The development of language was thus an evolutionary process like any other, creating complex design apparently out of nowhere. The early products of human sound copying changed the environment of memetic selection so that more complex sounds could find a niche. Just as multicellular organisms could arise only when single cells were already common, just as animals could appear only when plants were already producing oxygen, just as large predators could evolve only when there was plenty of small game about, so complex grammatically structured utterances could appear only when simpler ones were already common. A language with lots of words and well–defined structures would seem to be the natural result of memetic selection.
The next step is to understand how language itself was able to restructure the human brain and vocal system for its own propagation. This is meme–gene coevolution again and works as follows. I have assumed that people will both preferentially copy and preferentially mate with the people with the best memes – in this case the best language. These people then pass on
genetically
whatever it was about their brains that made them good at copying these particularly successful sounds. In this way, brains gradually become better and better able to make just these sounds. Grammatical language is not the direct result of any biological necessity, but of the way the memes changed the environment of genetic selection by increasing their own fidelity, fecundity, and longevity.
Note that this whole process is self–sustaining. Once language evolution begins, both the language itself and the brain on which it runs will continue to evolve under the combined pressure of memetic and genetic selection. This is not the only theory to treat language as ‘its own prime mover’, or as a self–sustaining process, but others have trouble with explaining how it ever began or why it takes the form it does. Deacon, for example, had to find a reason for crossing the ‘symbolic threshold’ in the first place. There is no such problem with the memetic theory of language origins. The critical step was the beginning of imitation – and there is no mystery about why natural selection would have favoured imitation. It is an obvious, if difficult to find, ‘good trick’, and one that is especially likely to arise in a species that already has good memory and problem–solving skills, reciprocal altruism, Machiavellian Intelligence and a complex social life. Once found, it sets in motion the evolution of a new replicator and its coevolution with the old.
I have done a lot of speculating and imagining here. Am I just making another equivalent of the ‘bow–wow’ or ‘heave–ho’ theories? Should I be reminded of the ban made by the Société de Linguistique de Paris?
I hope not. The difference here is that I am not suggesting that words arose because people heaving on heavy rocks made ‘heave–ho’ noises and began to speak – though I suppose the odd word might have come about that way. I am suggesting that verbal language is almost an inevitable consequence of memetic selection. First, sounds are a good candidate for high–fecundity transmission of behaviour. Second, words are an obvious way to digitise the process and so increase its fidelity. Third, grammar is a next step for increasing fidelity and fecundity yet again, and all of these will aid memorability and hence longevity. Once the second replicator arose, language was more or less inevitable.
The theory depends on a few basic assumptions, and these could be tested. One is that people preferentially copy the most articulate people. Social–psychological experiments show that people are more easily persuaded by ‘good talkers’ and ‘fast talkers’, but this needs more systematic research using tests of imitation.
Meme–gene coevolution assumes that people preferentially mated with the best meme–spreaders, in this case the most articulate people. We should remember that past selection for ‘good talkers’ may have used up most of the original variation, leaving most of us fairly articulate today. However, the preference may still be there, so that being highly articulate makes you sexually attractive. The history of love poems and love songs suggests as much, as does the sexual behaviour of politicians, writers and television stars (Miller 1993).
If the theory is right then human grammar should show signs of having been designed for transmitting memes with high fecundity, fidelity, and longevity, rather than to convey information about some particular topic such as hunting, foraging or the symbolic representation of social contracts. This is the memetic equivalent of adaptationist thinking in biology and I might be criticised for assuming that memetic evolution must always have found the best solution and for a kind of circular reasoning. Nevertheless, adaptationist thinking has been extremely effective in biology and may prove so in memetics.
Language continually evolves, and new words or expressions compete to be adopted, or co–opted from other languages. Again, we should expect the winners to be those of high fidelity, fecundity, and longevity. Wright (1998) has used memetics to study the translation of chemical terms such as acid, alcohol, or various elements, into Chinese, showing that alternative terms underwent intense competition for survival, with the
winners depending both on properties of the terms themselves and on the meme products already in existence at the time.
Whole languages also compete with one another for survival. Where languages have coexisted in the past we would expect the survivor to be the better replicator, and that languages with especially low–quality replication would most easily be destroyed. Now that so many languages are threatened with extinction this memetic approach might help us to understand what is happening. There is also a battle waging between the major world languages for dominance (or just survival) in industry, finance, transport, and information technology. Historical accidents have made some better placed than others, but we might usefully look at the evolution, competition, and extinction of languages with three things in mind – the fidelity, fecundity, and longevity of the memes they convey.
Finally, we should be able to predict how artificial languages could arise. There have been many attempts to get robots, or virtual robots, to use language. These usually begin by teaching the artificial systems a lot about natural languages, or by getting them to make associations between sounds and objects. The theory I have proposed suggests an entirely different approach that assumes no knowledge of any prior language, and no concept of symbolic reference.
Let’s imagine a group of simple robots, ambling about in some kind of relatively interesting and changing environment. We can call them copybots. Each copybot has a sensory system, a system for making variable sounds (perhaps dependent on its own position or some aspect of its sensory input), and a memory for the sounds it hears. Most importantly, it can imitate (though imperfectly) the sounds it hears. Now, imagine that all the copybots start roaming around squeaking and bleeping, and copying each other’s squeaks and bleeps.
The environment will soon become full of noise and the copybots will be unable to copy every sound they hear. Depending on how their perception and imitation systems work, they will inevitably ignore some sounds and imitate others. Everything is then in place for the evolutionary algorithm to run – there is heredity, variation, and selection – the sounds (or the stored instructions for making the sounds) are the replicator. What will happen now? Will there just be an awful cacophony, or will something interesting emerge? If the theory is correct then some sounds will have higher fidelity, longevity, and fecundity (depending on characteristics of the copybots) and these should be copied more and more accurately, and patterns begin to appear. Some sounds would be made more often, depending on events in the environment and the positions of the copybots themselves. I think this could be called
language. If so, it would not be the same as any language currently used by any natural or artificial systems.
If this worked, interesting questions would arise. Are the copybots really communicating? Are they talking
about
something? If so, symbolic reference would have arisen out of simply providing the robots with the capacity to imitate. In other words, the capacity to imitate is fundamental, not the capacity for symbolic reference. That is exactly what I would expect. The final question is, could we ever understand them?
To summarise, there is a memetic solution to the mystery of human language origins. Once imitation evolved, something like two and a half or three million years ago, a second replicator, the meme, was born. As people began to copy each other the highest–quality memes did the best -that is those with high fidelity, fecundity, and longevity. A spoken grammatical language resulted from the success of copyable sounds that were high in all three. The early speakers of this language not only copied the best speakers in their society but also mated with them, creating natural selection pressures on the genes to produce brains that were ever better and better at spreading the new memes. In this way, the memes and genes coevolved to produce just one species with the extraordinary properties of a large brain and language. The only essential step to starting this process was the beginning of imitation. The general principles of evolution are enough to account for the rest.
The answers to two difficult questions are now obvious, and the same. What is the big brain for? What is the function of language? – To spread memes.
CHAPTER 9
The limits of sociobiology
I have proposed two new theories – memetic theories – to account for human brain size and the origins of language. They both depend on the replicator power of the meme, and introduce some new principles into the way memes and genes interact – the processes I have called ‘meme–gene coevolution’ and ‘memetic driving’. I want now to set this memetic approach in context; to see how it compares with other theories and explain why theories based purely on biological advantage must fail. By exploring the different ways in which memes and genes can interact we will come up against the limits of sociobiology.
First, theories of ‘coevolution’ are not new. As I explained in
Chapter 3
there have been many, including those of Boyd and Richerson (1985), Deacon (1997), Donald (1991), Durham (1991) and Lumsden and Wilson (1981). What makes the present theory of meme–gene coevolution different is that both halves – the memes and the genes – are replicators in their own right, with equivalent status. Certainly, the two replicators are different. They differ in how they work, how they are copied, and the timescales over which they operate. There is also an important asymmetry between them in that memes can operate only by using the brains created by genes, whereas genes can (and do) operate perfectly well without memes. Nevertheless, both memes and genes have replicator power. They are essentially only out for themselves and if they can get copied they will – the rest follows from there.
Dawkins complained that his colleagues always wanted to go back to biological advantage. This theory does not go back only to biological advantage, but to memetic advantage as well. With two replicators working together things can get complicated, but not impossibly so, and with a bit of simplification we can tease out the three major types of interaction: gene–gene interactions, gene–meme interactions, and meme–meme interactions.
Gene–gene interactions
Gene–gene interactions are the stuff of biology. When white bears manage to stalk more seals on the arctic ice than brown bears do, genes for producing white fur spread at the expense of genes for producing brown fur. In this way, rival versions of genes (alleles) compete with each other. Genes also cooperate, however – otherwise we would not have organisms at all. In our own bodies, thousands of genes cooperate to produce muscles and nerves, liver and brain, and to result in a machine that effectively carries all the genes around inside it. Gene–gene cooperation means that genes for digesting meat cooperate with genes for hunting behaviour, while genes for digesting grass cooperate with genes for grazing and chewing the cud. Of course, they do not cooperate out of kindness but because it benefits their own replication to do so.
But these are not the only kinds of gene–gene interaction. Genes in one creature can affect genes in another. Mouse genes for fast running drive cat genes for pouncing quicker. Butterfly genes for camouflage drive better eyesight in birds. In this way, ‘arms races’ develop in which each creature tries to outwit the other. Many of the most beautiful creations of the natural world are the result of genetic arms races. Organisms exploit each other, as when ivy uses a tree to get height without building its own trunk, or parasites live inside the bodies of people and get their food for free. But others cooperate with each other in symbiotic relationships, such as ants and aphids that provide each other with protection and nourishment, or the many bacteria that live inside our own intestines and without which we could not digest certain kinds of food. It is even thought that the tiny mitochondria that provide the energy inside every living cell originated from symbiotic bacteria. They have their own genes, and these mitochondrial genes are passed on from mother to child in addition to all the more familiar human genes in the cell nucleus.
Another way of looking at the world is to see whole ecosystems as constructed by the interactions between selfish genes. Genes can have multiple effects (a single gene
for
a single effect is a rarity), and may be packaged inside different organisms. Dawkins (1982) provides many examples of what he calls the ‘extended phenotype’, by which he means all the effects of a gene on the world, not just on the organism in which it sits. Beavers build dams and those dams are as much an effect of genes as are spiders’ webs or snail shells or human bones. But the genes concerned need not even be part of the organism that builds the structure. For example, there is a parasitic fluke that lives inside snails and causes them
to grow thicker shells. Dawkins suggests that the thickness of a snail’s shell is a trade–off between growing a thick shell to protect it from birds, and saving the resources to make more baby snails. The fluke genes will not benefit from more baby snails, but they will benefit from a safe snail to live in – so fluke genes for making snails grow thicker shells are good replicators. This illustrates the important point that although the interests of a gene and the interests of the organism it sits in usually coincide, they do not always do so.