Quick answer
IQ test statistics describe different parts of measurement. A raw score records performance, norms provide a comparison, and measurement-error estimates describe uncertainty. No single statistic establishes that a test is appropriate for every purpose. Start by identifying the score type and reference group before interpreting how high or low a number appears.
Key takeaways
- A percentile is not the percentage of questions answered correctly.
- Reliability and evidence for an interpretation answer different questions.
- An interval is part of the report’s method, not a universal margin to invent.
1. Separate the raw result from the reported scale
Twelve correct answers is a raw count. Turning it into a standard score requires the assessment’s scoring procedure. A percentile describes relative standing in a reference group; it does not say how many questions were correct.
The APA describes deviation IQ as a standard-score approach with a conventional mean of 100 and a test-specific standard deviation. Check the actual scale rather than assuming every test uses identical units.
Sources: APA Dictionary: deviation IQ
2. Ask what each quality statistic is answering
A consistency statistic cannot by itself establish that a score supports the decision you want to make. Look for evidence about the proposed use, relevant examinees and conditions, as well as precision. Avoid treating a single impressive coefficient as a complete quality certificate.
For an illustrative check, ask whether a report intended to describe a specific skill is being used to claim a much broader ability. The extra claim needs its own support.
3. Read the reference information before comparing results
Find who was included in the norming sample, when it was collected and which age or other comparison applies. A score’s meaning depends on that reference. Different tests can report similar-looking numbers without supporting a direct comparison.
The table distinguishes the quantities a reader might encounter. It is a glossary for interpretation, not a scoring manual for converting raw answers.
| Quantity | What it describes | Common confusion |
|---|---|---|
| Raw score | Performance counted or scored by the test rules | Automatically an IQ |
| Standard score | A position on the assessment’s defined scale | The same units across all tests |
| Percentile | Relative standing in the specified group | Percentage of questions correct |
| Measurement error | Uncertainty associated with measurement | A fixed margin valid for every score |
| Confidence interval | A range constructed by a stated method | Proof that repeated scores must be identical |
4. A statistical model is not the same as an observation
For a hypothetical scale with mean 100 and standard deviation 15, a score of 115 is one standard deviation above the mean: (115 − 100) ÷ 15 = 1. This arithmetic does not tell us whether the test is valid, whether its norms fit or whether the input score is genuine.
Likewise, drawing a smooth bell curve is a modelling choice, not proof that every observed group follows it exactly. Use the published assessment documentation for actual score interpretation.
5. Read uncertainty in the report’s own terms
The testing standards explain that standard errors of measurement can inform confidence intervals and may differ across score levels. Pearson’s WAIS-IV reporting documentation illustrates that reports can include standard scores, percentile ranks and confidence intervals as separate fields.
Do not add a generic plus-or-minus-five band to every result or treat a small difference between tests as proven change. Ask which interval method and confidence level were used, and whether the comparison itself is supported.
Sources: AERA, APA and NCME: measurement error and confidence intervals · Pearson: WAIS-IV report fields
Frequently asked questions
Is a percentile the same as percent correct?
No. Percent correct concerns answers on a task; a percentile concerns standing relative to a specified reference group.
Does high reliability prove that a test is valid?
A consistency statistic alone does not establish evidence for every intended interpretation or use. Examine the particular purpose and population.
Should I calculate my own confidence interval?
Use the instrument’s documented reporting method or ask the assessor. A generic margin applied to any online score can create false precision.
Sources and next steps
Updated 19 October 2026. Sources support the distinctions explained here; they do not endorse CertifiedIQ.