The decoding approach chooses, from the whole brain, sets of regions or individual voxels that encode abstract intangible characteristics or psychological constructs while carefully guarding against overfitting.
The article by Chen, Nelson, and Hsu (2015) moves the field beyond questions that rely on spatial localization of brain data. It represents an advance in testing previously unaddressable research questions—notably, whether researchers can predict consumers’ thought processes on the basis of some spatially distributed pattern of fMRI activity on a small group of participants.