Experiment Design

Scientists follow a specific method for setting up experiments to test their hypotheses. In an experiment, we want to know if a change happens, and we want to be sure we can identify what caused a change to happen. Therefore, an experiment must include these basic parts:

Independent Variable:
This is the part of the experiment that you manipulate to see what happens. For example, let’s say we have a hypothesis that a tomato plant would grow just as quickly using organic compost as it would using Miracle-Gro ® water soluble tomato plant food. In an experiment designed to test this hypothesis the independent variables are the  Miracle-Gro ® and the organic compost. These are the soil treatments we are adding to see what happens with the tomato plants.

Dependent Variable:
This is the part of the experiment that you observe to see what happened. For example, in our tomato plant experiment the dependent variables are the tomato plants themselves, specifically, the growth rate of the tomato plants. This is what we will be measuring to see if there is any difference.

Control Variables:
These are the parts of the experiment that you make sure stay the same. This can be the most important part of the experiment. If you haven’t carefully controlled all the other variables, you cannot be sure if it is the independent variable you are testing that has actually produced the difference. For example, in the tomato plant experiment, we want to be sure that we are using the same base soil mixture, the same size and type of container, and the same variety of tomato seeds. We want to be sure all the plants are getting the same amount of water and light. We want to be as sure as possible that the ONLY difference between the tomato plants is the type of soil treatment we’ve added.

In addition to the 3 types of variables, a good experiment should also have the following:

Multiple trials:
If we were to design our tomato plant experiment to have only two plants, one with Miracle-Gro ® and one with organic compost, then it would not be a very good experiment. There are unaccounted for differences in the genes of any two specific plants (just like differences between my genes and your genes that make us appear different even though we are both the same species). One plant might just grow faster because it has the genes to, or for some other reason besides the soil treatment we’ve added. In order to reduce the impact of this random variation on our results, we want to apply our independent variable to multiple test subjects and measure the dependent variable multiple times. These repetitions are called trials. So we would plant 3 or more tomatoes in organic compost and 3 or more in Miracle-Gro ®. The more we plant, the more sure we can be of our results. But to be sure, we still need a...

Control group:
The control group is the group of trials that you do not apply the independent variable to. It provides the baseline against which you compare the experimental group (the group in which you are testing the independent variable). In our tomato plant experiment, the experimental group are all 6 plants, 3 with organic compost and 3 with Miracle-Gro ®. We would add another 3 tomato plants in our control group. These plants would get neither soil treatment. But we would meticulously ensure that all the control variables remained the same for this control group--the same base soil, the same tomato seeds, the same amount of light and water, etc. We would measure the dependent variable (the growth rate) for this control group whenever we measure the dependent variable of the experimental group.

Random selection:
Let’s say that we start this experiment with nine 14-day-old seedlings. We need to assign each of these seedlings to a group: organic compost group, Miracle-Gro ® group, or control group. If we assign all the biggest, healthiest seedlings to the organic compost group and all the scrawniest seedlings to the Miracle-Gro ® group because we want to support our hypothesis, we are biasing our experiment, and the results will not be reliable. We can even bias the experiment in this way without consciously trying to. So before we begin our experiment we need to devise a method to make sure we are selecting seedlings for each group randomly. Maybe we could ask a friend who doesn’t know our hypothesis to shuffle them up and divide them into three random groups for us. Thinking out how to randomize the groups is an important part of the experiment design.

Whenever you are asked to design an experiment for a science class, be sure you carefully think through and include all of the parts above.