by Ian J. Deary
What happened when Loehlin and his colleagues compared pairs of siblings? Biologically related children in the adoptive families (i.e. children born from the same pairs of birth parents) have intelligence test scores that correlate about 0.3 or a bit less. But when biologically unrelated children who spend their lives in the same family are compared, then the correlations are around zero: they do not come to resemble each other in intelligence after a lifetime spent in the same family. Taken together, all these results suggest an effect of genes on intelligence and not much effect of family environment. Do recall the sizeable effect of unshared environment.
The other few adoption studies in this area are not huge, and the results are not definitive. However, when John Loehlin and two of his colleagues recently summarized their years of work with the Texas adoption project here was what they concluded:
The results on IQ from the Texas Adoption Project are generally consistent with the results from other behavior-genetic methods, such as the comparison of identical and fraternal [non-identical] twins, or the study of twins reared apart. The major contributor to familial resemblance is the genes. Shared family environment has an appreciable effect when children are small, but this becomes minor by the time they are late adolescents. However, we found in our data some tantalising suggestions that the full story of family effects may prove to be more complicated than this, in a weak negative environmental association between mothers’ and childrens’ IQs. A particularly striking manifestation was that the birth mothers showed, if anything, higher IQ correlations with the children they had had no contact with since near birth than the adoptive parents did with their own biological children with whom they had lived all their lives. (p. 123)
Do genes and the environment tend to affect general intelligence or the more specific cognitive abilities mentioned in Chapter 1?
Key dataset 9
We already know from Chapter 1 that there is general ability and there are identifiable, though related, specific types of mental ability, like verbal and spatial ability, memory, and mental speed. So, just as we asked about the effects of ageing on these different aspects in Chapter 2, we can now ask about the generality or specificity of the genetic effects.
I want to address this type of question using another remarkable dataset. It’s the OctoTwin project in Sweden, featuring a group of identical and non-identical twins who have taken many intelligence tests. The remarkable thing is that they are all over 80 years old. In addition, they are relatively healthy and free from dementia. The project was following up previous studies that found that the genetic influence on specific mental abilities mostly came via genetic effects on general mental ability. That is, these previous studies found that: (1) general intelligence was quite strongly influenced by genes; (2) the group factors of ability were very strongly related to general ability; and (3) much of the differences among people in these group factors could be traced to the genetic effect on general mental ability. The OctoTwin project researchers wondered whether things were different as people grew much older.
23. The results from the OctoTwin study, which show that differences in group factors of intelligence are heavily influenced by the genetic contribution to general intelligence.
Look at Figure 23. The picture of general mental ability – general intelligence g – and the special abilities or group factors that are related to it is familiar. Here we have used the special abilities that were measured by Stephen Petrill and his co-workers when they reported the OctoTwin results in the leading journal Psychological Science. In order not to have a Figure that was festooned with off-putting numbers, I have used the following labels: very strong associations have very thick black arrows and four plus signs; strong associations get three pluses; medium associations get two pluses and a thinner arrow; weak associations get a single plus; and where there is not much of an association at all there is a dotted line. You can see, then, that all of the four specific abilities are associated with a hypothetical general mental ability, or g factor; all of the arrows are very thick and have three or four pluses. Next I want to bring in the convention for looking at genetic and environmental (common and unshared) contributions to intelligence that were described in Figures 15–18 and 22. How much do genes and the environment contribute to the OctoTwin people’s differences in general intelligence and to specific mental abilities?
Let’s start with general intelligence – g – first. There is a very strong effect of genes on general ability. In fact, the results in this study indicated that, in people over 80, genes – G – contribute about 76% of the effects that result in intelligence differences. The other appreciable effect is from unshared environment – U. It contributes about 20% of the effects on individual differences in general ability test scores. Common environment – C – contributes almost nothing at this age.
So much for general ability. What about the specific abilities? Well, they are very strongly related to general ability. Take verbal ability as an example. Since it is so strongly related to g we see that the big genetic influence on g ‘flows through’ to verbal ability. That is, genes have a big effect on general ability at that age, and general ability contributes most of the differences in verbal ability, so the genetic effects that contribute to general intelligence differences play a big part in verbal ability differences. Therefore, the genetic influence on general ability contributes a lot to individual differences in all the more specific/group ability factors (stratum II in Chapter 1).
But g is not the full story with respect to the more specific abilities. It relates highly to them, but they are also independent of general ability to an extent as well. So, what affects the rest of the differences in group factors in mental ability? The answer lies at the bottom of the diagram. In addition to a g effect, each separable group factor (stratum II) in mental ability has genetic and environmental influences not shared with the other group factors and not due to g. Again, look at verbal ability as an example. There are additional, but weak, effects of genes and common and unshared environment on verbal ability that have nothing to do with general ability. For spatial ability there is a moderate additional effect of the unshared environment. Memory is the most interesting here. Note that it has two pluses leading from g to it. Therefore, as compared with the other three specific mental abilities, there is a greater proportion of the differences between people in memory to be accounted for that is not related to general mental ability.
Memory, over age 80 anyway, seems to be the ability that is least dependent on general intelligence. We see from the arrows at the bottom of the page that there are moderately strong influences of genes and unshared environment on memory differences that have nothing to do with the influences of genes and environment on general intelligence.
Do we know yet which genes have an influence on intelligence test score levels?
No. Researchers have discovered that genes play a sizeable part in influencing differences in mental ability between people, but as yet they have no idea what those genes are. By contrast with the case of some illnesses, they cannot point to a gene and say that if you have this form of the gene you will have such and such a level of ability. And the fact is that, outside the area of mental handicap, such a direct association between genes and intelligence is not going to happen. The best guess among researchers is that mental abilities are influenced by an unquantifiable number of genes, each of which will have a small effect. In the last few years the search for these genes that influence human mental ability levels has just begun. Only recently have laboratories begun to collect people’s DNA and begun to ask which variants in DNA structure are associated with higher and lower levels of mental ability.
To follow this area up …
For general background on intelligence, genes, and environment, I found the following useful. Plomin’s piece is written for a lay audience and includes more discussion of the social implications of genetic studies of intelligence
Bouchard, T. J. (1998). Genetic
and environmental influences on adult intelligence and special mental abilities. Human Biology, 70, 257–79.
Plomin, R. (1999). Genetics and general cognitive ability. Nature, 402 (Suppl.), C25–C29.
Another good, and very straightforward and brief, article about intelligence, environment, and genes is the following.
Petrill, S. A. (1997). Molarity versus modularity of cognitive functioning? A behavioral genetic perspective. Current Directions in Psychological Science, 6, 96–9.
For a very clear description of the Minnesota twin study:
Bouchard, T. J., D. T. Lykken, M. McGue, N. L. Segal, & A. Tellegen (1990). Sources of human psychological differences: the Minnesota Study of Twins Reared Apart. Science, 250, 223–8.
Although the following book is aimed at the student and researcher, there are excellent descriptions of many key aspects of the genes and the environment and how they affect intelligence. See the chapter noted here for a good account of the Texas Adoption Project.
Loehlin, J. C., J. M. Horn & L. Willerman (1997). Heredity, environment, and IQ in the Texas Adoption Project. In R. J. Sternberg & E. Grigorenko (eds), Intelligence, Heredity and Environment. Cambridge: Cambridge University Press.
There were two good papers that I drew from in describing the Swedish study of old twins: both are technical, written for researchers.
McClearn, G. E. (et al.) (1997). Substantial genetic influence on cognitive abilities in twins 80 or more years old. Science, 276, 1560–3.
Petrill, S. A. (et al.) (1998). The genetic and environmental relationship between general and specific cognitive abilities in twins age 80 and older. Psychological Science, 9, 183–9.
There’s a lot more to genetics and the environment and how they contribute to intelligence test scores than I have been able to introduce here. For more on important issues like ‘gene–environment’ interaction and correlation and the ‘shared environment assumption’ and so forth, see
Plomin, R. (et al.) (1997, 3rd edn). Behavioral Genetics. New York: W. H. Freeman.
If you are puzzled or annoyed at the apparent lack of influence of family upbringing on intelligence (and other psychological characteristics, too, it seems), you must read the following book, which is devoted to this finding.
Harris, J. R. (1998). The Nurture Assumption: Why Children Turn Out the Way They Do. London: Bloomsbury.
Chapter 5 The (b)right man for the job
Does intelligence matter?
Entire books – popular and scholarly – are available that disparage the invention and application of intelligence tests. It is certainly true that intelligence tests were used inappropriately and over-zealously at times during the 20th century, and to the exclusion of other important human characteristics. They are a tool that may be misused, to be sure. All tools run this risk, but, as Queen Elizabeth I ripostes in Sir Walter Scott’s Kenilworth, ‘ “it is ill arguing against the use of anything from its abuse” ’, so let us move on and ask if they have utility. And think what we are asking. It is this: does a score on a short test of mental ability have any predictive power for some aspects of real-life achievement? We are not asking whether an intelligence score totally predicts human achievements – it never does, or anything near it – just whether intelligence test scores have some useful predictive power.
The first tests of human intelligence appeared in 1905. They were developed by Alfred Binet and Theophile Simon in Paris. These two researchers were given a practical problem: how might the authorities identify those children who would not benefit from the normal style of education? The IQ-type tests, which now number many hundreds, were their answer. Therefore, what we call tests of intelligence were invented to serve a practical purpose.
Currently, the main applications of intelligence tests are in education, in the workplace, and in medicine. Thus, mental tests are used to assess mental capability in the settings of school performance, work performance, and in looking at the effects of illnesses and medical treatments on the brain’s functions. It is well known that psychometric tests do a reasonable job of predicting educational attainment (see the Task Force report discussed in Chapter 7). There are other important factors too, but one’s score on a mental test has some moderately strong relation to future educational achievements. However, for the illustration of the potential impact of mental testing I am going to focus on an application from the field of work.
Key dataset 10
The work-related dataset I shall refer to is a remarkable compilation of findings by John Hunter, with his research colleagues Ronda Hunter and Frank Schmidt. Their interests are in job selection, in finding the right people to do a job well. They asked the following simple-seeming question: is it worthwhile for an employer to select people for a job on the basis of, among other things, a test of general mental ability (general intelligence)? The emphasis here is not on each of the individuals offering themselves for selection: it is on those using the test to make the selection and it is centred on a practical problem. That is, imagine you are an employer and you wish to select people to begin new jobs in your workplace. What is the best method of selecting the most productive new staff? How can you tell who will bring the most benefit to your organization? In essence, among the criteria that you compile in your selection portfolio, would it be worthwhile having a test of general mental ability?
Before going that far, Hunter and his colleagues point out some factors that you as a hirer might wish to consider. First, is there any variability in people’s performance on the jobs you are thinking about? If everyone does the job equally well, no matter what their personal qualities, strengths, and weaknesses, then why are you worrying about hiring decisions? If there is absolutely no difference in job performance between people then, with respect to productivity, you don’t have a problem. (At least, not one to do with productivity; you might reasonably want to select people you’ll enjoy working with.) That’s unlikely: in most jobs there will be some people who do the job better than others. And the bigger these differences, the more you have to be concerned about whom you hire. If there are huge differences in how well people do the job for you, then you want to get the people who will do the job best.
The second factor with respect to hiring that Hunter talks about is the amount of selectivity you can apply in choosing. That is, do you have the luxury of taking those whom you consider to be the very best people for the job, or must you take whoever appears for the job interview? Imagine a situation where you have 10 jobs to fill and 100 people apply. That gives you the luxury of picking the top 10% of applicants, and if you have a good method of selection you can get the cream of that 100 into your business. What if only 20 people apply? Instead of getting the top 10%, you have to take the top 50%. You’ll be selecting people who are not quite so good among the 10 successful applicants. If only 10 people apply you have to take all comers – those who will be good, mediocre, and poor at the job. Compared with the business that has the luxury of skimming off the cream, the top 10% of the best workers, you are going to lose productivity and income.
Help for aspiring employers there. But there is something missing so far. We’ve identified the fact that you only really need to worry about hiring decisions related to productivity when some people are better than others at the jobs you have vacant. Next, we’ve identified the fact that the more you are in a position to skim off the really best workers, the better are your chances of high production. The missing factor here is something that will identify the best workers. You need some basis for selection. You need some test that you can apply to your applicants so that you can pick out those people who will do the best job. You do not have infinite time or money to apply this test; the cheaper and quicker it is, the better. And, of course, the more accurately it can predict future job performance the better.
Does this really matter? Aren’t we really discussing some small, marginal difference in income here? Perhaps we should worry less about productivity and focus more on giving everyone a
n equal chance of being hired, no matter what their qualities for the job. John Hunter provided some back-of-the-envelope calculations. He based his numbers on the USA’s federal government around 1980. They hired about 460,000 people in any one year. The average tenure of their workers is 6 years. The average wage at that time was about $13,500. They were usually in the position of being able to select the top 10% of applicants; the jobs were popular and attracted many able people. Let’s assume they have some method of selection that relates quite highly to job performance – a correlation of just over 0.5 between the selection ‘test’ and later job performance. (By the way, it is not easy to measure job performance, and it is often based upon ratings by supervisors in the studies we shall discuss.) Given this setting and these assumptions, Hunter worked out the difference in cost, based on productivity differences, between applying and not applying the selection ‘test’. If you applied the selection ‘test’ in that situation, you would have a productivity gain of $15,610,000,000: over fifteen billion dollars (at 1980 prices).