Excerpt taken from Grant MacEwan Community College weg page.
Psychological studies vary in design. In correlational studies a researcher looks for associations among naturally occurring variables, whereas in experimental studies the researcher introduces a change and then monitors its effects. It is important to be able to distinguish between correlational and experimental designs, because only well-controlled experimental designs allow conclusions about cause and effect.
Consider the following example: A psychologist is hired by government to see whether there is any association between children watching violent television and an increase in their aggressive behavior. (question suitable for a correlational design) Does watching violent television make children aggressive? (question suitable for an experiment) Let's look at examples of the two types of studies.
Click here if you want to go directly to the quiz.
In a correlational study, the researcher would ask children (or their parents) to document the amount of violent television the child watches in a set time period (perhaps one week) and then the researcher would observe the children's behavior, recording instances of aggression. The researcher does no intervention. Suppose the researcher gets the following results, where each row in the table below corresponds to the scores of one child. Do you notice any pattern between the two columns of numbers? If you knew how much violent television a child watched, would you be able to predict the child's aggression score?
Hours per Week Spent Watching Violent TV | Aggression Score (least aggressive=1, most aggressive=10) |
---|---|
0 | 1 |
3 | 2 |
8 | 5 |
9 | 4 |
11 | 7 |
18 | 9 |
41 | 10 |
The table shows a strong positive correlation because an increase in the amount of violent television is usually associated with an increase in aggression score. We would predict that children who watch more violent television are likely to score higher on the aggression scale. But can we conclude that watching violent television causes the children to be more aggressive? Although this conclusion is certainly consistent with the results, it cannot be adopted because it is not the only plausible interpretation. It could be that children who are more aggressive choose to watch more violent television (aggression causes the television viewing habits rather than the other way around) or that television habits and aggression are not causally related, but appear to go together because both are associated with a third factor, such as amount of parent supervision. Perhaps children who are not supervised are more likely to watch more violent television, as well as score higher on an aggression scale. Correlational studies do not allow us to determine which of these interpretations is the correct one. To do that, we would have to do an experiment, as described in the next section.
In an experiment the researcher exposes the children to various types of television and then monitors their behavior for aggression. In a simple experiment, a researcher would create two groups of children (an experimental and a control group) through randomization (each participant has an equal chance of getting in either group). Randomization is used as a control so that the two groups are roughly equivalent at the start of the experiment. (see the section on Identifying variables and confounds for more information) The experimental group is shown violent television, and under identical conditions, the control group is shown nonviolent television. Afterwards the researcher sees whether children in the experimental group score higher on the aggression scale than children in the control group. If they do, the experimenter can conclude that watching violent television makes children aggressive (provided the researcher has controlled for confounds and extraneous variables.
This study was most likely correlational because an experiment would not be ethical. (In order to do an experiment, the researcher would have to control the students' drinking, forcing some students to drink heavily and then observing the effects of the drinking on their grades.) All we can conclude from the headline is that heavy drinking is associated with lower grades. We cannot conclude that drinking caused the lower grades because other plausible interpretations have not been ruled out. (Perhaps students drink more because they make lower grades. Or perhaps drinking and grades appear related only because they are both related to the degree of student commitment to being in school.)
Yes. An experiment could have been set up in two different ways. In one, called a between-subjects design, people are randomly assigned to groups. One group learns concrete words; the other learns abstract ones to see whether the group learning concrete words remembers more. In another experimental design, called a within-subjects design, all participants learn both the concrete and abstract words to see whether individuals learn concrete words better than abstract ones. (Of course, the order in which people learn the words would have to be controlled using a procedure called counterbalancing)
This question can only be addressed by correlational studies. (In order to do an experiment, the researcher would have to randomly assign some children to a group that gets abused; others to a group that does not get abused. Obviously, this cannot be done!) Thus, we must be cautious about assuming the cause of any association between experiencing abuse as a child and perpetrating it as an adult. Correlational studies do not support one interpretation over others.
Enough information to see whether it was a correlational study
or an experiment. In a correlational study, the researcher would
take advantage of the fact that some depressed people stay on
medication longer than others. (The researcher does not control
how long people are on medication). Suppose the researcher finds
that people on medication more than six months are less likely to
relapse. The researcher cannot conclude that the increased time
on medication improved the relapse rate because other
explanations have not been ruled out. (Perhaps people who stay on
medication longer differ from the others in ways that would
protect them from relapse. Maybe the people on medication longer
are also more likely to receive psychotherapy.)
In an experimental study, the researcher controls how long people
stay on medication. Half of a sample of depressed people is
randomly chosen to receive medication for less than six months;
the others receive medication for more than six months. The only
way the two groups differ is in the duration of the medication.
We cannot conclude anything about the cause of the difference in scores between the two groups. This is not an experiment because the researcher did not control group membership to ensure that the groups were roughly equivalent when they started school. (Imagine the reaction of parents if the researcher randomly assigned some children to attend private school and others to go to public school.). It is a quasi-experiment (resembles an experiment because it compares groups, but is not an experiment because the researcher did not control group membership) and should be interpreted like a correlational study. We cannot conclude that private school caused students to score higher on the math test. (Perhaps the students from private school are more likely to practice math on computer at home and it is this home activity, rather than experiences at school, that leads to the higher math scores.)