Mediating and Moderating Variables Explained

What is the difference between a mediator and a moderator? One of my former academic advisors used to always say “be a walking laboratory”. I think it’s a very poetic way of describing a core feature of psychological research—to come up with theories or explanations for various phenomena we observe. Sometimes there isn’t a clear-cut relation between a dependent and independent variable. In those cases, a mediating variable or a moderating variable can provide a more illustrative account of how dependent (criterion) variables are related to independent (predictor) variables.

A mediating variable explains the relation between the independent (predictor) and the dependent (criterion) variable. It explains how or why there is a relation between two variables. A mediator can be a potential mechanism by which an independent variable can produce changes on a dependent variable. When you fully account for the effect of the mediator, the relation between independent and dependent variables may go away. For instance, imagine that you find a positive association between note-taking and performance on an exam. This association may be explained by number of hours studying, which would be the mediating variable. 

A moderator is a variable that affects the strength of the relation between the predictor and criterion variable. Moderators specify when a relation will hold. It can be qualitative (e.g., sex, race, class…) or quantitative (e.g., drug dosage or level of reward). Moderating variable are typically an interaction term in statistical models. For instance, imagine researchers are evaluating the effects of a new cholesterol drug. The researchers vary the participants in minutes of daily exercise (predictor/independent variable) and measure their cholesterol levels after 30 days (criterion/dependent variable). They find that at low drug doses, there is a small association between exercise and cholesterol levels, but at high drug doses, there is a huge association between exercise and cholesterol levels. Drug dosage moderates the association between exercise and cholesterol levels. 

Let’s look at some examples in psychological research.

A recent paper by Frank, Amso, & Johnson (2014) examined the developmental relationship between early perceptual abilities and face perception in infancy.  In the study, the authors tested visual search abilities of 3-, 6-, and 9-month-old infants. Infants were shown panels of red rods against a black background. One of the rods was either slanted at a diagonal or moved back and forth. Accuracy at looking at the slanted or moving rod was calculated as “visual search accuracy”. Infants also viewed excerpts from Charlie Brown and Sesame Street and relative amount of time spent viewing faces was measured. They found that infants looked more at faces and were more accurate at identifying a moving target with age. This effect was fully mediated by visual search accuracy for moving rods. That is, developmental improvements in visual search accuracy fully accounted for the amount of time infants looked at faces.

A great example of a moderator comes from Cohen and Willis, 1985. In that study, the authors proposed a stress-buffering hypothesis. Prior research had suggested a main effect of social support on quality of life. However, Cohen and Willis demonstrated that the relation between social support and quality of life depends on an individual’s stress level. Someone who experiences a lot of stress, but has good social support, will show better outcomes (fewer symptoms of depression, anxiety, fatigue...) than someone with low social support. Social support is the moderating variable. 

These examples should clarify the difference between mediating and moderating variables. Both types of variables provide interesting explanatory means to describe psychological phenomena.