What is Construct validity?

Construct validity is the evidence that a test measures the psychological attribute it claims to measure, rather than something merely correlated with it. It is built from convergent evidence — the score tracks other measures of the same attribute — and discriminant evidence, that it does not track measures of different attributes.

Why it is the hard one

Nobody can observe conscientiousness, numerical reasoning or resilience directly. A test is a set of observable behaviours claimed to stand in for an unobservable attribute, and construct validity is the argument that the substitution is sound.

That argument is never settled by a single study. It accumulates: factor structure that reproduces across samples, correlations with related measures that are high, correlations with unrelated measures that are low, and score differences between groups that theory predicted in advance.

How to interrogate a vendor's claim

Ask what the construct is defined as, in one sentence, before asking anything statistical. A great many instruments cannot answer this cleanly, and a construct that cannot be stated cannot be validated.

Then ask for the discriminant evidence specifically. Convergent evidence is easy to produce and is what gets published in marketing material; discriminant evidence — showing the test does not simply measure verbal ability, or reading speed, or willingness to present well — is the part that distinguishes a measured instrument from a plausible questionnaire.

Cronbach and Meehl, in their own words

The term was defined in 1955, in a paper written for the APA’s own test standards committee, and its definition is narrower and more useful than the one that circulates: “Construct validation is involved whenever a test is to be interpreted as a measure of some attribute or quality which is not ‘operationally defined.’” The trigger, in the same passage, is stated as a condition rather than a preference — “Construct validity must be investigated whenever no criterion or universe of content is accepted as entirely adequate to define the quality to be measured.”

So construct validity is not a stronger grade of evidence to aspire to. It is what you are obliged to produce when there is no agreed criterion and no agreed content domain to appeal to instead, which is the situation for almost every personality and ability construct sold commercially.

The paper also supplies the standard that most vendor construct claims fail. A construct, it argues, has to sit inside a theory: “We shall refer to the interlocking system of laws which constitute a theory as a nomological network”, and “a necessary condition for a construct to be scientifically admissible is that it occur in a nomological net, at least some of whose laws involve observables.” A trait that appears in one publisher’s model, relates to nothing outside it, and predicts no observable is not yet a construct in this sense — it is a label on a scale.

The question this turns into for a buyer

Ask what else the construct is supposed to relate to, and in which direction, before you are shown a single correlation. A vendor with a nomological net can say: this should correlate positively with that, negatively with the other, and be unrelated to reading speed — and then show whether it did. A vendor without one can only show that the scale holds together internally, which is a statement about the items rather than about the attribute.

The 1955 paper is freely readable in full at the Classics in the History of Psychology archive, so this is a claim you can check rather than take on trust — including the claim being made here about what it says.

Not the same as face validity

Face validity is whether the test looks right to the person taking it. Construct validity is whether it measures what it says. The two are independent, and a test can have plenty of the first and none of the second.

Face validity

Sources

More terms beginning with C

Criterion validity Cronbach's alpha Cut score

Check a selection process against the four-fifths rule

Free, no signup, computed in your browser — with the remedy, not just the verdict.

Open the calculator

Last reviewed 2026-09-17