Reverse Inference

What it is

A brain scan normally runs in a forward direction. You give people a task and measure which regions become more active. Reverse inference runs the other way: it observes activity in a region and assigns it a single meaning. The nucleus accumbens, for example, is involved in reward, motivation and attention, so increased activity there does not by itself show craving.

The point was set out for a wide audience by Russell Poldrack in 2006, who argued that such inferences are only as strong as the region is specific to the process in question, and most regions are not very specific.

In effect

Claims in popular marketing books that a product or advert produced craving, love or trust because a particular region lit up should be read as interpretations, not findings. Behavioural measures, such as what people choose or do afterwards, are the better test.

The cigarette warning case shows why. A brain-scan claim that warnings stimulated craving was contradicted by published studies that measured craving and quit attempts directly.

What it does not say

It does not say brain imaging is useless. Forward inferences, from a known task to a measured response, are sound, and imaging can add to behavioural evidence.

It does not say every reverse inference is wrong, only that it needs support beyond the activity of one region.

---