Whether a research method measures what it claims to measure. About accuracy, not precision. A method can be reliable (consistent) but not valid (accurate) if it consistently measures the wrong thing.
Definition: Whether a research method measures what it claims to measure. About accuracy, not precision. A method can be reliable (consistent) but not valid (accurate) if it consistently measures the wrong thing.
Validity refers to whether a research method measures what it claims to measure. It is about accuracy: are your results a true reflection of the underlying phenomenon you are trying to understand?
Validity is not a single property. It is a family of related concepts, each addressing a different way your research can go wrong. They fall into two broad categories: validity of your study design, and validity of your measurements.
Internal validity: whether your conclusions about cause and effect hold within the study. If you claim X caused Y, are you sure it was not a confounding variable, a seasonal effect, or something else entirely?
External validity: whether your findings generalize beyond the study. Can you apply conclusions from your specific sample, setting, and timing to the broader population you care about?
Ecological validity: whether your study conditions reflect real-world use. A lab test with a facilitator watching is not how people use your product on a crowded train with one hand. A specific form of external validity.
Statistical conclusion validity: whether the statistical relationships you found are real — or artifacts of low power, violated assumptions, or multiple testing.
Construct validity: whether you are measuring the theoretical concept you intend to measure. The deepest form of measurement validity, requiring both theoretical clarity and empirical evidence.
Convergent and discriminant validity: the two sides of construct validity in practice. Your measure should correlate with other measures of the same construct (convergent) and not correlate too highly with measures of different constructs (discriminant).
Content validity: whether your instrument covers the full scope of the construct. Established through expert review, not statistics.
Criterion validity: whether your measure predicts or correlates with real-world outcomes. Comes as predictive validity (future outcomes) and concurrent validity (current criteria).
Face validity: whether the measure looks like it measures what it claims, on the surface. The weakest form — important for participant buy-in but proves nothing about actual measurement quality.
A method can be reliable without being valid — consistent but consistently wrong. However, a method cannot be valid without being reliable. Reliability is necessary but not sufficient for validity.
For a detailed breakdown of each type with practical UX examples, see Types of Validity in UX and Market Research.
The consistency of a research method—whether it produces similar results when repeated under the same conditions. About precision, not accuracy. A method can be reliable without being valid.
The degree to which research findings are independent of who conducts the study. If two researchers follow the same protocol and get different results, you have an objectivity problem.
Consistent, predictable bias that skews results in a known direction. Manageable because you can account for it in interpretation—far better than random, unsystematic error.
Systematic deviation from the true value in research findings. Cannot be eliminated, only managed through standardization and awareness. The goal is systematic bias (manageable) over unsystematic bias (chaos).
The science of measuring psychological constructs—attitudes, abilities, personality traits—through standardized instruments. The discipline behind every validated questionnaire in UX research.
This term is referenced in the following articles:
Standardized measurement instruments provide benchmarks and comparability. But using them effectively requires understanding what each one actually measures, and what it does not.
You will always introduce bias into your research, that is unavoidable. The goal is not elimination but management. Understanding the difference between systematic and unsystematic error is what makes findings trustworthy.
Good research does not happen by accident. The research plan is the single most important tool for avoiding unfocused, low-impact research, and for ensuring your work drives real decisions.
Translating a UI is easy; translating an experience is hard. How to use back-translation and local partners to avoid cultural blind spots.