A formal experiment has several fixed parts, regardless of what the experiment is about. First the experimenter must create a hypothesis, which is a testable theory from which the experiment is derived. Then an experimental population is chosen. This could be people or other living things, objects, or even atoms or molecules; the key is that no matter what the experimental population is composed of, is should be a reasonably large, uniform sample, and should be randomly divided into as many groups as there are levels of treatment, plus one for the control.
The treatment is the single variable or factor that the experimenter wishes to test. The control is a group that is not subject to the treatment. The control group is considered "normal" and is used as a baseline to compare the experimental group(s) to. It is important that only the factor being tested is different from group to group, and everything else is held constant.
Some treatments are done once, some are done in increments. For example, an experimenter who wanted to examine the effect of vitamin C supplementation on the mouse immune system could simply do a single treatment, in which case there would be a group of mice who were fed a standard diet and a similar test group who were fed the same diet plus a vitamin C supplement. Alternately, the experiment could include a control and several test groups, with each test group getting an incrementally larger supplementation of vitamin C.
As the experiment proceeds, the same data is gathered by the same methods and at the same frequency in each group. The data is then compared across groups by whatever statistical analysis is appropriate, and conclusions regarding the correctness of the hypothesis are drawn.