TODAY we view TV documentaries about identical twins who, despite being separated at birth, have had amazingly similar life experiences and grow up to have similar IQs.

But when we think about what those twins imply, idle entertainment turns into concern and anger. Must we believe that genes virtually determine IQ, and that IQ differences between racial groups are caused by genetic differences? For psychologists, there is a special cross to bear. The race and IQ debate has created a paradox about nature versus nurture that appears insoluble.

Ever since the American military tested conscripts during the First World War, it has been known that whites in the US outscore blacks by 15 points on IQ tests. In 1969, Arthur Jensen, an educational psychologist at the University of California at Berkeley, shocked liberal American public opinion by arguing that the racial IQ gap had a strong genetic component.

In 1973, Jensen bolstered his case by publishing a mathematical model. He plugged in two pieces of data: the 15-point black IQ deficit (one standard deviation), and a low estimate of the influence of environment on IQ (about 0.33). These implied that for environment to explain the IQ gap between blacks and whites, the environmental gap between the races would have to be immense (three standard deviations).

His evidence that environment was overshadowed by genes came from a wide variety of kinship studies. These included not only identical twins brought up apart, but also comparisons of identical and fraternal twins each brought up by their own parents, comparisons of adopted children with "natural" children and so forth. Most psychologists agreed (and would still agree) with him about the message of the kinship studies.

The mathematics of Jensen's model are impeccable. If environment is that weak, it will take a lot of it to cover a large IQ gap. The implication for blacks: their average environment would have to be worse--in its effect on IQ--than the environments of 99 per cent of whites. No social statistic, from income to occupation to school-ing, puts the average black person in such an impoverished environment.

Enter the paradox.

Jensen's model did more than imply that environment could not explain the IQ gap between the races. It ruled out an environmental explanation of a large IQ gap between any two groups. However, in 1987, one of us, James Flynn, did a worldwide survey of IQ trends over time and found that the current generation outscores the previous generation by between 9 and 20 IQ points. The size and the speed of these IQ gains virtually dictate an environmental explanation because genes just don't change that fast. So somehow, Jensen's model has led us astray.

To state the paradox starkly: how could solid evidence show both that environment was so feeble (kinship studies) and yet so potent (IQ gains over time)?

Two years ago, William Dickens of the Brookings Institution in Washington DC decided to do some modelling of his own. Building on the work of scholars such as Sandra Scarr, Uri Bronfenbrenner, Steve Ceci, Carmi Schooler, Judith Rich Harris and Eric Turkheimer, Dickens has constructed a model that we believe provides the solution.

This model retains all of the kinship data but reassesses it in the light of two assumptions. First, that those who have an advantage for a particular trait will become matched with superior environments for that trait. Secondly, that genes derive a great advantage from this because genetic differences are persistent. If your genes make you tall, you retain that advantage throughout your life. Environment is important, but, as we shall see, the environmental advantages that persist tend to be selected by our genes.

Take those born to be a bit taller and a bit quicker than average. When they start school, they are likely to be a bit better at, say, basketball. The advantage may be modest, but then reciprocal causation between the talent advantage and environment kicks in. Because you are better at basketball, you are likely to enjoy it more and play it more than someone who is a bit slow or short or overweight. Your genetic advantage is upgrading your environment, the amount of time you spend playing and practising. In turn, your enhanced environment upgrades your skill, so you are much more likely to be picked for your school team. There you get professional coaching, which makes you even more proficient.

So thanks to genes capitalising on the powerful multiplying effects of the feedback between talent and environment, a modest genetic advantage has turned into a huge performance advantage. And just as small genetic differences match people with very different environments, so identical genes tend to produce very similar environments.

Assume that everybody plays basketball and that your skill has as much influence on your fate as your success at school. In that case, identical twins, with the same height, weight, natural spring and quickness, would tend to "create" very similar basketball environments even when they grew up apart. Their environments would not, of course, be as similar as when they grew up together. Even so, their basketball histories would tend to be far more similar than those of randomly selected individuals who grew up apart, one of whom might be 15 centimetres taller than the other.

Kinship studies of basketball, whether they involved people with identical genes or different genes, would underestimate the potency of environmental factors. Playing, practising, being on a team, coaching, would all be attributed to genes simply because similarities and differences in them tend to accompany genetic similarities and differences between individuals.

This generation is far more skilled at basketball than the previous one, undoubtedly due to changing environment rather than genes. What would Jensen's model show? If he used kinship data, they would imply that those aspects of environment not matched with genes (which is all that environment gets credit for in kinship studies) are feeble--far too feeble to cover a large basketball performance gap. And his mathematics would demonstrate that the present generation would have to have an average environment for basketball better than that of 99 per cent of the previous generation. This, of course, would seem quite implausible, but now we know where the implausibility lies: with the conclusions generated by Jensen's model.

Powerful engine

The differences in cognitive ability that are measured by IQ tests may have the same dynamics. People whose genes send them into life with a small advantage for developing these abilities start with a modest performance advantage. Then genes begin to drive the powerful engine of reciprocal causation between ability and environment. You begin by being a bit better at school and are encouraged by this, while others who are a bit "slow" get discouraged. You study more (which upgrades your cognitive performance), earn praise for your grades, start haunting the library, get into a top stream, and so on. Another child finds that sport is his or her strong suit, does the minimum, does not read for pleasure, and gets into a lower stream. Both of you may go to the same school, but the environments you make for yourselves there will be radically different. The modest initial cognitive advantage conferred by genes becomes enormously multiplied.

Once again, just as different genes are matched with very different environments, so identical genes will be matched with very similar environments. Remember those TV documentaries. They show identical twins who, despite being separated at birth, joined the library club at the same age, won a school prize at the same age, married spouses with surprisingly similar backgrounds, practised the same professions. The sensational frequency of the "coincidences" obscures what is being revealed: the remarkable tendency of similar genes to get matched to similar environments.

The TV shows, the kinship studies, and Jensen's model have one thing in common: when genes get matched with potent environmental influences, genes get all the credit. The identical genes of identical twins get credit for all of the potent environmental influences they share. Our model shows why this is wrong. It hides or masks the true potency of environmental influences on IQ. It makes environment look too feeble to cause massive IQ gains over time.

The model's next task is to suggest just how environment performs its demanding role. Forces affecting the whole of society can provide something that an individual's life experiences usually cannot. They provide environmental trends that are just as persistent over time as the individual's genetic endowment, and environmental advantages that operate quite independently of genetic advantages. After all, the present generation has no better genes for IQ than the one before. Indeed, the reverse may be true because of lower birth rates among the more highly educated. Between generations, relatively small environmental differences gain enormous potency, just as small genetic differences between individuals did: they seize control of the powerful reciprocal causation that exists between cognitive ability and environment.

No one knows for certain what environmental trends caused massive IQ gains, but we can suggest a scenario. Massive gains in the cognitive abilities that IQ tests measure may have begun with the Industrial Revolution. Test results from Britain show gains beginning as far back as the generation born in 1872. The Industrial Revolution has steadily upgraded the quality of schooling, nutrition, disease control--trends that could have had a profound influence in raising IQ, at least up to about 1950.

After 1950, some nations such as the US and Britain have provided test information that allows us to break down global IQ gains into their components. The gains begin to show a new and peculiar pattern. They are missing or small on the kind of IQ tests closest to school-taught material (vocabulary, arithmetic and so forth), but they are huge on tests that emphasise on-the-spot problem solving (working out what verbal abstractions have in common, finding the missing piece of a matrices pattern, or arranging pictures to tell a story).

Perhaps the Industrial Revolution stopped demanding progress in the basics and started demanding that people take abstract problem solving more seriously. Children brought up in the affluence of the 1950s did not develop the Depression-era psychology of their parents, with its emphasis on the practical over the "frivolous". Affluence brought smaller families in which children's "whys" were taken more seriously, more professional and semi-professional jobs in which people were expected to take more initiative. It gave people the time and energy to make leisure more cognitively demanding, from chess, bridge or video games to conversation in which people were expected to take ideas and logic seriously.

We call these products of the Industrial Revolution "triggers". Models do not themselves specify ultimate causes and we suggest the above very tentatively. What the model does do is demonstrate the potency that triggers, whatever they may be, could gain from seizing control of reciprocal causation between cognitive ability and environment.

The most dramatic tool at the disposal of these triggers is the "social multiplier". This posits that when something raises the average performance of society, that rise becomes a powerful cause in its own right, and raises the average performance further and further until the original rise is greatly multiplied.

The most potent facet of our environment is other people. When something, perhaps the popularity basketball got from being televised, triggered greater participation in the sport, the average performance rose as one individual after another played more and got better. Initially, a few people learned to shoot with either hand, then others imitated them. The rise in average performance feeds back into a new challenge for each individual. Those who want to excel have to learn to pass with either hand and this raises the average performance once again. Every rise in individual performance raises the group average, which forces everyone to raise their individual performance a notch, which raises the group average a notch higher, and so on.

That's how even a modest environmental trigger of enhanced performance can become potent--by seizing control of the social multiplier--and cause huge performance gains in a relatively short time.

The same kind of reciprocal causation may explain IQ gains. Environmental triggers raise the cognitive demands of work, family interaction, leisure and everyday conversation. Those who respond by upgrading their cognitive performance raise the average cognitive performance. Then the rising average affects your employer, family and friends, and they demand or expect more. You (and many others) rise to meet their expectations, and the average cognitive performance jumps again.

Our model incorporates the social multiplier, which means it can show how quite modest initial environmental changes would be enough to explain huge IQ gains of 20 points over a single generation. Gains need not go on forever: the model builds in limits--and it is easy to see how brain capacity or an affluence that bred decadence could halt or even reverse IQ gains.

The model has a third task to perform. It might well help us make sense of a whole range of questions posed by IQ data. Why, for example, does the match between a genetic advantage for IQ and environmental advantages for IQ tend to become greater with age? Why do children make big IQ gains at the start of an enrichment programme, make little or no gain during the rest of the programme, and gradually lose those gains after the programme ends? Why do the IQs of adopted children, at adulthood, correlate better with the IQs of their natural parents than with those of their adoptive parents?

And to return to the race and IQ debate, the model suggests how environment might explain racial IQ differences just as it explains IQ differences between generations.

Finally, the model has acquired an overriding purpose. Since it applies--in principle--to the dynamics of any human ability where there is positive feedback between that ability and environment, we hope it will bridge the (sometimes bitter) division between hereditarians who think genes are dominant and environmentalists who think culture is dominant. They are both right: it all depends on whether genetic differences or environmental factors seize control of potent processes like the social multiplier.

From psychologists inspired by Hans Eysenck and Robert Plomin to anthropologists inspired by Franz Boas and Ruth Benedict, our model offers them all common ground. Perhaps we can all work together to advance our understanding of human intelligence and other important traits.