From Abracadabra to Zombies
In a nutshell: A control group is a group in a scientific study that is given nothing special to compare it with a group given something scientists are testing. Control group studies help scientists test a claim that one thing causes another.
Many scientists use control groups when testing new medicines. Let's say you want to find out if your creation, the Greatest Ever Medicine or GEM for short, really gets rid of poison oak rash and itch. You might give GEM to six of your friends who have had a bad camping experience and now have severe rashes because of cuddling a little dog who had been playing in some poison oak. How will you know whether GEM works? You might think that if your friends' rashes go away, then GEM did the job. But you might be wrong. Maybe the rashes would have gone away on their own if your friends had done nothing special to treat them.
A better test would be to give GEM to some people with a poison oak rash and compare them to a group of people who have the rash but who don't get any GEM. That's a start, but it's not enough for a good experiment. You want to control the groups in such a way that would rule out anything else but GEM from being the cause of getting rid of the rash. You don't want the non-GEM group to use some other medicine. Why not? Well, what if you then found that the rashes went away in everybody in both groups after two weeks? You wouldn't know whether both medicines work or whether neither medicine works. The rashes may have gone away on their own.
A better way to do this test would be to match your control group—the one not getting GEM—to your experimental group—the one getting GEM. The people in the two groups don't have to be exactly alike, but they should be matched up for things that might affect rashes. What affects rashes? Well, you have to have some medical knowledge to know that. To do a good job planning experiments you need to know about what you're studying, but some things should be obvious. You don't want to compare really sick people with really healthy people. Sick people might take much longer to heal than healthy people. Your experimental group kids might all heal in a week, while nobody in the control group heals in a week. This might make you think that GEM works, but if the control group kids were all much more sick than the experimental group to begin with, maybe GEM had nothing to do with the healing.
Scientists do two things to try to match their experimental and control groups. They make sure each group is large enough and they randomly assign people to each group. The number of people in the study (called the sample size) and randomization can make or break a study.
sample size and randomization
You shouldn't let the people in the study choose which group they will be in. Nor should you decide which group to put them in. This will lessen the chances of mismatching the groups and biasing your study. You might not mean to but you might put all the healthiest people in the experimental group. That would bias your study in favor of GEM. If the kids could choose which group to join they might bias the study in favor of the control group. Randomly putting people in either group gives everybody in the sample an equal chance of being put in either group. It is the least biased way of doing this kind of study. How this is done depends on several things that we don't have to worry about here. The main thing is that the study not be biased by having mismatched experimental and control groups.
By randomly putting people in the experimental or control group you lessen the chances of having all the really healthy people in one group and the unhealthy ones in the other. But if you start with a small sample, randomly assigning people to the groups won't help much to lessen the chances of bias. If you have 50 people who are randomly assigned to two groups, there is less chance that most of the healthiest ones will be in one group than if you have only 10 people in the sample.
an imaginary study
Let's imagine that 1,000 kids in your school get poison oak. You randomly select 100 from the 1,000 to be in your study. You then randomly put 50 of the 100 in the experimental group and 50 in the control group. You tell the control group to do nothing special for their poison oak. You tell the experimental group to use GEM every night before they go to bed. You might think that a scientist would then wait a certain amount of time, say one week, and then compare the results. But there's one more step a scientist will take to lessen the chances of bias. The scientist will not do this test alone. She will have another scientist help her. The scientist who keeps track of which kid is in which group will not be the one who looks at the rashes to see if there's been any change. Why? There is a chance that knowing which group a person is in will affect how the scientist sees the rash. If she knows you got GEM, she might see your rash as getting better when in fact it's the same as a person in the control group who she saw as getting worse. We call this process of not knowing who is in the experimental and control groups by the one looking at the rashes a single-blind study. After the evaluations are finished, the two scientists compare notes. One scientist tells the other who was in which group and then the results for the two groups are compared.
A double-blind experiment would give the control group fake GEM, something that looks like GEM and is used the same way (as a pill or salve) but is known not to have any effect on rashes. The fake medicine is called a placebo. The kids in the study would not be told whether they were getting real or fake GEM (the placebo). Why? To lessen the chance that their knowledge would affect their rash. You might not think that your beliefs would affect how a medicine works, but scientists have found that they sometimes do. This is called the placebo effect. On the off chance that knowing they're getting real medicine or knowing they're getting the placebo will affect how the medicine works, scientists will often do double-blind experiments.
The final step in a scientific study is to look at the results and see if one group did any better than the other. This is usually done using some mathematics known as statistics, which we won't talk about here. Let's say that after one week 40 of the 50 in the experimental group showed no sign of a rash, while 40 of the 50 in the control group still had a severe rash. Even without knowing any statistics, you should be able to see that it looks like GEM is a pretty good medicine for poison oak.
Should we stop there and start selling GEM on street corners? No. Scientists know that a single study can't prove that GEM works really well. Even though you took many steps to avoid biasing your study, it is still possible that something went on that you don't know about. An unknown something (an x-factor) might have tricked you into thinking GEM is better than it really is. If more studies find that GEM works much better than a placebo, then the scientist will say that GEM works. Even so, the scientist must remain open to the possibility that future studies will find some fault with earlier studies. As long as the evidence in its favor is strong, we will accept something as true. But, if evidence is found that shows we are wrong, we must be willing to change our minds.
(For another example of a control group study, see the entry on plant perception.)