When Alfred Binet was a schoolboy sitting his final school exams, an examiner annoyed by one of his answers told him he would never have a mind for philosophy. Binet still remembered it decades later as a grown psychologist. “Never! What a big word!” he wrote in 1909. He had heard teachers say such things to children far too often, and he thought it did harm.
So it is fitting that the first intelligence test was made to help children, not to rank them. By 1904 every child in France had to go to school, and teachers were finding some who couldn’t keep up. That year Binet and a young doctor, Théodore Simon, got the government to set up a commission on how to recognize those children and teach them. They needed a way to tell a child who was slow at school from one who was simply lazy or badly taught, and nobody had one. Within a year Binet and Simon had made it.
What they made, and what was later made of it, is one of the most useful and most abused tools psychology has. This chapter is about both.
What an IQ score is
Binet and Simon’s first scale, of 1905, was thirty small tasks, each a little harder than the one before. Near the end came one that marked the line between children who could manage ordinary school and those who would need help: put three words, Paris, fortune and stream, into one sentence. In 1908 they sorted the tasks by the age at which most children could do them. A child who passed the tasks most seven-year-olds pass, and no more, had a “mental age” of seven, whatever his real age was.
A German psychologist then suggested dividing the mental age by the real age. A ten-year-old who does as well as the average ten-year-old gets 1. A ten-year-old with a mental age of eight gets 0.8. In 1916 Lewis Terman, a psychologist at Stanford University in California, adapted Binet’s tasks for American children, tried them on about a thousand of them, and wrote the ratio as a percentage: 100 for exactly average. He called it the intelligence quotient, IQ, and his Stanford–Binet test made the word famous.
Dividing by age works for children, but not for adults. Nobody’s mental age keeps rising by a year every year, so a forty-year-old would look half as bright as a twenty-year-old. Modern tests work differently. A test is first given to a large group of people of each age, chosen to stand for the whole population, and a person’s score then says where they stand among people of their own age. The scores are set so that the average is 100 and they spread out in a fixed way.
So an IQ is not a quantity like your height in centimeters. It is a place in a crowd, and only the crowd it was measured against. A score of 115 means you did better than about 84 people in 100 of your age in the group the test was made for.
The best-known tests today are the Wechsler scales, first published in 1939, one for adults and another for children. Like the Stanford–Binet, they mix many kinds of question, words and pictures both: what a word means, which picture comes next in a series, which of several words doesn’t belong, and the score is built up from all of them. Another well-known test, Raven’s Progressive Matrices, made in Britain, uses no words and no clock: each puzzle is a grid of patterns with one missing, and you choose the piece that fits. Many psychologists regard it as a good measure of general intelligence.
Whatever the test, it is only useful if it passes three checks. It has to be standardized, given and scored the same way every time, with the same words and the same time allowed, or two scores can’t be compared. It has to be reliable: take it twice, a few weeks apart, and you should get nearly the same score. And it has to be valid, which means it measures what it claims to measure. For an intelligence test, validity is the hard part, because people don’t agree on what intelligence is.
What the tests measure
In 1904, the same year as Binet’s commission, the English psychologist Charles Spearman tested the children of a village school and of a fee-paying school in England. He compared their marks in Classics, French, English and mathematics, and even how finely they could hear the difference between two notes. Everything went together. A child who did well in one thing tended to do well in the others: not always, but much more often than chance. Spearman concluded that all kinds of mental work share one common ability, as well as skills of their own, and he called the shared part g, for general intelligence.
That pattern has turned up in test after test ever since, and it is the main reason IQ tests have one total score. What g actually is, nobody can say. It may be one ability, or many abilities that happen to grow together. Other psychologists split it up. One distinction that has lasted is between fluid intelligence, working out a problem you’ve never met, and crystallized intelligence, what you have learned and can call on: words, facts, ways of doing things. Fluid intelligence tends to slip as people get older, while crystallized intelligence holds up and can keep growing. The most detailed theory today has three levels: g at the top, a handful of broad abilities such as reasoning, memory and speed under it, and many narrow skills under those.
A task force of the American Psychological Association, which in 1995 set out what was really known about intelligence, didn’t try to define it. It said only that people differ in how well they understand complex ideas, adapt to the world around them, learn from experience, reason in different ways and get past obstacles by thinking. Intelligence tests, it said, are attempts to measure those differences.
And they measure something that matters. The task force found that IQ scores correlate with school marks at about 0.5. That is a strong link for psychology, and it still leaves three quarters of the differences in marks to other things, such as interest, effort, teaching and home. Scores go with how many years people stay in school, and, more loosely, with how well they do at work. But the same report listed what the tests leave out: creativity, wisdom, practical sense and understanding other people. It mentioned children who sold goods on the streets in Brazil and could do the arithmetic their business needed, though they had failed mathematics at school.
One use of the tests is the one Binet meant: finding children who need help. A score far below average, around 70 or lower, is one part of how an intellectual disability is recognized, but only one part. The other is real difficulty with everyday life, starting in childhood, and a doctor or psychologist needs both before using the word. A score far above average is how some schools pick children for extra challenge.
All of this belongs to a larger branch of psychology, the study of individual differences: how people differ from one another, in ability, personality and interests, why they differ, through the genes they inherit and the lives they lead together, and how to measure those differences fairly. Its strengths are that it asks questions whose answers are numbers anyone can check, and that a good test can find a child who needs help before years of failure show it. Its weaknesses are that a test measures only what it asks, that a test made in one culture can mislead in another, and that in the hands of people who forget this, a number can stick to a child like a label. Binet saw that danger early. A historian of his work records him warning that his scale was not like the weighing machine at a station that prints your weight on a ticket, and that it must not be allowed to fall into the hands of a fool.
A dark chapter
Others had very different uses in mind, and they came almost at once.
The English scientist Francis Galton had invented the word eugenics in 1883, for the idea of “improving” people by controlling who has children. Intelligence tests seemed to offer the idea a number to work with. Terman, who made IQ famous, wrote in 1916 that tests would soon bring “tens of thousands” of people under the state’s watch, and that this would end in “curtailing the reproduction of feeble-mindedness”. In the same book he predicted that tests would find large differences in intelligence between races, differences that no schooling could remove, and he wrote that children from those groups, “from a eugenic point of view”, were “a grave problem because of their unusually prolific breeding”.
When the United States entered the First World War, American psychologists offered to test the army’s recruits. By January 1919 they had tested 1,726,966 men, those who could read with one test, the Army Alpha, and those who couldn’t with a test of pictures, the Army Beta. Its makers believed the Alpha measured inborn ability at least as much as schooling. One of its parts asked men to finish sentences such as “The Wyandotte is a kind of” horse, fowl, cattle or granite (it is a breed of chicken), and “Crisco is a” patent medicine, disinfectant, toothpaste or food product (a brand of cooking fat). Those questions measured what a man had come across in American shops and magazines, not how well he could think. In the years that followed, test scores were used to keep people out of jobs, ranks in the army and colleges, and to argue for keeping immigrants out of the United States.
Eugenic laws were being written at the same time, in the same language of “feeble-mindedness” the testers used. Indiana had passed the first law allowing people to be sterilized, made unable to have children, on eugenic grounds in 1907. In 1927 the US Supreme Court was asked whether Virginia could sterilize Carrie Buck, a young woman held in a state home for the “feeble-minded”. It said yes. “Three generations of imbeciles are enough,” wrote the judge who gave the court’s decision, meaning Carrie, her mother and her baby daughter. Later research found that Carrie had become pregnant because a relative of her foster parents raped her. Her daughter, Vivian, went to school and was a solid B student who had been on the honor roll. More than 60,000 Americans were sterilized under such laws, in 33 states. Nazi Germany borrowed from an American model law for its own in 1933, under which more than 350,000 people were sterilized.
Binet died in 1911, before most of this happened. But he had already answered it. Some philosophers, he said, had lately given their support to teachers’ gloomy verdicts by claiming that a person’s intelligence is a fixed quantity that can’t be increased. “We must protest and react against this brutal pessimism,” he wrote. He set out to show that it was founded on nothing.
What changes a score
Binet turned out to be right. Scores do change, and three findings show it.
The first is the strangest, and it carries the name of the researcher James Flynn, whose work in the 1980s made it widely known: each generation does better on the same tests than the one before, so test makers have to keep resetting the average. In 1987 he gathered results from 14 countries and found gains of 5 to 25 points in a single generation. A later pooling of results from almost four million people in 31 countries, tested between 1909 and 2013, found scores rising by about three IQ points every ten years. Your great-grandparents, scored against today’s average, would come out well below 100, though they were plainly not a generation of fools. The rise was too fast for genes to explain, and the usual suspects are better food, better health, more schooling and a world full of the kind of abstract thinking the tests ask for. It has slowed in recent decades, and in a few countries, such as Norway, it has turned down. A study there of brothers born years apart found the fall within families too, which points to something in people’s lives, not their genes.
The second is school. In 2018 two psychologists gathered every kind of study that could tell whether school makes people better at tests, or whether brighter people simply stay in school longer. Some compared people of the same IQ at an earlier age; some used countries that raised the school leaving age; some compared children born just before and just after the cut-off date for starting school. Across 42 sets of data and more than 600,000 people, each extra year of education raised IQ by about 1 to 5 points, and the gain lasted into old age. They called schooling the most reliable way yet found of raising intelligence.
The third is the home a child grows up in. How much of the difference between people’s scores goes with their genes depends on their lives: in poor American families it was close to nothing (see the chapter on genes and upbringing). Heritable never meant fixed.
So a score tells you where someone stands now, in a test made for a particular crowd, after a particular life. That matters when a test made in one country is used in another. When researchers in Bhaktapur tested 600 babies with a standard American scale of development (not an IQ test, but built the same way), they first had to change its pictures. A photograph of someone vacuuming became a woman with a broom, a pool became bathing at a tap, and a child building a snowman became a child making a toy of clay. Even then, the authors warned that the babies’ language scores, unlike the others, needed caution when set beside the American averages.
Now imagine doing that for a whole country, with no such care. In 2002 two writers published a table that claimed to give the average IQ of nearly every nation on Earth, and within a few years more than twenty studies had used it. The writers never said how they searched for studies, or which ones they left out and why. When psychologists in Amsterdam went through the evidence for one region, Africa south of the Sahara, with written rules for which studies counted, they found that the table’s figures there rested mostly on groups of whoever was easy to test, not people chosen to stand for a country; that some higher-scoring studies had been left out with no reason given; and that some countries’ figures were simply worked out from their neighbors’. Their own careful estimate came out well above the table’s. Researchers from the same group showed that the table’s numbers go up and down with how rich, healthy and schooled a country is today, the very things thought to have raised scores in rich countries over the last century. Whatever such a number measures, it is tangled up with a country’s schools, food and health. No such table tells you anything about the people of Nepal, and this book doesn’t print one.
Other ways to be clever
If tests miss so much, are there other intelligences they ought to measure? The most famous answer, multiple intelligences, belongs to the next chapter. Another is emotional intelligence.
The idea came from two American psychologists in 1990: a set of skills for reading emotions in yourself and in others, managing them, and using feelings to plan and get things done. It became famous in 1995, when the writer Daniel Goleman argued in a bestselling book that emotional intelligence predicts success better than IQ does.
The evidence is more modest. Part of the trouble is measuring it. Some tests check the skill itself, asking you, say, what a face shows or what a person in a story would feel, with right and wrong answers. Others ask you to rate yourself, which mostly measures how you see yourself, and overlaps a good deal with personality. In 2020 a pooling of 158 studies of more than 42,000 students found that emotional intelligence went with better marks, a modest link of 0.2, twice as strong when the skill was tested as when students rated themselves. After intelligence and personality were allowed for, it added only one or two percent more to the prediction, though it still came third, after intelligence and conscientiousness. Nobody could say which came first: whether reading feelings well helps students do well, or doing well makes school feel easier. At work, the tested skill went with better performance in jobs that deal with people’s feelings all day, like nursing or sales, and with worse performance in others. So emotional intelligence is real enough to measure, and useful, but it doesn’t beat IQ.
Another question about who is clever has a clearer answer. Are boys and girls different in ability? In 2005 the American psychologist Janet Hyde gathered 46 large reviews of differences between the sexes, in everything from mathematics to self-esteem. Of the 124 differences in them, 78% were small or close to zero. A few were not, such as how far and how fast people can throw, and how often they hit someone, where men came out higher. In mathematics, a review of 242 studies of more than a million people found almost no difference between the average boy and girl, and the scores were spread out almost equally. On ability, boys and girls are far more alike than different.
So intelligence, as psychologists measure it, is real: people do differ in how quickly they learn and how well they reason, the differences can be measured fairly reliably, and they matter for school and work. But an IQ score is a place in a crowd, not a quantity fixed inside you. It leaves out a great deal, it rises with schooling and a better life, and it means little when a test made for one people is used on another. The same tool that was made to find children who needed help was soon used to argue for stopping “the feeble-minded” from having children and for shutting out immigrants. Binet’s warning still stands: a score can be useful, but only in the hands of people who remember what it can’t say.
Found a mistake? Please tell us.
