The radish study has two groups. It is pretty obvious to everyone involved which is the test group and which is the control, since one group has to eat radishes every day and one group does nothing except have a dental checkup after a year.
Suppose people in the R group, recognizing that their decay rate would influence the outcome of the test of radishes, might be concerned about spoiling the test if they eat sugary foods or don't brush and floss their teeth regularly. As a result, the R group might fare better because of better eating habits and dental hygiene, but our researchers would think it was because of the radishes. It's not clear whether the subjects would actually change their habits, but in if the experiment succeeded we could not be sure whether it was the radishes or "knowledge of being in the test group" that was responsible for the results.
The way to avoid this problem is to disguise which group is which by having the control group go through a procedure similar enough to that of the test group that it's impossible for the subjects to tell whether they are in the group that is really being tested. To do this we might have the control group eat carrots each day instead of radishes.
It is known that people's expectations can often
influence their behavior and perceptions. People will often report
substantial pain relief or improvement in medical symptoms as a result
of taking fake medicine, called a "placebo", that has no active
The process of disguising which group the subjects belong to is called "blinding", because the subjects are blind to which condition they are being given. Blinding is common with tests of taste preferences (blind taste tests) because people's liking for a product is so greatly influenced by their knowledge of the brand name. In the radish vs. carrot test, "blinding" may be a somewhat misleading word, since the subjects do know whether they are getting radishes or carrots, but the crucial point is that they don't know which one the researchers think is worthwhile to test.
We have assumed here that carrots don't have an effect on tooth decay. Obviously a placebo must be something that doesn't have an effect of its own. I recall reading about one study to test whether certain fish oils reduced blood cholesterol. The placebo used was olive oil. The study found the fish oil ineffective, but later other studies showed that olive oil is helpful in reducing cholesterol. The fish oil may have been effective but showed no benefit when compared to a placebo that was also effective. It is obviously important that placebos should not have a significant effect of their own, either good or bad.