A Biologically Grounded Framework for Testing Competing Models of Hallucinations

Hallucinations are false percepts that are common across clinical and nonclinical populations. Clinical hallucinations can be debilitating, difficult to treat, and often predict poor functioning. The limited success of current treatments for hallucinations is due in large part to our immature understanding of their underlying circuit and cognitive mechanisms. While several mechanistic theories of hallucinations have been proposed, individual theories rarely account for robust empirical findings while also generating quantitative and testable predictions. These elements are often sacrificed (including in our previous work) to develop theories that are broad enough to accommodate multiple areas of research. However, as a consequence, these theories have reduced cognitive and biological detail and limited explanatory value because they are difficult to test and nearly impossible to falsify – at odds with a core premise of computational psychiatry. Here, we present a mathematical framework that helps formalize a range of hallucination theories via biologically grounded computational models. We constrain the space of candidate models using well-replicated findings from the basic, preclinical, and clinical neuroscience literatures. We then demonstrate how this framework can inform the development of richer behavioral paradigms and analytical approaches that reveal unique signatures of each model, allowing experimenters to directly pit competing models against each other. Overall, we present a general framework to compare and falsify candidate mechanisms underlying hallucinations, a key prerequisite to advancing our current understanding of this psychotic phenomenon and ultimately identifying more precise treatment targets.