At Our Wits' End, page 3
For example, somebody might be very verbally skilled and use lots of big words—meaning they will superficially appear to be highly intelligent—while, overall, this is not the case. You certainly wouldn’t want them trying to fix your computer or diagnosing some potential illness. This is seemingly true of many people who study humanities subjects at university. In general, the student reading a subject like cultural anthropology will have high verbal intelligence, but they will be much less intelligent, overall, than the stereotypical tongue-tied, shy physics student.[11] That is not to say that there are not highly intelligent and inquisitive students who study cultural anthropology, but it is simply a fact that their average IQ is lower than that of those who study physics.11 IQ tests typically measure three main forms of intelligence: verbal, numerical (mathematical), and spatial (geometric). On the WAIS IV test, a typical test of verbal comprehension (known as ‘similarities’), for example, might be: ‘What is the connection between an apple and an orange?’ The correct answer is that they are both kinds of fruit. This would receive 2 points whereas the less nuanced ‘They’re both food’ would receive 1 point.[12] Some people are higher in one manifestation of intelligence than another, and, rarely, they may have above average scores in one measure of intelligence and below average in another. Einstein, for example, had such fantastic mathematical skills that he worked out an original proof of Pythagoras’s theorem at the age of 12. However, his linguistic skills were so poor that he failed the entrance examination for the Federal Institute of Technology in Zurich.[13] But, Einstein is an extreme case; an outlier. Overall, in group studies, the many different measures of cognitive ability always positively correlate with one another. It is consistently found that, within-groups, high ability in one task goes with high ability in other tasks.
This is why the computer comparison works so well. A faster processing computer will be better at pretty much any task you set it. It will always run more efficiently and will be able to easily cope with tasks which would overwhelm slower computers, causing them to simply crash. The positive correlations that exist between many different cognitive ability measures mean that we can talk about a ‘general factor’ that underpins performance in all of them. This was first described statistically by Charles Spearman (1863–1945). Spearman was a pioneering English psychologist and a very unusual man. He joined the army, became an officer, and then suddenly left, in 1897, intent on pursuing an academic career. While studying for his degree at University College London, he published a seminal paper, in 1904, in which he showed that the ability of school children in tests on different subjects inter-correlated.[14] This can be seen in Table 2.[15]
Table 2. Spearman’s (1904) matrix of correlations showing that all of the scholastic aptitudes correlate positively with one another—so high-level performance in maths goes hand-in-hand with high-level performance in classics, French, English, etc.
Classics
French
English
Maths
Pitch
Music
Classics
1
French
0.83
1
English
0.78
0.67
1
Maths
0.7
0.67
0.64
1
Pitch
0.66
0.65
0.54
0.45
1
Music
0.63
0.57
0.51
0.51
0.4
1
With this, Spearman argued that there was a g-factor—a general factor —which underpinned the relationship between how well people did in all these diverse subjects. But let’s pause to examine Spearman’s ‘correlation matrix’ in more detail. It can be seen that ability in French —remember this is with a British sample—very strongly correlates with ability in classics, which is no surprise because they both involve the learning of foreign languages, and French, Latin and, to a lesser extent, Ancient Greek are closely related. The correlation between classics and English is a bit weaker but even with maths the relationship is strong, at 0.7. We can see that the ability to learn a foreign language is more strongly related to maths than it is to English. And pitch discrimination also correlates with these academic subjects. This is consistent with an idea first postulated by Sir Francis Galton (1822–1911)[16]—who we will also meet later. Galton proposed that intelligence is driven in part by the ability to notice subtle differences among sensory inputs (pitch, colour, etc.)—so intelligent brains have more ‘bandwidth’ as they can take in more information, which can in turn be used for solving problems more effectively. For example, if the problem was building a waterproof roof, an intelligent individual might notice that a particular material behaved very slightly differently from another and so choose the superior material. This ability to deal with subtlety would mean that they had solved the problem better.
IQ is a statistical construct which measures Spearman’s g factor. So, the IQ test is measuring g. However, the IQ test is not a perfect measurer of g, just as many a school maths test will also, to some extent, measure vocabulary ability along with mathematical ability, meaning it is not a perfect measurer of mathematical ability. The IQ test —in testing, let’s say, the ability to place an animal in the correct category—is measuring g, because g predicts the ability to do this. But there are other mental abilities that also predict the ability to do this. So, the test also tests these measures, measures which influence performance in very specific cognitive domains. Spearman termed these s for specialised abilities—there being many different s’s that are unique and that only influence maths ability or language-learning ability and so forth. So, the IQ test measures a series of cognitive abilities. These abilities are predicted by g but also by specialised abilities which are independent of g.
Things have moved on from Spearman’s somewhat crude two-factor or g and s model of intelligence. Today, researchers typically conceive of intelligence as a pyramid. At the pinnacle is Spearman’s g factor. Beneath this are what are called group factors or sometimes primary mental abilities. These would include determinants of performance that may be shared among groupings of cognitive abilities, but which are nonetheless independent of the g factor. Examples of these ‘mini-g’s’ would include the Verbal, Perceptual, and Rotational primary ability groupings.[17] These ability groupings strongly correlate with each other, but performance in these domains is also in part a reflection of other things; specific narrow abilities. Beneath these are even more specific narrow skills—i.e. s’s that predict performance within the sub-domains of the primary ability clusters. These, likewise, correlate with g but are also substantially measuring other highly specialised cognitive processes. This is why somebody might not have a particularly high score in any of three main ability clusters but, nevertheless, be brilliant at darts or realistic drawing. They have a very specific cognitive skill. Finally, at the bottom of the pyramid are the numerous test-specificities —extremely narrow skills that influence performance on specific ability tests, but do not help in others—being really proficient at pattern recognition may help on tests that rely on this for problem solving, but won’t help much on tests that don’t (like a vocabulary test).
It is important to note that not all indicators of intelligence are equally g-loaded—that is, the degree to which they measure the g factor varies.[18] To give an example, how well you do at school in English or in maths are both strongly predicted by how intelligent you are because they both measure the g factor. However, maths is more closely related to g than English. It is a better proxy for g. English, in turn, is likely to be a better proxy for general intelligence than art. But if you were a university lecturer faced with a huge pile of university applications and you wanted the most intelligent students you could save a lot of time by simply turning to how well the applicant scored in GCSE maths,[19] even if they were applying to read modern history. This would, of course, be extreme and simplistic. But it seems fairly likely that most admissions tutors will take GCSE maths into account even if their subject is modern history, as the essence of it is the ability to think logically.
Intelligence and IQ Tests
IQ compares your intelligence with those of your own age. Intelligence increases throughout childhood. Problems which an average adult could solve would be beyond even the brightest three-year-old. But we would not say that the three-year-old who was already beginning to learn to read was not very intelligent. In comparison to adults, all three-year-olds are not especially intelligent. But their IQ is a different matter. Intelligence increases up to middle age and then decreases from middle age onwards.[20] As such, IQ is a comparative measure—comparing the individual with a group sample of the same age.
The IQ number is a way of expressing the individual’s position in a rank ordering of IQ test scores for their age group; hence the term ‘intelligence quotient’ (IQ). The average IQ is set at 100. Larger numbers represent above average IQ and lower numbers represent below average. IQ is ‘normally distributed’ in a so-called bell curve and, in that respect, it is rather like height. IQ distribution, on the bell curve, is conventionally divided into ‘standard deviations’ of 15 IQ points. Most people have an IQ of somewhere around the average of 100, just as most people’s height is clustered around the average with smaller and smaller percentages of people being either very short or very tall. Indeed, 68% of the population have an IQ that is between 85 and 115. Only 14% of the population have an IQ that is either between 115 and 130 or between 70 and 85. And—you can see where we’re going with this!—as a consequence 95% of the population have an IQ that is between 70 and 130. This can be considered the ‘normal’ IQ range because those who are below 70 would be classified as having learning difficulties and those above 130 would be exceptionally bright. So, 2% of the population have an IQ either between 55 and 70 or between 130 and 145. These are the people of either very low or very high intelligence respectively. Just 0.1% of the population, at either end of the bell curve, are higher or lower still. This can be seen in Figure 1.
Figure 1. The IQ bell curve.[21]
We measure weight with a weighing-scale and height with a tape measure. The IQ test is simply the instrument through which we measure IQ. No instrument is perfect. Some weighing-scales will give a much more accurate and subtle reading than others and it’s the same with IQ tests.[22] The test of the accuracy of an instrument is if its results strongly correlate with other instruments that are measuring the same thing or measuring something very similar. So, one measure of weight might involve getting people to physically lift different people and estimate which was the heaviest. If this correlated with what the weighing-scale indicated, we could have faith in the scales. The results of IQ tests strongly correlate with intuitive measures of thinking ability (such as school exams) and they are not merely culturally influenced. We know that IQ testing is valid and robust, because culture-fair IQ tests have similar predictive power across cultures.[23] This is exactly the opposite of what we would predict if the tests were poor-quality instruments that were highly subject to cultural bias.
Also, IQ test results correlate positively with something objective—that is, with differences in reaction times.[24] It is widely accepted among leading psychometricians such as Arthur Jensen,[25] Hans Eysenck,[26] and Ian Deary[27] that IQ tests correlate with this objective neurological measure. There are many different tests of reaction times. An American psychologist called Arthur Jensen (1923–2012) developed one of the most commonly used. Jensen, who was Professor of Psychology at the University of California at Berkeley, spent his entire academic career studying intelligence, though he was a social worker before that. Jensen pioneered the ‘odd man out test’, in which people were seated in front of a bank of lights. They observed which light went on, lifted their finger from a ‘home’ button and pressed the button closest to the illuminated light. Studies have found correlations of between 0.3 and 0.4 between IQ and how quickly they did this. On average, the quicker your reaction times are, the cleverer you are. In other words, IQ is, in part, a measure of processing speed—once again, the computer analogy shows its relevance. This correlation with reaction times means that a significant part of being intelligent is simply having a high functioning nervous system. So, there is every reason to be confident in the validity of IQ tests.
A number of other criticisms have been levelled against the objectivity of IQ tests. The most fashionable is so-called ‘stereotype threat’. This is the idea that certain groups are stereotyped to do badly on IQ tests. As such, when they take the test they become worried about doing badly and therefore they do badly—thus conforming to the stereotype, presumably because they’re stressed. However, studies using large samples have shown no evidence for this effect whatsoever. In some cases, the opposite effect has been observed—those who are told they will do badly in something become more motivated to do well and so they do better than they otherwise would. In addition, there is clear evidence of publication bias around stereotype threat. When a study proves it, then it’s published. If it disproves it, it gathers dust as an unpublished dissertation or sitting in some academic’s file drawer.[28]
Another criticism of IQ tests is that motivation plays a part in test scores. This is obviously true. If we give you some complex sum and tell you to solve it, most of you aren’t going to bother unless we can persuade you that doing so is vitally important to understanding the rest of the book (don’t worry—we’re not that cruel!). Motivation plays a role in IQ tests and so do a host of other things, like suffering from flu on the day of the test. But if nobody gets a prize for doing well in the test then people will only be equally unmotivated to try their hardest. The solution to ensuring that these factors don’t skew the results is drawing upon lots of samples and drawing upon large samples and we have these. For example, the US National Longitudinal Study of Youth is a sample of over 12,000 individuals—a sample that is in fact fairly representative of the population of the USA. A ‘longitudinal’ study involves repeated observations of the same set of individuals over time. So, the lives of the participants are tracked over time to see how they unfold.
As already noted, intelligence is a vital predictor of life outcomes, correlating with school results at 0.7, university performance at 0.5, and postgraduate performance at 0.4. It correlates with salary at 0.3 and is an important predictor of occupational status.[29] It has been found that those in less-selective professions, such as nurses, have an IQ of about 110, while the average is 120 for doctors and lawyers, and higher still for those who rise to the top of these kinds of profession.[30] The average PhD student in an education department has an IQ of around 117, while the average PhD student in a physics department has an IQ of 130.[31] The more intelligent are more likely to engage in civic activities such as voting, and are less likely to endorse extreme political parties or opinions. They are, presumably, less extreme because they are better able to foresee the negative consequences of extreme action and are better at perceiving nuance. They are more likely to engage in civic activity because they are more cooperative and trusting. They can also better understand the positive outcomes of doing so, such as living in a nice environment or being politically free.[32]
The Causes of Intelligence Differences
Why are there differences in intelligence? We know from studies of identical twins—who share roughly 100% of their genes in common—that intelligence is strongly heritable. Heritability is in essence a measure of how closely parents resemble their children in a group study. The number refers to how closely the parents’ IQ predicts the child’s. A heritability of 1 would mean that children’s IQs were wholly a product of their parents’ IQs; zero would mean the statistical relationship between parent and offspring’s IQ was random. In these twin studies, researchers test identical twins, meaning that the only variable causing them to diverge is the environment. Intelligence is seemingly 80% heritable—meaning that 80% of the variation among individuals is due to genetic factors and overwhelmingly, therefore, people resemble their parents in terms of intelligence.[33]
Environmental factors that are relevant include access to good nutrition and a sufficiently cognitively stimulating environment during childhood. But an intellectually stimulating childhood is not enough. Your adult environment also impacts your intelligence. Those with higher intelligence will tend to create a more intellectually stimulating environment for themselves when they grow up, surrounding themselves with other highly intelligent people, for example. For this reason, among others, the heritability of IQ during childhood is not very high, as the child’s environment will be a reflection of factors that are beyond its control. Only as the child becomes an adult will its environment start to reflect its own intelligence. This ultimately leads to a heritability of about 80%.[34] Think of the super-intelligent character Matilda in Roald Dahl’s children’s book of the same name. Her parents have no interest in intellectual matters and there are hardly any books in their house. This kind of environment will reduce Matilda’s IQ because it reflects her parents’ intelligence rather than hers. As she moves in with Miss Honey, and eventually becomes independent of her, she will start to create an environment which reflects her own innate intelligence. This will lead to a substantial increase in her knowledge (but not necessarily her abstract reasoning ability), which will be reflected in an increase in her measured IQ.[35]
