Child geniuses may not be ahead forever
Recent research suggests that intelligence may not be as stable across the lifespan as we assume.
18 March 2024
By Emma Young
Stories of child geniuses seem to be perennially popular. One of the latest appeared at the end of January 2023, featuring a boy who taught himself to read at the age of two. By age three, he had become the UK's youngest member of Mensa, the society for people with a high IQ.
Mensa admits people who score in the top 2 per cent on standard intelligence tests, including the well-known Stanford–Binet test. Admittance is for life — members don't have to retake the tests to prove themselves again.
But while we often think of intelligence as being stable (in the absence of certain illnesses or injuries), a major new systematic meta-analysis of data from 205 longitudinal studies on a total of more than 87,000 people has found that it's more variable across the lifespan than we might think.
Moritz Breit at the University of Trier and colleagues looked at studies that assessed 'general intelligence' (the Stanford–Binet test measures this) and/or abilities that fall within the Cattell–Horn–Carroll (CHC) model of cognitive ability. The list of CHC abilities includes decision-making speed, working memory capacity, learning efficiency (a measure of the speed and accuracy with which someone learns, stores, and consolidates new information), fluid reasoning (which is needed to solve novel problems) and 'comprehension knowledge' (a measure of the depth and breadth of knowledge and skills developed through experience and learning — including words and general knowledge).
Their set of studies included participants that ranged in age from 1–88 who lived across several countries (though most of the data came from those based in North America and Europe), and whose cognitive ability was re-tested anywhere from a day to 79 years after an initial assessment.
The team found that, for adults, cognitive abilities — and general intelligence, in particular — were in fact pretty stable, at least over time periods of about six years. This stability ranged from 0.65 for the 'least' stable abilities, which included fluid reasoning and learning efficiency, to 0.8 for general intelligence. (A stability of '1' would have meant identical scores on the initial test and a subsequent assessment.)
The lower stability of the more fundamental cognitive abilities might seem surprising, but studies have found that, with age, processing power wanes more than knowledge-based abilities, and that it also varies more day-to-day. However, the team did also find that the tests of knowledge-based abilities used in these studies were more reliable than the tests of the more effortful-processing based abilities. This discrepancy could have helped to explain the differences in stability.
When it came to children, though, there were bigger differences between results on the same type of test conducted at two different time points. For children aged under four, stability didn't reach higher than a "low" 0.7, even if the two tests were conducted less than a year apart. For 8-year-olds, a stability of 0.8 was observed, but only if the two tests were conducted within two years. For 12-year-olds, the researchers write, a test result could be assumed to have a 0.8 stability for four years (after which the results on two tests became less similar). At age 18+, a score of 0.8 could be assumed for about six years, but after this, re-testing would be recommended, they write.
There are some limitations even to this comprehensive meta-analysis, which considered a range of variables, including test reliability, when analysing the data. For example, no one study investigated the stability of cognitive abilities from early childhood to late adulthood. Instead, the team had to combine different information from different studies to arrive at the age trends.
But the study also has some clear strengths, and has raised a number of interesting findings; differences in the stability of different types of cognitive abilities, of course, but also the greater variability in test-re-test scores for children — and younger children, especially — compared with adults. The team did explore whether the nature of testing children versus adults might account for this, but this did not seem to be the explanation. Though the authors don't delve into what exactly might explain this result, it could relate to the fact that children's brains are developing rapidly, and potentially also reflects changes in classroom environments and learning.
Overall, though, while the common conception that intelligence is stable seems to be the case (or at least roughly the case) for healthy adults, for children it's a different story. What this means in practice is that IQ test scores should have a good 'shelf life' for adults, but this is not the case for children, and especially not for children under the age of seven. So, a child who gains a Mensa-level score on an intelligence test at the age of three may potentially get a sub-Mensa — or, alternatively, even better — score at 5, 7 or 17.
