful or not. A crucial aspect of this is that who gets to spread their genes will increasingly come to depend on whether they can pick up the memes that happen to be successful. Some memes may be easier to pick up for certain people than for others, for reasons other than general imitative ability. For example, if what is judged to be “good at imitation” involves dancing, this will favor those with genes for good coordination, given the principle that we should mate with the best imitators. Note that this also means that genes producing meme filters which for some reason judge dancing memes to be bad will be at a disadvantage, since people with those filters will be less disposed to pick up the dancing memes that would make them more attractive as partners. So which memes happen to be successful has an impact both on how our meme filters evolve, and on who gets to pass on their genes.
For Blackmore, it is this runaway scenario that creates our big brains. Trying to build better and better meme filters as memes evolve to get past them leads to extreme demands on our brains, leading to it becoming so big that childbirth becomes dangerous. It’s our genes that have changed the way our brains work in a drastic fashion, but the reason it has happened is due to the memes, and the result is to make brains which are better spreaders of memes.
I’m not really sure what to make of Blackmore’s claim here. The idea obviously depends on imitation being very processing intensive, and that being better at imitation and discrimination of what to imitate requires a bigger brain. These are at least in principle testable claims, which predict that animals who are good at imitation have bigger brains than genetically similar species who are worse at imitation. I haven’t been able to find any research supporting this claim, and there are some methodological issues in determining a good measure of brain size – just pure size clearly doesn’t work, since most of the variance in brain size is explained by bodily size, and even more refined measures such as the encephalization quotient face serious issues.
Whatever the outcome of her specific theory of our large brains, it demonstrates the principle that memes evolve for their own propagation, and not just in service of our genes. It also shows how it is possible for memes to come to drive the genes, in that the results of memetic selection will impact the selection of genes, possibly in drastic ways. Let us now turn to language, to see another example of this.
2. The origins of language
The origins of language is a famously thorny field. It has even been called “the hardest problem in science” by people who work on the evolution of language, though I suppose we should take this with a grain of salt, as who doesn’t want their own field of study to be the hardest? Blackmore holds that we must answer the following two questions if we are to understand the origin of language:
- What is the function of language? That is, what were the selection pressures that gave rise to it?
- Why do only humans have language? This relates to the first question, in that language cannot be an obvious result of too common selectional pressures, or else more animals would have developed it. So how come we have it, while no other animals do?
Traditionally, the two main approaches to the origins of language have been adaptationism and non-adaptationism. Non-adaptationists simply reject the first question, and hold that language didn’t evolve due to selectional pressures at all, but rather as a side-effect (or “spandrel”) of how our cognition and mental processing evolved. This view has been advocated by some prominent names, including Noam Chomsky and Stephen Jay Gould.
The issue with this view is that it cannot account for the adaptive complexity of language. As Pinker & Bloom point out, adaptive complexity or design in nature can only be accounted for by natural selection, since it is vanishingly unlikely (even miraculous) that “any system composed of many interacting parts where the details of the parts' structure and arrangement suggest design to fulfill some function” should arise through non-selectional means (such as genetic drift). For example, the eye is made for seeing, and without understanding this “what for” or telos of the eye, we would be at a loss in accounting for how its various complex parts came together. Pinker & Bloom argue that language shows the same kind of adaptive complexity, and therefore that the non-adaptationists have to posit miracles in order for their account of language to work. Of course, this is by no means a settled matter, but the consensus is that the non-adaptationist route cannot account for the origins of language (for an amusingly vitriolic debate concerning the merits of adaptationism in biology, see thisexchange between Gould and Dennett).
Pinker & Bloom instead take the adaptationist route, and try to give an answer to the question of what the function of language is. For Pinker & Bloom, the function is “communication of propositional structures over a serial channel”, but there are many different adaptationist stories, including that the function is gossiping (as a substitute for grooming practices among apes), or the ability to use displaced reference (talking about things that aren’t visible to you).
But Blackmore holds that these theories fail to account for our second question, that of why only humans have language. If there is an adaptive advantage to communication of propositional structures, why haven’t other species found the same trick? Also, there is the question of how things like gossip and communication of propositional structures could be useful before language had already developed enough to sustain them (a classic instance of a chicken-and-egg problem). Dennett quotes Bickerton, who writes that gossip couldn’t spread
until language had undergone a considerable degree of development. In other words, the earliest stages of language could not have expressed any gossip of the slightest interest, and since some other motive would have had to drive those stages, gossip [...] cannot have constituted the selection pressure underlying the origin of language.
Now, similar issues arise for other instances of complex design. The eye cannot appear fully formed, so we cannot account for the emergence of the eye by saying that a fully formed eye is clearly useful to us. But, staying true to the spirit of Darwinian gradualism, such chicken and egg problems can be resolved by appealing to very gradual development which confers an ever greater selective advantage (in the case of the eye, first being able to react to light, then to discriminate different sorts of light input, etc).
In this spirit, Blackmore offers a memetic theory of what allowed language to gradually develop enough for it to be useful to us (including being used for things like gossip and communication). The idea is that once imitation has gotten underway, we will mimic each other's sounds. There are more sounds than can possibly get copied, so there is memetic competition between these sounds. Which are the sounds that will spread the most? The answer is, vacuously, those who are best at getting themselves replicated. But what then makes for a good replicator? Blackmore’s answer is that it is those with the highest fidelity (those that get copied with most accuracy), fecundity (those that most copies are made of) and longevity (those whose copies persits the longest).
And Blackmore argues that language improves all of these aspects. Language increases fecundity, since more people can copy someone speaking (sound travels further, and around corners etc) than if they have to watch the gestures of a specific person. Language also increases fidelity, in that, through separating sounds into words, it is easier to copy these “digitized” representations than just a long stream of sound (so sounds that are digitized will get copied more). Language also improves memorability (we cannot easily remember or repeat random noises, but can easily do so with language).
The point of all of this is that, given that there is memetic competition to get copied between different sounds, the winners of this competition will be those with high fidelity, fecundity and longevity. And these are features language has, meaning that sounds which are more “language-like” will be at an advantage. Blackmore also claims that grammaticality further improves fidelity, fecundity and longevity, and so sounds with the properties of grammar may begin simply as the result of this memetic selection (again, because sounds that have those properties will spread more easily).
This means that in the beginning of language, the sounds we made weren’t understood, and weren’t for communicating at all. Indeed, they may have even incurred costs in terms of energy and opportunity cost (there may have been better memes to imitate), but were copied simply because they were good replicators. Language may have begun as a parasite, and only later turned into a symbiont. Due to the principle of memetic driving, those who had the best language were preferentially mated with, and copied the most. This means that genes for picking up and speaking these memes were selected for, creating another runaway process, where language could spread even easier, and in turn giving rise to more pressure to be able to speak and so on, causing the restructuring of our vocal apparatus in the process. In this sense, Blackmore offers an answer to the question of the function of language – it is to spread memes – that is what it memetically evolved for, and since language changed the environment in which selection of genes took place, we had to adapt to it.
Blackmore’s theory is similar to the one later offered by Christiansen and Chater, where they take the view that language is an organism that has had to adapt to the way our brains learn and process information in order to spread. By applying evolution not just to us, but also to language itself, we can explain things like why many words share common endings, and which translations of new words will be the most successful (typically, in order to be successful a word has to be either very common, or make use of some regularity found among other words, so called Regularity x Frequency interactions). This echoes an insight that Darwin himself had, namely that “the survival and preservation of certain favoured words in the struggle for existence is natural selection”.
Is this view of the origins of language plausible? It is certainly a theory worth taking very seriously, as it avoids the issues of the adaptationist accounts which have to explain how the first utterances we made were useful enough to us so that the complex structure of language could gradually emerge (for on this account, words don’t have to be useful to us, just good at replicating). The proposed mechanism of the co-evolution of grammar, where utterances evolve some features of grammaticality purely through memetic selection, which then causes our genes to become more attuned to such structures (in a positive feedback loop) is also tantalizing.
But is all of this really anything more than a “just so” story? Well, in a sense no, but as both Dennett and Gould agree (though the scope they think it applies to differs), such stories are a necessary component in any science concerned with the past, whether it be history or the evolutionary sciences. There really isn’t any alternative way of giving answers to these questions than to tell plausible stories about how the past may have looked like given the evidence of the present. This does not mean that “just so”-stories should be unconstrained speculation – different stories predict different things, and there are often ingenious methods for falsifying such predictions. Only by judging which “just so”-stories do best given theoretical and experimental considerations can we get an understanding of the origins of language, and Blackmore’s theory is one such story.
We’ve then seen that Blackmore’s answer to our first question is that language evolved to spread memes. What about the second question, why only we have it? For Blackmore, the answer lies in the fact that she thinks that our imitative abilities are unique, and so that only we have memes which could evolve in the way required for language.
There are two issues with this. The first is that it merely pushes the problem back one step – why should we alone in nature be able to imitate? Blackmore does not provide a clear answer. The second issue is that it is far from clear whether our imitative abilities are as unique as Blackmore thinks. There is a lot of evidence that birds, primates and cetaceans are all capable of “true imitation” and that “animal traditions” (see Avital and Jablonka) exist in nature.
For example, you may have heard of the orcas that have started attacking and removing the rudders of boats off the coast of Spain, a behavior believed to have been picked up through imitation from a female orca known as White Gladys. This is clearly far from any sort of innate behavior in orcas, and it shows how culture can spread in animals, with acquired behaviors continuing to be passed on to newer generations.
This poses a big problem for Blackmore. If animals can also spread memes, how come they don’t have language and accumulation of culture like we do? To me, it seems like the lack of language is the best explanation for the lack of an explosion of culture – we’ve seen how language improves the fidelity, fecundity and longevity of memes, and without it, a critical point of cultural takeoff may never happen. This echoes Dennett’s sentiment that language is the “launching pad” of culture, without which “there is none of the snowballing accumulation” that has marked human progress. But Blackmore’s account of language suggests that it arises only due to memetic selection, and this leaves her unable to explain why other species capable of imitation don’t have language (and in extension culture) such as ours.
Now, this doesn’t mean that her memetic account of language is wholly false – it may still be that words and grammar evolved for their own benefit, and not for ours. But it requires that there is some other factor or filter that prevents this type of memetic language evolution from taking off in other species that are capable of imitation. Until such a factor is identified, Blackmore’s account must remain incomplete.
Note that this issue also affects her account of the big brain – if memes are all it takes, why haven’t we seen a similar explosion in brain size amongst other species that imitate? Now, the situation is far from hopeless for Blackmore, since several possible filters which would explain our uniqueness in these questions exist, but all of them are controversial. Given this, the part of Blackmore’s book that deals with the rise of the meme through memetic selection bootstrapping our language and our big brains is the one that feels most incomplete.
3. Consciousness and the self
Blackmore is famous for experimenting with mind-altering substances, having once published an article in the Daily Telegraph with the headline “I take illegal drugs for inspiration”. As philosopher Thomas Metzinger has pointed out, having out-of-body experiences and being in the strange mental states that psychedelic drugs induce make it all too easy to become a dualist, that is, to think that reality is divided into two types of “stuff”, matter and mind, or body and soul, which are wholly different and separate from each other.
This was famously the doctrine of René Descartes’ Meditations on First Philosophy, where he argued that, since he could conceive of body and mind as separate, they had to be two different substances. The chief issue for this view is to explain how the substances interact with each other (such as how mind comes to affect the body), especially given that physical causes seem to account for all physical happenings in the world (philosophers speak of the “causal closure” of the physical). For Descartes, the answer was that communication happened in the pineal gland in the brain (somehow, magically), which has fueled New Age speculation about the “third eye” and the pineal gland producing endogenous DMT ever since.
But despite her out-of-body and drug experiences, Blackmore remains a skeptic and rejects ontological dualism. Instead, she explicitly endorses the meme-based account of consciousness given by Dennett in Consciousness Explained. Let’s look a bit at what this theory says, and how it ties in with Blackmore’s speculation about the self in The Meme Machine.
At first glance, it may seem strange to think that memes should have anything to do with consciousness – after all, memes haven’t been around for all that long, evolutionarily speaking. Shouldn’t we primarily look at the evolution and function of the brain if we want to explain consciousness? Dennett denies this. The answer he gives in Consciousness Explained (somewhat amended later, as we shall see) is that consciousness is a product of cultural evolution. That is, without memes, there is no such thing as consciousness.
This is a rather staggering claim. Among the implications is that no animals have consciousness, and that even humans lacked it before sufficient culture came along. In this respect, it is reminiscent of psychologist Julian Jaynes, who claimed that the ancient Greeks at the time of the Iliad had no consciousness (instead acting based on auditory hallucinations of gods giving them commands). Jaynes was widely ridiculed for his theory, with philosophers like Ned Block charging that he made a simple use/mention-error in confusing consciousness itself with our cultural concept of consciousness. Dennett, however, was more sympathetic (while dismissing the part about auditory hallucinations), writing a review titled “Julian Jaynes’ Software Archeology” where he points out that many phenomena do not exist independently of our concepts of them (using morality and history as examples).
The term “software archeology” can help us bring Dennett’s own theory of consciousness into focus. For Dennett, there is a hardware/software distinction with regards to our minds, with the hardware being the brain and the software, essentially, being the memes that reside in our brains. Software archeology, then, is tracing which memes, including concepts, appeared at what time in our history, and what impact they would have had on us. It is worth mentioning that the software/hardware distinction isn’t a lapse back into dualism – there is nothing non-physical about an iPhone running an app, and similarly there is nothing non-physical about the effects of memes in the brain[66].
And there is little doubt memes have huge effects on the brain, especially the memes of language. Learning to speak and to read has significant effects on the plastic human brain. In a Bayesian brain context, where the brain is viewed as continually trying to predict its incoming sensory input, Andy Clark has argued that “expectations induced by exposure to words and phrases are especially strong” and that they help us decide which sensory input to attend to/trust the most (so called “precision-weighing”). Clark writes that “Language [...] provides a finely tuned means of artificially manipulating the precision of prediction error at different levels of neuronal processing”.
What Dennett is claiming is that cultural evolution, and language in particular, have given us memes which have such a profound effect on our brains’ organization that they enable the existence of consciousness. But this still leaves us with two questions: (1) which memes are required for us to be conscious? And (2) how do these memes give rise to consciousness?
As for (1), Dennett believes memes concerning self-talk and autostimulation are among the most important – by putting a question to ourselves, we may unexpectedly find out we also have an answer to it, since other parts of our brains may react upon hearing a question compared to when posing it, allowing knowledge to be shared by different parts of the brain (we shall soon see why this is important). Memes concerning reason-giving, and accounting for one’s own actions, and memes relating to theory of mind (and the concept of mind in general) are no doubt also important. In general, Dennett claims that a lot of cultural evolution needs to have taken place for the sufficient memes for consciousness to arise, thinking it requires “not just language and social interaction, but writing and diagramming as well”[67].
To turn to (2), why then should these memes, this software, give rise to consciousness? In order to understand Dennett’s answer, we first need to discuss the hardware of the brain in a bit more detail. A central point is that processing in the brain is massively parallel – unlike an ordinary computer, processing doesn’t take place in a stepwise fashion, with the next operation being executed only after the previous one has finished. Rather, processing of all sensory modalities (and other processing as well) happens simultaneously, with continuous revising and editing of the mental representations they involve.
Nowhere in the brain are these processing streams united into a finalized version presented to a conscious observer. The fallacy of thinking that they are – what Dennett calls the “Cartesian theater” – imagines a spectator (perhaps a little man?) watching the finished products of the brain’s activity as if on a movie screen. Dennett’s point is that there is no such spectator, and there is no such thing as a “finished product” of the brain’s processing – there is just the continuous revising and re-revising of representations (what Dennett calls the “Multiple Drafts” model).
For Dennett, as for modern global workspace theories of consciousness (which may be seen as heirs to Dennett’s project in Consciousness Explained, see for instance Dehaene 2014), the criterion for which of all the things that get processed in the brain that can be said to be conscious is whether they reach system-wide influence, what Dennett calls “fame in the brain”. That is, the processes that are conscious are the ones that come to be broadcast to and influence other processes in the brain, including memory, reasoning, verbal reporting and planning/action. The idea here is not that once some threshold system-wide influence is reached, this causes consciousness in the sense that some determinate contents are presented to an observer. Rather, the idea is that all there is to consciousness is for representations in the brain to come to have such an effect on subsequent processing[68].
So now we have a criterion for consciousness – system-wide influence, or fame in the brain. But how does this relate to the memes? The answer is that Dennett thinks that it is in virtue of the memes we mentioned in response to (1) that anything can come to have system-wide influence in the brain. He claims that those memes amount to software simulating a serial computer, built on top of the parallel hardware/wetware of the brain, maintained through the habits (of self-talk, of introspection) that the memes have infected us with. That is, he thinks that the memes of our culture “install” a virtual machine with just a single step-wise stream of processing, which he calls the Joycean machine, after James Joyce’s “stream of consciousness”-concept (perhaps it could equally well be called a “Jaynesian Machine”).
The Joycean machine attempts to order and account for the parallel underlying events in the brain. This gives rise to a “user illusion”, since the summaries of the underlying events may be simplified, mistaken, and ordered in ways that do not correspond to the order in which they actually occurred. We do not have perfect knowledge of why we do the things we do, but our cultural practices demand that we are able to provide reasons, and so the Joycean machine will often confabulate and make mistakes. But in some ways, it is able to provide a single stream summary of the underlying parallel activity of the brain.
According to Dennett, it is this integration of different brain systems, engendered by the effects of the Joycean machine, that is required in order for anything to reach system-wide influence. Once the memeplex of the Joycean machine has been formed in culture, it continues to get passed on, due to the cultural environment children are exposed to (including demands for reasons, such as “why did you do that?” and preferences, such as “which do you like best?”). Essentially, Dennett’s whole account of consciousness states that the things that enter the Joycean machine are the things we are conscious of, since those are the ones that achieve fame in the brain.
No doubt the consciousness enabled by the Joycean machine is useful to us in many ways – especially when it comes to higher reasoning (system 2 thinking) and planning, being able to integrate different information is crucial. But given that this is a memetic explanation for consciousness, Dennett doesn’t rule out that some aspects of our consciousness (perhaps some of the habits the memes install in us) aren’t adaptive for us at all, but merely exist since the memes that make up those parts have been good at getting themselves copied.
All of this may seem very strange, and no doubt the sheer unintuitiveness is part of why Consciousness Explained has been so misunderstood. Part of the blame falls on Dennett, since he can be rather coy about what he actually means, as if he’s worried that people will dismiss him out of hand if he just clearly states his conclusions. But it’s hard to fault him for this. After all, claiming that consciousness is a recent cultural evolution depending on selfish replicators that don’t necessarily act for our benefit can be a tough sell.
I think Dennett overstates the importance of memes for consciousness somewhat. One of the things in Consciousness Explained which haven’t stood the test of time is the claim that the Joycean machine is required for fame in the brain to occur (and in effect that animals don’t have any form of consciousness). More recent work on global workspace theories of consciousness shows that system-wide influence often does occur even among animals, simply due to the biological wiring of their brains.
But this does not mean that the memes that make up the Joycean machine are unimportant. In a retrospective article about Consciousness Explained that has been officially endorsed by Dennett, philosopher Jared Warren amends Dennett’s account somewhat to allow for some forms of biological, non-memetic forms of fame in the brain. Nevertheless, he holds that the Joycean machine significantly alters the richness and contents of consciousness, in that it improves the efficiency of communication between different parts of the brain, and the types of contents that are broadcast (which can now be linguistic). As Warren writes:
Even the most robust biological broadcasting systems fall short of the powers of the Joycean machine. This leaves most or all non-human animals with, from our perspective, limited experiences. They can feel pain, but they cannot suffer. They can be sad, but they cannot feel despair. The Joycean machine enables a wider and richer range of conscious experiences, to the point where we might even say that Joycean consciousness is different in character and kind from non-Joycean consciousness.
With this amended version, memes aren’t required for consciousness to exist, but they are necessary for our type of consciousness, including our serial stream of consciousness that animals lack. Memes are, essentially, what separates what it’s like to be me from what it’s like to be a bat.
Personally, I’m quite drawn to Dennett’s account of consciousness, but of course, it remains controversial. Critics hold that in all of his discussions concerning consciousness, Dennett misses the very essence of why we care about consciousness, namely our felt subjective experiences, the intrinsic what-it-is-likeness of having a conscious experience, the qualia that defines what consciousness is. He doesn’t address the hard problem. It’s become a trope to say that Dennett doesn’t explain consciousness, he explains it away.
But I’m not so sure. As Warren points out, Dennett’s theory allows us to talk about what it is like to experience something, and gives criteria for when that takes place (when there’s fame in the brain). What he denies is not subjective experience altogether[69], but the philosopher’s notion of qualia as private, intrinsic, ineffable (see Quining Qualia). And that notion of qualia seems almost impossible to entertain without drifting into the dualism of Descartes, where our minds are made of special stuff that is different to all other things (in the case of qualia, of “phenomenal properties” which don’t have the power to cause anything).
Dennett’s theory, especially in its modern day form of neuronal global workspace accounts, is certainly taken seriously, and goes a long way to show just how big an impact memetic theory has if it is right – without memes, there wouldn’t be any such thing as a stream of consciousness.
Blackmore’s book often makes reference to Dennett’s theory, and this especially comes to fore in the penultimate chapter, titled The Ultimate Memeplex, which discusses the self. Traditionally, philosophy has thought of the self as the thing that has conscious experience, which also accounts for what makes someone the same person over time. But given Dennett’s theory, which denies the Cartesian Theater and therefore also any observer for which conscious experiences are had by or presented to, Blackmore argues that there cannot be any such thing.
So for her, the real question to ask (rather than speculate about the metaphysical basis of the self) is why so many of us seem to think there is a self, when in fact there isn’t. Here, Blackmore naturally appeals to a memetic explanation. She claims that the self is constituted by a memeplex she terms the selfplex, which consists partly of picking up the concept of the self, and partly of all the memes that manage to get us to identify with them, that is, those memes that influence what we take ourselves to be. Memes that get the status of personal beliefs will be favored over other memes (a song you think of as your favorite song will do better than other songs), and in turn reinforce the notion that there is someone that has those beliefs. As Blackmore writes:
The more you take sides, get involved, argue your case, protect your possessions, and have strong opinions, the more you strengthen the false idea that there is not only a person (body and brain) talking, but an inner self with esoteric things called beliefs.
So according to this explanation, all there is to the self is a set of memes working together for their own propagation, in the process making you believe in something that simply isn’t there. It is almost impossible not to get infected with a selfplex, and its (literally) self-protecting and self-perpetuating nature ensures that most of us are stuck with the notion that it is ‘I’ that have experiences and make decisions.
But since this view is false, Blackmore also holds that we have to give up the idea of free will, since she thinks free will would require a self to consciously make decisions. This leads Blackmore to some rather extreme views about creativity and the role of consciousness in shaping history – for her, consciousness and creativity literally “cannot do anything” – both are just the result of interactions between memes and genes, and they have no power in and of themselves.
For example, Blackmore says of someone writing a book, that “I would say that the book was a combined product of the genes and the memes playing out their competition in [the author’s] life”. Similarly, for free will, she claims that a man saying they made a conscious decision to have cereal for breakfast actually had cereal for breakfast due to genetic and memetic factors (having a genetic makeup that inclines him towards carbohydrates in the morning, responding positively to ads for the brand of cereal etc), and that saying it was due to a conscious decision “adds nothing. It is just a story [the man] tells after the fact”.
This is a pretty Zen way to think about things. The Buddha is claimed to have said “actions exist, and also their consequences, but the person that acts does not”, and this is echoed by Blackmore’s perspective that we don’t have to talk about selves or consciousness at all in order to understand the world.
But while I can see the attraction of this perspective, I do think it is overly reductive. For one thing, saying our actions are influenced by genetic and memetic factors isn’t any more of a problem for the idea that we have free will than what determinism is, and famously many philosophers believe there is no direct clash between free will and determinism (they are compatibilists). Dennett, for instance, argues that we have free will, because we need such a concept to account for different patterns of behavior between agents and non-agents. Both a boulder rolling down a hill and a skier going down a slope will ultimately be predictable given complete knowledge of microphysical properties and starting conditions, but at that level of description we will miss important patterns that the skier exhibits, such as choosing to veer away from obstacles, for which we need a higher-level explanation in terms of what the skier chose to do (and reasons for doing it, not just a causal explanation).
The claim that creativity and consciousness “do nothing” also threatens to miss out on real patterns. For this amounts to denying there is any difference at all between the blind variation and selection of evolution and the human top-down intelligent design that we revere in our artists, composers and scientists. Steven Pinker, a critic of memetics, has written that
The striking features of cultural products, namely their ingenuity, beauty and truth (analogous to organisms’ complex adaptive design), come from the mental computations that direct – that is invent – the “mutations” and that “acquire” – that is understand – the “characteristics”.
This is a claim about how biological and cultural evolution differs, and while I think Pinker overstates his case (surely, much cultural design has been perfected not by any specific designer or inventor, but rather through memetic evolution, as with the boats we discussed in section 1), there is nevertheless something right about pointing out that human design can often be highly directed, in contrast with the “blind watchmaker” of evolution.
However, this ability to direct our search for variations on design, and the advanced capabilities we have of understanding purposes and functions of cultural products, have only been made possible due to memetic evolution. As Dennett says, just as one cannot do much carpentry with one’s bare hands, one cannot do much thinking with one’s bare brain – one needs the tools for thinking. And these tools have evolved gradually after the rise of language, originally in a completely blind, randomly mutational way, but in the process giving us ever-increasing powers of comprehension and ability to design intelligently. Without the memetic selection that operates without the directedness of human inventions, we wouldn’t have those powers of invention. Nevertheless, simply saying that “it’s all due to memetics and genetics” misses out on real differences in patterns between how we design things now that we have all these tools for thinking, and the way that non-directed evolution works.
With that said, Blackmore is certainly right to insist that there isn’t a transcendent immaterial self that is somehow able to act independently of our genetics and culture. While this may seem like an obvious point, even people like Dawkins can sometimes veer dangerously close to falling into the trap of endorsing a self that’s completely unconstrained. The last line in The Selfish Gene reads: “We, alone on earth, can rebel against the tyranny of the selfish replicators.”.
This suggests that, just as there are increasing opportunities to direct genetic evolution, we can also choose which memes we are exposed to in order to direct which memes get to thrive in us. In one sense, it is certainly true that we can, say, stop reading a particular newspaper in order to avoid a certain kind of meme. But, and this is Blackmore’s point, such choices aren’t made independently of our memes and genes – only against the backdrop of other memes can we take decisions about which memes we want to filter out. She writes: “The choices made will all be a product of my genetic and memetic history in a given environment, not of some separate self that can ‘have’ a life purpose and overrule the memes that make it up”. This is also true of genetic engineering – the direction in which we will take biological evolution will be highly dependent on which memes flourish (for instance, what is considered attractive, useful and good in a given culture).
Blackmore seems to think that the closest we can get to “stepping out of the meme race” (as the title of the last chapter in the book reads) is to practice meditation as a kind of “meme-weeding” that lessens the harmful effects of the selfplex and the false consciousness it brings. But of course, from her perspective, no one can really choose to meditate – you can only happen to be of the right genetic makeup, and happen to be exposed to the meditation meme at a fortuitous time (perhaps through her book) so that you are driven to start doing it.
4. Institutions
One of the things that has always struck me about institutions (in the broad sense of public institutions, churches, corporations etc) is how they often prioritize acting for their self-preservation over their stated purpose. For example, I recently spoke to a union representative, who was disappointed that their union had moved forward with a suggestion that would make it their top priority to get workers to negotiate on their own behalf (for free) instead of sending trained negotiators (who achieve much better outcomes, but who are expensive for the union). There was little doubt this would make the union worse at actually representing the workers, but it would save a lot of money, and probably help the organization thrive in the sense that there’d be more money to advertise and poach people from other unions.
Another testament to this type of effect is the so-called Shirky principle, which states that “institutions will try to preserve the problem to which they are the solution”, with one example being a bus company taking legal action against car-poolers because they eliminated the need for buses.
Institutions also seem to be good at getting the people they employ to take actions they wouldn’t take in any other setting – people “just following orders” can do horrific things, ranging from industrialized genocides to punishing employees reporting safety violations to covering up sexual abuse taking place within the institution. The worst episodes in human history, the Holocausts and Gulags, are all like this, with people acting in horrifying ways due to the workings of an institution which encourages or demands such actions.
I think memetics can help explain what’s going on here. The main thing that separates memetics from other co-evolutionary theories of culture is that the memes evolve for their own sake – they are not beholden to being beneficial to our genes. Inventions like contraception, and even things like higher education are definitely bad for your genetic fitness (higher education correlates with having fewer children). Now, there are limits to just how bad for our genes the memes can be without also hurting themselves, since they wouldn’t have anywhere to spread if they completely wiped us out (similar to a virus that is both too lethal and too good at spreading). Nevertheless, memes are even less beholden to our general well-being than they are to our genetic fitness, and there’s no guarantee that memes that promote human flourishing will be favored in the memetic selection, as the institutional horrors we discussed above show.
This principle that memes evolve for their own sake applies to institutions in a very simple way – the institutions that survive will be the ones that are best at preserving themselves, not necessarily the ones that perform their function the best, or that increase utility the most. Naturally, any institution which completely eradicates the problem it’s meant to solve will no longer be useful for anything, and may therefore cease to exist. Institutions which manage to preserve the problem they appear to solve will be more successful in continuing to exist, explaining the Shirky principle.
This is a simple insight, but it isn’t always taken heed of. For instance, this article by Currie et al. discusses the cultural evolution of institutions, and uses game-theoretic simulations to show how simple punishment mechanisms can help reduce free-riding and tragedies of the commons-type scenarios. They suggest that the function of institutions is to “help facilitate potential solutions to persistent or recurring cooperation or coordination problems”.
No doubt some institutions perform such functions, with varying degrees of success. But this neglects the fact that an institution’s usefulness to us or to society isn’t the only thing its memetic fitness depends on (though it may certainly play a part), and whatever these other factors are, they will also shape which effects the institution has on society.
Currie et al. are aware not all institutions are beneficial for us. They write that:
[...] institutions can exist without a clear function, in the same way that other cultural traits can be neutral or by-products of other adaptations. Second, institutions can enforce socially detrimental behaviours if these institutions are badly designed, or designed by a minority for a minority”
But this still fails to appreciate the memetic point that institutions evolve for their own sake. They don’t have to be designed to benefit humans (though typically every institution will have its beneficiaries, this may be so only because such a structure benefits the organization), and crucially, they don’t really have to be designed by anyone at all.
Though there is certainly planning and top-down design of institutions, many of their features build on preceding cultural evolution, and imitate successful organizations of the past (and present). From the memetic perspective, the first institutions were almost certainly not designed or planned in any way, and they weren’t understood by the people who made them up. Rather, they took hold in society simply because they were memeplexes that were good at getting in people’s heads. In modern society, we have more comprehension, and better abilities to direct the evolution of institutions, but I’d wager a lot of features of institutions still come about through memetic evolution, with gradual adaptations arising in response to a changing cultural ecosystem. And even where there is “intelligent design” of institutions, it’s still a question of natural selection which organizations manage to survive, and again, these will be the organizations that are best adapted to their environment, not necessarily those that best serve the purpose their designers intended (even a well-intentioned designer can impart unintentional side-effects through their design, as with the cobra effect)
This perspective also applies to things like Durkheimian functionalism, which views society as an organism, and its institutions as different organs of it which all serve some function for the benefit of the whole. While this is misguided given the perspective I’m advocating for here (since institutions only have to benefit themselves, and may be in fierce competition for survival among each other, society is more like an ecosystem where different organisms compete than a unified organism), it is undeniable that a lot of practices, institutions, taboos and rituals can be seen to serve clear societal purposes. The problem for functionalism which ultimately led to its decline was the lack of a credible mechanism for how these institutions came to be designed to serve such purposes (as Dennett also mentions). But this is readily explained by the memetic perspective, which is compatible with the idea that memes which promote such functions can be selected for over long periods of time (perhaps because they are judged to be good), allowing such institutions to arise without any designer[70].
So, according to memetics, institutions can be viewed as memeplexes that have been adapted for their own survival, not necessarily to serve our purposes (though some institutions certainly do). What then are some of the adaptive features that institutions evolve to aid their survival? Blackmore is mostly concerned with the institutions of religion (holding similar atheist views as Dawkins and Dennett), and argues that several features of religion may be explained through memetic evolution. Things like punishing non-believers, encouraging producing as much offspring as possible (through taboos on masturbation and contraception), co-opting memes about truth and morality, and promoting ideas that the religion should be spread through conquest (such as the crusades) are all traits that will help the memes of religion to spread.
The point about having many children is especially interesting, since it works due to parents transmitting their religion to their children. This is called vertical transmission, and works similarly to how genes are passed on from one generation to the next. However, with memes there is also the possibility of horizontal transmission, where we pick up memes not from our parents but from people of our own, or even later, generations.
As long as memes are transmitted vertically, there cannot be much clash between genes and memes, because the success of the memes depends on how many children they can be vertically transmitted to (hence, the anti-contraception memes of Christianity serve the genes as well as the memes). However, this is not at all the case with horizontal transmission, and as cultural complexity increases, horizontal transmission starts becoming the dominant form of transmission. Now that we have internet, radio and TV, the memes of our parents matter much less than they did in the past, and this may account for the (slowly) waning influence of religion world-wide. We shall discuss the implications of this more later.
As for other institutions, the ones that will do best are the ones that evolve incentive structures that get people to take actions which promote the good of the organization. This may include things like CEO bonuses, a culture of employees believing in the stated mission, ostracizing those who aren’t “team-players” etc.
In particular, it’s often good for the survival of institutions if they can decrease the agency of individual humans within the organization (and the impact of their values, morality etc) and promote conformity. If the CEO of a large oil company that engages in drilling in a conflict area starts having moral qualms about this, it could destroy the company if his or her morality allowed them to change the course of the company on their own. Institutions which have safeguards against this, like a board of directors who can replace the CEO, will do better than ones that don’t. The same goes for institutions which are good at imparting memes in the leadership that causes them to identify with the organization, and being willing to make sacrifices in terms of their private lives and what they think is right in order to promote its success. There may even be an advantage for institutions which promote people without moral scruples to positions of power, which may explain the prominence of psychopaths at the upper echelons of power (it’s not just that psychopaths strive for positions of power, organizations also evolve to put them there).
As for conformity, this also serves the purpose of not letting people’s individuality intervene with what’s beneficial for the organization. This may explain things like hazing rituals, silencing dissenting or critical voices, and a culture valuing “chain-of-command” (if you’ve ever seen The Wire, this type of phenomenon is a large part of what the show is about).
Of course, no one ever sat down and decided that it would be good to have hazing rituals and to promote psychopaths to be CEOs. The point is just that institutions which have such a culture are at an advantage compared to other institutions, and it’s the victors that perpetuate their memes. This reveals the emptiness in people saying that “corporations are just made up of people” who can do things in a moral way if they only choose to. While that is of course true in some sense, institutions have evolved specifically to get the people that make them up to act in ways that don’t harm the institution (through promoting the spread of certain memes among leadership and employees), in that it is institutions that have succeeded with this that have survived in the past (and therefore being the ones that are copied in the present). We are also born into a society full of institutions, which affect us and how we think from birth – it wouldn’t surprise me if many of the memes we are imparted with from early childhood do more good for our institutions than they do for us.
A final adaptive feature of institutions that I think is worth mentioning is that they do something analogous to niche construction, where organisms alter their environment, often to make it more beneficial for themselves. In the institutional context, this may include things like lobbying to pass laws which benefit you, regulatory capture, awarding stipends to attract the best talent, or using your power over the media to control which memes flourish to suppress critical voices.
It’s worth emphasizing that none of this adaptation requires institutions to have any foresight, or will, and even less so life or consciousness. In one sense, one may speak of memeplexes as organisms, in that they adapt to their environment through natural selection, but we should be careful not to take the analogy too far. Just how far it can be taken may be interesting to look into – certainly, we can talk about the interactions between both institutions and individuals and among different institutions in similar terms as to how we describe interactions between organisms. There are mutualist institutions who help each other, there are commensals who despite existing close to each other have little impact on each others’ memetic fitness, and parasitical institutions who are detrimental to other institutions and individuals. One can even imagine a taxonomy of institutions with various phyla, clades and species of different types of corporations, religious organizations, governments etc. Whether this would be at all useful, I do not know, but it may clarify where the disanalogies between the ecology of biological organisms and the ecology of institutions reside.
The perspective I’ve sketched here may seem rather bleak. It just feels wrong to think about institutions evolving in ways that decrease the impact of our personal morals and make us more conformist, and how they may thrive even if they are detrimental to human values. I will admit that I take a dim view of how some of our institutions function, but I’ve also chosen to focus on the ways in which institutions may be harmful to us because I think that’s where a memetic explanation really shines, and where theories like Durkheimian functionalism fall short. But we shouldn’t forget that there are many institutions that are good for humanity (at least given my memes about what ‘good’ is).
One can view politics as a way of trying to promote such organizations, and impose constraints on more parasitical institutions. We have tremendous powers of comprehension and intelligent design, and can use these to try to design institutions which minimize the problems I’ve pointed to. I do not think that this is easy, given that even organizations that are well-designed for human purposes to begin with may evolve according to the pressures of their environment to become less beneficial to us. Nevertheless, the study of memes and the view it takes of institutions may be helpful here, and a better science of how institutions act, and for whom they act, could be very important in the future.
This is especially so given that the memes of the future may look very different from what they do today. Technology has always shaped which memes can spread – after all, the medium is the message – and the inventions of newspapers, radio, TV and the internet have all been great leaps forward for memes in general, enabling ever greater horizontal transmission of memes (at the expense of memes that have typically spread vertically). But we may soon have technology which doesn’t just spread more memes, but which gives rise to an entirely new kind of replicator. This third replicator is what Blackmore calls tremes.
We already have bots and large language models which selectively copy content they’re fed with, and as AI advances and allows such programs to take more actions on the basis of the memes that they encounter, we may soon have memes spreading entirely between programs at a significant scale, with the best replicators being those that are best at being reproduced and copied by such agents. The reason Blackmore thinks this constitutes a new replicator is because the selection differs from that of memes – which memes flourish in humans is to a large degree the product of our genes, which determine how our senses and attention works, and which determines which evolutionary buttons memes can push to be successful (food, sex, power etc). But which tremes flourish is not at all constrained by human biology, but rather by how we’ve constructed the agents that do the copying. And we may not even know exactly how a given architecture for an AI impacts which tremes are best at spreading.
We noted before that it would be bad for the memes if they wiped out humanity, since they’d have nowhere left to spread. More generally, our genes have always imposed some constraints on how harmful memes can be, given that they’ve evolved to prevent harmful memes from spreading (in that genes for brains that won’t let the worst kinds of memes in will have an advantage, and so get passed on). This is not so for tremes, since their success only depends on finding a home inside the AI. Our brains and our genes are irrelevant to their replicative success, and if the tremes that are best at getting replicated by AIs happen to be harmful to us, then we better pray the AI doesn’t have too much power to affect our lives.
One can view the alignment problem in AI as a special case of the misalignment that already exists between institutions and what’s good for humanity. Our AI companies are often driven by profit (a very successful meme for institutions), and as we know from all too many cases, the profit motive often clashes with taking a course of action that benefits humanity or is moral (by the standards we judge such things). Profit-seeking corporations cause opioid epidemics, devastate rainforests to make room for farms (not unlike how an AI might make room for paperclip factories), suppress research into climate science and the dangers of smoking, and use their influence within governments to start wars against countries trying to implement land reform. Much like a paperclip maximiser might, they employ power-seeking strategies, such as lobbying, bribes and propaganda, just to ensure their future profitability. None of this is done out of sheer malice (though we are certainly right to blame the responsible executives), but rather because of the misalignment between human flourishing and the memes that make up such institutions (in particular the profit motive). Given this, I’m very skeptical that the institutions which stand to benefit from advanced AI will stop its development before it reaches the level where which tremes spread has a big impact on our lives. And at that point, we are in very dangerous territory.
I’m not saying it’s impossible that AI will turn out to bring more good than harm, and certainly not that all technology is bad (a lot of technological progress has definitely been good both for the memes and for “us”), but if a replicator arises whose interests are entirely unconstrained by the interests of our genes, and if such a replicator comes to have large amounts of power over our lives, then I think we need to be acutely aware of the potential dangers of this.
5. Conclusion
We have now seen how memes play a role in several big questions, from the origin of language in the far past to the possible rise of a third replicator in the future; from the evolutionary puzzle of our big brains to philosophical questions about consciousness and the self. The meme concept is highly fertile, offering more explanations than I can address in this review. Just to mention a few things I’ve omitted, Blackmore also thinks memetics holds the key to understanding altruism, why we believe in the paranormal, and it has implications both for the Jungian collective unconscious[71] and for Nietzschean theories of morality[72].
Despite this wealth of explanations, memetics is still very much a struggling field, and you won’t hear the grandiose claims we mentioned in the introduction made on its behalf these days. So what’s the problem with memetics? Doubtlessly, some of the resistance is ideological, in that the meme remains associated with Dawkins, who was one of the key combatants in the “Darwin wars” of the 90s, where evolutionary theorists fought a surprisingly heated battle about the role of natural selection in evolution. However, there are also substantive criticisms of the memetics.
The first criticism comes from Richerson and Boyd, leading proponents of the dual inheritance theory of cultural co-evolution that some argue has superseded memetics. They argue that much of cultural transmission does not rely on “discrete, faithfully replicating, genelike bits of information”. That is to say, they think that the gene-meme analogy doesn’t hold, and that memetics cannot account for the spread of all culture, only that which is discrete and faithfully transmitted.
It’s worth pointing out that even if we agree that not all cultural transmission meets these criteria, a lot of it does. In particular, words are prime examples of discrete, usually faithfully transmitted bits of culture. And perhaps we shouldn’t expect memetics to account for all of culture. But for Blackmore, it kind of has to, given that she thinks everything about human behavior is the result of complex interactions between memes and genes in a given environment. So her response to the criticism of Richerson & Boyd is that we shouldn’t expect the gene-meme analogy to be perfect, insisting that what is crucial about memetics is that it takes memes to be a replicator in its own right, one that’s not beholden to the interests of the genes. So while she agrees that “the copying fidelity of most memes is very low, there is often no right way of deciding where one meme begins and another ends, and most memes do not appear to be particulate”, she writes that
This does not disqualify songs, stories, scientific theories, or technologies from being replicators; it simply means that these memes are rather poor-quality replicators—as we might expect from an evolutionary process that began only a few million years ago, at most.
So for her, the relationship between memes and genes is not that of a perfect analogy. Both are replicators, and there may be other analogies, but we shouldn’t expect all biological concepts about genes to be transferable onto memes. This also applies to another one of the ubiquitous criticisms of memetics, namely that it is “Lamarckian”. Lamarck famously thought that traits acquired during an organism's lifetime could get passed on to its offspring (so, if you amputated a leg, that might have an impact on the appearance of your child). This is a completely false idea, and fell out of favor long before we knew anything about DNA due to Weismann’s barrier.
But memes are a bit Lamarckian. For example, if you watch me make a soup, and then copy the way I did it, any accidental “acquired characteristics” that I gave the soup, such as too much salt, or a faulty ingredient, will be passed on to the soup you make. Blackmore here distinguishes between “copying-the-product”, and “copying-the-instruction”. In the former case, you copy my soup by observing how I do it. But in the latter case, you copy the recipe for making the soup, and make it based on that recipe, and that is not Lamarckian, since the characteristics my soup acquires won’t change what the recipe says. Copying the instruction often yields better copying fidelity, and is more closely analogous to genes, in that we can talk of a “memotype” (the recipe for the soup) and a meme-phenotype (the soup made on the basis of that recipe). However, the real reply to the Lamarckianism objection, according to Blackmore, is that we simply shouldn’t expect such biological concepts to apply neatly to memes. Memetic inheritance is both Lamarckian and non-Lamarckian (Weismannian) depending on whether we copy the product or the instruction, and this shows that this replicator is simply different from the gene.
Another objection against memetics is that natural selection must operate on random variations, but that the variations we produce of memes aren’t random at all. As Gould writes: “variation must be random, or at least not preferentially inclined toward adaptation. For, if variation comes prepackaged in the right direction, then selection plays no creative role, but merely eliminates the unlucky individuals who do not vary in the appropriate way”. But we’ve seen that humans are able to combine the memes they are infected with through intelligent design, often to make them better at spreading (this is what advertising and rhetoric is all about).
However, as we discussed in the section on consciousness, these capabilities are themselves the result of previous memetic selection, which has worked blindly on more or less random mutations. As Dennett says, we can view cultural evolution as a cumulative process that begins in a completely Darwinian fashion, where all memetic adaptation is the result of natural selection on memes, which then gradually bootstraps our comprehension up to the point where we can now often apply “memetic engineering” techniques to steer the course taken by memetic evolution. There’s been a “de-Darwinization” of cultural evolution, but this doesn’t mean evolutionary logic doesn’t still apply to culture. It’s also an open question how much of our current culture can be explained through the intelligent design of comprehending actors, and how many of the features of our institutions and culture are better explained as adaptations that have arisen through the natural selection of memes. My bet is that there is less comprehension compared to what we would normally think.
A final criticism against the field of memetics is that it isn’t predictive, and that all it is doing is recasting points from other disciplines (sociology, anthropology etc) using memetic jargon. But I think Blackmore does a very good job of showing that memetics does make falsifiable predictions, and The Meme Machine is littered with them, ranging from theories about bartering with memes to ways of testing her language hypothesis. Among her predictions is that “linguists should find signs that grammar is optimized for transmitting memes with high fecundity, fidelity and longevity, rather than for conveying information on specific topics such as hunting or for forming social contracts. Social psychology experiments should show that people preferentially copy more articulate people and find them more sexually attractive than less eloquent people”. Given this, I think we can lay this objection to rest, and instead focus in the future on trying to see if these predictions actually pan out.
I mentioned Richerson and Boyd earlier, and their account of cultural evolution is currently much more in favor than memetics. While there is a lot of agreement between memetics and their dual inheritance theory, they differ in that the latter doesn’t take culture to involve a second replicator, and that they ultimately think that “culture is on a leash” held by the genes, even though there are some “rogue cultural variants” which spread for their own sake (like chain letters). We’ve seen that a lot hinges on this – if culture is ultimately always constrained by genetic benefit, then we should presumably be less worried about the ways technology develops in the future. It also means the question of cui bono? – who benefits, genes or memes? – matters less when evaluating the function of cultural items and institutions. For Richerson and Boyd, ultimately, over time, culture spreading for its own sake will be constrained by our genes.
It’s premature to try to definitively establish whether memetics or dual inheritance theory fits better with reality. We need to test the predictions of memetics, and become clearer about just what differences in predictions exist between it and dual inheritance theory. However, I do think an argument needs to be given as to why memes wouldn’t constitute a second replicator, if we are to accept Richerson and Boyd’s claim that our genes holds culture on a leash. The objections we’ve looked at here remain unconvincing. Given this, I think it’s an open question, with a lot at stake, the degree to which elements of our culture, including institutions, exist to perpetuate themselves, irrespective of how they affect our genes.
Much of the value of memetics lies in its application of evolutionary theory on culture, breaking down our prejudices that wherever there is design and function in culture, there must also be a designer. Most of the points made in this essay could probably be recast in the language of dual inheritance theory, and explanations similar to those offered by memetics for language and other phenomena already exist using different terms. But we have also seen that there is a difference between memetics and other theories of cultural evolution, in that they offer different answers to the Cui bono?-question. It is not out of the question that the fate of our species hinges on that question.
Therefore, I hope the future will see a serious science of memetics, which gives the predictions of memetics the attention they deserve. In The Meme Machine, Susan Blackmore has laid the foundation for such a science, and anyone interested in these questions would do well to read it.