If case studies and correlational studies cannot tell us much, if anything, about causes, how can we identify causal factors? The most direct way is to perform an experimental study.
In order to understand what experimental studies are, first we must be clear about what is meant by a causal relationship (also see Section 1-6, What is a Cause?). A causal relationship between two variables is a relationship in which changes in one variable (the causal variable) produce changes in a second variable (the affected variable). For example, increases in the number of alcoholic drinks consumed cause increases in reaction times (the amount of time it takes to react to a stimulus, such as a stoplight that changes from red to green).
In order to test for a causal relationship between two variables — in order to show that changes in one variable produce changes in a second variable — we need to control for the effects of extraneous variables, including "third variables" (the third-variable problem), as well as for the direction of causation (the directionality problem). Experimental studies are designed to control for these factors, thereby excluding all possible causal explanations for the results but the one being tested.
Researchers experimentally test the claim that a particular causal relationship exists between two variables by doing two things. First, they manipulate the causal variable directly instead of allowing it to vary "on its own" (as occurs in uncontrolled research situations, such as case studies and correlational studies). That is, the researchers set up a situation in which a group of participants is subjected deliberately to one value of the causal variable and a second group of participants is subjected to another value. Manipulating the causal variable controls for the directionality problem because, by changing the causal variable themselves, researchers know which of the two variables changed first.
As an example, let's say that researchers want to test the hypothesis that using caffeine while studying for tests causes increased test scores. In this case, the causal variable is whether or not caffeine is ingested while studying, and researchers can manipulate this variable directly by having, say, one group of students drink regular coffee (with caffeine) and having a second group of students drink decaffeinated coffee (see Table 1) while both groups study for a test. At some later time, students will take the test in order to see which group does better.
| Causal Variable | Affected Variable | |
|---|---|---|
| Group 1 | Caffeine |
Test Scores |
| Group 2 | No Caffeine |
Test Scores |
Table 1. Design of an experiment testing the hypothesis that using caffeine while studying for tests causes increased test scores
In order to be sure that it is the causal variable that is causing changes in the affected variable, it is essential that the two groups be as similar as possible at the start of the experiment. That is, there can be no average differences between the two groups with respect to any extraneous variables. For example, if the students are allowed to choose which group they want to be in — the caffeine or the no-caffeine group — the researchers probably will end up with one group of students who mostly like caffeinated drinks (and perhaps drink them regularly) and a second group of students who mostly do not like caffeinated drinks (and drink them only occasionally). In this case, the two groups are different even before they have been exposed to the causal variable.
In order to make the two groups in an experiment as similar as possible at the start of the experiment, it is best to randomly assign participants to the two groups. A random process (such as flipping a coin) should be used to decide which group each participant is assigned to so that neither the researchers nor the participants are responsible for group assignments. Random assignment controls for the effects of extraneous variables, including third variables, by making sure that, if extraneous variables are having effects, those effects are equal in the two groups. And thus, if the two groups differ at the end of the experiment, this difference can be attributed to the effects of the causal variable alone.
To continue the caffeine/no-caffeine example started above, let's say that 100 college students are randomly assigned to one of two groups by drawing cards from a box: fifty cards have a "1" written on them and fifty cards have a "2" written on them. Those drawing the 1 cards are assigned to Group 1, whereas those drawing the 2 cards are assigned to Group 2. The students in Group 1 then are given a caffeine pill and the students in Group 2 are given a placebo pill (a pill that looks like the caffeine pill but that has no caffeine in it). After taking their pills (and it is important that they don't know whether they have taken the caffeine or placebo pill since "suggestion" could affect their studying or test-taking; see Section 2-3 on the effects of suggestion), both groups are given one hour to read five pages from a psychology textbook and then answer a set of study questions. The next day, all participants take a 20-item multiple-choice test that assesses their knowledge of the material. Figure 1 shows fictional results from this research. The mean test score for Group 1 (the caffeine group) was nearly 63%, whereas the mean test score for Group 2 (the placebo group) was nearly 53%. Based on this result, the researchers can conclude that using caffeine while studying causes higher test scores.
Figure 1. Mean test scores (fictional) for those who do and those who don't use caffeine while studying
Study Questions for Section 3-6
- How would you define a causal relationship in your own words?
- If we find a drug that reduces the activity of a particular brain chemical (Variable A) and also leads to a decrease in the symptoms of a particular mental disorder (Variable B), have we shown that there is a causal relationship between Variable A and Variable B? Why or why not? (NOTE: If you are not sure of your answer, please ask me.)
- How is a correlational study similar to and different from an experimental study?
- How does the manipulation of the causal variable control for the directionality problem in an experimental study?
- How does random assignment control for the effects of extraneous variables in an experimental study?
- Why can we make a cause-and-effect interpretation of the results of a single experimental study but not of the results of a single correlational study?