-----

Neuroimaging studies attempt to assign function to brain regions. The neuroscientific practice of assigning functions to active brain areas involves two general patterns of inference, so-called "forward inference" and "reverse inference". In this talk, Roskies will evaluate these inferential patterns. Neuroscientists have recognized and discussed the logical problems attending reverse inference, but have viewed forward inference as straightforward. Here Roskies argues that reverse inference is overly maligned, whereas forward inference is even more problematic than previously recognized. Both forms of inference are closely tied to the problem of cognitive ontology. Roskies situates this problem in contemporary views of philosophy of neuroscience, and suggests avenues using artificial neural networks to address these problems.

-----