Treatments, Controls, and Placebos
Treatment and Control Groups
In some experiments, the independent variable is whether or not subjects receive some treatment intended to produce some benefit for the subject. The treatment could be a medical treatment, psychotherapy, or almost anything else. For example, it could be whether or not college students take a weekly quiz. In these experiments, the group that gets the treatment is referred to as the treatment group or the experimental group, and t he group that does not get the treatment is referred to as the control group.
What is the Point of the Control Group?
A claim about whether or not a treatment works is a claim about a statistical relationship between two variables. One (the independent variable) is whether or not subjects receive the treatment, and the other (the dependent variable) is some other is something that the treatment is supposed to improve (e.g., depression, memory). To assess such a relationship, you have to compare the average score on the outcome variable across levels of the treatment variable. That is, you have to compare those who got the treatment to those who did not.
Another way to look at this is that you always have to worry about whether subjects who did not receive the treatment might turned out exactly like subjects who did. The only way you can know for sure is to include a control group.
For example, imagine that a college professor gives a group of students a weekly quiz, and he finds that on average their final course grade is a B+. Does this constitute evidence that taking the quizzes had a positive effect on students’ performance in the course? No way! You would also have to know the average grade of a group of students who did not take a weekly quiz. Maybe it is such an easy course that students who did not take a weekly quiz would also have averaged a B+. To show that the quizzes worked, then, you would have to show that those who took them earned higher grades on average than those who did not take them.
Does There Have to Be a Control Group?
Sometimes students get confused and think that every experiment must include a control group. It is true that every experiment must include a comparison across levels of the independent variable, but it is not true that one of the conditions must be a control group. Consider, for example, an experiment on the effects of noise level on concentration. The experimenter might test people under a quiet condition and under a noisy condition, but she does not necessarily have to test them under a condition with absolutely no noise at all. (It is not even clear that this would be possible.) Similarly, a college professor might compare the grades of students who have taken difficult weekly quizzes with the grades of students who have taken easy weekly quizzes. But he does not necessarily have to compare these to another group that takes no quizzes at all. In both cases, the control condition might be interesting, but it is not necessary to draw conclusions about a causal relationship between the variables.
Placebos and Placebo-Control Conditions
A placebo is a fake treatment that appears to be a real treatment from the subject’s perspective. For example, in studies of drug effectiveness, some subjects might be given a placebo that is nothing more than a sugar pill. In studies of psychotherapy effectiveness, some subjects might come in and talk to a therapist for an hour even though the therapist does not use the technique of interest.
In many experiments, subjects in the control condition are given a placebo. This is sometimes called a placebo control condition to distinguish it clearly from a no-treatment control condition (i.e., the kind of control condition discussed above).
The reason for doing this is as follows. Subjects in a treatment condition are often expecting an effect of the treatment, whereas subjects in a no-treatment control condition are not. That is, subjects’ expectations are a potential confounding variable. I f there is a difference between treatment and no-treatment control conditions, it could be caused by differences in subjects’ expectations rather than by the treatment itself. This is called a placebo effect. But note that if subjects in the control group receive a placebo, then they have the same expectations about improvement as subjects in the treatment group. Their expectations have been controlled so that now any difference between the two groups must be due to the treatment itself and not to subjects’ expectations.
Many experiments are designed with both a no-treatment control group and a placebo control group. Comparing these two control groups allows the researchers to see whether there is a placebo effect. Comparing the placebo control group to the treatment group allows them to see whether the treatment had an effect over and above that of the placebo.