One of the most important things for you to learn in this course is the difference between correlational research and experimental research. Like correlational research, experimental research concerns relationships between variables. Unlike correlational research, however, experimental research provides strong evidence for causal interpretations. Here we will focus on the two most important features of experimental research.
Manipulation of an Independent Variable
One important feature that distinguishes experimental research from correlational research is that instead of simply measuring two variables, the researcher manipulates one of them. This means that the experimenter actually changes the value of that variable in a systematic way. This variable, which is called the independent variable, is the one that the researcher believes is the cause. The other variable, which the researcher believes is the effect, is called the dependent variable.
For example, you could do a correlational study on the relationship between noise level and concentration by going to a variety of places, measuring the noise levels there, and giving people a task that requires concentration. Or you could do an experiment by setting up a situation in which you could manipulate the noise level—perhaps by making it really loud for one group of people and really soft for another. And of course you could give them a task that requires concentration, and their performance on this task would be the dependent variable.
But what does this have to do with causation? The idea is that if changes in one variable occur reliably right after you manipulate another variable, then you can be fairly sure that your manipulation is causing the changes. But if you are not manipulating anything, it is harder to know what the cause is. Back to the noise example. If you turn up the noise level and people lose their concentration, then it seems clear that your turning up the noise level caused them to lose their concentration. But what if you do a correlational study and find that people tend to be less able to concentrate in noisier places? Well, it could be that the noise is causing the loss of concentration or it could be that people with good concentration tend to avoid noisy places (the directionality problem).
Control of Extraneous Variables
The second feature that distinguishes experimental research from correlational research is the control of extraneous variables. Extraneous variables are basically all variables other than those you are interested in for purposes of your research. In an experiment on the effects of noise on concentration, there is an infinite number of extraneous variables: age and sex of the research participants, whether or not they have eaten recently, the temperature of the room, the time of day, ….
To control extraneous variables means to keep their values or levels as similar as possible across the different values or levels of your independent variable. The different values or levels of your independent variable are also called conditions. For example, your noise experiment might have two conditions—noisy and quiet—and what you want is for things to be as similar as possible under these two conditions. You do not want old people in one condition and young people in the other, or men in one condition and women in the other, or hungry people in one condition and full people in the other, …. Why not? Because if there is a relationship between your independent and dependent variables, you cannot tell if it was because of your manipulation or if it was because of one of these extraneous variables. (It is the third-variable problem again.) If people in the noisy condition were hungry and people in the quiet condition were not, then perhaps that is why those in the noisy condition had more trouble concentrating. But if things are highly similar under the different conditions, then you cannot "blame" the effect on anything other than the independent variable.
An extraneous variable that differs systematically across conditions is called a confounding variable. It is important to see the difference between extraneous variables and confounding variables. For example, in an experiment on the effectiveness of cognitive psychotherapy for treating depression, the independent variable is whether or not patients get the psychotherapy, and the dependent variable is how much they improve. Certainly, there will be lots of extraneous variables that are not really of primary interest: the sex of the patients, the age of the patients, their occupations, what they typically eat for breakfast, …. These extraneous variables are not a problem if they are similar in the two conditions. If both conditions contain men and women, young patients and old, doctors and plumbers, pancake eaters and coffee drinkers … then there is no problem. In fact, in many situations, this kind of within-condition variability is considered good because then the results of the experiment can be said to apply to a wide variety of people. What is bad, however, is if any of these extraneous variables are confounding variables, in the sense that their average level varies across conditions. For example, if one condition contains a lot more men than women, then sex of the patients is a confounding variable. Now, if there is a difference between the conditions, it could be due to the psychotherapy or it could be a sex difference (e.g., women tend to improve over time but men do not).
The Limitations of Experiments
The obvious advantage of experimental research is that it provides stronger evidence for causal claims. It does, however, have at least two limitations.
The first is that sometimes you cannot do an experiment because you cannot manipulate the independent variable, either for practical or ethical reasons. For example, if you are interested in the effects of a person’s culture on their tendency to help strangers, you cannot do an experiment. Why not? You cannot manipulate a person’s culture. Or if you are interested in how damage to a certain part of the brain affects behavior, you cannot do an experiment. Why not? You cannot go around damaging people’s brains to see what happens. In such cases, correlational research is the only alternative.
The second limitation of experimental research is that sometimes controlling extraneous variables means creating situations that are somewhat artificial. A good example is provided research on the effect of smiling on first impressions. To control extraneous variables, people are typically brought into a laboratory and asked standard questions about a small number of posed stimulus photographs. It is legitimate to ask, however, whether the effect of smiling is likely to be the same out in the "real world" where people are actually interacting with each other. For a good discussion of why this is not always a problem, though, see Stanovich's (2007) discussion of the "artificiality criticism" of psychological research.