by Ian J. Deary
The statement was expanded upon, with details of its history and a useful bibliography, as an editorial in the journal Intelligence:
Gottfredson, L. S. (1997). Mainstream science on intelligence: an editorial with 52 signatories, history, and bibliography. Intelligence, 24, 13–23.
Another source of guidance for those who want some verbal formulation of human intelligence is the following book.
Snyderman, M. & S. Rothman (1988). The IQ Controversy, the Media and Public Policy. New Brunswick: Transaction Books.
Snyderman and Rothman polled experts concerning their views on human intelligence and its measurement. The majority of the experts agreed – a substantial minority disagreed – that there was a consensus among psychologists and educators as to the kinds of behaviours that are labelled ‘intelligent’. What emerged, though, was near unanimity about the core aspects of intelligence, and a tapering-off in agreement on some facets of human performance that I already noted to be problematic. Here are the aspects of human behaviour they were asked to rate, with the percentage of experts who thought each aspect was an important element of intelligence.
Compared with this near-unanimous agreement on aspects of intelligence, ‘memory’ was endorsed as an important element of intelligence by 80.5%, mental speed by 71.7%, general knowledge by 62.4%, creativity by 59.6%, and achievement motivation by only 18.9%.
Chapter 2
Ageing and intelligence: senility or sagacity?
What happens to mental abilities as we grow older?
Most people of middle and old age are willing to concede that their physical prowess in many areas was not what it was when they were in their 20s and 30s. Similarly, they sometimes complain with a hint of humour that their memory is not what it was. It’s an interesting fact of life, though, that one hardly ever hears people complain about their worsening intelligence as they grow older.
Asking about the ageing of human intelligence means asking at least two different questions. First, how stable are the individual differences among people as they move from childhood through adulthood to old age? The interest here is in whether those at the top and the bottom stay there or whether there is more change, with people who used to perform poorly in early life doing better later on, and vice versa. That is, is the rank order of people’s mental ability in our school classroom still the same when that class meets again at retirement or even later?
Second, is there on average a tendency for people to lose mental power as they grow old? This is a question about what old age does to people on the whole. That is, on average, do the people in our school classroom decline to a degree below their peak mental function in their young adulthood?
Key dataset 3
Have a look at Figure 5. It’s called a ‘scattergram’, a diagram with a scatter of points. Each point on the diagram represents a person. In fact each is a combination of two numbers relating to that person. The horizontal shows the score that a given person obtained the first time they took a certain mental test. The vertical shows the score that the person obtained the second time they took the test. The diagonal line in the diagram is the line along which all the points would rest if mental test scores were perfectly stable over time – that is, if every person got the same score the first and second time they tried the test all the points would fall on that line.
5. A scattergram graph that compares people’s scores on the Moray House Test in 1932 (at age 11) and in 1998 (at age 77). Some crosses represent more than one person.
They don’t. The points deviate from that pattern of perfect stability in two ways. First, notice that most of the points fall above the diagonal line. That means that most people scored better the second time they took the test: the group, on average, improved over time. (That could just mean that they had more practice on the second occasion, and that they recalled some of the questions and were generally better prepared and less anxious about the test. When more details about these data are revealed it will become clear that this is unlikely.) Second, note that there is a general pattern of points flowing from the bottom left of the graph to the top right. There aren’t many points in the top left or bottom right of the graph’s area. But notice, too, that there is some spread, so that the correlation is not perfect. Some people did better than their first score and some did a bit worse, but there is still quite a strong tendency for the people who did well first time round to do better on the second test. We’ll discuss a bit more about what these data mean after describing the way in which the researchers did the testing.
On the morning of 1 June 1932 everyone in Scotland who was born in 1921 and who was at school sat a mental ability test. In a massive national exercise that has never been repeated in any other country in the world, the entire population of 10- to 11-year-olds took the same intelligence test under the same conditions. It was organized by the Scottish Council for Research in Education and it was called the Scottish Mental Survey 1932. The survey data were to assist in educational provision and to measure the amounts of mental handicap in schools. Teachers did the testing and the scoring of the tests. The test was provided by the then-famous educational psychologist Sir Godfrey Thomson from the University of Edinburgh, the originator of the Moray House Tests which were used in the UK as ‘11-plus’ tests for selection into different types of secondary school. The test used in the Scottish Mental Survey 1932 was a variation on one of the Moray House Tests. So, for 45 minutes on a summer’s morning in 1932, exactly 87,498 children applied their brainpower to questions about words, sentences, numbers, shapes, codes, instructions, and other miscellaneous mental tasks.
Until about the 1960s the Scottish Mental Surveys (another one was done in 1947, on the people born in 1936) were famous for their completeness, and several scholarly books were published with the statistical data from them. However, as the 11-year-olds from 1932 reached middle and old age, the data gathered dust in a series of Edinburgh attics and basements. Psychologists interested in the study of human intelligence differences had all but forgotten about the results of the Scottish Mental Surveys.
In the summer of 1996, Lawrence Whalley (from the Department of Mental Health, University of Aberdeen) called me (at the Department of Psychology, University of Edinburgh) to ask whether there might be some mileage in administering some mental tests to the Aberdeen Birth Cohort 1921. This group of people was being studied for signs of cardiovascular disease, and Professor Whalley wanted to know whether heart disease reduced intelligence levels. Not really, was my reply, because, without prior mental test data on these people that was collected before their illness started, information on current mental ability would not tell us about relative changes in cognition as a result of disease. However, coincidentally, at that time I was reading Richard Lynn’s book Dysgenics (1996, published by Praeger), which referred to a study conducted by the Scottish Council for Research in Education (SCRE) on people born in 1921. I called Lawrence Whalley back: the Aberdeen cohort might indeed be worth retesting, because many of them probably did have their mental ability tested at age 11 years in the Scottish Mental Survey 1932. This was the sequence of accidents that led to his tracking down the Survey data to a safe bunker in SCRE’s offices in St John’s Street, Edinburgh. In a series of heavy ledgers and brown paper packages tied up with string, the more than 60-year-old data were preserved, recorded in the neat copperplate script of the 1930s teachers. Each region of Scotland had its own ledgers. Each of the region’s schools had its own pages in the ledger. And each line of each ledger contained a pupil’s name, date of birth, and mental ability test score.
As we literally blew dust from these ledgers, it began to sink in just how valuable were these data. In recent years the Western world’s populations have changed, with a higher proportion of older people. It has emerged that one of the determinants of high quality of life in old age is avoiding cognitive decline. But to find out whether people have or have not retained their intellectual abilities one needs to know what people used
to be like. Though there are some studies that have followed up people as they grow older, none has been able to relate ability in childhood to ability in old age. Before the re-emergence of the data from the Scottish Mental Survey 1932, there were almost no data looking at change in mental ability over the whole human lifespan.
Our first aim was now to discover how stable intelligence was from age 11 to age 77. We set about trying to find some of the still-healthy people who took part in the Scottish Mental Survey 1932. Advertisements were placed in the media and our researchers contacted people via their general medical practitioners. We hired Aberdeen’s Music Hall for the morning of 1 June 1998 and set it out as an examination hall. Crucially, we obtained a copy of the original Moray House Test that was used in 1932 and had it reprinted. Only a couple of tiny changes to two questions were needed to prevent anachronisms in the test. Exactly 66 years to the day after they first sat the test, 73 people came along to re-sit the test that they had last seen as young schoolchildren (Figures 6 and 7). A meeting some weeks later increased our numbers to 101. The instructions were read out exactly as they had been in 1932 and the same time limit was applied.
The results we found are summarized in Figure 5. Most people scored better at age 77 than they had done in 1932 at age 11, but that is not the main point. The important finding is that, largely speaking, the people who did well in 1932 also tended to do well in 1998. Those who did more poorly as children tended to stay near the bottom. Most people fall along a line which indicates general consistency in scores. The
6. Cover of the Moray House Test used in the Scottish Mental Survey 1932.
7. Photograph of members of the Scottish Mental Survey 1932 returning after 66 years to sit the same mental test they had taken at age 11. The venue is the Music Hall in Aberdeen, the date 1 June 1998.
correlation was greater than 0.6, which is large. As an added technicality, the people we got back to take the test in 1998 were not fully representative of the whole population. On average they were better test scorers than the full population and they tended not to have such a wide spread in scores. This narrowing of the range of scores lowers our correlation from the true value, and it seems that 0.7 might be nearer to the mark than 0.6 over 66 years.
Note the points marked A, B, and C in Figure 5. Person A has about an average score at age 11 and again at age 77. B and C are the two people who show the most dramatic inconsistencies between 1932 and 1998. Person B has an average score in 1998, but a very poor score when tested in school 66 years earlier. Person C had an average score in 1932 but was equal-lowest when we tested people in 1998, representing a dramatic drop in relative performance.
What these results mean is that there is a large amount of stability overall, and about an equal amount of change, in our relative levels of mental ability from early adolescence to old age. Looking again at Figure 5 we see that the points are far from describing an exact straight line. That means that there is a considerable amount of change. Some people improve and some decline with age; many change their rank order in the group. The sources of these continuities and changes in this important area of our mental lives are being sought by many current researchers. Sadly, for example, we discovered later that Person C in Figure 5 was in the early stages of Alzheimer’s dementia.
In summary, the answer to the first question about age and intelligence is that there is a substantial stability in the rank-ordering of human intelligence across the human lifespan, and also substantial change. There is substantial change because the correlation coefficient measuring stability is far from perfect.
Key dataset 4
The psychologist K. Werner Schaie wanted to find out whether people’s mental abilities changed with age: not so much whether individual people stayed at the same level in a rank order, but whether the average level of ability improved or declined with age. He used a complicated procedure of testing in order to find out. It requires some digression to explain why less elaborate studies were inadequate. Obviously, the Scottish data I described above are unsuitable: at age 11 years intelligence has still not reached its adult peak and so, at age 77, the people from the Scottish Mental Survey 1932 did better but were not being compared with their best-ever performance.
One way to find out whether people of different ages have different levels of mental ability is to go out and test thousands of people of different ages on the same battery of mental tests. It would be important to make sure that people at each age level were representative of the whole population. It would be inappropriate just to examine brighter older people versus duller younger ones. With this proviso, the great benefit of this type of test is that it can be done quickly, at one point in time. In fact, there are many data of this type. The norms from the well-known test batteries, like the Wechsler test, for example (see Chapter 1), tell a clear story. For some of the tests in the Wechsler battery there is little change with age: older people do just as well as younger people on some tests. On those tests that measure vocabulary, general information, or verbal reasoning, there is little or no age-related decrement in ability. For those tests that are timed or time restricted, more abstract, or require reasoning about spatial relationships, there are marked decrements with age: younger people out-perform older ones. The USA’s Department of Labor tested over 30,000 people in the workplace on the General Aptitude Test Battery, and found much the same – that there was little change between age 20 and 60+ on abilities like vocabulary, and that there was a straight decline from age 20 to age 60 on tests of abstract and spatial reasoning, especially when these had to be done at speed.
This type of study is called a ‘cross-sectional’ study, when different ages are measured at the same time. It faces major problems of interpretation. The people of different ages who are being compared do not share the same educational, nutritional, medical, or cultural histories. Any differences between the ages might well be caused by these factors rather than age per se.
So, some psychologists have taken on the challenge of doing longitudinal studies: that is, they test people when they are young and then again when they are old(er). Two of the most remarkable studies are from North America and are related to the testing that went on during military recruitment for the two world wars. W. A. Owens tested over 100 American men in 1950 and 1961 after finding their ‘Army alpha’ scores from 1919. The Army alpha was the first-ever group test of intelligence devised for adults, and was developed for recruitment of American men into the army in World War I. Owens found that, 30 and 40 years later, the men were just as good at verbal ability and almost as good at numerical ability, but had slipped quite a bit from their young adult scores on abstract reasoning. These are similar to the findings of the cross-sectional studies. Comparable results are found, too, in the Concordia University study which has retested a few hundred men 40 years after their original test on recruitment to the Canadian armed services in World War II. They were just as good at age 65 as they were at 25 on verbal ability, but much poorer on non-verbal ability. Reasoning under speed concerning the logical relations of abstract shapes declines especially clearly as people grow older.
However, these longitudinal studies have problems too. They might be much harder to carry out than cross-sectional studies, not least because you might have to wait several decades to conduct one. Inevitably, not everyone comes back to be tested some years later. Some people die, some become ill, some move away, some cannot be found, and others just refuse to take part. The ones that do come back are not a representative group, and the results obtained from them cannot be generalized to the whole population. Another problem is that any group of people who were born in about the same calendar year go through a sequence of human experiences – medical, cultural, educational, and so forth – that are unique to that cohort of people, again making their results not necessarily generalizable. Whatever decline is found might apply only to that group of people undergoing their particular life histories. Last and greatest, though, of the problems that
beset the great effort that goes into longitudinal studies is that of practice. When people take a test for a second time they might be doing better than we should otherwise expect because they have done it before. This can mask any effects of ageing.
To counter some of the problems of cross-sectional and longitudinal studies, Schaie began a cross-sequential study in Seattle. Figure 8 illustrates the design of this study. His participants were members of a medical insurance scheme. The Figure’s left-hand side has the dates when people were tested, in 1956 and every 7 years thereafter until 1991. There are 6 grey columns in the Figure, each shorter than the last. In the first of these columns in the Figure, the number 500 at the bottom shows that in 1956 Schaie recruited 500 people. These people ranged in age from the late teens to 80s. They were tested on various mental tests to cover key mental abilities at the stratum II level (see Figure 4). This is a standard cross-sectional study: a number of people with different ages are tested on some intelligence tests at one point in time and they are compared to see whether older people differ from younger people in the scores they obtain.
8. A chart that illustrates how K. Werner Schaie set up the Seattle Longitudinal Study to examine the effects of age on intelligence.
Progressing up this first column in Figure 8, we see that Schaie called back these people every 7 years to retake the same tests. Thus, he used his original cross-sectional study of 500 people to conduct a longitudinal study. Note that, as time moves through the 1960s to the early 1990s the numbers fall from several hundred in 1963 to only 71 in 1991. As noted above, some people die, some get ill, some move away, and some just can’t, or don’t want to, come back.