Whether the instrument measures what it claims to.
9 topics · 15 curated works
No prior grounding assumed.
Comparison of Classical Test Theory and Item Response Theory and Their Applications to Test Development
Ronald K. Hambleton & Russell W. Jones · 1993
Contrasts classical test theory, where a score's meaning depends on the particular sample and items used, with item response theory, which places a…
+2 more at this level
Assumes you know the vocabulary.
Survey Research
Jon A. Krosnick · 1999
Reviews how question wording, order and response-scale design measurably change survey answers, and argues satisficing - giving a plausible-enough…
+5 more at this level
Primary sources and full treatments.
The Vectors of Mind
L. L. Thurstone · 1934
Recasts mental abilities as vectors in a multi-dimensional space and develops the multiple-factor-analysis methods to locate them, an alternative to…
+5 more at this level
12 of 15 works
Ronald K. Hambleton & Russell W. Jones
Contrasts classical test theory, where a score's meaning depends on the particular sample and items used, with item response theory, which places a person's ability and an item's difficulty on the same scale independent of which items were given.
link checked 17 Sept 2026Taciano L. Milfont & Ronald Fischer
A tutorial on testing whether a scale measures the same construct the same way across groups, walking through configural, metric and scalar invariance and what each licenses a researcher to compare.
Godfred O. Boateng et al.
Lays out a step-by-step protocol - item generation, expert review, pilot testing, factor analysis - for building a psychometric scale, aimed at researchers outside psychometrics who need to construct a defensible measure.
Jon A. Krosnick
Reviews how question wording, order and response-scale design measurably change survey answers, and argues satisficing - giving a plausible-enough answer rather than the most accurate one - explains many of these effects.
link checked 17 Sept 2026Norbert Schwarz
Demonstrates that survey answers are constructed on the spot from the wording, order and response format of a question, so trivial changes in how a question is asked can reverse what a survey appears to find.
Frank B. Baker
A worked introduction to item response theory - the one, two and three-parameter logistic models - showing how item difficulty and discrimination can be estimated on the same scale as a respondent's ability, independent of which items were used.
Mohsen Tavakol and Reg Dennick
Argues that alpha measures the interrelatedness of items rather than unidimensionality, so a high value can mask a multidimensional scale and is routinely over-interpreted as proof of reliability.
link checked 17 Sept 2026Daniel Lakens
Walks through how to calculate and correctly interpret standardised effect sizes for the most common psychology study designs, arguing that reporting a p-value alone hides how large or small the underlying effect actually is.
link checked 17 Sept 2026Lee Anna Clark & David Watson
Revisits their own 1995 guide to argue test construction should start from a clear theoretical account of the construct being measured, updating the practical steps - item pool generation, item selection, validation - for what has changed in the field since.
link checked 17 Sept 2026L. L. Thurstone
Recasts mental abilities as vectors in a multi-dimensional space and develops the multiple-factor-analysis methods to locate them, an alternative to Spearman's single general factor that argued for several independent primary mental abilities instead.
link checked 17 Sept 2026Lee J. Cronbach
Derives a single coefficient, alpha, that estimates a test's internal-consistency reliability from one administration, generalising and unifying earlier split-half reliability formulas into the measure now reported for almost every psychological scale.
link checked 17 Sept 2026Lee J. Cronbach & Paul E. Meehl
Defines construct validity as the case where a test cannot be validated against any single external criterion because the thing it measures - like anxiety or intelligence - is itself a theoretical construct, and lays out how a network of predicted relationships can validate it instead.
link checked 17 Sept 2026