Direct comparison
Reliability vs Validity in Research: Differences & Examples | CASRAI
Reliability is consistency of measurement — the same results repeatedly. Validity is accuracy — measuring what is intended. A measure can be reliable but invalid.
Side-by-side comparison
| Dimension | Reliability | Validity |
|---|---|---|
| What it is | Consistency of measurement across time or raters. | Accuracy — measuring what is intended. |
| Core question | "Does the measure give the same result each time?" | "Does the measure capture what it claims to?" |
| Types | Test-retest, inter-rater, internal consistency (Cronbach's α). | Content, construct (convergent, discriminant), criterion validity. |
| Related error | Random error — variability in measurement. | Systematic error — consistent mismeasurement of a construct. |
| Target analogy | Shots cluster tightly together. | Shots land near the bullseye. |
| Can have one without the other? | Reliable but invalid: tight cluster, wrong place. | Valid but unreliable: scattered around the correct target. |
| Logical relationship | Necessary but not sufficient for validity. | Requires reliability — cannot be valid without consistency. |
| How assessed | Correlation coefficients (e.g. ICC, Cronbach's α). | Expert panels, factor analysis, criterion correlations. |
Common questions
FAQ
Can a measurement be reliable but not valid?+
Yes — this is the key relationship to understand. A bathroom scale that consistently reads 2 kg too high is perfectly reliable (same result every time) but invalid (not measuring true weight). Tight consistency tells you the error is systematic rather than random, but it does not tell you whether the right thing is being measured.
What is Cronbach's alpha and what does it measure?+
Cronbach's alpha (α) is a coefficient of internal consistency used in questionnaire and scale development. It ranges from 0 to 1 and estimates how well a set of items all measure the same underlying construct. A value above 0.70 is generally considered acceptable, though the appropriate threshold varies by context. It is a measure of reliability, not validity.
What are the main types of validity?+
Content validity asks whether the measure covers the full domain of the construct. Construct validity (including convergent and discriminant validity) assesses whether the measure relates as theory predicts to other constructs. Criterion validity (concurrent and predictive) assesses whether the measure correlates with external criterion measures. Together these build a validity argument for an instrument.
Going deeper
