Test-retest reliability refers to the consistency of a measurement over time. It assesses whether a test or measurement tool produces similar results when administered to the same people on two or more occasions, assuming the underlying trait being measured hasn’t changed. In short, does the same instrument used on the same person to measure the same trait produce the same (or very similar) result?
This form of reliability is especially important for tests that aim to measure stable characteristics such as intelligence, personality traits, or reading ability. High test-retest reliability indicates that the test yields repeatable and dependable results across time.
To assess test-retest reliability, researchers typically:
- Administer the same test to the same people at two different times.
- Calculate the correlation coefficient or Cronbach’s alpha between the two sets of scores (often using Pearson’s r or an intraclass correlation coefficient).
Values range from –1 to +1:
- A value near +1 indicates high reliability—scores are consistent over time.
- A value near 0 indicates low reliability, suggesting instability or inconsistency.
- A negative value may indicate a systematic problem with the test or the measurement process.
For example, a psychology researcher gives a personality inventory to a group of participants in January and again in February. If the test-retest reliability is 0.88, it suggests the test produces highly consistent results over time and is likely measuring a stable personality trait.
« Back to Glossary Index