Construct

In language testing, a construct is the specific ability, knowledge, or competence that a test claims to measure. It is an abstract theoretical entity — not directly observable — that must be operationalised through test tasks and scoring criteria. Construct definition is the foundation of all validity arguments.

Construct Validity

A test has construct validity to the extent that test scores can be interpreted as meaningful indicators of the target construct. Messick (1989) unified all validity types under the construct validity umbrella, arguing that validity is a property of score interpretations, not of the test itself.

Two primary threats:

Construct-irrelevant variance (CIV) — Test scores are influenced by factors extraneous to the construct. Two subtypes:

• Construct-irrelevant difficulty — Features that make the test harder for reasons unrelated to the construct (e.g., culturally unfamiliar topics in a reading test, time pressure penalising slow writers)
• Construct-irrelevant easiness — Features that inflate scores (e.g., test-wise strategies, predictable item formats)

Construct underrepresentation (CU) — The test is too narrow and fails to sample the full breadth of the construct. A writing test that only elicits argumentative essays underrepresents "writing ability." A speaking test with no interaction underrepresents "speaking ability."

Construct in Practice

• Rating scales operationalise the construct through descriptors — each criterion should map to a dimension of the construct
• Test specifications should explicitly state the construct definition before item writing begins
• Washback is shaped by the construct — a well-defined construct produces teaching that develops the target ability
• IELTS defines its writing construct through four criteria: Task Achievement/Response, Coherence & Cohesion, Lexical Resource, Grammatical Range & Accuracy

Key Scholars

• Messick (1989) — unified validity theory centred on construct validity
• Bachman (1990) — communicative language ability as construct for language tests
• Bachman & Palmer (1996) — test usefulness framework integrating construct with other qualities