This site is not maintained. Click here for the new website of Richard Dawkins.


← Meme Theory, Zahavi's Handicap, and the Baldwin Effect

Zeuglodon's Avatar Jump to comment 38 by Zeuglodon

Comment 32 by phil rimmer

Comment 31 by Zeuglodon

I think it's an error to assign the designation of "copying" too readily


it must follow from the genetic need for brains to replicate information.

I fail to see your semantic nicety.

My point is similar to one about group selection - it presupposes that the genes have brought about certain qualities in their host organisms. In the case of memes, memetics presupposes that there is a replication mechanism set up in brains by genes. In the case of group selection, it presupposes that animals have social behaviours. In both cases, I think it would be more interesting and more productive to ask why those conditions have come about in the first place. After all, group selection has been rendered obsolete by such investigations. Memetics may or may not go the same way, but either way those previous mechanisms must be explored first. This is why genes are the topic of my interest.

I qualified "copying" in a very detailed way, aligning it with "RNA World Soup" replication, which we understand to be very poor with huge lateral leakage and little to help us define entities at all.

RNA is not a terrible replicator - it still has the ability to preserve its genes for multiple generations before the first mutation kicks in. Memes cannot be bootstrapped by a weak analogy, because their mutation rate is severe by contrast.

I did suggest mechanical and expressive actions to be candidates for gene-like memes (the video illustrated the potential fidelity of this copying) and that these can constitute a (formal) substrate for more complex transfers (?)

I'm not sure what you mean here, so I'll assume it's a lead-in to your next paragraph.

O'Hooligan details the mechanical nature of ritual, rhyme, rhythm and music and the like. These formal processes, mirror neuron copied with good accuracy, can become the cultural machine that transcribes more complex stuff.

My point, though, is that neurons don't copy with good accuracy. A single fragment of information (say, the original copy of a sheet of music) may endure, but the copies that come from that one are the first generation, and the copies of those become very Xerox-like in that a copy of a copy quickly degenerates. Try copying out a passage from a book, and then copy out that passage from the copy, and do this several times, and in every generation the spelling mistakes, minor punctuation, and the like will accrue, word substitutions may occur, and people are not above "improving" the text. Now, this feeble Xerox-of-Xerox style copying may have enough juice for the weakest of evolutionary processes, but it does not last multiple generations without change, so how can it fixate in a meme pool?

Comment 33 by OHooligan

"imperfect" makes it sound as though the replication process could be as awful as I describe, but if the mutation rate is so high, it can barely be called replication from brain to brain

By "imperfect" I meant a non-zero mutation rate. If I'd meant "wildly inaccurate" - as you seem to assume - I'd have said so.

I think you minimize the difference and don't realize it. A gene can literally be atom-for-atom exact for hundreds or thousands of generations before the first mutation hits it. If you're going to be so wide with the designation "imperfect", you could call anything an imperfect evolutionary system because everything would have a non-zero mutation rate. Mutation is, when you get down to it, a change of structure or composition.

You assert that idea transfer from mind to mind is too error prone to serve as a replication mechanism in the evolutionary sense. I disagree. There's a lot of error-correction going on that maintains the core essentials of a viable meme.

I talk about the correction mechanism further down this post, but to focus on it for a moment: I already pointed out that two brains trying to replicate an idea or components of an idea between them are more likely to make it less error-prone if they repeatedly go over the idea i.e. have a discussion. The trouble is that the change then works both ways, so the original meme is mutated too. My scepticism comes from the fact that, to compare a meme with a gene, invoking essentials is a sign that the replication system is not up to scratch. Nobody says the essentials of a gene are maintained. A gene passes through literally identical at all points, until it mutates, at which point it is no longer the same gene. It becomes a different one, though that different one will still play the evolution game as its precursor did.

This is a long way from what happens to a meme. An idea shared among people is always different in some detail, and this is only the first or second generation. It's changed and altered in so many ways that it can soon diverge and even branch in wave after wave of subcultural change. It is true that the ideas have an epidemiology much like a virus', and fads grow and diminish, change and sometimes return altered.

However, once it's changed, it's not the same idea. Every iteration changes, so every iteration, in a strict sense, is not the same idea. This will come across as pedantic, but either this means that bits of ideas are memes as opposed to whole ideas (like bits of genome or genes are the true replicating entity of interest as opposed to whole genomes) or that memes are not evolving entities. You may wish to loosen the term replicator to include ideas, but it's the fact that such a loosening has to occur that would make a scientist suspicious of the concept.

Just because an information channel is noisy, doesn't mean that it can't carry a signal. For an example right under your nose: the internet.

I did not say the channel could not carry a signal, so this point of yours is true but addresses a straw man. What I was saying was that the signal could not be part of a replication system because the very existence of noise that can alter the process (even on the first generation) is enough to raise scepticism about meme theory. I am more supportive of the idea that memes would be bits of ideas rather than whole ones, to get around the noise problem. I even said so in my last comment:

It is possible that bits of the message, or the gist, are what endure, and that the whole idea is more like a genome than like an individual gene. In the jargon, it could be a memeplex, with some memes surviving the transition better than others. I think this would be a stronger angle to stress, as the gist of a speech would be like those bits of it that must translate well, and a piece of speech would have higher person-to-person fidelity than a whole speech. Nor do I mean that you'd remember fragments of speech like "and then", "evoluti", "Selection b", or "Evolution by natural selection". I mean that, if you and your interlocutor have a long enough chat about an idea, the odds are greater that your ideas will align with each other and be more faithful copies.

This is similar to your point about the "essence" of an idea getting across, but I think it is much more rigorous because it keeps us with what's physically going on in brains. The imparting of an idea, then, would be like making a copy of a memeplex - bits will fail to be copied, but any one meme could be copied exactly, and copied multiple times. This would be a legitimate way to get around the imperfect copying problem, because then bits of an idea could be exact copies for long enough to enable an evolutionary mechanism to occur.

My main criticism for this counter would be that any attempt to make sense of information coming in must already rely on a built-in ability to reverse-engineer what the other speaker said, and must already rely on some built-in ideas. In other words, discussing an idea with someone is not so much about creating a new pattern in their heads as about activating one already in there.

But then in what sense is this an independent replication system? Those parts of the brain coding for such innate ideas must already have been set up by genes, just as the legs of a stick insect clone must have been set up by the genes. The replicator is the gene or set of genes that install that part of the brain. An independent replicator is not needed to explain the similarities.

I suppose I could make it more explicit if I refer to an example. Think of subcultures of recent decades, like rebels, outlaws, wild ones, bohemians, punks, shock jocks, mau-maus, bad boys, gangstas, sex divas, bitch goddesses, vamps, tramps, and material girls (not my own wording - taken from How The Mind Works, page 502). They all differ in many ways, but an underlying thread connects each one: defiance of mainstream or of authority. Quentin Bell calls this conspicuous outrage, and it possibly follows the logic of the Zahavi-Grafen handicap principle, but applied to seeking allies and social integration. The image says "Look at me, I'm so secure and strong in my position that I can afford not to cooperate with you".

Followers without such confidence naturally might mimic it to get the same response from other people (fear and respect), but doing so dilutes the effectiveness of that subculture because now everybody - weak and strong alike - is doing it. So, a new subculture with new details arises among the socially powerful to make them stand out. It's discussed in more detail in the aforementioned book.

My point is that the common thread underlying the idea or ideas, far from being a clue to a replicator system exclusive to culture, is actually something that's set up in individuals by genes building brains to behave in certain ways. It doesn't itself replicate because it is the phenotype of genes that do the replicating. When a new trend is set off, it's noisy window dressing to the "essence", which is actually not a replicator. The reason we can understand each other in the first place is because the same genes in both our bodies set us up identically, at least in part, to begin with. Even if I've never heard of postmodernism before, the reason I can grasp the idea in the first place is because I have a brain that's like the brain that told me about the idea, not because the idea has copied itself from one to the other.

Chinese whispers show a counter-example: when there's no error-correction, what's transferred isn't a meme, it's just a Confused Noise (to borrow from E.E. Milne once again).

See my point above about this error-correction. In any case, I would have thought that Chinese Whispers would at least make you raise questions about a replication system that doesn't get it right on the first try. DNA has nearly never experienced such a problem in all the millions of years it's been around. Chinese Whispers shows why memes are awful replicators - because the ratio of mutations to number of generations can distort a message beyond recognition of the original.

Yes, but the copying doesn't need to be at all accurate in its internal details, just as long as it results in a recognizable copy at the macro level.

This is a confusion, akin to saying that genes can mutate highly so long as the overall genome is preserved. If anything, the instability of the small detail would quickly add up to instability of the big picture, because the catchment area of mutations increases with size. And any unifying thread between the fashions of subcultures, as mentioned above, would most likely be an innate thread owing their existence to genes in the first place, defeating the point of memetics.

Imagine reverse-engineering a product to make an equivalent one, as opposed to getting hold of the original design and knocking out identical replicas. A bike is a bike, even if every one is a bit different, and it's bike-like ideas that make for viable memes. The design, plans, tools in my bike factory may be totally different from those in yours, but both factories make things that people recognize as bikes.

I think this is disingenuous. A bike could have an extra wheel at the rear and be called a tricycle. Giving it a motor makes it a motorbike. Two extra wheels might make a quad bike, and before you know it you've got something similar to a small car. If we invented hovering technology, we'd next get hoverbikes. At which point do we draw the boundary of "bike", and why? In any case, this tells us nothing about bikes or about the ideas behind them getting replicated - only that the ideas endure in some fashion across genetic generations - nor does it suggest that the only possible origin for bikes would be the replication or evolution of ideas. The mistake is to think evolution is the only algorithm by which a good design comes about. There's nothing necessary about this connection at all.

The main unifying thread is that we use a bike as a sort of transport, which is a basic need for a creature that evolved to get about, possibly over long distances, for food or to find new habitat to settle into. We don't follow the basic mechanics of bikes because there are memes for bike-ness, but because anybody designing a transport device has to follow the laws of physics, such as those involving gears and wheels, and some work better than others.

It is tempting to point and say that this falls into line with memetics, but does it really? Any system that zeroed in on a better thing than others could be recruited for memetics in this way. Memetics requires some ideas to be better than others for some purposes (e.g. for making an efficient transport), but if some ideas are better than others for some purposes, nothing about it suggests ideas are memes. At most, you've got a decision-making or problem-solving system, which is what brains are.

Most modifications of a basic idea can and usually are made to the prototype. For instance, if something is wrong with the bike's steering, we don't get rid of the bike or artificially select generations of better bike designs. We go to a mechanic to install the new parts or to fix the old ones, to give it an upgrade, or to buy a different bike (whereupon the old one may be sold to someone else needing it for another purpose).

You may wish to go back to the drawing board and look at the ideas behind bikes through the ages and claim that those are the replicating entities, but most of these ideas come from artifacts that can outlast any one biological generation. We preserve them, and copy them only as a last resort when these artifacts start to decay - say, an old photo begins to wear and tear. Each copy is like a Xerox of a Xerox, which is why original copies of manuscripts and of ancient tools are so valuable in the first place. Ideas preserved in this way don't replicate. They endure.

Error correction is thus a basic requirement for an idea to be a meme - in a circular, self-serving definition, memes are the ideas that have what it takes to survive in their environment.

But the error-correction you describe, just as with the replicating mechanism, had to come from genetic evolution in the first place. In any case, see my point above about how the error-correction could actually be a problem for the meme idea.

Comment 34 by Schrodinger's Cat

I think the problem with identifying memetic replication arises solely because one is trying to identify it at too high and broad a level. One needs a closer analogy to genes...

Yes, I agree completely, which is why the macroscopic approach strikes me as being flawed.

At the most basic level, one can find an extremely accurate meme in the letters of the alphabet. The alphabet is a structure that daily gets copied with pretty much 100% accuracy. This is of course analogous to the 4 letters of DNA code.

Then one has words. These are long lasting memes.....which do in fact 'evolve' over time. Its interesting to note how many new words are alterations to existing words, either in form or context...similar to genes.

To take the analogy further, sentences or short pieces of prose are the chromosomes. In that context, consider the huge number of short 'sayings' or quotable quotes that exist within culture. In fact its interesting to note how most political or religious organisations have short 'slogans' and religions have short stories....parables. The brevity itself makes the slogan easier to remember......less likely to 'mutate'.

While I probably would question the exact one-to-one parallels between the two hierarchies, I would not deny you've got the point. If memes exist at all, they cannot be the gross structures such as sentences or general ideas. They would have to be much smaller entities.

And so on. So the fact that someone might not have the exact brain state as Marx on reading about dialectic materialism is sort of missing the point. The level of complexity plays a huge factor in ability to 'copy'....and clearly there is a simpler level at which memes most certainly do get copied with a high degree of accuracy.

I agree, up to a point, but you have mischaracterised my position with that comment about "exact brain state". I mean that, if a unit called a meme is to be copied at all, it must be a physical copy in the brain - whether by neural nets of hundreds or thousands - that matches. In other words, something somewhere must be replicating with total accuracy.

My problem is that this steers away from the Charybdis of gross mutation rates and towards the Scylla of genes doing all the work. Any similar structures between our brains at the neural net level is very likely to have already been set up by genes in the first place. We would all have an ear for phonemes even if we came to identify different phonemes in our particular language. But in that case, the longevity of any particular phoneme - say, the short "a" - is explained by genes setting up the mechanism for recognizing one when we hear one, defeating the purpose of invoking an independent replicator in the first place. I go into more depth above.

I appreciate this is the weakest point of my counterargument to memetics, largely because it depends on the parameters both for the generation/mutation rate and for how much this structure owes its set-up to genes, but even if it were refuted, it does not change the fact that at the macro level, ideas are unlikely to be replicating, much less evolving, entities.

Comment 35 by jimblake

You assert that replication is the key mechanism and that information is not the thing to focus on.

I'm pretty much paraphrasing what Dawkins has said. The whole of the life sciences is rooted in there being a replicator, both to kick-start and to maintain it. Dawkins explains it well enough in The Selfish Gene. No replicator, no life, so no life sciences.

I dispute that. As you know, a DNA molecule with no meaningful information is not a gene, but the replication of the molecule will not produce evolution because the molecule itself will have no effect on the number of copies produced.

I think you've confused yourself. A DNA molecule, and by necessity a gene, is a replicator whether it effects the number of copies produced or not. A gene is a piece of DNA, so how could it be otherwise? The point you contest over is actually about phenotypes, which are the other things that genes produce. They do this simply by manufacturing amino acids, which produce polypeptides and the bases for proteins. Information is not a magic quality added to make it possible. It's the inevitable outcome of such a replication-phenotype system, because a gene that codes for a protein that, via a very long chain of causation, results in more blubber under the skin, could be a gene with information about the cold environment in which a whale lives if a geneticist was given the corpse and told to glean as much about it as possible. Replication is necessary for the information to arise in the first place.

By contrast, brains carry masses of information, but only in a few species, and most dramatically in humans, has there even been a candidate for memetics. This must be because some structure or structures exist in the brain which make replication possible.

However, a DNA molecule with genetic information written into its structure can produce evolution as it replicates because the meaning of that information may have an affect on the number of copies. That is why I say that the gene is the information and DNA is the medium.

It's certainly true that a DNA molecule can be transcribed to an RNA molecule and create the same "gene", and probably has done in biological history, but the value of such an information transfer would be down to physical properties of both, and one of those properties would be replication. Sooner or later, you simply can't escape replication's role in the process.

I would say that replication is not the key mechanism and that information IS the thing to focus on. If something in the information in the gene or the meme has an affect on the number of copies produced, then mutation of that information can allow evolution to take place.

Again, I think you're confusing phenotypes with information. The phenotypes have an effect on the genes, it is true, but it is as an outcome of the replication effect that information enters the picture. After all, if all DNA did was manufacture proteins, nothing special would happen. You wouldn't even get bodies.

The genes for adaptations to hot climates that make up the phenotypes of a kangaroo or of a camel came from a random mutation generator, and the variants only failed to survive because they did not match onto the environment's needs as well as their allelic rivals did. The reason any useful gene can spread is because of replication. Even neutral drift requires a replicator.

Comment 36 by phil rimmer

Most anglophones of a certain age knew you meant A.A.Milne. How did we know? Multiple noisy channels. Execute the chinese whisper routine down enough (not necessarily contemporaneous) channels and we will be able to extract useful information.

And suppose neither of us had checked the original primary source and had gone on and described "E.E.Milne"? I wouldn't even have known it was the same person until phil rimmer mentioned it, by which point I could have passed it on to someone else, and they could have passed it on to someone else etc. To speak metaphorically, the mistake could be halfway around the world before the truth got its boots on, and yet never compete with A.A.Milne because people could easily think they were two different people.

The main reason you can verify the spelling in the first place is because the mass-produced books (one generation, remember) written by him still have his name on. The other main reason is that spelling is generally digital, so it's harder to get a name wrong than to get a biography about the man wrong. Nobody is suggesting that the name "A.A.Milne" arose when it competed with alleles for the slot of "name" among human brains and reached fixation by following selection pressures. Instead, the name was decided upon by a single brain (say, the mother or the father). The structure representing it in the brain did not replicate against its alleles until it reached fixation. "A.A.Milne" did not crowd out "B.B.Milne" or produce better "A.A.Milne" copies by comparison.

Again, all the language machinery and mental mechanisms were set up by genes beforehand. Most of the quibbling over a name is like quibbling over a tool to use for a specific purpose rather than like genes noisily crowding their rivals out of existence (his parents are unlikely to forget the alternative names they thought up for A.A.Milne any more than an evolution-supporter is likely to forget what creationism is). This is one reason why I think memetics is more likely to be a problematic rather than a helpful analogy in the long run.

Brains are wired and memories encoded using the associative Hebbian mechanism, cells that fire together wire together, thus, combined with a simple Bayesian predictor, have the structure to become pattern maximising discriminators extracting the most likely information from those channels.

My point is that there are alternative mechanisms to account for cultural changes that don't involve an "evolution by natural selection" analogy.

memes are the ideas that have what it takes to survive in their environment

This is the basis of a nice functional definition.

There are a mountain of problems with this definition. What is an "idea", and how would you distinguish it from a non-idea? What is it physically? What do you mean by "survive", and what is the "environment" in which they must live or die? How is this different from straightforward change and having a beginning, middle, and end? Does calling such a thing a "meme" lead to confusion, akin to calling groupiness "group selection"? And where do replicators appear in the definition?

This is one reason I stick to my group selection analogy, because it is too, too easy to get sloppy with the idea precisely because there are bits in it that make sense on their own - the spread of ideas, for instance. You have to ask yourself how a scientist would go about the problem.

I like this very much. You are describing what I have liked to describe in the past as "cultural machines" the processes of which are wired into brains in the very earliest of years. These machines (cultural processes) are concerned with the precise packaging of information and its formal organising permitting amongst of things noisier channels.

Aphoristic knowledge is certainly what we aspire to. It is highly portable and robust.

It may be worth considering that aesthetics could be key to memetic robustness.

See my reply to SC above.

I'll stop there for now, as I'm already pushing my luck writing a reply so long. Hopefully, it will contribute a little more to the discussion.

Fri, 27 Jul 2012 13:45:45 UTC | #950159