What does test-retest reliability assess?

What does test-retest reliability assess?

• A. The breadth of content coverage.
• B. Stability of results over time.
• C. Agreement among different raters.
• D. The extent to which test results relate to a criterion measured outside the test.

Answer: (B)

The main idea is how consistently a test measures something across time. Test-retest reliability looks at whether the same people would get similar scores if you administer the exact same test again later under similar conditions. You give the test once, wait a while, then give it again and compare the two sets of scores. If the scores line up closely, the test is stable over time, indicating good test-retest reliability. If they don’t line up, there may be measurement error or actual changes in what’s being measured between administrations. The time interval matters: too short an interval can inflate stability because people remember items or pathways to answers, while too long a interval allows real changes in the trait or situation to alter performance. This type of reliability is different from asking whether the test covers enough content (that’s about breadth of coverage), or whether different raters would score the same observation the same way (inter-rater reliability), or whether the test relates to an external criterion (validity, such as predictive or concurrent validity).

The main idea is how consistently a test measures something across time. Test-retest reliability looks at whether the same people would get similar scores if you administer the exact same test again later under similar conditions. You give the test once, wait a while, then give it again and compare the two sets of scores. If the scores line up closely, the test is stable over time, indicating good test-retest reliability. If they don’t line up, there may be measurement error or actual changes in what’s being measured between administrations.

The time interval matters: too short an interval can inflate stability because people remember items or pathways to answers, while too long a interval allows real changes in the trait or situation to alter performance. This type of reliability is different from asking whether the test covers enough content (that’s about breadth of coverage), or whether different raters would score the same observation the same way (inter-rater reliability), or whether the test relates to an external criterion (validity, such as predictive or concurrent validity).