School of Psychology

Confusion about the methods

Confusion seems to exist about precisely what each of these two techniques do and what they do not do. There are a number of reasons for this.

First, since they have similar sounding names, most people assume that they are related and similar, and possibly derive from the same source. It should be understood that they do not derive from the same source, and their similar sounding names come from lay language terms that are often confused with each other (e.g., does one 'mediate' or 'moderate' a meeting?). Further, these common-use terms do not map precisely upon the meanings intended in the statistical context.

Second, as noted above, statistics textbooks typically do not do a very good job of explaining these two approaches. In my experience, I have found very few that discuss these two techniques together, thereby drawing out their similarities and differences.

Third, reports of moderation and mediation in the empirical literature are not always clear. It may be ambiguous what the researchers did in performing their particular test. Rarely do researchers perform both moderation and mediation on the same dataset, because many researchers believe (wrongly) that these methods are exclusive of each other, so examples of this type of work are rare.

Fourth, both are special cases of two different statistical approaches (moderation is a special type of ANOVA interaction, and mediation is a special type of path model), and therefore they do not receive as much attention and coverage as mainstream statistical approaches.

Several articles have appeared over the last decade or so to try to rectify the misperceptions and confusions surrounding these techniques. Baron and Kenny (1986) wrote an article over two decades ago from the perspective of social psychological research, and it stands as the seminal article that researchers use to try to disambiguate these two techniques. Subsequently, excellent work by Holmbeck (1997; 2002) has extended these views to clinical psychology. Dave Kenny's web site is recommended and a recent book by David McKinnon (2008) on mediation and the Research in Prevention Laboratory web-site are also very helpful. Andrew Hayes has just published a book on mediation, moderation, and conditional process analysis (2013). And I have just published a book named Doing mediation and moderation (Jose, 2013) that is a good introduction to these two topics. I strongly recommend reading these works and examining these web-sites because they present considerable context for the use of these techniques, and they point out common misunderstandings and pitfalls to which researchers may fall prey.