Monday, May 4, 2009

The risk of neuroscience degenrating into phrenology

Interesting article here from Perspectives on Psychological Science, by Edward Vul, Christine Harris, Piotr Winkielman, & Harold Pashler (h/t New Scientist).

Basically, a lot of the very strong correlations that fMRI scans are showing between parts of the brain "lighting up" and emotions felt by subjects, are likely to be spurious. This article claims that the experimenters are making some basic statistical mistakes and are over-stating the links they see. To quote the abstract:

We show how this nonindependent analysis inflates correlations while yielding reassuring-looking scattergrams.

No comments: