LAUSR

Abstract Measurement concepts and vocabulary have become commonplace in the human sciences. To this end, the field of educational measurement has largely concerned itself with the development and use of a family of mathematical, “psychometric” models collectively known as latent variable models. These models contain parameters that are commonly interpreted as referring to properties of individuals and test items, and estimates of these parameters are accordingly interpreted as measured values of these properties. Such interpretations are based on the erroneous assumption that the correspondence between these parameters and properties has been substantiated, whereas the educational measurement literature and discourse frequently conflates or confuses the two; different instantiations of a pervasive representational fallacy. This fallacy drives an unscientific approach to modelling where the nature of educational phenomena and their causal role in measurement is ignored. To advance beyond the mythology of “latent variables”, the human sciences need to eradicate this fallacy and establish scientific models for measurement.