Use of Measurements, Controls, and Statistics

Suppose you wanted to test the hypothesis that a regular exercise program causes people to have a lower resting heart rate. First, you would have to decide on the nature of the exercise program. Then, you would have to decide how the heart rate (or pulse rate) would be measured. This is a typical problem in physiology research, because the testing of most physiological hypotheses requires quantitative measurements.
The group that is subject to the testing condition—in this case, exercise—is called the experimental group. A measurement of the heart rate for this group would only be meaningful if it is compared to that of another group, known as the control group. How shall this control group be chosen? Perhaps the subjects could serve as their own controls—that is, a person’s resting heart rate could be measured before and after the exercise regimen. If this isn’t possible, a control group could be other people who do not follow the exercise program. The choice of control groups is often a controversial aspect of physiology studies. In this example, did the people in the control group really refrain from any exercise? Were they comparable to the people in the experimental group with regard to age, sex, ethnicity, body weight, health status, and so on? You can see how difficult it could be in practice to get a control group that could satisfy any potential criticism.
Another potential criticism could be bias in the way that the scientists perform the measurements. This bias could be completely unintentional; scientists are human, after all, and they may have invested months or years in this project! Thus, the person doing the measurements often does not know if a subject is part of the experimental or the control group. This is known as a blind measurement.
Now suppose the data are in, and it looks like the experimental group indeed has a lower average resting heart rate than the control group. But there is overlap—some people in the control group have measurements that are lower than some people in the experimental group. Now, is the difference in the average measurements of the groups due to a real, physiological difference, or is it due to chance variations in the measurements? Scientists attempt to test the null hypothesis (the hypothesis that the difference is due to chance) by employing the mathematical tools of statistics. If the statistical results so warrant, the null hypothesis can be rejected and the experimental hypothesis can be deemed to be supported by this study.
The statistical test chosen will depend upon the design of the experiment, and it can also be a source of contention among scientists in evaluating the validity of the results. Because of the nature of the scientific method, “proof” in science is always provisional. Some other researchers, employing the scientific method in a different way (with different measuring techniques, experimental procedures, choice of control groups, statistical tests, and so on) may later obtain different results. The scientific method is thus an ongoing enterprise.
The results of the scientific enterprise are written up as research articles, and these must be reviewed by other scientists who work in the same field before they can be published in peer-reviewed journals. More often than not, the reviewers will suggest that certain changes be made in the articles before they can be accepted for publication.