“Perceptual Decision-Making” and Neuroscience: A Brief Anatomy of a Pseudoscience

A significant swathe of the activity labelled “neuroscience” today concerns “perceptual decision-making.” The term putatively refers to a real-world process whose features may be revealed via experiment. But the experiments in question artificially and severely constrain, a priori, the form and content of the “data,” ensuring that it is binary, categorical, and forms a rough sigmoid when correct “decisions” are plotted against stimuli treated as varying along a single, arbitrary dimension. These by-design sigmoidal datasets are then interpreted as imperfect reflections of corresponding binary, categorical, unidimensional processes in the brain. The job of the practitioner is to recover the “true” sigmoid hiding in the data - the true sigmoid being whichever of a number of standard sigmoids the practitioner chooses to fit, post hoc, with the help of free parameters. This procedure, self-insulated from real-world pushback, lacks methodological, functional and empirical validity, and its associated ad hoc assumptions cannot be reconciled with the most basic mental facts. Results have low to no replicability and reproducibility. The paradigm has been extended to individual neurons using p-hacking-type exploration of data.