'g' is for ability but it's also for genes

In our series on Big Science Questions, Harriet Swain looks at controversies over intelligence and Robert Plomin discusses the crucial role of genetics

The word "intelligence" means so many different things that it is best to use other terms to avoid confusion. What I mean by intelligence is "general cognitive ability", or g, which refers to the substantial overlap that exists between different cognitive processes. This overlap is one of the most consistent findings of research into individual differences in human cognitive abilities over the past century. It has been found even in tests of processes that seem to have little in common.

For example, general reasoning is assessed through tests such as Raven's Progressive Matrices, in which subjects need to detect logical progression in a series of matrices consisting of geometric forms. Spatial ability is assessed through solving mazes, identifying simple geometric figures hidden in more complex shapes and deciding whether or not one figure is a rotated version of another. Vocabulary tests assess the product of previous learning, and tests of memory typically involve presenting digits or pictures to see how well they are recalled.

Despite this diversity, individuals who do well in one test tend to do well in others. In a meta-analysis by John Carroll in 1993 of 322 studies that included hundreds of different kinds of cognitive tests, the average correlation was about 0.30, which is highly significant. This overlap emerges not only for traditional measures of reasoning and of spatial, verbal and memory abilities, such as those mentioned above, but also for rates of learning and for information-processing tasks that rely on reaction time.

Psychologist Charles Spearman recognised the overlap in cognitive abilities nearly a century ago. He called it g to create a neutral signifier of general cognitive ability that avoided the many connotations of the word intelligence. This g is best assessed by a statistical technique called principal components analysis. This identifies a composite dimension that represents what diverse cognitive measures have in common.

Such analysis indicates that g accounts for about 40 per cent of the total variance in people's performance of cognitive tests. The rest of the variance is accounted for by factors such as spatial, verbal and memory abilities and by variance unique to each test. The more complex the test, the higher the importance appears to be of the g factor. For example, g accounts for a high amount of variance in Raven's Progressive Matrices, while it is less important in simple memory, reaction or processing speed tests.

But g is not just a statistical abstraction. A look at a matrix of correlations between cognitive measures will show all overlap strongly while g is also indexed reasonably well by a simple total score on a diverse set of cognitive measures, as is done in IQ tests. In fact, g is one of the most reliable and valid traits in the behavioural domain; its long-term stability after childhood is greater than for any other trait, it is better at predicting important social outcomes such as educational and occupational levels, and it is a key factor in cognitive ageing. There are, of course, many other important non-cognitive abilities, such as athletic ability, and g by no means guarantees success in school or in the workplace. Achievement also requires personality, motivation and social skills now referred to as "emotional intelligence". But nothing seems to be gained by lumping all such abilities together as the popular notion of "multiple intelligences" does. For me, g is what intelligence is about.

Although the evidence for g in the human species is widely accepted, acceptance is not universal. Arguments against it involve ideological issues, such as concerns that g merely reflects knowledge and skills that happen to be valued by the dominant culture, and objections of a more scientific nature. These include theories that focus on specific abilities, such as Howard Gardner's theory of multiple intelligences, which argues for very specific abilities, including non-cognitive abilities such as dance, and Robert Sternberg's "componential" theory of cognitive processing, which tries to identify cognitive processes that underlie cognitive abilities. When these theories are examined empirically, however, g shines through. For example, Sternberg concluded: "We interpret the preponderance of evidence as overwhelmingly supporting the existence of some kind of general factor in human intelligence. Indeed, we are unable to find any convincing evidence at all that militates against this view." Concepts such as working memory are only just beginning to be assessed using traditional measures of cognitive ability, but so far they support that view. Of course g is not the whole story - factors representing specific abilities are also important - but trying to tell the story of cognitive abilities without g loses the plot entirely.

The existence of g appears to go against the tide of current cognitive neuroscience, which views cognitive processes as specific and independent. But research in cognitive neuroscience focuses on average performance - such as what bit of the brain lights up with neuroimaging when a particular task is performed - and g is not about average performance, it is about the individual differences in performance and the fact that individuals who perform well on some tasks tend to perform well on most tasks. In this kind of analysis, the data clearly point to g.

But the fact that g exists does not imply that its source must be a single general physical process, such as the complexity of a neuron's dendrites or the extent of the myelin sheath surrounding neurons' axons. Nor is its source solely physiological, involving synaptic plasticity or speed of nerve conduction. Nor is it a single psychological process such as working memory. Instead, it represents a chain of such physical, physiological and psychological processes, all enlisted together to solve functional problems. As an analogy, athletic ability depends on psychological processes such as motivation, physiological processes such as oxygen transport and physical processes such as bone structure. But athletic ability is not one of these things, it is all of these things.

Genetic research is also important for the story of g because g is substantially heritable. There are more studies addressing the genetics of g than there are for any other human characteristic. Studies including more than 8,000 parent-offspring pairs, 25,000 pairs of siblings, 10,000 twin pairs, and hundreds of adoptive families all come to the conclusion that genetic factors contribute substantially to g. Estimates of heritability vary from 40 to 80 per cent in each study, but estimates based on the entire body of data are about 50 per cent, indicating that genetic variation accounts for about half of the variance in g. Now research has begun to look for the specific genes responsible for this heritability - research that is likely to be accelerated by the publication of the draft of the human genome sequence earlier this year.

But genetic research has gone beyond merely demonstrating that g is substantially heritable. A particularly important finding from multivariate genetic research, which analyses genetic and environmental sources of covariance among traits, is that g is where the genetic action is. While g accounts for about 40 per cent of the total variance of cognitive tests, multivariate genetic research indicates that g accounts for nearly all the genetic variance of cognitive tests. That is, what is held in common among cognitive abilities is almost completely genetic in origin. What makes us good at all tests is largely genetic, but what makes us better at some tests than others is largely environmental. This suggests that genetic links among cognitive processes may have been forged by evolution to coordinate effective problem-solving across the modules of mind.

Genetic research also shows a strong genetic overlap between school achievement, as assessed by tests or grades, and g. In other words, the same genetic factors that contribute to individual differences in g are responsible for many individual differences in school achievement. Conversely, what is different between achievement and ability or g is largely environmental. These findings suggest that assessing achievement "corrected" for g could bypass genetic influence, with far-reaching implications for educational selection, assessment and issues of added value in schools.

Another surprising finding is that the strength of genetic influence on g increases from early childhood, through middle childhood, to adolescence. This is counterintuitive to the notion that environmental influences accumulate as life goes by. It suggests to me that children select, modify and even create environments conducive to the development of their genetic proclivities. For this reason, I think about g as an appetite rather than an aptitude. At the simplest level, this means that children like to learn what they find easy to learn. But I think that there is more to it than that: the mechanisms by which genes affect the learning process may have as much to do with motivation as with the brain's hardwiring.

Robert Plomin is MRC research professor, Social, Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, King's College London.

<P align=left> Big Science Questions contents page </a>