Parents ask us a version of this question more than any other. A child takes a test at six, scores in the middle, and the parent wants to know whether that number is a verdict. Another child scores well at nine and the parent wants to know whether it will hold. Both questions have the same answer, and it is more interesting than either a yes or a no.
A child’s IQ score is stable enough to mean something and unstable enough that no single number should ever be treated as settled. Those two facts sound contradictory. They are not, and the reason comes down to what psychologists mean by the word stable.
Stability does not mean getting the same number twice
When researchers say intelligence is stable, they are not claiming a child who scores 112 will score 112 again. They are talking about rank order: whether children who scored above their peers at one age tend to score above their peers at a later age. That is a question about a group, and it is measured with a correlation.
A correlation of 1.0 would mean the ranking never changes at all. A correlation of 0 would mean a childhood score tells you nothing about a later one. Real numbers sit between the two, and where they sit depends almost entirely on two things: how old the child was at the first test, and how long you wait before the second.
Preschool scores predict poorly. This is not a flaw in the tests so much as a fact about four year olds, whose attention, language and willingness to cooperate vary enormously from one morning to the next, and whose abilities are genuinely still forming. By about age eleven the picture is different, and by the mid teens a score behaves much more like an adult’s.
What a sixty-six year follow-up found
The strongest evidence on this comes from Scotland, which did something no other country did. In 1932 and again in 1947, almost every eleven year old in the country sat the same mental ability test on the same day. Decades later, researchers tracked down the survivors and tested them again.
In the 1947 cohort, 1,391 people who had been tested at age eleven were retested in old age. The correlation between the two scores was 0.68, with a standard error of 0.01 (Deary et al., 2012). Two thirds of the ranking had survived sixty-six years, a world war, every life event those people lived through, and the ordinary decline of ageing.
| Gap between the two tests | Typical correlation | What that feels like in practice |
|---|---|---|
| A few weeks | 0.85 to 0.95 | Nearly the same number, give or take a few points |
| One year | 0.75 to 0.90 | Usually the same band, occasionally the one next door |
| Five years in childhood | 0.60 to 0.80 | Same broad region, individual children move noticeably |
| Age 11 to old age | 0.68 | Two thirds of the ranking survives sixty-six years |
The final row is measured: Deary and colleagues (2012) followed 1,391 people from the 1947 Scottish Mental Survey and found a correlation of 0.68 (standard error 0.01) between test scores at age 11 and in old age.
Read that table the other way, though, and the same finding says something else. A correlation of 0.68 leaves a great deal unexplained. Plenty of individuals in that study moved a long way up or down relative to their peers. The group pattern was strong. Any one person’s path was not guaranteed by it.
Why younger children’s scores move more
Three things make a young child’s score less fixed than an older child’s.
The first is measurement. Every test carries error, and shorter tests carry more of it. A twenty question test taken once gives a rougher estimate than a two hour assessment given by a psychologist, which is one of the real differences between an online test and a formal assessment.
The second is the child. A six year old who is tired, hungry, anxious or simply bored will score below what they can do. That is not a small effect, and anxiety in particular can take several points off a score without touching the underlying ability at all.
The third is real development. Children do not grow their abilities at a uniform rate. A child can spend a year consolidating and then move quickly. Schooling drives some of this directly: a meta-analysis by Ritchie and Tucker-Drob (2018) found that each additional year of education was worth roughly one to five IQ points, and that the gain persisted rather than fading. What actually moves a child’s score is a question with real answers.
In my experience
My own daughter took a nonverbal reasoning test at seven, in a school hall, on an afternoon when she had fallen out with a friend at lunchtime. The result put her squarely in the middle. I remember feeling the small, unworthy lurch of disappointment that I suspect most parents will recognise and few will admit to.
She took a similar test at eleven and scored fourteen points higher. Nothing dramatic had happened in between. She had grown up, she had four more years of school, and she was not, on that particular day, thinking about a friend who had been unkind to her. The second number was not the truth revealed. It was simply a second sample, taken under better conditions, of something that was always a range rather than a point.
That is the habit I would pass on. Treat a score as a reading taken on one day, in the way you would treat a single blood pressure measurement, and the question stops being what is my child’s IQ and becomes the more useful what range does my child seem to sit in.
What this means for reading your child’s result
If your child is under seven, hold the number loosely. It is real information, but its predictive power at that age is modest, and it will be a better estimate in a few years.
If your child is eleven or older, the score is more informative, though still a range rather than a point. Knowing where a score sits in the population tells you more than the number itself does.
If the score is lower than you expected, the single most common reason is not ability. It is conditions on the day, and there is a long list of things a low score does not mean.
And if you want a second reading, take one. Leave a few months so the questions are not fresh, and run it under decent conditions: morning, fed, rested, unhurried. You can take the free kids IQ test again at any time, and there is an adult version if you want to sit the equivalent yourself. Comparing two readings taken properly will tell you considerably more than agonising over one.
References
- Deary, I. J., Yang, J., Davies, G., Harris, S. E., Tenesa, A., Liewald, D., Luciano, M., Lopez, L. M., Gow, A. J., Corley, J., Redmond, P., Fox, H. C., Rowe, S. J., Haggarty, P., McNeill, G., Goddard, M. E., Porteous, D. J., Whalley, L. J., Starr, J. M., & Visscher, P. M. (2012). Genetic contributions to stability and change in intelligence from childhood to old age. Nature, 482(7384), 212-215. doi:10.1038/nature10781
- Deary, I. J., Whalley, L. J., Lemmon, H., Crawford, J. R., & Starr, J. M. (2000). The stability of individual differences in mental ability from childhood to old age: Follow-up of the 1932 Scottish Mental Survey. Intelligence, 28(1), 49-55. doi:10.1016/S0160-2896(99)00031-8
- Ritchie, S. J., & Tucker-Drob, E. M. (2018). How much does education improve intelligence? A meta-analysis. Psychological Science, 29(8), 1358-1369. doi:10.1177/0956797618774253