Percentile Rank in AP Psychology

The claim: a raw score means little until you rank it

A raw score, like the number of items someone got right, tells you almost nothing by itself. Percentile rank turns that number into a position: how this person did compared to the people the test was standardized on. That comparison group is called the norm group.

This is why percentile rank lives in Unit 2: Cognition, Topic 2.8: Intelligence and Achievement, under the objective about how intelligence is measured. The course frames intelligence and achievement tests as useful only when they follow psychometric principles, and standardization is one of them. Percentile rank is the payoff of standardization. Once a test has been given to a large group under uniform procedures, any new score can be placed within that group's distribution.

IQ scores themselves work the same way. Modern IQ tests no longer use the old mental age divided by chronological age formula. A score of 100 is set at the middle of the norm group, which makes it the 50th percentile.

Evidence: finding percentiles with the 68-95-99.7 rule

No calculator is allowed on the AP Psychology exam, so you need the normal curve shortcuts. These only work when scores are approximately normally distributed. In a normal distribution, about 68% of scores fall within 1 standard deviation (SD) of the mean and about 95% fall within 2 SD. Split those in half and you get the percentiles for each landmark:

• At the mean: 50th percentile
• 1 SD above: about 84th percentile
• 2 SD above: about 98th percentile
• 1 SD below: about 16th percentile
• 2 SD below: about 2nd percentile

Worked example: a normally distributed test has a mean of 100 and SD of 15. A score of 130 is 2 SD above the mean, so it's about the 98th percentile. A score of 85 is 1 SD below, so it's about the 16th. The same logic works with any mean and SD, as long as the distribution is roughly normal. On a normally distributed test with a mean of 115 and SD of 10, a score of 105 is 1 SD below, putting it near the 16th percentile.

With a small list of raw scores, just count. If 15 people took a test and one person's 80 is the 8th score from the bottom, then 7 people scored lower. That's 7 out of 15, or roughly the 47th percentile.

Limits: where a percentile rank can mislead

Percentile is not percent correct. Someone at the 60th percentile might have answered 90% of the items correctly on an easy test or 40% on a hard one. The percentile only reflects how others did.

Percentiles also stretch and squeeze the real distance between scores. Most people cluster near the mean, so moving from 100 to 115 jumps a person from the 50th to about the 84th percentile, a 34-point gain. Moving the same 15 points from 115 to 130 only gains about 14 percentile points. Two people a few IQ score points apart in the middle can look far apart in percentile terms.

A percentile rank is only as good as its norm group. Because of the Flynn Effect, IQ scores have generally risen over time in much of the world, so comparing someone today to an old norm group can make their percentile look higher than it would against current peers. That's one reason tests get restandardized. And if the norm group doesn't reflect the background of the person being tested, or the test carries cultural bias, the percentile can carry that bias into decisions about school placement, jobs, or other opportunities, which is exactly the kind of systemic issue Topic 2.8 asks you to explain.

Placing a score on the curve in multiple-choice and Article Analysis items

Say a multiple-choice item gives a test's mean of 100 and SD of 15, notes that scores are approximately normally distributed, and asks for the approximate percentile of someone who scored 115. A question could also add a second step, such as what an aptitude score at the 84th percentile predicts about that person's performance compared to peers. Find how many SDs the score sits from the mean, then match it to the landmarks above. Without the normal-distribution information, the mean and SD alone can't pin down a percentile.

The Article Analysis question asks you to interpret basic statistics from a summarized study. If the summary supplies a percentile, or gives enough information to estimate one (for example, means and SDs plus an approximately normal distribution), state the percentile and then say what it means in plain terms ("scored higher than about 84% of participants").

One classic wrong answer treats percentile as the share of questions answered correctly. Another mixes up direction: a score 1 SD below the mean lands near the 16th percentile, not the 34th, because you have to count everyone below the score, including the whole bottom tail.