Defining variables in relationship to other variables: When interactions suddenly turn out to be main effects

It is frequently the case that independent variables in experimental designs in social psychology are defined in relationship to levels of other independent variables. For instance, an experimental design in social cognition research might examine the effects of two different expectations on memory for information that is either consistent with the first expectation (and inconsistent with the second) or consistent with the second expectation. When the information factor is defined in this way, its interaction with expectation is perfectly confounded with the main effect of the stimulus factor defined in an absolute rather than a relative manner. Because we are trained to interpret interactions as qualifications of main effects, this alternative, and frequently more parsimonious, interpretation may be ignored. We illustrate the general issue and then review literature where interpretational ambiguities have resulted. (C) 2000 Academic Press.