Psychometrics in experimental psychology: A case for calibration

Abstract

Psychometrics is historically grounded in the study of individual differences. Consequently, common metrics such as quantitative validity and reliability require between-person variance in a psychological variable to be meaningful. Experimental psychology, in contrast, deals with variance between treatments, and experiments often strive to minimise within-group person variance. In this article, I ask whether and how psychometric evaluation can be performed in experimental psychology. A commonly used strategy is to harness between-person variance in the treatment effect. Using simulated data, I show that this approach can be misleading when between-person variance is low, and in the face of methods variance. I argue that this situation is common in experimental psychology, because low between-person variance is desirable, and because methods variance is no more problematic in experimental settings than any other source of between-person variance. By relating validity and reliability with the corresponding concepts in measurement science outside psychology, I show how experiment-based calibration can serve to compare the psychometric quality of different measurement methods in experimental psychology.