IIF Certified Assessment
Start IQ Test
IIF Certified Assessment
An IQ score is not an absolute measurement. It is a comparison against whoever the test was calibrated on, in the year it was calibrated. Population performance keeps drifting upward, so an old test quietly flatters you — the same way an old salary sounds bigger than it was.
Trahan et al. 2014 meta-analysis, all studies pooled. The best-supported single figure, and our default.
The same performance, scored against different years
Each column is what your result would read as if the test had been normed in that year. Above the line means an older, more flattering yardstick. The thin bar on each column is the range across published rates.
| Normed in | Reads as | Range across rates | Difference |
|---|---|---|---|
| 1960 | 121.9 | 118.6–124 | +6.9 |
| 1970 | 119.6 | 117.4–121 | +4.6 |
| 1980 | 117.3 | 116.2–118 | +2.3 |
| 1990 | 115 | 115–115 | ±0 |
| 2000 | 112.7 | 112–113.8 | −2.3 |
| 2010 | 110.4 | 109–112.6 | −4.6 |
| 2020 | 108.1 | 106–111.4 | −6.9 |
| 2026 (today) | 106.7 | 104.2–110.7 | −8.3 |
A score of 115 from a test normed in 1990 is being measured against the population as it was 36 years ago. On 2026 norms the same performance is worth about 106.7 — a drop of 8.3 points, which moves it from the 84.1th percentile to roughly the 67.2th. Depending on which published rate you accept, the honest answer is somewhere between 104.2 and 110.7.
This does not apply to a test you take today. A current, properly normed test already compares you with the present-day population — there is nothing to correct, and applying a drift adjustment to it would invent an error that is not there. Use this only for a score you are holding from an older instrument.
An old score is not wrong. It is dated. Norms are recalibrated as populations shift, so a result from 1995 was measured against 1995. The adjustment above estimates the drift — but an estimate of an old number is still not a current measurement.
Nobody needs a psychometrics lecture for this one. A salary of £20,000 in 1990 was a different thing from £20,000 today — same number, different purchasing power, because the yardstick moved underneath it. IQ scores do the same, for the same kind of reason.
Before a test is sold, it is given to a large sample and the results are set so that this sample averages 100. That sample is a snapshot of a particular population in a particular year. Everything the test reports afterwards is a comparison against those people.
Across the twentieth century, average raw performance on these tests rose steadily — better schooling, better nutrition, more test-like thinking in everyday life. This is called the Flynn effect, after the researcher who documented it. The tests did not get easier; people got better at them.
Take a 1990-normed test today and you are ranked against 1990 people, who on average performed slightly worse than today’s. Your number comes out higher than it would on a current test. Publishers fix this by re-norming every decade or so — which is exactly why your old certificate and a new result disagree.
A lower adjusted score does not mean you became less intelligent, and it does not mean your original test was wrong. Both numbers are correct answers to different questions. “115 on 1990 norms” and “about 107 on today’s norms” describe the identical performance, priced in two different currencies. If you are comparing an old result with a new one — yours or somebody else’s — you have to convert before the comparison means anything, and even then, check whether the gap clears the margin of error.
Most calculators that attempt this pick a rate, hide it, and present the output as fact. The rate is the single most disputed quantity in the whole topic, so the selector above is not a power-user option — it is the honest part of the tool.
Roughly three points per decade, averaged across many countries and test types. The number that made the phenomenon famous, and still the one most often quoted second-hand.
Trahan and colleagues pooled the available studies and found a smaller effect than the headline figure — 2.31 overall, rising to 2.93 when restricted to modern Wechsler and Stanford-Binet tests. This is our default because it rests on the widest evidence base.
Measured directly across the move from WAIS-IV to WAIS-5, the drift came out at about 1.2 points per decade — less than half the classic figure, which suggests the effect has slowed considerably in recent years.
A 2018 study in PNAS using Norwegian conscript data found scores peaked with the 1975 birth cohort and have declined since. If that pattern holds where you are, a score from the 1990s may need little or no adjustment — and one from the 2000s could arguably need adjusting the other way.
The further back your test, the wider the honest answer gets — because a longer gap multiplies whatever disagreement exists about the rate. That is why every column on the chart carries its uncertainty bar even when you have picked a single rate, and why the panel shows a range next to the point estimate. If you take one figure from this page, take the range.
The adjustment is real, but it is not always relevant. These are the cases where it changes a decision and the cases where it is noise.
| Test normed in | Years to today | Adjustment | A reported 115 becomes | A reported 130 becomes |
|---|---|---|---|---|
| 2020 | 6 | −1.4 | 113.6 | 128.6 |
| 2010 | 16 | −3.7 | 111.3 | 126.3 |
| 2000 | 26 | −6.0 | 109.0 | 124.0 |
| 1990 | 36 | −8.3 | 106.7 | 121.7 |
| 1980 | 46 | −10.6 | 104.4 | 119.4 |
| 1970 | 56 | −12.9 | 102.1 | 117.1 |
| 1960 | 66 | −15.2 | 99.8 | 114.8 |
Two things to keep in mind when reading that table. First, these are population-level corrections — they describe how the reference group moved, not how any individual changed. Second, the adjustment sits on top of the test’s own measurement error, so the total uncertainty around an old score is wider than either alone. A 1990 score of 115 is best read as “somewhere in the low hundreds on today’s scale”, not as 106.7 exactly. The percentile calculator shows how wide that error is on its own.
Every correction on this page is an estimate layered on top of an estimate: a drift rate that is still disputed, applied to a number taken years ago under conditions nobody recorded. It is the right thing to do with a score you already hold. It is a poor substitute for one taken today against today’s norms.
Normed on current data and scored on the standard scale — mean 100, standard deviation 15.
The score does not expire, but the norms behind it go out of date. Your result was set against the population as it was when the test was calibrated, and average performance has drifted since. A score from a test normed in 1990 typically reads about 8 points higher than the same performance would on today’s scale. The number is still a valid statement about 1990; it is just not a like-for-like comparison with a modern result. The IQ score converter handles the other half of the problem — scales that differ between tests.
Most likely because the two tests were normed decades apart. The older test compares you with an older, on average slightly lower-performing reference group, so it returns a higher number for identical performance. Before concluding that anything changed, convert both scores to the same norms using the tool above, and then check whether what is left of the gap clears the margin of error on the comparison tool. Most apparent drops survive neither step.
The observed rise in average raw performance on intelligence tests over the twentieth century, named after James Flynn, who documented it across many countries. Because tests are re-centred on 100 at each re-norming, the rise does not show up in reported scores — it shows up as the need to re-norm. Estimates of its size range from about 1.2 to 3.0 points per decade depending on the test, the country and the period studied. The IQ bell curve shows what a few points of drift does to your share of the population.
Roughly 2.3 points for every decade since the test was normed, using the best-supported meta-analytic figure — so about 8 points for a test normed in 1990. But there is genuine disagreement: the published range runs from about 1.2 to 3.0 points per decade, and some recent evidence suggests the effect has stalled or reversed. Treat the result as a range rather than an exact figure, which is why the tool above shows one. The IQ percentile calculator converts any adjusted score into its standing.
Possibly, in some populations. A 2018 study of Norwegian conscription data found that scores peaked with the 1975 birth cohort and have been falling since, and the most recent Wechsler transition measured a much smaller drift than the classic figure. Evidence differs by country and by ability tested, so the honest position is that the rate is uncertain and probably smaller now than it was — which is exactly why this tool lets you choose it rather than deciding for you. Our page on how accurate IQ tests are covers the measurement error underneath.
No. A current, properly normed test already compares you with today’s population, so there is nothing to correct. Applying a drift adjustment to a fresh result would introduce an error rather than remove one. This tool is only for interpreting a score you already hold from an older instrument. IQ and age covers the other thing that changed between the two sittings — you.
Two corrections, in this order, and they are independent. First convert the score onto the scale you want, since older tests often used SD 16 or SD 24 rather than SD 15 — that is what the IQ score converter is for. Then apply the norm drift above for the years elapsed. Doing only one of the two is the usual source of a badly wrong figure.
It is a separate correction and easy to conflate with this one. The year affects which population you were normed against; your age affects which age band within it. A score from 1990 taken at 18 and the same score taken at 55 are different statements. IQ and age covers the age side.
Because unusually high and unusually low first results both tend to drift back toward the middle on a second sitting — a property of measuring anything imperfectly, not a change in you. Regression to the mean explains it, and the IQ comparison tool tells you whether the drop clears measurement error at all.
More from the toolkit: work out what any score means as a percentile, see where it sits on the distribution, understand why children of high-scoring parents score lower, or browse all IQ tools.
No drift to correct, no norming year to look up. A score measured against today’s population, with its percentile and confidence range stated up front.
