An IQ percentile describes relative standing in a norm group. It is not the percentage of questions answered correctly and not the percentage of intelligence a person possesses. An IQ score of 115 is often described as being around the 84th percentile on a scale with a mean of 100 and standard deviation of 15. That statement is easy to misread. It does not mean the person answered 84 percent of the test correctly. It does not mean they have 84 percent of some maximum amount of intelligence. It means that, under the relevant normative distribution, their standardized score is higher than the scores of roughly 84 percent of the comparison population. A percentile is a rank position, not an amount.
Put 100 fictional people in a room
Imagine 100 people sampled so that they perfectly represent the norm population, arranged from lowest to highest standardized score. Someone at the 50th percentile would stand around the middle. Someone at the 84th would have about 84 people below their position and about 16 above it. That image captures what percentiles are for: they translate an unfamiliar standardized score into an intuitive statement about relative standing. But real norm groups contain many more than 100 people, scores are estimated with uncertainty, and percentile conversions depend on the test’s normative system. The room is an illustration, not how an actual test is normed.
Why percentile gaps behave strangely
Percentiles are not equal-interval units. Moving from the 50th to the 60th percentile does not represent the same standardized-score change as moving from the 90th to the 100th—indeed the 100th percentile is not a finite top score on an ideal continuous normal distribution. On the common IQ metric:
- 100 corresponds approximately to the 50th percentile;
- 115 corresponds approximately to the 84th;
- 130 corresponds approximately to the 98th.
Notice what happens. Fifteen IQ points from 100 to 115 move through about 34 percentile points. Another fifteen points from 115 to 130 move through only about 14. The standardized-score intervals are equal; the percentile intervals are not. This is why percentiles are excellent for communicating rank and poor for arithmetic. You should not average percentile ranks as though they were ordinary equal-interval scores.
Percentile of what population?
Every percentile is incomplete without a reference group. A child’s score may be compared with same-age peers in the test’s standardization sample. Other assessments may use grade-based or other norms. If the norm group is inappropriate, the percentile can be misleading even when the arithmetic is correct. This is why standardized testing depends on representative norming samples, clearly specified comparison groups and periodic renorming.
What about uncertainty?
An obtained IQ score is an estimate affected by measurement error. If a report gives a confidence interval around a standardized score, the corresponding percentile standing is uncertain too. Suppose a fictional assessment estimates an IQ of 115 with a confidence interval extending several points on either side. Reporting “84th percentile” as though it were a perfectly known location hides that uncertainty. The most defensible interpretation is a range of plausible standing, informed by the instrument’s reliability and scoring guidance.
Percentile is not probability
Another common mistake is to interpret the 84th percentile as an 84 percent probability that the person is intelligent, successful or will outperform another specific person. A percentile says none of those things. It summarizes relative performance on the measured construct in a normative distribution. Predictions about school, work or any other outcome require separate evidence and remain probabilistic.
The simplest correct translation
When you see an IQ percentile, translate it this way:
Where does this standardized test performance sit relative to the relevant norm group?
That is all the percentile is trying to answer. It is often a useful answer. The trouble begins when rank is mistaken for amount, certainty, destiny or a percentage of the test that somebody “got right.”
References
- AERA, APA, & NCME. (2014). Standards for Educational and Psychological Testing. AERA.
- National Academies of Sciences, Engineering, and Medicine. (2015). Psychological Testing in the Service of Disability Determination. National Academies Press. DOI
- Wechsler, D. (2008). WAIS–IV Technical and Interpretive Manual. Pearson
