This paper makes some basic distinctions, that are essential for the correct use and understanding of the term `interaction' in studies of the effects of drugs on driving. The first and most important distinction is between the descriptive and explanatory use of `interaction'; most of the confusion in the literature is due to misunderstanding of this distinction. Experimental designs and statistical evaluations to prove interactions are only descriptive and explain nothing, even if they are tested by statistical inference. It is necessary here to distinguish between descriptive interactions concerning factors and observables. Thus `multivariate' can mean `multifactorial', as used in analysis of variance (ANOVA), or `multiobservations', as used in multivariate analysis of variance (MANOVA). Explanatory interaction must avoid confusing correlative and causal descriptions. The corresponding models for causal interactions in pharmacology are called `coergisms'. There is often confusion between the three ways of approaching coergisms: (1) evaluative, where it is crucial to distinguish (desired) main effects of a compound and (undesired) side effects; (2) descriptive, about the direction of change of a main effect after introducing an additional condition; (3) explanatory, about a coergism's possible causes.
Abstract