What Is a Controlled Experiment?

A controlled experiment is one in which everything is held constant except for one variable. Usually, a set of data is taken to be a control group, which is commonly the normal or usual state, and one or more other groups are examined where all conditions are identical to the control group and to each other except for one variable.

Sometimes it's necessary to change more than one variable, but all of the other experimental conditions will be controlled so that only the variables being examined change. And what is measured is the variables' amount or the way in which they change.

Example of a Controlled Experiment

Let's say you want to know if the type of soil affects how long it takes a seed to germinate, and you decide to set up a controlled experiment to answer the question. You might take five identical pots, fill each with a different type of soil, plant identical bean seeds in each pot, place the pots in a sunny window, water them equally, and measure how long it takes for the seeds in each pot to sprout.

This is a controlled experiment because your goal is to keep every variable constant except the type of soil you use. You control these features.

Read More

The Role of a Controlled Variable in an Experiment

By Anne Marie Helmenstine, Ph.D.

Why Controlled Experiments Are Important

The big advantage of a controlled experiment is that you can eliminate much of the uncertainty about your results. If you couldn't control each variable, you might end up with a confusing outcome.

For example, if you planted different types of seeds in each of the pots, trying to determine if soil type affected germination, you might find some types of seeds germinate faster than others. You wouldn't be able to say, with any degree of certainty, that the rate of germination was due to the type of soil. It might as well have been due to the type of seeds.

Or, if you had placed some pots in a sunny window and some in the shade or watered some pots more than others, you could get mixed results. The value of a controlled experiment is that it yields a high degree of confidence in the outcome. You know which variable caused or did not cause a change.

Are All Experiments Controlled?

No, they are not. It's still possible to obtain useful data from uncontrolled experiments, but it's harder to draw conclusions based on the data.

An example of an area where controlled experiments are difficult is human testing. Say you want to know if a new diet pill helps with weight loss. You can collect a sample of people, give each of them the pill, and measure their weight. You can try to control as many variables as possible, such as how much exercise they get or how many calories they eat.

However, you will have several uncontrolled variables, which may include age, gender, genetic predisposition toward a high or low metabolism, how overweight they were before starting the test, whether they inadvertently eat something that interacts with the drug, etc.

Scientists try to record as much data as possible when conducting uncontrolled experiments, so they can see additional factors that may be affecting their results. Although it is harder to draw conclusions from uncontrolled experiments, new patterns often emerge that would not have been observable in a controlled experiment.

For example, you may notice the diet drug seems to work for female subjects, but not for male subjects, and this may lead to further experimentation and a possible breakthrough. If you had only been able to perform a controlled experiment, perhaps on male clones alone, you would have missed this connection.

Sources

• Box, George E. P., et al. Statistics for Experimenters: Design, Innovation, and Discovery. Wiley-Interscience, a John Wiley & Soncs, Inc., Publication, 2005. 
• Creswell, John W. Educational Research: Planning, Conducting, and Evaluating Quantitative and Qualitative Research. Pearson/Merrill Prentice Hall, 2008.
• Pronzato, L. "Optimal experimental design and some related control problems". Automatica. 2008.
• Robbins, H. "Some Aspects of the Sequential Design of Experiments". Bulletin of the American Mathematical Society. 1952.