Earlier topics (see here) noted the importance of controlling factors which could potentially confound the results of your experiment. In addition, it may be necessary to have treatment groups which serve as controls. There are three main functions of these controls or control groups.
- Serve as a comparison for treatment groups. Suppose we are doing clinical trials on the effectiveness of a new anti-cancer drug. We give the drug to cancer patients, and then monitor their progress, but how do we judge the drug’s success? It may produce a spectacular cure, but this is unlikely. And even a drug which only acts to slow the disease is a substantial benefit for patients. We need to be able to compare those who receive the drug, to those who did not. Those who received no treatment act as a comparison. If the disease proceeds slower and survival is longer in those who received the drug than those who did not, we begin to gain confidence that the new treatment is beneficial.
- Ensure experimental design is appropriate. Experiments give allow us to test our hypothesis under controlled circumstances, but often at the expense of being slightly different from natural conditions (see also field vs. lab). A control can help ensure that our experimental conditions have not seriously changed the thing we are trying to measure. For example, consider an experiment to see if a new pesticide can lower the fertility of aphids (a way of controlling the species). For convenience and cost-effectiveness, we rear our aphids in containers maintained in a lab. Our treatment group receives the pesticide, while the control group does not, and we measure the number of eggs laid per female in each group. In this case our control group acts as a comparison for the treatment group, but we can also compare fertility in this group to that seen in natural conditions. If it is much higher or much lower, we are obviously not reproducing natural conditions well enough, and we must question whether our treatment will have the same effect outside the lab.
- Help to rule out confounding factors. Sometimes we are unable to isolate our treatment from all other factors which could affect the outcome. For example a fertilizer may need to be dissolved in water for application, and the water might also affect plant growth. More subtly, providing a patient with a drug may not only provide the real benefit of the drug, but a boost to the patient’s morale which could also improve his health (known as the placebo effect). In these cases a control group will need to be set up which receives everything the treatment group receives, except the factor being tested. In the fertilizer experiment the control group must be receive the same amount of water at the same time as the treatment group, but with no fertilizer added. And in the drug test, the control group must receive a sugar pill disguised as the real drug (a placebo).
Positive and Negative Controls:
Our goal in every experiment is to ensure that if we see some effect in our response variable, the only possible explanation is the change we introduced in the manipulated variable. There are two conclusions we can reach in our experiment, and need to rule out errors in both. We may reach a positive conclusion (there is an effect of the treatment), or a negative conclusion (there is no effect of the treatment). Either conclusion may be an error, and controls may be needed to rule out both types of error. Typically, we use a control which should give a positive result if everything is working well (positive control) and a control which should give a negative result if everything is working well (negative control).
For example, if we are using a test to determine whether protein is present in an unknown sample we could conclude either that protein is present (positive result) or that it is not present (negative result). To ensure that the test does not indicate protein is present when it is really absent (called a “false positive”), we would test a sample with no protein (negative control). To ensure that the test does not miss detecting protein when it is really present (false negative), we would test a sample known to contain protein (positive control). If both controls come out as expected, we can be confident in our experimental results, whichever way they turn out.
Controls not always needed:
In some experiments, a distinct control group is not needed as a comparison. Different treatment groups are compared against one another directly. For example an experiment testing the attractiveness of different coloured traps on insects cannot have a control group with no colour. Similarly, an experiment testing the effect of temperature on growth rates in turtles would not need a control without temperature. Note that these experiments would still be controlled, assuming confounding factors were held constant.