Why Are My IQ Scores Different on Different Tests?

You took an IQ test and got 128. A few weeks later another test says 141, or 112. Which one is you?

Usually both are telling you something true, just with different amounts of noise and on different rulers. Below are the seven reasons IQ scores differ, how big each effect tends to be, and a checker that tells you whether your own two numbers really disagree.

Is it normal to get different IQ scores?

Yes. Even the best clinical tests do not give the same number twice. A small gap between two good tests is expected. A big gap usually has a specific cause you can find.

Reason	Typical size of the difference
Measurement error	About ±5 points on a top clinical test, more on short or online tests
Different scale (standard deviation)	148 on a scale with SD 24 equals 130 on the usual SD 15 scale
Different abilities tested	A few points to 10 or more, depending on your strengths
Practice and retesting	About 5 points on a second attempt, on average
Old norms (the Flynn effect)	About 2 to 3 points for every decade the norms are out of date
Regression to the mean	Very high or low first scores tend to come back towards 100
Generous online scoring	Can be well over 10 points; there is no published standard

Check your own two scores

Do your two scores really disagree?

Enter both scores. We put them on the same scale and check whether the gap is bigger than normal test error.

First score

Score

Scale

SD 15 (Wechsler, Stanford-Binet 5, most tests)SD 16 (older Stanford-Binet)SD 24 (Cattell, some high-IQ societies)

Where you took it

Psychologist or supervised testOnline or unsupervised test

Second score

Score

Scale

SD 15 (Wechsler, Stanford-Binet 5, most tests)SD 16 (older Stanford-Binet)SD 24 (Cattell, some high-IQ societies)

Where you took it

Psychologist or supervised testOnline or unsupervised test

70100130160

First, on the usual scale

128 ± 6

Second, on the usual scale

146 ± 12

Gap

18 points

This gap is bigger than test error alone explains. Errors of up to about 13 points are normal here, so something else is likely at work: different abilities tested, old or unrepresentative norms, practice, or an online test that scores generously.

Assumes a typical standard error of 3 points for supervised tests and 6 for online ones, with 95% bands. Real tests vary, so treat this as a guide, not a verdict.

What is the margin of error on an IQ score?

Every test score is an estimate. Psychologists describe the wobble with the standard error of measurement, which depends on how consistent the test is.

The WAIS-IV, one of the most carefully built tests in the world, has a full-scale reliability of about 0.97 to 0.98. That gives a standard error of roughly 2 to 3 points, so a 95% confidence band runs about 5 points either side of your score (Canivez, 2010). Even that is a best case: it ignores day-to-day changes in how you feel and small errors by the examiner.

Shorter tests have wider bands. A test with a reliability of 0.80 has a standard error of about 7 points, so its 95% band is about 13 points either side. That is why a psychologist's report gives you a range, not just a number.

Why is my Cattell IQ higher than my Wechsler IQ?

Because the two tests use different rulers. Every IQ scale puts the average at 100, but they stretch the scores differently:

• Wechsler tests (WAIS, WISC) and the current Stanford-Binet use a standard deviation of 15.
• The older Stanford-Binet (Form L-M) used 16.
• The Cattell Culture Fair test and some high-IQ societies use 24.

The same position on the bell curve, the top 2%, is about 130 on a scale with a standard deviation of 15 and about 148 on a scale of 24. To convert, take the distance from 100, divide by the old standard deviation and multiply by the new one.

Cattell (SD 24)	Old Stanford-Binet (SD 16)	Wechsler (SD 15)	Percentile
124	116	115	84th
132	121	120	91st
148	132	130	98th
160	140	137.5	99.4th
172	148	145	99.9th

If your two scores came from tests with different scales, convert them first. Many apparent disagreements disappear. You can see where any score sits on the usual scale with our IQ percentile calculator.

Do different IQ tests measure different things?

Yes. All IQ tests aim at general reasoning ability, but each samples it through different tasks.

• Matrix tests, such as Raven's matrices and our own test, use visual puzzles. They lean on fluid reasoning.
• Full clinical batteries such as the WAIS add vocabulary, general knowledge, working memory and processing speed.
• Verbal and numerical tests reward schooling, reading and language more.

If your verbal ability is much stronger than your visual reasoning, or the reverse, you will score differently depending on the mix. Scores from good tests still agree closely: different well-built tests of reasoning typically correlate strongly with each other, but not perfectly, and that leftover difference shows up as a gap of several points for an individual. Our guide to fluid and crystallized intelligence explains the two main abilities behind this.

Why do online IQ tests give higher scores?

Some do, for three reasons:

1. No real norms. A proper IQ score compares you with a large, representative sample of people your age. Many online tests compare you with nobody, or with a group of self-selected test-takers who are not typical of the population.
2. Flattery sells. A high number is more likely to be shared, and more likely to lead to a paid certificate. Nothing stops a site from scoring generously.
3. Easy, untimed questions. Without a firm time limit, anyone can solve a puzzle slowly. Without hard items, the test cannot tell 130 from 145, so scores bunch at the top.

There is no published standard for how much online scores are inflated, and it varies by site. Good online tests do exist: researchers have built public reasoning tests that correlate well with established ones (Condon and Revelle, 2014). Our guide to how accurate online IQ tests are explains what to look for.

Does practice raise IQ scores?

Yes. Taking the same kind of test again raises scores by about a third of a standard deviation on average, roughly 5 IQ points, and a third attempt adds a little more (Scharfen, Peters and Holling, 2018). In a study of young adults retaking the WAIS-IV after three or six months, full-scale IQ rose by about 7 points (Estevis and colleagues, 2012).

So if your second score is higher, practice is the first suspect. It reflects familiarity with the format, not a jump in ability. See how to prepare for an IQ test for what practice does and does not change.

Why did my IQ score go down?

The most common reasons are:

• Regression to the mean. If your first score was very high, a second test will usually come out a little lower, even if nothing about you changed. Any single score mixes your true ability with some luck, and luck rarely repeats. With two tests that correlate at 0.8, someone who scored 130 on the first would be expected, on average, to score about 124 on the second.
• Newer norms. Average performance on IQ tests rose through much of the twentieth century, by about 2 to 3 points a decade (Trahan and colleagues, 2014). A test with old norms compares you with an easier past population and reads high. A newly normed test reads lower.
• The day itself. Poor sleep, illness, stress or distraction can cost a few points, especially on timed tests.
• A harder or broader test. If the second test covered abilities that are not your strongest, your overall score will be lower.

Does IQ change from childhood to adulthood?

Your rank compared with people your age is fairly stable, but individual scores do move. In the Lothian Birth Cohort, people tested at age 11 and again at age 90 had scores that correlated at 0.54, or about 0.67 after a statistical correction (Deary, Pattie and Starr, 2013). That is a strong link across nearly 80 years, but it still leaves plenty of room for an individual to move by 10 points or more.

Childhood scores also come from children's tests with their own norms, so a number from age 8 and a number from age 30 are not directly comparable.

Which IQ score is right?

Usually, the truth sits somewhere between your scores. To decide which to trust more:

1. Put them on the same scale. Convert scores that use a standard deviation of 16 or 24.
2. Check whether the bands overlap. Use the checker above. If they do, the tests agree.
3. Prefer the more careful test. A supervised test with current, representative norms beats an untimed online quiz.
4. Prefer the first rested attempt over a third or fourth try.
5. If you need a decision, for school, work or a diagnosis, get an individual assessment from a psychologist and use that.

Where does the LifeScore test fit?

The LifeScore IQ test is 25 visual reasoning puzzles with a firm 12-minute timer. It measures one important ability, matrix reasoning, and it is an estimate rather than a clinical IQ. If your other scores disagree, it gives you another carefully timed reading of your reasoning that you can compare with them. Take it once, when you are rested, and read the result alongside your other evidence rather than instead of it.

Common questions

How much can IQ vary between tests?

Between two good, supervised tests on the same scale, gaps of up to about 10 points are common. Gaps of 20 or more usually point to a specific cause: a different scale, very different content, heavy practice or a poorly normed test.

Can your IQ score change?

Your score can change from test to test because of error, practice and conditions. Your underlying ability, compared with people your age, is fairly stable in adulthood, though education and health can move it over years.

Is a 10 point IQ difference normal?

Often, yes. Two tests with standard errors of 3 points each can differ by up to about 8 points through error alone, before you add different content or practice.

Is an online IQ test accurate?

Some are useful estimates. Many are not. Look for a fixed set of questions, a firm time limit and an honest statement that it is not a clinical test.