Intelligence_A Very Short Introduction Read online

  Between them, Rosalind Arden and Shelley Cox flattered me into thinking I could write something accessible about human intelligence differences. Linda Gottfredson, Shelley Cox, Tracy Miller, and Alan Bedford made good suggestions on earlier drafts. I thank those whose datasets are the fabric of this short treatment. An author must have an audience in his mind’s eye. Mine was focused on my intelligent and incredulous mother, Isobelle.

  A word about correlation

  This series of Very Short Introduction books, and this particular book on human intelligence, is intended for the general, interested reader. The material aims to be accessible but still intellectually pithy. I have tried to avoid patronizing generalizations in favour of demonstrating what a real research project in this field looks like, and what it can and cannot tell us. My reason for taking this course was that, among popular accounts of intelligence research, one can find diametrically opposed views about the same sets of data. Therefore, I wanted the reader to think about actual findings, not the Chinese whispers issuing from several-times-digested summaries of the research.

  The approach adopted here erects one hurdle that I have to clear. The use of statistics is central to research on intelligence. Researchers typically test large numbers of people on a variety of mental tests, and discovering the pattern and significance of the differences between people cannot be done without statistical examination of the data. Some of the key debates in human intelligence are about statistical matters. Further, the statistics we employ in intelligence research are among the more complicated in the discipline of psychology. Now, there was no point in trying to fashion a general book that was replete with statistics: no one would read it. In the end I decided that there was no escaping one type of statistic: correlation. This is easy to understand. If you know what correlation is, just skip the rest of this section and move on to Chapter 1. If you don’t, read the following non-technical explanation.

  Correlation is a way of describing how closely two things relate to each other. It is expressed as a number called a correlation coefficient. The range of values that a correlation coefficient can take is from –1 through 0 to 1.

  Take an example. Say that I stop the first 100 adult women I meet in the street and measure their heights and weights. I am curious to know, let’s suppose, whether being taller also means being heavier. A correlation coefficient can be calculated according to a formula and it will tell me how strongly the two are related. Imagine that everyone who was taller than someone else was also heavier than them. There would be a perfect association between the two: the correlation would be 1. That’s not going to happen. The situation in real life is that we all know some short fat people and some tall thin people. On the whole the taller people are heavier, but there are many exceptions. Therefore, there is a strong trend toward taller people weighing more, but it is not perfect. The correlation is probably around 0.5, a highish positive correlation.

  Extend that example. Say I also decided to measure the length of their hair. I am curious to know whether the taller people grow their hair longer. I am almost certain that there is no tendency whatever for tall people to have their hair either longer or shorter than smaller people. My guess is that height would have absolutely no association with hair length at all. If I am correct, the correlation coefficient would be 0. The two things have no tendency to go together.

  One more extension to the example. Let’s say that in addition to measuring people’s heights we ask them to walk a measured distance, say 20 metres. We count the number of steps it takes them. I am curious to know whether there is any association between height and the number of steps it takes to cross this distance. My guess is that taller people would on the whole take fewer strides. The correlation coefficient would probably confirm this; but note that it would find that being taller would go along with a smaller number of steps. So the correlation would be negative; as one value (height) goes up, the other one (steps taken to cover 20 metres) goes down. It might be about –0.4. However, the value is not the important thing here. The point I want to get across is that important, strong correlations can have negative or positive values. It’s when the value of the correlation is zero that there is no relationship between two things.

  A correlation can describe for us whether one thing tends to go up or down with another thing, or whether there is no relation at all between the two.

  Next, a word about the sizes of correlations. I mentioned above that height and weight probably had a fairly high correlation, about 0.5 or thereabouts, or maybe more. (In fact, I got the 0.5 value by calculating it from heights and weights of a number of people’s data that I happened to have on my computer.) In psychology and other sciences that look at social phenomena, we do not often find correlations beyond about the 0.5 level. There is a convention that correlation coefficients above about 0.5 are called large or strong effects. Those between about 0.2 and 0.5 are called medium, modest, or moderate. Those below 0.2 are called small or weak.

  Last, a word about the nomenclature I shall use. For variety I shall not always refer to correlations between two things. Sometimes I shall say the ‘relation’ or the ‘relationship’ and at other times I shall say the ‘association’. When I use these words I am referring to a correlation. And if I qualify any of these terms with the adjectives large, medium, or small, these will refer to the sizes of coefficient mentioned in the previous paragraph.

  In much of the material that follows, we shall use the correlation coefficient to describe how strongly intelligence test scores relate to other things. Sometimes I shall be looking at whether one type of intelligence test has a high correlation with another type. Sometimes I shall be asking whether intelligence test scores correlate with anything about our achievements in real life. And sometimes I shall be asking whether anything about our brains and brain functions correlates with intelligence test scores.

  It is important to emphasize that correlations describe the relation between two things that we have measured in a group of people. Indeed, the larger that group is the more confident we can be that the correlation value is the correct one. So, the value applies across a group that we have measured. But people make the common error of applying the correlation to themselves personally. Let’s say we announce that we have tested heights and weights of people and we say that there is a strong correlation, such that taller people tend to be heavier. A short, portly person might well look at themselves and exclaim that we are talking nonsense, that they are living proof that there’s no such association. We must recall that in any situation where a correlation is not +1 or –1 (i.e. almost all the time), we shall find exceptions to the association that we have found. The lower the correlation, the more exceptions we shall find as we meet up with individual people.

  So, correlations are summaries that tell us about the association between two things in a given sample of people. They don’t tell us about individuals. Moreover, they don’t necessarily tell us that we shall find the same association in other samples of people. If we find a correlation between two things in adult men we cannot assume that we should find the same correlation in children or in women, for example.

  Here is a practical example from the world of intelligence research that makes the point again about groups of people versus individuals.

  There’s a modest correlation between scores on intelligence tests and job status. The UK government has produced a book in which researchers can grade people’s jobs according to a scale. At one end of this scale there are professional jobs, like those of lawyers and doctors, and at the other end there are tasks like manual labouring. As I said, there is a modest correlation between intelligence test scores and job status, perhaps about 0.4 or a bit above that. That tells us something about a group of people: it says that, in general, there will be a tendency for the people with higher intelligence test scores to get more skilled and professional jobs. But, because the correlation is not very strong, it means there will be many exceptions. When we start to lo
ok at individual people, we shall find some lower scorers who ended up in professional jobs and higher scorers who are working with their hands. Therefore, correlations – even quite strong ones – do not tell us about individuals: a correlation is a description of a tendency in a group of people.

  And there’s another lesson from this. Take the correlation between intelligence and job status. Because that correlation is not especially high, it means that there’s a lot more to getting a good job and a high salary than high intelligence. This is what we shall see all along the line: intelligence might have some influence on things, but there is always a lot more to any human story than just intelligence.

  To follow this area up …

  There are good descriptions of correlation, in settings related to intelligence, in the following books.

  Cooper, C. (1999). Individual Differences. London: Arnold.

  Herrnstein, R. J. & C. Murray (1994). The Bell Curve. New York: Free Press.

  The following is a good and accessible introduction if you want to read more about the conceptual and statistical issues related to measuring intelligence (‘psychometrics’) and other aspects of the human psyche.

  Kline, P. (2000, 2nd edn). Handbook of Psychological Testing. London: Routledge.

  Chapter 1

  To see ‘g’ or not to see ‘g’ …

  How many types of intelligence are there?

  The first question I want to address is simple. Should we talk about human intelligence – human mental abilities – as one thing or as many things: intelligence or intelligences? This question of how to conceive of human mental capacities is a vexed one. Psychologists argued about it for most of the 20th century, and the debate continues. From the non-specialist’s viewpoint, they appear to do little more than coat opinion with statistical opacity. The nub of the issue is that discussions about human mental ability are a commonplace. Yet in the frequent appellations of people’s being ‘clever’, ‘smart’, ‘intelligent’, ‘bright’, and ‘sharp’, there often exists a tension. On the one hand, we are sometimes referring to people as being generally mentally able or less so: ‘What a bright guy!’ Contrariwise, we sometimes pick out a special mental ability that a person has in some abundance, that appears to contrast with their otherwise modest arrangements: ‘He’s good with figures, but he can never remember where he puts things and he has no common sense.’

  It’s probably better to get in at the start a proclamation of incompleteness. In psychology we tend to measure that which can be measured. Therefore, when we discuss the mental abilities and their relations, it must be kept in mind that, if there are some qualities that we value but we feel cannot easily be measured, then our account of intelligence will be limited. For example, we are relatively poor at measuring things like creativity and wisdom, some of the most valued human attributes. What I want to do now is give an indication of the sorts of things measured in some well-known intelligence tests and ask whether these different skills are related to each other or whether they are largely distinct.

  Key dataset 1

  The first research story here concerns the decision by a large international psychological company to update its most comprehensive intelligence test. The job involved recruiting and testing over 2000 people in 28 American cities. Each person was tested on 13 mental tests over a total time of an hour or two. Using this dataset, the question I want to address here is: do people tend to be good at some tests and poor at others, or are people just generally good or bad at mental tests?

  Before proceeding, let’s be clear about the sorts of mental tasks that people were asked to do in these intelligence tests. Look at Figure 1. The first thing to notice is the 13 rectangular boxes around the bottom of the diagram. Each of these boxes has the name of a different mental test. Together the 13 tests make up a collection of tests called the Wechsler Adult Intelligence Scale, version III. This is usually just shortened to the WAIS-III. It costs many hundreds of pounds to buy and may only be bought by people with the proper credentials, for example, educational, clinical, and occupational psychologists. It can only be administered by a trained psychological tester, working one-to-one with the person being tested for up to a couple of hours. The 13 individual tests involve a wide range of mental effort for the person being tested. It is useful to describe the individual tests and some of the items so that we are not discussing this topic too abstractly. Because the tests are copyright, I describe items like those that appear in the test but not the actual items themselves.

  1. The hierarchy of mental ability test scores from the Wechsler Adult Intelligence Scale III.

  If you were to sit the WAIS-III test, the types of mental task you would be asked to do are as follows:

  Vocabulary. Tell the examiner what certain words mean. For example: chair (easy), hesitant (medium), presumptuous (hard). (33 words) Similarities. Say what two words have in common. For example: In what way are an apple and a pear alike? In what way are a painting and a symphony alike? (19 questions)

  Information. General knowledge questions covering people, places, and events. For example: How many days are in a week? What is the capital of France? Name three oceans. Who wrote Inferno? (28 questions)

  Comprehension. Questions about everyday-life problems, aspects of society, and proverbs. For example: Tell me some reasons why we put food in a refrigerator. Why do people require driving licences? What does it mean to say ‘a bird in the hand is worth two in the bush’? (18 questions)

  Picture completion. Spot the missing element in a series of colour drawings. For example: that spokes are missing from one wheel in a picture of a bicycle; that one buttonhole is missing from a jacket in a picture of a person. As in the earlier tests in the collection, the questions become progressively more difficult. (25 drawings)

  Block design. After looking at two-dimensional patterns made up of red and white squares and triangles, you have to reproduce these patterns using cubes with red and white faces. (14 patterns)

  Picture arrangement. Given a series of cartoon drawings you must put them in an order that tells a logical story. (14 of these series)

  Matrix reasoning. Find the missing element in a pattern that is built up in a logical manner. An example of this type of task is shown in Figure 2. (26 questions)

  Arithmetic. Mental arithmetic problems. (20 questions)

  Digit span. Repeating a sequence of numbers to the examiner. Sequences run from 2 to 9 numbers in length. An easy example is to repeat 3–7–4; harder is 3–9–1–7–4–5–3–9. In the second part of this test the sequences must be repeated in reverse order (maximum of 16 forward and 14 reversed sequences).

  2. An example of a matrix reasoning item. This was not taken from the Wechsler Adult Intelligence Scale-III because their test materials are protected by copyright. It is an item developed for but not used in the revision to the famous Raven’s Progressive Matrices test. I thank John Raven (son of the inventor of the original test) for allowing me to use this item.

  Letter–number sequencing. The examiner reads a series of alternate letters and numbers. You must repeat them, putting the numbers first and in numerical order, followed by the letters in alphabetical order. For example, you would repeat ‘W-4-G-8-L-3’ as ‘3–4–8-G-L-W’ (maximum of 21 trials).

  Digit-symbol coding. You write down the number that corresponds to a given symbol. An example of this type of task is shown in Figure 3 (as many as you can in 90 seconds).

  Symbol search. You identify from a list of abstract symbols which symbol in a given pair is contained in the list (as many as you can in 2 minutes).

  3. Part of a test that is quite like the digit-symbol coding test of the Wechsler Adult Intelligence Scale-III. The idea is to enter the code that corresponds to each number in the empty space provided. The score is the number completed in 90 seconds. In the real test there would be far more items available for completion.

  Some of these tests involve knowledge picked up from education, and some don’t. Some i
nvolve language, some numbers, some shapes, and some are more abstract. Some are done at speed, within time limits, and some not. Some involve memory and some don’t. Some involve reasoning with information given by the tester; some involve discovering rules; some involve articulating abstract principles; some involve practical knowledge. The tests are tapping quite a wide range of our mental functions: seeing similarities and differences, drawing inferences, working out and applying rules, remembering and manipulating mental material, working out how to construct shapes, processing information at speed, articulating the meaning of words, recalling general knowledge, explaining practical actions in everyday life, working with numbers, attending to details, and so forth. They are reasonably representative of the spread of contents scoured by IQ-type tests. Arguably, certain sorts of mental functions do seem to be poorly represented here, or not represented at all, but it is true to insist that a reasonably wide range of thinking skills gets a look-in. And, for those who wish to write these tests off as mere ‘paper-and-pencil’ tests, only 3 of the 13 tests require the examinee to write anything down, and none requires writing words.

  The WAIS-III is developed and marketed by the Psychological Corporation in the USA and the UK. This large company develops and markets a wide range of psychological tests around the world. When they were gathering information about the WAIS-III in the USA, they tested 2450 people. These people were a fair sample of ordinary American citizens: there were equal numbers of men and women, there was a representative sample from age 16 to age 89, the ethnic and regional mix was like that of America as a whole, and there was a good spread of educational level among the people tested. Every person sat the 13 tests described above. The results of this big testing exercise saw a repeat of one of psychology’s most surprising and most reproduced findings.