Reverse Correlation

Reverse correlation is a technique that allows a researcher to infer the collection of “features” used by an information processing system (e.g., human vision). Probably the best way to understand what reverse correlation is and how it works is through an example: Let's suppose you want to know what facial features a person uses to identify a human face. How might you go about solving this problem? A first guess might be to monitor an observer's eye movements as they perform a task that involves identifying human faces. However, this requires a strong assumption: Namely, that attention and the subsequent use of information are both perfectly correlated with eye movements. It is easy to show that this assumption is in fact not always true; for ...