Step 4: Conduct an Experiment
Many people think of an experiment as something that takes place in a lab. While this can be true, experiments don't have to involve laboratory workbenches, Bunsen burners or test tubes. They do, however, have to be set up to test a specific hypothesis and they must be controlled. Controlling an experiment means controlling all of the variables so that only a single variable is studied. The independent variable is the one that's controlled and manipulated by the experimenter, whereas the dependent variable is not. As the independent variable is manipulated, the dependent variable is measured for variation. In our car example, the independent variable is the shape of the car's body. The dependent variable -- what we measure as the effect of the car's profile -- could be speed, gas mileage or a direct measure of the amount of air pressure exerted on the car.
Controlling an experiment also means setting it up so it has a control group and an experimental group. The control group allows the experimenter to compare his test results against a baseline measurement so he can feel confident that those results are not due to chance. For example, in the Pasteur experiment described earlier, what would have happened if Pasteur used only a curved-neck flask? Would he have known for sure that the lack of bacteria growth in the flask was because of its design? No, he needed to be able to compare the results of his experimental group against a control group. Pasteur's control was the flask with the straight neck.
Now consider our air-resistance example. If we wanted to run this experiment, we would need at least two cars -- one with a streamlined, birdlike shape and another shaped like a box. The former would be the experimental group, the latter the control. All other variables -- the weight of the cars, the tires, even the paint on the cars -- should be identical. Even the track and the conditions on the track should be controlled as much as possible.
Step 5: Analyze Data and Draw a Conclusion
During an experiment, scientists collect both quantitative and qualitative data. Buried in that information, hopefully, is evidence to support or reject the hypothesis. The amount of analysis required to come to a satisfactory conclusion can vary tremendously. Because Pasteur's experiment relied on qualitative observations about the appearance of the broth, his analysis was fairly straightforward. Sometimes, sophisticated statistical tools have to be used to analyze data. Either way, the ultimate goal is to prove or disprove the hypothesis and, in doing so, answer the original question.