Dissociable prototype learning systems: evidence from brain imaging and behavior.

The neural underpinnings of prototype learning are not well understood. A major source of confusion is that two versions of the prototype learning task have been used interchangeably in the literature; one where participants learn to categorize exemplars derived from two prototypes (A/B task), and one where participants learn to categorize exemplars derived from one prototype and noncategorical exemplars (A/non-A). We report results from an fMRI study of A/B and A/non-A prototype learning that allows for a direct contrast of the two learning methods. Accuracy in the two tasks did not correlate within subject despite equivalent average difficulty. The fMRI results revealed neural activation in a network of regions consistent with episodic memory retrieval for the A/B task while greater activation of a nondeclarative learning network was observed for the A/non-A task. The results demonstrate that learning in these two tasks is mediated by different neural systems and that recruitment of each system is dictated by the context of learning rather than the actual category structure.