Can correlation show causality?

11 Mar

This week I’m looking at whether or not finding a correlation in a experiment can and does show causality ie: does finding a link/correlation between two variables prove that variable A is responsible for the change in variable B? The short answer….well no it doesn’t. There is no way of proving that some other extraneous variable in the experiment wasn’t the reason for the change. However this doesn’t mean that the two aren’t in some way linked for example it’s impossible to find a causal relationship if there wasn’t some sort of correlation between the two variables therefore correlation is essential for finding a causative relationship but it doesn’t mean that everytime there is one causation is proved. This is of course asssuming that the varibale relationship on linear. If the relationship in non linear a correlation may not be present.

In an experiment in order to find causation there needs to be high levels of control, control that is so tight that it probably would be highly difficult, if not impossible, to acheive. Think how would you be able to control the movements, life etc of participanst without breaking every ethical code in the handbook? Cause and effect means that you have a definite ‘this led to this’ relationship whereas causation is more of a ‘well they’re connected’ kind of link. This is no where near as recise and definite as what is needed for a casuative relationship to be determined. A non psychology example of when correlation doesn’t show causation is arm length. There is a high correlation between the length of people’s right arm and their left arm, the length of the persons right arm however didn’t cause the left arm to be the length it is. A psychology version of this is the example of Reiche et al’s study (2004) they looked into whetehr or not extreme psychological stress causes cancer. They found no casusative relationship between the two they did however find a strong suggestion of a correlation between extreme stress and a weakened immune system which in turn  makes people more suseptible to virus associated cancers.

In conclsuion though causation is associated with correlation and is often precluded by a correlative relationship, a correlation doesn’t determine cause and effect and therefore should not be viewed as proving a cause/effect relationship.


2 Responses to “Can correlation show causality?”

  1. psue76 March 14, 2012 at 11:48 am #

    Another way in which you can see whether there one variable affects the other is by doing a single case design. Using this method, you can see whether by including an intervention (hence changing a variable) you can see if there is an effect. For example, using an AB design, participants who suffer from chronic back pain partake in a single case design where they are given a “cognitive- behavioural graded exposure in vivo” in order to try and reduce the symptoms of pain-related fears (Vlaeyen, 2001). Participants go through a baseline period and are then given the intervention treatment which is the cognitive- behavioural approach.

    Vlaeyen, 2001.

  2. sigmafreud March 14, 2012 at 5:00 pm #

    The question of correlation and causation is one that is often tackled by psychologists. Wrongly interpreting data to conclude that variables share a causational relationship is the focal point of many discussions and, indeed, pseudoscience.
    Exhaustive variable testing has to be conducted, in which possible confounding variables have to be eliminated and isolated one by one, to endure that no latent “third variable” exists which could be influence both the independent and dependent variables. Of course, an initial degree of control helps ensure that as few extraneous variables are having an impact on the data than possible. However, to achieve the level of control needed to ensure causation is next to impossible. Simon (1954) believed that the most efficient way to test correlations was to continuously add parameters and variables, and to take a holistic approach to the conducted research, observing any changes which each variable yields.
    Experimenters tend to jump to the causation conclusion if the one variable has a strong relationship with another. Kaplan and Saccuzzo (2009) suggested that this misinterpretation can lead to wrong or misleading conclusions of the gathered results. This misinterpretation could take the form of believing the results to be more significant, or conclusive than they actually are.
    The fact is that correlation can never guarantee causation. Correlation shows a relationship, that may or may not be directly caused by one of the variables upon the other.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: