Controlled Experiments (page 3)
In experimenting, the term control has special meaning. Here's an example of a very simple controlled experiment about making lemonade. Start with two glasses containing the same amounts of water. Into the first glass, put one teaspoonful each of sugar and lemon juice. Into the second glass, put two teaspoonfuls of each. The first glass is called the "control;' and the second glass is called the "experimental." By tasting one and then the other, you can make a better judgment about the changes than had you used only one mixture. You can continue to change the amounts of sugar and lemon juice in the experimental glass while the control glass stays the same.
Growing Popcorn Plants
An experiment with growing popcorn plants can demonstrate a more advanced controlled experimental design. It requires more materials, equipment, and time, but it can be made much more meaningful:
Use two identical planters with the same kind of soil in each one. Plant 30 popcorn seeds in each planter. To water the experimental planter, you will add baking soda to the water in carefully measured amounts, such as one level teaspoon of baking soda for each quart of water. To water the control planter, you will use plain tap water. Use the same amounts of water in each planter. Try to keep all other conditions—for example, temperature and light—the same for both planters.
In setting up a controlled experiment, many decisions must be made. For example, which seeds will you use for the experimental and which seeds for the control planter? You will take all of the seeds from the same package, of course. Of course? There, you see, is an important decision. You want the seeds to be as much alike as you can get them—as alike in inherited traits as possible. As you pick the seeds you should be careful to choose uniform, sound seeds (e.g., not broken, undersized, or discolored). See appendix A for instructions on how to do a seed germination test to make sure the seeds you have are viable.
Even with all this care, it is possible for prejudice to begin to affect results, so do not decide yet which planter will be experimental and which will be control. You will use a random choice method for making this decision, as you will in making others to come. Put the two groups of seeds into separate containers, then decide which group goes in which planter by random choice, "blind;' as in taking names out of a hat or in tossing a coin.
At this point, some people might say. "Why all this bother? Why not just plant the seeds and get on with it?" Of course, if one were just planting seeds for the fun of it, or for growing food, you would not worry about doing it by random choice, in an unprejudiced manner. We must look ahead, however, to the final judgment—your report—that will depend very much on the making of random decisions at this early stage, and when you report your findings, you want to have full confidence in your methods.
Continue to make as many decisions as you can randomly: The soil you use in each planter must be mixed thoroughly in one large container and then transferred alternately to each container in small amounts so as to not favor either. Plant the seeds by spreading them in as nearly the same pattern around each planter as you can, and take care to cover them uniformly with soil.
Now you are ready to choose which planter is to be the control and which is to be the experimental. To make an unprejudiced decision, a random choice, you might flip a coin.
The rule is to arrange to make as many random decisions as you can—by tossing a coin, pulling numbers or names out of a hat, using a list of random digits, or by other means. Even though you held constant as many variables as you could, there are probably unknown variables that you could not make constant, for example, differences in the genetic traits of the seeds, or differences in the soil about which you may not know. Therefore, you will need to take your chances on these unknown variables. They may work for or against your hoped-for result—your hypothesis.
Can we ever make the conditions exactly alike in the two parts of a controlled experiment—aside from the independent variable? Not likely. Nevertheless, we can count on a controlled experiment to give more valid results than a simple, uncontrolled experiment would give us, and certainly more than the usual mixed-up conditions of ordinary, everyday life.
When the plants have grown to a point that you feel you can judge the results, you end the experiment. Carefully measure the heights of the plants in each planter to compare them, keeping careful records of the measurements for each planter. If a camera is available, take photographs, always with the planters marked in such a way that it is easy to see in the photo which one is experimental and which one is control.
An even better way to measure the plants for comparison is to weigh each one on a laboratory balance scale. Dig up each plant, taking care not to damage the roots. Carefully wash the soil off of the roots and layout each plant on newspaper to dry, keeping the plants for each planter separate while you do the washing and drying. Next, weigh each plant and carefully record the weights for the experimental and for the control.
Now the statistics: For each planter, calculate an average, or mean, of the weights (or measure the lengths if you did not use a balance). You can now compare the averages to determine whether the independent Variable—the baking soda in the water—made a difference in the growth of the plants.
Here a most important problem comes up. We can assume that the mean weights of the plants in each planter will differ somewhat, anyway. So how much difference do you need to show to be able to claim that the baking soda in the water (the independent variable) made a significant difference? Of course, if the plants all died in the experimental planter, you have no problem. Most often, however, the differences are not that obvious. In this case you need to use statistical methods for deciding how much difference is meaningful or significant. The term significant difference should become part of your thinking about any investigation you do.