Extraneous Variables, Control of

For many researchers, one of the hallmarks of scientific discovery is the establishment of causal relationships; that is, identifying consistent and robust associations between one or more independent variables (IVs), or the cause of an observed outcome, and a target dependent variable (DV), or the effect or observed outcome itself. The ability to describe, explain, predict, and control such effects is a core goal of any social scientific research paradigm.

One of the most difficult tasks in establishing causal relationships, however, is isolating the expected IV–DV relationship from extraneous (or unexpected) [Page 484]relationships that are not core to one’s research program. An example of such a scenario might be a researcher looking to understand how violent television content (IV) might cause an individual to be physically and verbally hostile toward other individuals in a room (DV), without considering a preexisting antagonistic relationship between the people in the room (extraneous variable). At best, a failure to consider and control for the influence of extraneous variables can either mask or inflate expected causal relationships; at worst, ignoring such variables can cause researchers to accept erroneous causal associations or miss important relationships.

By its nature, the practice of research design—in particular, experimental research design—is about controlling variance, or identifying and removing variables not key to one’s research that might have an unexpected impact. A common saying among social scientists is MAXMINCON—a portmanteau referring to the tripartite goals of maximizing systematic (desired) variance, minimizing error (random) variance, and controlling for extraneous (or unexpected) variance. Three common strategies for controlling extraneous variance are outlined in this entry, followed by a discussion of the differences between extraneous and integral variance.