Differences are about the most obvious feature of human intelligence, and have been recognized since antiquity. The study of cognitive differences among humans can be seen as three main topics: the structure and measurement of human intelligence; the predictive power of mental ability tests; and the causes of human intelligence differences. These topics are usefully and accessibly reviewed by Neisser et al. (1996) and Deary (2001).
1. Structure and measurement of human intelligence differences
2. What do intelligence differences predict?
3. The causes of intelligence differences
1. Structure and measurement of human intelligence differences
In 1904 the English ex-army officer Charles Spearman made one of the most important discoveries in the history of psychology. Taking their school examination scores, and testing schoolchildren on various assessments of sensory acuity, he found that all assessments of mental performance showed positive correlations. Those who were good at one test of mental function tended to be good at all the others. Moreover, he thought the correlations showed a particular pattern of positive associations, allowing him to conclude that what was common to the tests was just one underlying trait, which he famously called 'general intelligence,' usually denoted g. His 'two-factor theory' of intelligence stated that, when we perform a mental test, we employ just two abilities: general intelligence and a highly specific ability applicable only to the test being performed. Because g was a part of all mental performances, Spearman suggested that, to measure it, one need not worry what basket of mental tests was used, so long as they were varied, because the general element would be g. This is Spearman's dictum, 'the indifference of the indicator'. Spearman's ideas about the bases of g were articulated in cognitive, sensory, and physiological terms. Thus, respectively, he suggested that differences in g might be accounted for by 'eductive' cognitive skills, fineness of sensory discrimination, and differences in mental energy. All of these ideas find a place in modern research on intelligence differences (Deary 2000).Spearman's breakthrough happened without the aid of a single mental test as we would know it today. The first mental test appeared a year after, when the Frenchman Alfred Binet (1857–1911) devised a series of tests to help in finding those children who would not progress in normal school circumstances. Sans theory, Binet put together a series of high-level mental tasks and found that the average child at any given chronological age could perform certain tasks but not others. That the Binet tests took off around the world was largely due to the evangelizing of the Americans Henry Goddard and Lewis Terman. From these 'hotch pots' (Spearman's term) of mental tasks emerged the idea of a child's mental age: the mental tasks he or she could manage compared with other children — as compared with a chronological age. Divide mental age by chronological age and multiply by 100 and we have the well-known 'intelligence quotient' or IQ. These tests were so successful because they identified children with learning difficulties more accurately and in a much shorter time than did extensive observations and medical examinations. The next step in the massive expansion of mental testing came with the application of mental tests to American recruits in the First World War. By the end of the war millions of Americans had been tested using the first group tests of mental abilities (see Zenderland 1998 for a history of mental testing in this period).
Ever since Spearman's seminal paper in 1904 there are two aspects of human intelligence differences that have caused tension in research. First, from then until now, there have been arguments about the number of concepts one needs to understand human intelligence differences. Second, there has been a tension between those researchers who, like Binet, proceeded, apparently successfully, to measure intelligence differences and those, like Spearman, who tried to understand the bases of intelligence differences. The former question is addressed now, and the latter is addressed in the last section.
How many dimensions of intelligence are there along which humans differ? The standard method of answering this question has been to devise multifarious mental tasks, apply them to a group of people, and employ statistical techniques like factor analysis to find how many underlying traits are needed to account for the associations among the test scores. Over the 20th century the suggestions ranged as follows: one (Spearman), a huge number (Thomson), about seven unrelated intelligences (Thurstone), perhaps 120 distinct abilities (Guilford), seven to nine-and-a-half (Gardner) (Neisser et al. 1996). The answer that most researchers accept today was available in the first half of the 20th century, from the British psychologists Philip E. Vernon and Sir Cyril Burt. Both suggested that human intelligences formed a hierarchy, and that it was not necessary to conceive human intelligences exclusively in terms of general or specific abilities.
The accuracy of the Burt–Vernon view was established when, in 1993, John Carroll published his reanalysis of hundreds of the most impressive and best-known studies of human intelligence conducted throughout the 20th century. He gathered data from the laboratories of many famous researchers, often those with disparate views about the structure of human intelligence, and submitted them to a common mode of factor analysis. He found that almost all studies found a general intelligence factor, often accounting for 40–50 per cent of the variance in test battery scores. Next, he found that there were separable, but correlated, 'group factors' of intelligence. These are mental capabilities less general than g, such as reasoning, verbal ability, memory, mental speed, and so forth. Finally, he found that there were separable, but correlated, highly specific mental abilities. Therefore, a partial consensus exists in the century-long debate about the number and type of human intelligence differences. There is a general intelligence factor; there are correlated but separable abilities that are broad in scope but are more specific than g; and there is a large number of correlated but distinct specific mental abilities.
There are still dissenters to this workable but imperfect descriptive scheme. The best known is Howard Gardner (1999) who has found large audiences, largely outside academic psychology, for his theory of multiple intelligences. His proposal is that the following separate intelligences exist: linguistic, logical–mathematical, musical, spatial, bodily kinaesthetic, interpersonal, intrapersonal, and naturalistic. Gardner is criticized for having done relatively little empirical work to validate these ideas. The first four of his intelligences are well known to be correlated, physical abilities are not usually considered to be a type of intelligence, and the personal skills are usually viewed as aspects of personality.
Returning to the hierarchy of intelligence, it is clear that this is a descriptive scheme for mental test performances. It may be asked whether these test scores can predict anything and whether we understand the bases of intelligence differences in terms of brain function.
2. What do intelligence differences predict?
The results of a brief encounter with an intelligence test have some predictive power in education, in work, and in life more generally. This area is vast, but three examples make these points.Intelligence test scores correlate positively with educational outcomes: people who score higher on mental tests do better in school and college examinations. The typical correlation is in the region of 0.5 (Neisser et al. 1996). A large-scale example of this is the intelligence testing in English schools (Smith, Fernandes, and Strand 2001). In the 1990s, over 20,000 pupils took the Cognitive Abilities Test (CAT) at entry to secondary school. They were from six education authorities and 176 schools, and were representative of pupils nationwide. The CAT has verbal, quantitative, and non-verbal sections. The mean score on the CAT at entry to secondary school correlated highly with GCSE scores five years later, as follows (with correlations in parentheses): total GCSE performance score (0.74), art and design (0.45), business studies (0.56), creative arts (0.53), design and technology (0.51), English language (0.68), English literature (0.63), French (0.66), geography (0.68), German (0.59), history (0.65), information technology (0.47), mathematics (0.76), physical education (0.54), science (0.72), Spanish (0.61). The correlations with the non-verbal section for the CAT test, which involves neither verbal nor numerical skills, were still high, typically being about 0.1 lower for each subject. Therefore, in this massive, representative, and longitudinal study it is clear that performance on a relative brief test of intelligence has impressive associations with school performance several years later. This applies to everything from art to physical education, in addition to the more traditional academic subjects.
In the world of work it is important for an employer to make some assessment of how effectively a new worker will perform. Over the past 80 years thousands of studies of this type of question have involved various assessments of potential new recruits: interviews, references, personality tests, graphology, intelligence tests, and so forth. The field's thousands of studies were reviewed by Schmidt and Hunter (1998). They found that a psychometric test of general intelligence had an average correlation of 0.51 with work performance. Work performance is assessed in various ways, but one common method is ratings by supervisors. The only assessment that did better was a 'work sample' test, where people carry out some of the work they will be doing in the job. This type of assessment is expensive and not suitable to many situations, whereas a mental test is cheap and can be given to all applicants. By comparison with mental tests, the following correlated more poorly with job performance: unstructured interview (0.38), conscientiousness test (0.31), reference checks (0.26), job experience (0.18), years of education (0.10), and graphology (0.02). Thus a cheap and fast mental test is almost universally applicable in recruitment for employment and has surprisingly high utility.
In medical settings, tests of mental abilities are used to assess people's cognitive functions in response to illness and to treatments. By comparison with the fields of education and work their application here attracts little controversy: they are merely useful tools for assessing outcomes. Here, though, two unusual medical applications of intelligence tests are mentioned. First, cognitive impairment with ageing, especially dementia, is a common clinical problem. It is important to assess a potential sufferer's mental capabilities. However, these are interpretable only against an assessment of the person's previous mental function, something that is rarely available. A British invention, the National Adult Reading Test (NART), provides an assessment of 'premorbid IQ'. The test involves reading a series of English words that do not follow normal pronunciation rules. The ability to pronounce these words correlates very highly with long tests of intelligence in healthy people and, more importantly, the ability is preserved while others fade in early dementia. Therefore, the NART provides an invaluable archaeological record of the person's prior ability even in the face of cognitive decline associated with dementia. Another surprising finding using IQ tests is that people with higher IQ scores at age 11 are significantly more likely to be alive at age 77 (Whalley and Deary 2001).
3. The causes of intelligence differences
Intelligence test scores from tests assessing disparate-seeming abilities co-vary and form a hierarchy, and the scores are useful in various fields of human endeavour and problems. However, there is still the question of the origins of intelligence. Family, adoption, and twin studies demonstrate beyond doubt that intelligence differences are substantially heritable. Across all studies, the contribution of genes to human cognitive ability differences in the population is about 40–50 per cent, though the actual figure is not considered important (Plomin et al. 2001). The small print of this finding is of more interest. Heritability probably changes with age, from as low as 20 per cent in childhood to over 60 per cent and even as high as 80 per cent in very old age. Given that the remainder of the 100 per cent must come from environmental influences and error, the main source of environmental influence is a person's individual experience, not the environment shared with the rest of the family. Most of the heritability of specific mental tests (such as verbal ability, spatial ability, and memory, say) arises from the genetic contribution to general intelligence. Though it is still in its infancy, the main enterprise of the coming years will be the search for the individual genes that contribute to intelligence differences, using molecular genetic techniques.Other research on the origins of intelligence differences seeks a more mechanistic account (Deary 2000 reviewed this research). Thus, echoing ideas that came from the Victorian polymath Sir Francis Galton and from Spearman, researchers have found associations between intelligence test scores and several types of assessment. It is now well established that, in healthy adults, the correlation between intelligence and brain size is about 0.4. Intelligence test scores correlate significantly with speed of reaction and with the efficiency of processing visual and auditory information. And people with better scores on mental tests are notable for having more efficient working memory. Though there are many such associations, the main limitation of this type of research is that the fields of cognitive science and neuroscience have not provided validated models of mental function that afford an understanding of the brain-processing measures that are used to correlate with test scores.
(Published 2004)
— Ian Deary
- Bibliography
- Carroll, J. B. (1993). Human Cognitive Abilities: A Survey of Factor Analytic Studies.
- Deary, I. J. (2000). Looking down on Human Intelligence: From Psychometrics to the Brain.
- — — (2001). Intelligence: A Very Short Introduction.
- Gardner, H. (1999). Intelligence Reframed: Multiple Intelligences for the 21st Century.
- Neisser, U., Boodoo, G., Bouchard, T. J., et al. (1996). 'Intelligence: knowns and unknowns'. American Psychologist, 51.
- Novartis Foundation (2000). The Nature of Intelligence: Novartis Foundation Symposium 233.
- Plomin, R., DeFries, J. C., McClearn, G. E., and McGuffin, P. (2001). Behavioral Genetics (4th edn.).
- Schmidt, F. L., and Hunter, J. E. (1998). 'The validity and utility of selection methods in personnel psychology: practical and theoretical implications of 85 years of research findings'. Psychological Bulletin, 124.
- Smith, P., Fernandes, C., and Strand, S. (2001). Cognitive Abilities Test 3: Technical Manual.
- Whalley, L. J., and Deary, I. J. (2001). 'Longitudinal cohort study of childhood IQ and survival up to age 76'. British Medical Journal, 322.
- Zenderland, L. (1998). Measuring Minds.
