Perception of Covariation
E stimates of the degree to which variations in two variables are coupled have been investigated under several topics, including covariation, co-occurrence, contingency, correlation and joint probabilities. Here I will use these terms as they were used in the studies cited, without making distinctions among them, but will tend to use covariation as the default generic term.
The ability to detect covariation is widely recognized to be an important one for any creature. As Alloy and Tabachnik (1984) put it, “Information about the relationships or covariations between events in the world provides people and animals with a means of explaining the past, controlling the present, and predicting the future, thereby maximizing the likelihood that they can obtain desired outcomes and avoid aversive ones” (p. 112). Several experimental tasks have been used to study the estimation of covariation. These include inspecting graphical representations of the relationship between two variables (correlation scatterplots), inspecting series of number pairs, inspecting full or partial two-by-two contingency tables, and looking for evidence of covariation in case records (e.g., looking for the coupling of smoking and lung cancer in medical records).
Some investigators have questioned whether people untutored in statistics have an abstract concept of contingency or correlation (L. J. Chapman & J. P.