Reliability is the consistency of a data collection method: repeating it under similar conditions should produce similar results. Validity is whether the method actually measures what the investigation intends to measure.
This distinction is HL-only content from AHL 4.12: Data collection, reliability and validity tests in IB Mathematics: Applications and Interpretation.
How the Difference Works
A method can be reliable without being valid. Bathroom scales that always add three kilograms give consistent readings, but they do not validly measure actual mass. Poor reliability also limits validity because inconsistent measurements cannot accurately represent the intended variable; however, consistency alone does not establish accuracy or relevance.
| Concept | Core question | Example or improvement |
|---|---|---|
| Reliability | Would repetition produce similar results? | Standardize instructions, wording, equipment, and conditions. |
| Validity | Does the method measure the intended variable? | Replace school attendance as a measure of academic ability with a direct performance measure. |
| Test-retest reliability | Are results consistent over time? | Give the same questionnaire to the same group twice and compare results. |
| Parallel-forms reliability | Do equivalent versions produce similar results? | Compare outcomes from two questionnaires designed to measure the same characteristic. |
| Content validity | Does the method cover the complete concept? | A mathematics test containing only algebra does not assess overall mathematical achievement. |
| Criterion-related validity | Does the result agree with an accepted external measure? | Compare a new fitness assessment with an established fitness test. |
A common misconception is that reliable data must be valid. This is false: a biased or incorrectly calibrated instrument can produce highly consistent but systematically incorrect results.
Exam Technique
In an HL response, do not merely label a method “unreliable” or “invalid.” Identify the specific weakness, explain whether it affects consistency or measurement of the intended variable, and suggest a targeted improvement. This can be tested in Paper 1 short-response questions or within a Paper 2 investigation.
