Bettmann/Contributor/Bettmann/Getty Images
Although humans are good at determining patterns, a statistical technique would allow for better precision. Such a technique is the correlation coefficient. The correlation coefficient gives both the strength of the relationship and its direction. If the number of friends on Facebook was associated with more health center visits, then this relationship would be called a positive correlation. However, if fewer friends were associated with more visits, then this relationship would be called a negative correlation.
Figure 3.3 Scatterplot Diagram Showing Negative Relationship Between Two Measures
In the real world, few relationships are perfectly related to one another. Thus, the correlation coefficient is also able to reflect the degree of an association between two variables. Technically, the degree of relationship determined by the correlation coefficient is denoted by the letter r. Whether the relationship between the variables is positive or negative is denoted by the + and − signs. The correlation coefficient can range from −1 to +1. A perfect positive relationship would be r = +1, and a perfect negative relationship would be r = −1. If there was no relationship, it would be r = 0. Figure 3.4 shows a variety of relationships that differ in degree and direction; Figure 3.5 shows a variety of relationships and their corresponding r values.
positive correlation: an association between two variables where an increase in one variable correlates to an increase in the other
negative correlation: an association between two variables where a decrease in one variable correlates to a decrease in the other
What are the basic ideas of correlational studies? In correlational studies, the researcher is interested in asking whether there is an association between two variables, but he or she does not attempt to establish how one variable influences the other, only that a relationship exists. Establishing that such an association exists may be the first step in dealing with a complex problem.
Figure 3.4 Scatterplot Diagrams Showing Various Relationships That Differ in Degree and Direction
Figure 3.5 Scatterplot Diagrams of Different Relationships and Their Corresponding r Value
Source: Landis, R. S. (2007). Measures of association/correlation coefficient. In S. G. Rogelberg (Ed.), Encyclopedia of industrial and organizational psychology (pp. 471–474, Figure 1). Thousand Oaks, CA: SAGE.
My colleagues and I used a correlational technique to understand whether being tortured influenced the brain differently if one experienced dissociative experiences, in which one feels detached from reality (Ray et al., 2006). Although not much is known concerning neuroscience measurements of individuals who have been tortured in their native country, there has been some suggestion that torture leads to different types of psychopathological disorders. There is also some evidence to suggest that torture victims adopt psychological mechanisms to escape the experience of the situation. Dissociation is one such mechanism in which individuals are able to distance themselves from such extreme negative experiences as torture or rape. We developed a dissociative scale that asked individuals about dissociative experiences they had experienced. Since some other neuroscience work had shown that non-normal MEG (magnetoencephalography) activity was seen in different areas of the brain in disorders such as schizophrenia, depression, and PTSD (e.g., Rockstroh, Wienbruch, Ray, & Elbert, 2007), we used MEG as the variable to correlate with dissociative experiences. We found a positive correlation of .60 between our measure of MEG activity and the dissociation scale for the left hemisphere as a whole and a negative correlation of -.61 for the right hemisphere as a whole. More precise analysis showed the involvement of the left ventral region of the anterior cortical areas.
What we cannot know from correlational research is whether either variable influences the other directly. There are a variety of ways that one can see a high correlation. It might be that there is a direct relationship. That is, having lots of friends might make you feel better and make you less susceptible to disease. However, it might also be true that if you went to the health center often, you might not have time for friends or not feel like being with others. In our torture study, one might logically assume that the torture influenced the brain. However, it might have been the case that those individuals who had particular types of MEG activity would be those that later developed dissociative experiences.
It is also possible that a third unspecified variable actually may have influenced the two variables in a correlational study. For example, there is a strong correlation between eating ice cream and wearing bathing suits. However, in this case it is a third variable, the warm weather, which produces the relationship. In the case of health and friends, it could be genetic makeup or having a job, both of which could influence the time one had for friends as well as health. Thus, the nature of a correlational study is to describe the relationships but not to suggest which variable influences which other variable.
It is often said that correlation does not imply causality. For example, a researcher might want to know whether a relationship exists between the type of food a child eats and the likelihood of the child having a particular mental disorder such as hyperactivity. One approach would be to examine the diets of children who show hyperactivity and those who do not. What if there was a high association between eating foods with sugar, for example, and hyperactivity? You could conclude little other than that there was an association or correlation between the two variables. There are at least three ways to understand this relationship. (1) It might be that sugar is associated with hyperactivity. (2) It might also be that those who are hyperactive seek sugar. (3) A third variable such as a specific gene or neurotransmitter or sleep pattern might lead to both eating foods with sugar and showing hyperactivity.
The association of two factors does not in itself imply that one influences the other. However, if there is a low correlation between the events, you can infer that one event does not cause the other. A high degree of association is always necessary for establishing that one variable influences another; a correlational study is often the first step to providing the needed support for later experimental research, especially in complex areas.
Concept Check
What are four characteristics of the scientific approach? What are two additional key ingredients in psychological research?
What are the three stages of the scientific process?
Why has the case study been particularly valuable in psychopathology research?
What are the four characteristics of the naturalistic observation method?
If a research study reports that two variables are correlated, what do you know about their relationship? What don’t you know about their relationship?
The Experimental Method: Making It Happen
Just finding that a relationship exists between two events does not allow us to determine exactly what that relationship is, much less that one event actually caused the other to happen. If we want to state that one event produced another event, we need to develop a much stronger case for our position.
To do this,
