A confound is an extraneous variable that is beyond the control of the experimenter. Confounds affect the cause-and-effect relationship between the independent and dependent variables. In this way, they make it difficult for the researcher to determine the effects of the independent variable on the dependent ones. According to Brennan, in experiments that involve human subjects, it is difficult to control or eliminate confounding due to human variability.
An example is an experiment investigating the effect of a stress-reduction intervention on depression levels over a period of six months. This experimental study has three possible confounds, namely, history, maturation, and testing effects. The history effect relates to unrelated factors that reduce or increase stress levels during the course of the study, e.g., exams. The maturation effect relates to the cognitive growth that affects the dependent variable during the study period. On the other hand, repeated measures over this period lead to testing effects, whereby participants get used to a particular test, affecting their performance.
The study design can be altered to minimize the confounding effect in a study. These include matching, restriction, and randomization. Restriction entails limiting sampling to ensure that only subjects with similar confounds are recruited. This approach is effective and easy to undertake. However, it reduces generalizability and sample size and can cause residual confounding. Brennan writes that, in matching, the researcher matches the study groups to ensure that they are equivalent with respect to confounders. This strategy is useful in case-control studies that involve multifaceted variables. However, it reduces the sample size and is costly.
The third method is randomization, where the subjects are randomly selected into any of the treatment groups. It facilitates comparability of the different groups through baseline data (controls). However, randomization cannot control multiple confounders at the same time.