"The mirage of big-data phrenology"
The idea of mapping psychological functions to brain structures has a venerable history. Since the advent of neuroimaging techniques, the prospect of finding one-to-one correlations between psychological functions and brain structures regained vigor, and became the main goal of the relatively recent research field known as cognitive neuroscience. Unfortunately, as cognitive neuroscience develops, the ideal picture of finding one-to-one mappings from psychological functions to brain areas looks unrealistic. In a recent paper, Genon and colleagues (2018) argue that these difficulties stem from the fact that, up to date, scientists have followed a flawed top-down approach: the strategy of starting from a set of accepted psychological categories to then work their way down to the brain areas with which they are supposed to be correlated. To overcome the shortcomings of the top-down approach, Genon and colleagues argue instead for a bottom-up approach, whereby researchers begin with the “a priori defined construct of the brain region”, and work their way up to the psychological categories associated with it. Moreover, given the limitations of traditional piecemeal experimental methods from top-down approaches (e.g., lesion deficits, single fMRI studies, etc.), they favor using a “big-data” approach to identify the function that best characterizes a given brain region. In this talk, I will argue that despite its promise, their proposed bottom-up approach faces serious conceptual and empirical challenges.