Matt Ridley

We human beings may be determined to a surprising extent by the dictates of our genes, but we are determined even more by what we have learnt in our lifetimes. The genome is an information-processing computer that extracts useful information from the world by natural selection and embodies that information in its designs.

Evolution is just terribly slow at processing the information, needing several generations for every change. Little wonder that the genome has found it helpful to invent a much faster machine, whose job is to extract information from the world in a matter of minutes or seconds and embody that information in behaviour - the brain.

Your genome supplies you with the nerves to tell when your hand is hot. Your brain supplies you with the action to remove your hand from the stove-top.

The subject of learning lies in the provinces of neurosciences and psychology. It is the opposite of instinct. Instinct is genetically-determined behaviour; learning is behaviour modified by experience.

The two have little in common, or so the behaviourist school of psychology would have had us all believe during much of the twentieth century. But why are some things learnt and others instinctive?

Why is language an instinct, while dialect and vocabulary are learnt?

James Mark Baldwin, the hero of this chapter, was an obscure American evolutionary theorist of the last century, who wrote an article in 1896 summarising a dense and philosophical argument that had little influence at the time, or indeed at any time during the subsequent ninety-one years. But by a stroke of good fortune, he was plucked from obscurity by a group of computer scientists in the late 1980s, who decided his argument was of great relevance to their problem of teaching computers how to learn.1

What Baldwin wrestled with was the question of why something is learnt by an individual in his lifetime rather than pre-programmed as an instinct. There is a commonly held belief that learning is good, instinct bad - or, rather, that learning is advanced and instinct primitive. It is therefore a mark of human rank that we need to M E M O R Y 2 2 1

learn all sorts of things that come naturally to animals. Artificial-Intelligence researchers, following this tradition, quickly placed learning on a pinnacle: their goal was the general-purpose learning machine. But this is just a factual mistake. Human beings achieve by instinct the same things that animals do. We crawl, stand, walk, cry and blink in just as instinctive a way as a chick. We employ learning only for the extra things we have grafted on to the animal instincts: things like reading, driving, banking and shopping. 'The main function of consciousness', wrote Baldwin, 'is to enable [the child] to learn things which natural heredity fails to transmit.'

And by forcing ourselves to learn something, we place ourselves in a selective environment that puts a premium on a future instinctive solution to the problem. Thus, learning gradually gives way to instinct. In just the same way, as I suggested in the chapter on chromosome 13, the invention of dairy farming presented the body with the problem of the indigestibility of lactose. The first solution was cultural - to make cheese — but later the body evolved an innate solution by retaining lactase production into adulthood. Perhaps even literacy would become innate eventually if illiterate people were at a reproductive disadvantage for long enough. In effect, since the process of natural selection is one of extracting useful information from the environment and encoding it in the genes, there is a sense in which you can look on the human genome as four billion years'

worth of accumulated learning.

However, there comes a limit to the advantages of making things innate. In the case of spoken language, where we have a strong instinct, but a flexible one, it would clearly be madness for natural selection to go the whole hog and make even the vocabulary of the language instinctive. That way language would have been far too inflexible a tool: lacking a word for computer, we would have to describe it as 'the thing that thinks when you communicate with it'.

Likewise, natural selection has taken care (forgive the teleological shorthand) to equip migratory birds with a star-navigation system that is not fully assembled. Because of the precession of the equi-noxes, which gradually changes the direction of North, it is vital 2 2 2 G E N O M E

that birds recalibrate their star compass in every generation through learning.

The Baldwin effect is about the delicate balance between cultural and genetic evolution. They are not opposites, but comrades, trading influence with each other to get best results. An eagle can afford to learn its trade from its parents the better to adapt to local conditions; a cuckoo, by contrast, must build everything into instinct because it will never meet its parents. It must expel its foster siblings from the nest within hours of hatching; migrate to the right part of Africa in its youth with no parents to guide it; discover how to find and eat caterpillars; return to its birthplace the following spring; acquire a mate; locate the nest of a suitable host bird - all by a series of instinctive behaviours with judicious bouts of learning from experience.

Just as we underestimate the degree to which human brains rely upon instincts, so we have generally underestimated the degree to which other animals are capable of learning. Bumble bees, for instance, have been shown to learn a great deal from experience about how to gather nectar from different types of flowers. Trained on one kind, they are incompetent at another until they have had practice; but once they know how to deal with, say, monkshood, they are also better at dealing with similar-shaped flowers such as lousewort - thus proving that they have done more than memorise individual flowers, but have generalised some abstract principles.

Another famous example of animal learning in an equally simple creature is the case of the sea slug. A more contemptibly basic animal is hard to imagine. It is slothful, small, simple and silent. It has a minute brain and it lives its life of eating and sex with an enviable lack of neurosis. It cannot migrate, communicate, fly or think. It just exists. Compared with, say, a cuckoo or even a bumble bee, its life is a cinch. If the idea that simple animals use instincts and complicated ones learn is right, then the sea slug has no need of learning.

Yet learn it can. If a jet of water is blown upon its gill, it withdraws the gill. But if the jet of water is repeatedly blown on the gill, the M E M O R Y 2 2 3

withdrawal gradually ceases. The sea slug stops responding to what it now recognises as a false alarm. It 'habituates'. This is hardly learning the differential calculus, but it is learning all the same.

Conversely, if given an electric shock once, before water is blown on the gill, the sea slug learns to withdraw its gill even further than usual - a phenomenon called sensitisation. It can also be 'classically conditioned', like Pavlov's famous dogs, to withdraw its gill when it receives only a very gentle puff of water if that gentle puff is paired with an electric shock: thereafter, the gentle puff alone, normally insufficient to make the sea slug withdraw its gill, results in a rapid gill withdrawal. Sea slugs, in other words, are capable of the same kinds of learning as dogs or people: habituation, sensitisation and associative learning. Yet they do not even use their brains. These reflexes and the learning that modifies them occur in the abdominal ganglion, a small nervous substation in the belly of the