IQ Raw Score vs. Percentile Rank: What Each Means

What does each number measure?

A raw score records the points assigned to responses under a test’s scoring rules. On a simple right-or-wrong quiz, it may be the number of answers correct. On another assessment, the raw result could combine points, count errors, or record completion time. The scoring rules determine what gets counted; the label “raw” does not guarantee that the number is a plain count.

A percentile rank adds a comparison: it describes the percentage of scores in a specified group that fall below or at a given score, according to the report’s convention for ties. The people whose results provide that comparison are the norm group. The American Educational Research Association, American Psychological Association, and National Council on Measurement in Education describe percentile norms as a person’s standing within a defined population. A University of Minnesota assessment glossary makes the distinction concrete: a raw score is an initial, unadjusted result, while a percentile rank places a score in a specified distribution.

Consider a purely illustrative example, not test data: someone answers a certain set of items correctly and receives a raw count under the test’s rules. That count alone tells you what the scoring process recorded. To call it, say, a particular percentile, you would also need scores from a defined comparison group and a stated method for locating the count within that distribution. The same raw count can therefore have different percentile ranks in different groups.

The quick reading rule is simple: raw score asks, “What did the scoring rules record on this test?” Percentile rank asks, “Where does that result sit relative to this group?” Both concern performance on an assessment. Neither is a verdict on identity, worth, diagnosis, or fixed potential.

Sources: Standards for Educational and Psychological Testing (2014); Scales, Norms, and Equivalent Scores; A Consumer’s Guide to Tests in Print

Why isn’t a percentile the percent correct?

Percent correct and percentile rank use different denominators. Percent correct divides points earned by the total possible points. Percentile rank compares a score with scores from other people in a specified distribution. A person could answer a modest share of a difficult form correctly yet rank relatively high in a particular group; someone could answer many items on an easy form and rank lower in a group whose scores are higher. Those are possible relationships, not claims about any actual IQ test or group.

A percentile also is not an equal-interval ruler. A ten-point change in percentile rank does not necessarily represent the same change in test performance everywhere along the scale. The ETS technical report by William H. Angoff treats percentile ranks as one kind of score scale, distinct from raw and standard scores. A University of Minnesota glossary similarly cautions that percentile positions vary with the group used and distinguishes percentile ranks from equal-interval standard scores.

This matters when reading small differences. A percentile may look precise because it is printed as a number, but its meaning depends on the observed score distribution, how ties are handled, and whether the test has enough score resolution to support that rank. In a short test, one additional response can correspond to a noticeable jump in a reported percentile if few raw-score values are available. The percentile’s numerical step should not be mistaken for an equivalent-sized improvement in the underlying performance.

So a percentile helps answer a relative-standing question, but does not tell you how many items were correct, how large the performance gap is, or how the person would do on a different assessment. Ask for the test and group behind the rank before interpreting it.

Sources: Scales, Norms, and Equivalent Scores; A Consumer’s Guide to Tests in Print

When is the comparison group fair?

A percentile is useful only for the comparison its reference group can support. The joint testing standards say norm-referenced interpretations depend partly on whether the reference group is appropriate to the intended inference, clearly described, and accurately represented. If a sample is used, the standards call for technically sound sampling of sufficient size; they also note that norms can lose usefulness over time and may need review. The relevant group depends on the question, rather than on a universal default.

For example, a report might compare a person with people of a similar age because age is relevant to its interpretation. That does not automatically make the comparison right for every other purpose. A group used to describe one population may not support a conclusion about another. A percentile without the group’s description, test edition, and norming information leaves out the facts needed to judge the comparison.

Raw scores also have limits. The ETS report describes scaling, norming, equating, and calibration as separate tools for giving scores meaning and comparability. The joint standards explain that equating can place raw scores from alternate forms on a common scale when those forms were built to the same content and statistical specifications and administered under the same conditions. Matching item counts alone is not enough: two forms can differ in difficulty, content, scoring, and administration. A raw score on one quiz should not be compared directly with the same number on another simply because both have, for example, the same number of questions.

Before treating a percentile as evidence, look for the assessment name and form, the reference group and its date, and a clear description of how scores were converted. For a raw result, first identify what was counted and whether items carried equal or different weight. These details determine whether the comparison answers your question. Even when the comparison is well supported, it describes relative performance on that assessment, not every ability or future outcome.

Sources: Standards for Educational and Psychological Testing (2014); Scales, Norms, and Equivalent Scores; Minnesota Statewide Testing Program Glossary

What should you do with an online result?

Test IQ Free’s 50-question adaptive reasoning test covers pattern reasoning, number patterns, and spatial reasoning. Its free result includes domain feedback and a provisional estimated IQ range. The publication brief states that the conversion uses placeholder parameters and has no validated adult reference population; the displayed band is not a confidence interval. The optional $9 report uses that same conversion for its point estimate and percentile. It can explain the result, but it does not add validation or accuracy. These are product facts from the brief, not findings established by the sources cited above.

That makes the test most useful as a low-stakes practice session: try the original questions, review the worked answers, and choose a reasoning operation to practice. Treat the result as a record of performance on those tasks under those conditions. It cannot establish a change in general intelligence. If an educational, clinical, or employment decision has material consequences, ask a qualified professional about an assessment suited to that decision. Reviewing an explanation can show which step or rule to revisit on a similar task; it does not supply the comparison data needed to interpret a percentile. Keep practice feedback and population comparisons as separate uses of a result. It is useful feedback about this particular practice session.

Sources: Standards for Educational and Psychological Testing (2014)

Questions readers ask