What is IQ? What intelligence tests measure, and don't

An IQ test measures how well someone does on a set of mental tasks, such as reasoning, vocabulary, working memory and processing speed, compared with a large sample of people their own age. The score, an intelligence quotient, is scaled so that 100 is the average and most people fall between 85 and 115. It predicts school results well, and many other outcomes less well, and it does not capture everything people mean by intelligence.

In short

• An IQ is a ranking, not a quantity: it says where your score sits relative to others your age, with 100 as the middle.
• Scores track one another across very different tasks, which is the evidence for a general ability, and they predict grades better than almost any other single measure.
• Average scores on the same tests rose about 2 to 3 points per decade through much of the 20th century, the Flynn effect, which shows that environment shapes what IQ tests measure.

What does an IQ test actually ask you to do?

A modern test, such as the Wechsler scales used by psychologists, is a battery of short subtests given one-on-one, often over an hour or more. Some ask for definitions of words or how two things are alike. Some show a pattern of shapes with a piece missing. Some ask you to repeat a string of digits backward, or to copy symbols as fast as you can. The subtests are grouped into indexes, typically verbal comprehension, reasoning with new problems, visual and spatial thinking, working memory and processing speed.

The striking finding, first noticed by the psychologist Charles Spearman in 1904, is that people who do well on one kind of task tend to do well on the others. Someone with a large vocabulary is, on average, also quicker at completing visual patterns. Psychologists call that shared factor g, for general intelligence, and the full-scale IQ is mostly an estimate of it. IQ measures how well you solve the kinds of problems on the test, relative to your age group, and that ability turns out to be broad but not all-encompassing.

How is an IQ score calculated?

The first tests worked differently from today's. In 1905, Alfred Binet and Théodore Simon in France published a test to find schoolchildren who needed extra help. Each task was matched to the age at which a typical child could do it, giving a child a "mental age." In 1912 the German psychologist William Stern turned that into a quotient: mental age divided by real age, times 100. An 8-year-old who did as well as a typical 10-year-old scored 10 ÷ 8 × 100 = 125. Lewis Terman's 1916 American version, the Stanford-Binet, made the idea famous.

The ratio breaks down for adults, since a 40-year-old doesn't perform like twice a 20-year-old. So in 1939 David Wechsler replaced it with the method used today, the deviation IQ. A test is first given to a large sample chosen to match the population, its norm group. Your raw score is then compared with theirs, and converted so that the average is 100 and the standard deviation, a measure of typical spread, is 15.

A worked example shows the conversion. Suppose people your age in the norm group average 50 correct answers on a test, with a standard deviation of 8. You get 58: exactly one standard deviation above the average. That becomes 100 + 15 = 115. Because scores follow a bell curve, the scale maps to percentiles like this:

IQ score	Standard deviations from average	Share of people scoring lower
70	−2	about 2%
85	−1	about 16%
100	0	50%
115	+1	about 84%
130	+2	about 98%

So an IQ of 115 doesn't mean 15% more intelligence than average. It means a higher score than about 84 out of every 100 people your age. Two-thirds of people fall between 85 and 115.

What does IQ predict?

More than most single measures in psychology, but far from everything. In 1996 a task force of the American Psychological Association, led by Ulric Neisser, reviewed the evidence in "Intelligence: Knowns and Unknowns." It found that IQ scores correlate with school grades at about 0.50, among the strongest links in the field, and that test scores were the best single predictor of how many years of education a person completes. Squaring that correlation gives about 25%: a quarter of the differences in grades between students line up with differences in IQ, and three-quarters come from other things, such as effort, interest, teaching and circumstance.

Links to job performance are real but smaller: the task force put the correlations with measures such as supervisor ratings and work samples at about 0.3 to 0.5, which means IQ lines up with roughly 10% to 25% of the differences in how well people do their jobs. Scores also correlate with income and social status later in life, though more weakly. The task force stressed that IQ explains only a modest share of who succeeds, and that within a group of people with similar scores, outcomes vary enormously. That is the general point about any score of this kind, the MBTI included: it can predict averages for groups well and individuals much less.

What does IQ miss?

The tests sample a narrow slice of mental life under timed, quiet conditions. They don't directly measure creativity, practical judgment, wisdom, social and emotional skill, persistence or curiosity, all of which matter for how a life goes. Some psychologists have proposed broader models: Howard Gardner's "multiple intelligences," and Robert Sternberg's theory adding creative and practical abilities to analytic ones. Those have their own critics, since some of their abilities are hard to measure reliably. But most researchers agree that IQ captures an important ability, not the whole of intelligence.

A score is also a snapshot. Illness, anxiety, poor sleep, an unfamiliar language or culture, or simply not trying can lower it. Psychologists treat a single result as an estimate with a margin of error, often several points either way. And people are poor judges of their own scores: the tendency of the least skilled to overrate themselves, the Dunning-Kruger effect, was first shown on tests of humor, logical reasoning and grammar.

Why did IQ scores rise over time?

When tests are re-normed, new samples do better on the old questions. James Flynn, a political scientist working in New Zealand, documented this in the 1980s, and it now bears his name. A 2014 meta-analysis of 285 studies found average gains of about 2.3 points per decade since 1951, and about 2.9 per decade on modern Wechsler and Stanford-Binet tests. Across studies, the biggest gains have been on abstract reasoning puzzles, while school-taught skills such as vocabulary and arithmetic rose much less.

These changes are too fast to be genetic, so the causes are environmental. The leading candidates are better nutrition and health, more and longer schooling, smaller families, and a modern world that trains people to think in abstract categories. In some countries the rise has stopped or reversed. Using test scores from 736,808 Norwegian military conscripts born from 1962 to 1991, the economists Bernt Bratsberg and Ole Rogeberg found in a 2018 study that scores rose until the 1975 birth cohort and then fell. Because the decline appeared between brothers in the same families, they concluded that its causes were also environmental, not a change in who was having children.

Still argued

What exactly g is, a single property of the brain or a statistical summary of many abilities that tend to go together, is still argued among researchers. So is what the Flynn effect means: whether people really became more intelligent, or became better at the specific kinds of abstract thinking IQ tests reward. Flynn himself changed his view over his career. Researchers agree that genes and environment both matter for individual differences in IQ; how they interact is an active field, and findings about one population don't automatically carry over to another.

The brain behind a score keeps changing through life. An IQ result describes how someone did on one set of tasks, at one time, against one group, and is most useful when it is read that way.

Check yourself

• Without looking back: why does an IQ of 115 not mean someone is 15% smarter than average?
• What is the difference between the old ratio IQ and today's deviation IQ, and why did the change happen?
• Why do researchers think the Flynn effect has environmental, not genetic, causes?

Sources

• Intelligence: knowns and unknowns — Neisser et al., American Psychologist, American Psychological Association, 1996
• The Flynn effect: a meta-analysis — Trahan, Stuebing, Hiscock and Fletcher, Psychological Bulletin, 2014
• Flynn effect and its reversal are both environmentally caused — Bratsberg and Rogeberg, Proceedings of the National Academy of Sciences, 2018
• Intelligence quotient — Wikipedia
• Flynn effect — Wikipedia