How IQ changes with age

Last reviewed 9 September 2026

Why your score barely moves

People are often surprised that a 70-year-old and a 25-year-old can both have an IQ of 110 while performing quite differently on the raw test. The explanation is that IQ scores are age-normed.

Clinical tests band their standardisation samples by age and score you against your own band.⁶ Your IQ is not "how much reasoning you did" - it is "how you compare with other people your age". Since the whole reference group ages with you, the number can stay flat while the underlying raw performance changes considerably.

Two curves, opposite directions

Averaging "cognitive ability" across the lifespan hides the most important fact, which is that its components move in opposite directions.

The distinction goes back to work in the 1960s separating fluid ability - reasoning on novel problems - from crystallised ability - accumulated knowledge and vocabulary. Fluid ability peaked early and declined; crystallised ability held up or continued rising well into later life.³

This is why an older person can be markedly better at their job while being measurably slower on a matrix item. Most real work leans on accumulated knowledge and pattern recognition, which is exactly the part that holds up. See fluid vs crystallised intelligence for the full distinction.

When decline starts - a real dispute

Here the literature genuinely disagrees, and the disagreement is instructive because it turns entirely on study design.

Cross-sectional studies compare people of different ages at one moment. These consistently show decline in reasoning, speed and memory beginning in the twenties or thirties.¹

Longitudinal studies follow the same people for decades. These typically show little decline before the sixties.²

Forty years apart is not a rounding error. Each design has a distinct flaw:

• Cross-sectional studies confound age with cohort. A 70-year-old tested today grew up with less schooling and different nutrition than a 30-year-old. Some of the apparent "decline" is the Flynn effect seen sideways - later cohorts simply score higher.⁷
• Longitudinal studies are inflated by practice and attrition. Participants get better at the tests through repeated exposure, which masks decline. And people who drop out are disproportionately those declining fastest, so the surviving sample looks healthier than the population.

The reasonable synthesis: some genuine decline in fluid abilities probably begins earlier than the longitudinal picture suggests and later than the cross-sectional one implies, it is gradual, and it is small relative to the differences between individuals of the same age.

Different abilities peak at different ages

A large study using web-based testing of tens of thousands of participants found there is no single age at which "cognitive functioning" peaks, because different abilities peak at markedly different times.⁵

Processing speed peaked earliest, around the late teens to early twenties. Short-term memory peaked in the mid-thirties. Emotion recognition peaked later still, in the forties or fifties. Vocabulary peaked latest of all - in some analyses not until the late sixties or beyond.

The practical implication is that "peak mental age" is not a meaningful concept. At 22 you are near your best on raw speed and considerably short of your best on knowledge. At 55 the reverse. Which matters depends entirely on the task.

What stays remarkably stable

The most striking finding in this whole area concerns not the level but the ranking.

Scotland tested almost every eleven-year-old in the country in 1932. Decades later, researchers traced surviving participants and retested them at age 77 on the same instrument. The correlation between scores taken 66 years apart was around 0.6, rising above 0.7 once corrected for measurement error.⁴

That is an extraordinary degree of stability for anything measured across two thirds of a century. Absolute performance changes a great deal; your position relative to your peers changes far less.

Two caveats keep it honest. A correlation of 0.6 still leaves substantial individual movement - plenty of people shifted considerably. And rank stability is not permanence: it is compatible with everyone declining together.

What this means for your score here

This test does not age-norm. It has one set of provisional item parameters applied to everyone, because age-banded norms would require a standardisation sample we do not have.

The consequence is straightforward and worth knowing: if you are past your forties, this test will tend to underestimate you relative to a properly age-normed clinical test, because 23 of its 30 items tap fluid reasoning and spatial ability - the abilities that decline - and only seven tap verbal knowledge, which does not. If you are in your twenties, the reverse bias applies mildly. The methodology page states this among the other limitations.

Common questions

References

1. Salthouse, T. A. (2009). When does age-related cognitive decline begin? Neurobiology of Aging, 30(4), 507-514. doi:10.1016/j.neurobiolaging.2008.09.023
2. Schaie, K. W. (1994). The course of adult intellectual development. American Psychologist, 49(4), 304-313.
3. Horn, J. L., & Cattell, R. B. (1967). Age differences in fluid and crystallized intelligence. Acta Psychologica, 26, 107-129.
4. 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.
5. Hartshorne, J. K., & Germine, L. T. (2015). When does cognitive functioning peak? The asynchronous rise and fall of different cognitive abilities across the life span. Psychological Science, 26(4), 433-443. doi:10.1177/0956797614567339
6. Wechsler, D. (2008). Wechsler Adult Intelligence Scale - Fourth Edition: Technical and Interpretive Manual. Pearson.
7. Pietschnig, J., & Voracek, M. (2015). One century of global IQ gains: A formal meta-analysis of the Flynn effect (1909-2013). Perspectives on Psychological Science, 10(3), 282-306. doi:10.1177/1745691615577701

Related reading