Visual perception in schizophrenia: assessing predictive processing in the earliest stages of the visual cortical hierarchy

PROJECT SUMMARY Schizophrenia is a serious psychiatric disorder characterized by hallucinations, delusions, and disordered thinking and behavior that impairs daily functioning. As per the NIH’s Research Domain Criteria framework, we might gain a deeper understanding of this disorder by focusing on basic dimensions of functioning and how they relate to behavior and neural systems. With this in mind, research in recent years has sought to link psychosis with dysfunction in “predictive perception”. According to this framework, perception is an intrinsically predictive process that involves inferring the causes of our sensory inputs by combining those inputs with prior beliefs about the states of the world. And it has been proposed that psychosis is the result of maladaptive inferences based on inappropriately weighting prior beliefs relative to sensory input. However, testing this idea is complicated by the challenge involved in interpreting neurophysiological measures of perception in terms of feedforward sensory activity vs top-down predictions, and by the heterogeneity seen across individuals with schizophrenia. With these issues in mind, we propose to use innovative stimuli to derive EEG responses from the earliest stages of human visual cortex that are not overly complex in terms of their hierarchical generative architecture, and to relate those responses to measures of psychosis across healthy individuals, patients with schizophrenia, and patients with bipolar disorder. Specifically, we will collect high-density EEG from two innovative experiments. The first experiment aims to derive so-called “perceptual echoes” from human visual cortex. These “echoes” are novel measures of recurrent activity in early visual cortex that have been modeled as outputs from a predictive coding architecture. The second aims to leverage a powerful visual illusion to dissociate weak bottom-up stimulus changes from strong top-down predictions. Using the data from these experiments, we will compute impulse response functions that characterize early visual processing. To isolate an index of top-down feedback for each experiment, we will use comparative analyses (comparing impulse response function characteristics across latencies for Aim 1 and across experimental conditions for Aim 2). We will supplement our understanding of these indices – and produce additional dependent measures – by modeling the causal connectivity architecture underlying our impulse response functions. Finally, we will use linear mixed effects modeling to test the hypothesis that top-down predictive processing (as indexed by both impulse response functions and connectivity) relates to positive symptoms of psychosis transdiagnostically. Overall, the project aims to produce robust and interpretable indices of top-down predictive perception that will deepen our understanding of the symptomatology and underlying pathophysiology of psychotic disorders, which in turn, could improve early detection or provide biologically inspired alternatives to the current DSM nosology.