Robert Wherry Speaker Series

2025: Daniel J. Bauer
Rethinking moderation: Beyond Bilinear Interactions 

Psychology is deeply infatuated with moderation effects. Conventionally, moderation is said to exist when the relation between a predictor and an outcome depends in magnitude or direction on the value of a third variable, the moderator. Almost invariably, moderation hypotheses are evaluated through the specification and testing of product-interaction models, for instance a linear regression model in which y is regressed on x, z, and xz. The partialed product represents the interaction of x and z, testing the hypothesis that the effect of x depends on z (and vice versa). More specifically, the model implies the conditional effect of x on y to be a linear function of z, a form referred to as a bilinearinteraction. Over time, fitting such models has become deeply engrained within the field, used seemingly without reflection whenever moderation is hypothesized. The purpose of this talk is to interrogate this practice. From whence did it come? Why are we so enamored of it? Has our habitual use of the product-interaction model blinded us to other possibilities? Can we expand our conceptualization of moderation to articulate and test a richer array of research hypotheses?