By Sylvain Magne - 2016
In her influential 1999 book, The Meme Machine (chapter 5) Susan blackmore raised three important problems about memetics. Each problem was titled as follow:
And then she added a last chapter titled:
These three problems are still largely relevant today, as progress with memetics is proving to be slow. However I think my proposed views on memetics, which I call the code model, could help answering or clearing up some of those points. As you may have noticed, these problems were formulated as clear, rather negative, statements of facts. They show how Blackmore felt strongly that no clear solution could be found just yet. Today I personally feel a bit more optimistic about it all, so I would like to reformulate those points as the following questions:
For every one of these points I will briefly introduce Blackmore’s position and I will then offer my own answers. Please note that Blackmore’s views may have changed since then, so these are in response to a Susan Blackmore from 1999.
Blackmore dives quickly into this problem and starts by asking:
“Is Beethoven’s Fifth Symphony a meme, or only the first four notes?”
The underlying question here is to know where a meme starts and where it ends, and whether the question is worth asking or not. To put it simply, Blackmore’s answer to this question could be summed up like this: we don’t need to know. She makes a very good point that, when it comes to genes, we don’t actually have a clear answer to this question. Depending on how we are studying genes it may be more convenient to consider genes to be the same as cistrons, or any length of DNA material that gets copied successfully. This fact, she argues, hasn’t stopped genetics from making great progress.
So, I suppose we could satisfy ourselves with this sort of status quo, but let’s note that this doesn’t necessarily have to remain an unsolved problem, whether it is with genetics or memetics. Also there may be much to gain in finding out the answer to this question, if there is one. I would like to argue that many units of memes can indeed be identified, and that this is the case for the majority of memes.
Finally, Blackmore seems to think that there is nonetheless a lower limit to the size of replicators. She seems to think that words are too small to be memes:
“A single word is too short to copyright and an entire library too long, but we can and do copyright anything from a clever advertising jingle to a 100 000-word book. Any of these can count as memes – there is no right answer to the question – ‘What really is the unit of the meme’.”
Here is what Blackmore says about short bits of DNA:
“A sequence of DNA that is too short is meaningless – it lasts almost indefinitely, being passed on identically from generation to generation but taking part in countless different kinds of protein synthesis and countless different phenotypic effects.”
In my opinion there is no need to deny words, or letters, the status of memes. The vast majority of words are cultural entities that are perfectly copied through the ages with a steady flow of variations and selections that display the iconic hallmarks of evolution. We have vast amounts of historic records to back it up. The same goes for letters, numbers and any sign that we commonly use to speak and write. Furthermore, words, and especially written words, have a high degree of fidelity, making them strongly digital. What more do we need to be convinced? If anything words are the perfect candidates for making a case for memetic cultural evolution. The question of meme units becomes extremely clear when we realise that letters and words are perfectly discreet entities with very well defined beginnings and ends. In my view, we could not hope for a better example of meme units.
Meme units can be combined to create larger codes and those larger codes can also be considered as units. This means that memes can overlap. We have letters as basic units. Those units can combine to form larger units such as words. Words can combine to form even larger units such as idioms, expressions, quotes, songs, all the way to entire book-size units. For example, i is a meme, it is a meme and It takes two to tango is also a meme. It just so happens that it takes two to tango carries the meme it which carries the meme i. This is a form of memetic collaboration.
This said, is any and every sub-part of it takes two to tango a meme? No. Each word it, takes, two, to and tango are memes. Each individual letters are also memes and even the spaces between the words are memes. However, random parts such as kes tw and o tang are not memes. How can we determine that? Thanks to grammar. It is the rules of grammar that will allow a reader to determine which parts of the code is a unit. That's all it takes. Unfortunately, grammar rules do not come neatly packaged with each meme. Grammar rules are known by the readers of the code, the code simply complies with the rules. When it comes to our own languages, we know the rules, so it is quite easy for us to work out the units. It becomes difficult when studying other codes for which we don't have the grammar rules. The challenge is then to guess them using statistical analysis or trial and error. That's what Jean-François Champollion had to do when he deciphered the Egyptian hieroglyphs. That's what geneticists are working on with genetic language.
When it comes to genes, the grammatical rules of DNA language, if they exist, could determine where the units are. To me it is quite obvious that nucleotides are the letters of the genetic alphabet, and that the codons are the basic terms (words) of the DNA language. Cistrons also resemble sentences with their ponctuations. I wonder if there are larger grammatical structures in DNA code.
Units are quite obvious to us when it comes to verbal languages, but what about "body language"? What about "design language"? These are indeed a lot less standardised and in a constant state of change. However, there are nonetheless very strongly defined units in body language and design as well. I put body language and design together because they are both light memes. They are what we recognise with our eyes. Just like our ears are sensitive to sound patterns, our eyes are sensitive to light patterns. There are many shapes that our brains are ready made to recognise easily, such as facial features and expressions, but also geometrical shapes and symmetry for example. This means that we are very good at distinguishing between a smile and a frown or between a circle and a square. Our eyes are good enough to recognise a vast amount of patterns and we use this ability to differentiate between a man dancing and another breaking an egg, or between a roman arch and a gothic arch. There are many shapes (and movements) that can easily qualify as units. And again they can combine, just like words, to form larger units. It is by combining them that we can create a rich and complex culture of designs and behaviours.
It seems Blackmore suggests that because short codes such as words or codons appear everywhere, it makes them meaningless, but I think it might be quite the opposite. I can understand why a geneticist may not care about codons because they just never seem to change and therefore seem to have no impact on their research. But from a theoretical point of view, if it gets copied, it is a replicator, and the way DNA replication works involves copying the sequence of nucleotides, codons included. To me, it looks like codons are everywhere precisely because they are the best replicators and by far. They are in the replicators’ superleague. They have made it to a point where the competition has been eliminated and they go virtually unchallenged. They are the super replicators among replicators.
Also, let’s not forget that they have evolved into what they are. Codons and nucleotides didn’t come out of nowhere ready made. There would have been a lot of pre-DNA-RNA candidates, with possibly competing forms of nucleotides and codons that have today disappeared.
All this considered, if the smallest bits of codes can be replicators then how small can they really be? Is half a letter a meme? Is half a nucleotide a gene? For every meme or gene there is indeed a limit. Every code has a medium to carry it and the code is simply a particular arrangement or pattern of that medium. What we want to determine is the limit between the smallest piece of code and the medium itself.
Let’s imagine a binary code made out of a line of black and white pebbles. The pebbles would be the medium, and the sequence of pebbles would be the code. Black would mean “0” and white would mean “1”. A black pebble is just a black pebble, it’s not a code, but the presence of a black pebble somewhere in the sequence is a code indeed, and it’s a code that says “0”. That is in essence where the limit is.
Another way to differentiate between code and medium is to establish which part is being copied and which part is not. For that we need to know which part the copying process is actually paying attention to. In this pebble example, what we want to copy is a sequence of pebbles but not the pebbles themselves. When we are copying the pebble sequence, the internal structure of the pebbles is irrelevant and ignored, as long as the pebble does its job of being a pebble. When copying the sequence, new pebbles will be used with possible differences in size for example. But these differences don’t matter because they are not the target of the copying process.
With regard to DNA, the replication process does not copy the structure of nucleotides, it only uses them as building material. However, their sequence is what is being meticulously copied. So we may say that the pebbles are the nucleotides and the code is their sequencing. Therefore, a genetic code could not be shorter than one single nucleotide. The inner structure of the nucleotide is irrelevant as long as it does its job as a “pebble”.
With regard to sounds for example, it is pretty straight forward. The medium is the air and the code is the sequence of sound waves. The minimum sound bit available would be a sound bit that is simply not too short to be noticeable, such as a single note.
I fear that if we ignore smaller bits of codes, we might be missing out on a lot of perfectly valid memes. I also think that most memes come in very clear units indeed and knowing them is important to understand culture.
Blackmore’s answer to this question is unequivocal when she says:
“We do not know.”
However, Blackmore makes a similar argument as before and points out that Darwin’s ideas allowed to make great scientific progress without knowing anything about DNA and its mechanism of replication. Blackmore remains careful and wishes not to commit to a particular view on this matter. To her, memes may be found in brains, in artefacts or behaviours, and she does not rule out any other possibility. This said she does seem to strongly suspect that memes probably live inside brains, if nowhere else. With regard to how memes are mechanically being copied, Blackmore simply admits that her “imitation based” model does not offer any answer:
“[...] memes depend on being transmitted from one person to another and, by definition, this is done by imitation. We have already seen how poorly understood imitation is [...]”
Blackmore then concludes:
“[...] it is true that we do not understand in detail how memes are stored and transmitted. But we have plenty of clues and we certainly know enough to get started.”
I won’t say that I have all the answers to this problem, for sure, but I think that my “code based” model can answer parts of it.
To the question of where memes are stored, my answer is that they are not inside brains, not in artefacts, and not in behaviours. I claim that memes are in the codes that we exchange when we communicate and those codes are made essentially of light waves and sound waves. With technological memes the list of media includes also electric currents, radio signals, wifi, fiber optics, etc. However, it does not include a computer’s built-in memory just like it does not include human brains.
I realise this sounds rather counterintuitive, but hear me out. For a start, the very concept of replicator forbids the fact that memes could be made of some kind of brain structures. Indeed, brain structures are not copied from brain to brain, simply because when we express ourselves we simply do not send any data that describes the structure of our brain. No brain structures copied means no replicators inside our brains. On the other hand, we can know where memes should be. Replicators are all about communication. Indeed when we communicate we send out signals in the shape of sound waves coming out of our mouths or light waves bouncing off our bodies. The only thing that travels between two people are those sounds and that light, and that is exactly where, if anywhere, memes should be. Memes have to be surfing on those sound and light waves, otherwise there would be no communication. That is why I claim that memes are patterns, made of the media that we use to communicate. I call those patterns, codes.
So what is in our brain if it is not memes? The brain structures that are formed inside our brain when we interact with memes are simply phenotypes. Memes build our brains to some extend and make us their meme machines. In that sense I totally concur with Blackmore’s main focus of her book. This goes as well for artefacts and behaviours, those are also on the phenotypic side of memetic evolution.
This doesn’t mean that the copying process is discontinued when it goes through the brain, certainly not. It just means that the memetic code is transcoded into some kind of brain code. The brain code is the brain’s strategy to recreate the original memetic code when needed. But each brain may find its own way to store that code. This is true of computers too. Two computers for example may not use the same components, or the same language or the same recording device. So when they receive a signal they will store it with whatever protocol they use. However, when those computers prepare to send a copy of that signal they have in place a strategy to transcode their own code back into a copy of the original version. There is nothing crazy about that.
So my claim is that brains only carry memetic phenotypes as strategies to recreate the original memetic codes. While the true memetic codes only exist between people (or between computers, or between computers and people).
Blackmore reminds us of what Lamarckism is:
“‘Lamarckism’ is the principle of the inheritance of acquired characteristics. That is, if you learn something or undergo some change during your lifetime, you can pass it on to your offspring.”
Blackmore eventually prefers to simply do without the concept of Lamarckism for she feels it may not be relevant to ask the question in the first place.
I tend to generally agree with Blackmore’s position. I feel the reason we’re asking the question in the first place has more to do with politics than science. The real interest that we have in this question stems from the fact that Lamarck’s ideas failed strongly against the better Darwinian model. As a result, Lamarckism has become some sort of symbol of “exactly what evolution is not”. In a way, it is unfair. The thing is that Lamarckism is not impossible at all in principle, it’s just not the way biological evolution works. However, when it comes to cultural evolution, the rules of replication and selection are so different that it could be said to allow Lamarckian evolution to take place. Well so what? There is no need to call it heresy.
I believe there are real life examples of this. Large social structures such as religious organisations, companies, governments, football clubs and the like, seem to do exactly that. They change and reinvent themselves over the ages, update some of their own rules, and pass them on to future generations of social structures. There is nothing dramatic or exceptional about Lamarckism in cultural evolution. It is just something that can happen, and it does.
This said, there is a limit to what this Lamarckian self programming can do, and it is a direct consequence of the concept of replicators. Replicators can change other replicators, that’s allowed, but a replicator cannot change itself. That is “forbidden” because a replicator that would change itself would effectively kill itself. The modification would create a mutated replicator which could then compete with its parent replicator for a place in the pool. Therefore it can’t be said to be the same code. So again, I agree with Blackmore that it is probably best not to burden ourselves with the concept of Lamarckism as I have doubts that it may be particularly useful.
Beyond Lamarckism, this question brings Blackmore to introduce a new concept to her “imitation” model. She imagined two types of transmission which she named “copy-the-product” and “copy-the-instructions”. She gives the following illustration:
“Let us suppose that my daughter plays a beautiful piece of music for her friends and one of them wants to learn to play it too. Emily could either play it many times until her friend can copy it accurately (copy-the-product), or simply hand her the written music in a book (copy-the-instructions). In the first case, any changes Emily makes will be passed on, and if there follows a series of pianists copying each other, the composition may gradually change, incorporating the errors or embellishments of each player. In the second case, the individual playing styles of each pianist will not have any effect because copies of the (unembellished) written music are passed on.”
Blackmore makes an interesting point here. She brings to our attention the fact that the strategy “copy-the-instructions” is often more reliable than the strategy “copy-the-product”. It is an interesting way to look at it but I think the code model may offer a better perspective to understand what is going on.
We have indeed two different channels through which the piece of music can be passed on and copied. In my code view, I see two separate and very distinct memes here. There is a sound meme and there is a light meme. The sound meme is the pattern of sound waves created by Emily when she performs the piece of music. The light meme is the pattern of light waves that are bouncing off the piece of paper where the music score is written.
Let’s consider the sound meme first. In this case, Emily’s friend Sarah, has to rely entirely on her ears and her brain memory to recall the melody. As she hears the music and the sound waves are hitting her eardrums, these signals are transcoded and absorbed by the brain which then triggers a chain of brain programming. The programming is not instantaneous and will be improved with repeated exposures to the sound meme. This said, the speed of the programming will be greatly increased thanks to Sarah’s pre-existing understanding of music. Eventually, Sarah’s brain will have built a brain programme able to reproduce a copy of the sound meme. When the moment comes for Sarah to reproduce the music, her brain will “run” the programme that was created and will use it to transcode that brain programme back into sound waves.
Let’s now consider the light meme. In this case, Sarah’s ability to use the music score relies on Sarah’s ability to read that score and to perform accordingly. We are assuming that Sarah already has acquired those cultural skills. Thanks to that, Sarah doesn’t actually need to make any specific effort other than make a copy of the score and carry the piece of paper with her. In this case, her skills combined with the piece of paper provide a ready made strategy to reproduce the piece of music. When the moment comes for Sarah to play the music, she will expose herself to the light meme by looking at the paper. When entering her eyes, the light meme will be transcoded to some brain code. This brain code will interact with her existing brain programmes, those that “understand” music, and will transcode the brain code back into sound code.
So in both cases, Sarah gets equipped with a strategy to reproduce the piece of music. Both are genuine copying processes but there is a major difference between the two. As Blackmore points out, we can see clearly that the second case offers a more effective and more reliable way to copy the music. Why is that? In this case it is not because one is more digital than the other. Indeed both are highly digital because of the nature of music. The reason the paper one is more successful is because it is aided by an artefact. The paper based memory device is simply more reliable, faster and easier to use than the brain based memory. Even though it involves more transcoding steps between light codes, brain codes, and sound codes.
In this code view there is no need to call for “products” or “instructions”, there are simply two sets of memes competing for the same chance to be replicated.
Beyond this particular example, there can be an interesting relationship between different memes. For example, the sound meme and the light meme that we mentioned are mutually transcodable. This means that one can be turned into the other and vice versa, with no loss of data. One can hear the music and write the corresponding score just like one can read the score and perform the music. This doesn’t mean however that they are the same meme. Indeed, natural selection still applies and there is a chance that one will be better at surviving than the other. They compete and they collaborate at the same time.
On one hand we have a sound meme that is painfully short lived, because of the speed of sound which defuses the meme in a split second. On the other hand we have a light meme that is actually even shorter lived. The light waves bouncing off the piece of paper get lost into the surrounding environment at the speed of light. But in this case there is a major difference. While the light waves dissipate, the piece of paper remains and will keep bouncing light as long as the sun is shining. The piece of paper is an artefact that behaves as a broadcasting device for the light meme. It is extremely efficient and cheap to run such an artefact. There is no equivalent trick that can be used for sounds. This said, it is interesting to notice how bats can use echos to “visualise” their environment. If we had those sorts of skills, maybe we could imagine a kind of “sound equivalent” of paper and ink.
Finally, Blackmore mentions that, in the case where Sarah copies the “product”, there is a high risk of copying errors and because Blackmore sees the music as a product and not as a meme, she concludes that the accumulation of errors is a kind of Lamarckian evolution, where acquired characteristics of the product are passed on. I my view, because I see the music as a sound meme, when errors happen I see a case of memetic mutation. As a result the incorrectly copied song is potentially a whole new meme. To me this example is not a case of Lamarckian evolution, but a case of simple mutation.
As a conclusion, taking the code view on this example allows us to account for what is happening, in memetic terms, without having to call for new concepts such as “copy-the-product” or “copy-the-instructions”, or even Lamarckian evolution. To me the code model is simpler to apply and has more explanatory power.
In this last chapter Blackmore expresses the lack of consensus around memetic terminology, highlighting at the same time our lack of clear understanding of the subject. After reviewing different opinions on the subject, she takes the following stand:
“I shall use the term ‘meme’ indiscriminately to refer to memetic information in any of its many forms; including ideas, the brain structures that instantiate those ideas, the behaviours these brain structures produce, and their versions in books, recipes, maps and written music. As long as that information can be copied by a process we may broadly call ‘imitation’, then it counts as a meme.”
This view is close to Daniel Dennett’s view of memes as informational entities, where information is an abstract entity that can hop from medium to medium. Dennett and Blackmore both see the same memes existing in all their instantiations. In response to this view, I have written an article entitled “Objections to Daniel Dennett’s Informational Meme”, in order to explain why this view of memes is flawed. If you wish to have a more developed argumentation, you can read the article at this address:
As I have mentioned briefly previously, my code-centered views are quite different. I consider memes to be codes and not information, and to be travelling with the media that we exchange in order to communicate, such as sound and light waves. In this code view, artefacts, behaviours, brain structures, social structures and other cultural products are all meme machines, they are on the phenotypic side of memetic evolution.
Blackmore presents quickly different views of memes by Dennett, Dawkins, etc. but none of them seem to have seriously considered those sound waves and light waves as being a good candidate as meme carrying media. Blackmore and other memeticists do mention sounds being copied, behaviours, writing, etc, but they seem to always fall short of considering that memes have to actually travel from one person to another and that the space between people could be a good place to look for memes.
Out of curiosity, I went and searched for the words “light waves” and “sound waves” inside Blackmore’s book and also inside Dennett’s book, Darwin’s Dangerous Idea. Neither of those terms appeared even once. It seems to suggest that they really didn’t look at all in that direction, not even to try and disprove this idea. But memes don’t teleport themselves.
One of the reasons why people missed or ignored this is probably because, intuitively we tend to turn our attention to the brain, thinking that it must be in this mysterious black box that memes are hiding. Or simply, by analogy with genes we assume memes should be found within their vehicles. But when some people pointed out that memes in brains did not really make sense, they turned to artefacts or behaviours, but few are convinced by this option. Maybe the reason why they didn’t consider light waves and sound waves is because those are somewhat invisible. It is true for sounds of course, but to some extent it is true for light as well. Of course we see light, otherwise we would be blind but we don’t see the photons themselves shooting through the air. Our experience of vision ignores completely the mechanism by which we actually see. When looking at an apple we just “see” an apple, but we are in no way informed clearly about the fact that enormous amounts of photons continuously hit the apple, bounce off it, travel through air, through our retinas and then land at the back of our eyes. In the same way, we are naturally oblivious to the fact that if we can hear, it is because there is a fast moving sequence of minute pressure changes in the atmosphere, which finds its way inside our ears and makes our eardrums vibrate.
Imagine this. If instead of speaking out words, a series of paper planes came out of our mouths. Each of them would have a word written on them, and they would land inside our ears. Imagine if instead of showing to someone how to juggle, paper planes would shoot out of our hands. Each plane would be a folded photograph portraying an instant of the juggling and they would find their way into our eyes. Now, if we actually saw all those paper planes flying around constantly, carrying their bits of codes, wouldn’t it be suddenly quite obvious that memes are indeed travelling on those paper planes?
This is, in a nutshell, how my code model differs from Blackmore’s imitation model and Dennett’s informational model: memes are surfing the waves that are travelling between people.
If you wish to know more about my code model and the arguments I put forward, visit my blog: