Control Groups

Returning to the relationship between television viewing habits and aggressive
behavior in children, suppose that correlational evidence indicates that
high levels of the former variable predict high levels of the latter. Now the
researcher wants to test the hypothesis that there is a cause-effect relationship
between the two variables. She decides to manipulate exposure to television
violence (the independent variable) to see what effect might be produced
in the aggressiveness of her subjects (the dependent variable). She
might choose two levels of the independent variable and have twenty children
watch fifteen minutes of a violent detective show, while another twenty
children are subjected to thirty minutes of the same show.
If an objective rating of playground aggressiveness later reveals more hostility
in the thirty-minute group than in the fifteen-minute group, she still
cannot be confident that higher levels of television violence cause higher
levels of aggressive behavior. More information is needed, especially with regard
to issues of control. To begin with, how does the researcher know that it
is the violent content of the program that is promoting aggressiveness? Perhaps it is the case that the more time they spend watching television, regardless
of subject matter, the more aggressive children become.
This study needs a control group: a group of subjects identical to the experimental
subjects with the exception that they do not experience the independent
variable. In fact, two control groups might be employed, one
that watches fifteen minutes and another that watches thirty minutes of nonviolent
programming. The control groups serve as a basis against which the
behavior of the experimental groups can be compared. If it is found that the
two control groups aggress to the same extent, and to a lesser extent than
the experimental groups, the researcher can be more confident that violent
programming promotes relatively higher levels of aggressiveness.
The experimenter also needs to be sure that the children in the thirtyminute
experimental group were not naturally more aggressive to begin
with. One need not be too concerned with this possibility if one randomly
assigns subjects to the experimental and control groups. There are certainly
individual differences among subjects in factors, such as personality and intelligence,
but with random assignment (a technique for creating groups of
subjects across which individual differences will be evenly dispersed) one
can be reasonably sure that those individual differences are evenly dispersed
among the experimental and control groups.