Correlation and causation are two terms that are often misunderstood, especially in the context of the sciences. Correlation is simply a relationship between variables, whereas causation means that one variable is directly affected or influenced by another variable (i.e. a cause and effect relationship). Causation is one way in which two variables can correlate, but associations between variables can take many forms and be due to a number of reasons that are not causative.
Variables or events that are studied can be related in a number of ways - they may have a common cause, they may be indirectly associated via intermediates or they may be coincidental. The direction of association may be positive or negative. Positive correlation is when the variables change in the same direction (e.g., both increase). Negative correlation is when the variables change in opposite directions (e.g., one increases and the other decreases).
Studies that find correlation between two or more variables do so mathematically. Statistically the changes in the variables are related, but this does not mean that they are physically interacting and/or have an effect on one another. Correlations that have unclear steps to relate the variables physically are indicative of a non-causal relationship. For example, a hypothetical study finds that more car accident victims have red hair. The study then concludes that having red hair and being in an accident are correlated, but it does not mean that having red hair causes more accidents. How the conclusions are stated in a scientific study (use of “correlated” or “associated”) is a clue as to what can be stated about the relationship between the variables.
Causation is often shown by controlled studies after correlations were previously made between two or more variables or events. A direct relationship is established by showing an interaction (i.e. having a blood alcohol level of a certain level directly causes car accidents by impairing the driver). Many correlations are mistakenly considered causation, but one variable causing an event or changes in another variable is a very specific finding with very specific details of how the two variables or events are related. Thus, causation is a much more specific and certain relationship between the variables being studied.
Epidemiological studies are considered the best way to establish general causation for human health effects as they confirm the "criteria of causation". This is especially important in toxicology. General causation means that the effect of the variable on the other is expected under certain conditions. If the conditions are changed, the cause and effect relationship may no longer exist. Other definitions of causation depend on the field of study, some consider individual vs. conceptual, legal vs. proximate, and actual vs. factual – all of which vary by the extent to which the direct effect of the variable on the other is shown. Most psychological research finds correlation, but not causation, because of the rigorous standards and testing needed to confirm causation.
Substance Use In Minnesota (SUMN) provides a number of examples of correlation versus causation in regards to their data and statistics that may be helpful in understanding the difference between these two terms. Remember - correlation does not equal causation.