Research network for the Interdisciplinary Study of Predictive Processing in Memory and Perception (PPiMP)

• Applicant: Dr. Helen Blank
• Subject Area: General, Cognitive and Mathematical Psychology ; Human Cognitive and Systems Neuroscience ; Cognitive, Systems and Behavioural Neurobiology
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 442810744

It has long been proposed that predictive processes form a fundamental principle of cognitive function in Psychology (von Helmholtz & Nagel, 1909). Accordingly, to achieve successful perception, the brain combines prior predictions about the environment with often noisy or ambiguous sensory inputs. The current interest in the field to investigate the neural principles of predictive processing has been revived by recent advances in cognitive and computational neuroscience. There is a vibrant debate on how such predictive processes are employed by the human brain and how they are implemented in a neural architecture (Aitchison & Lengyel, 2017). Predictive Coding theories present an influential framework in this debate (Rao & Ballard, 1999) and are applied to clinical research (Sterzer et al 2018). According to these theories, cognitive functioning is hierarchically organised: higher levels process more abstract information and lower levels more concrete sensory data. Predictions are fed backwards from higher to lower levels in this hierarchy, while prediction errors (i.e., discrepancy between prediction and evidence) are passed forward to higher levels to modify future predictions and minimize errors. Over the past decade, Predictive Coding has gained significant traction across different cognitive domains, including perception, memory, and language, and has recently been suggested as a unifying theory and global cognitive architecture for the brain (Friston, 2010). However, to date little empirical research has taken into account more than one cognitive domain. As a result, current models of predictive processing are implemented and used differently across domains, e.g., as cognitive models in learning and memory, or as neuronal models in perception. How can these different applications be brought together? More specific theories are needed to explain where relevant priors come from and how priors are combined with sensory input. Furthermore, if Predictive Coding is understood as a unifying theory, it needs to be challenged comprehensively in the light of alternative accounts, such as sharpening or tuning, in which neural representations of expected sensory signals are enhanced without relying on feedback mechanism of error signals (Lee & Mumford, 2003). The PPiMP network will bring together early-stage researchers from a variety of different domains (i.e., perception, memory, language, and clinical samples) with complementing expertise in predictive processing to tackle these issues. The PPiMP has the following three key objectives: 1) to challenge predictive processing accounts across cognitive domains and against other theories; 2) to facilitate new infrastructure and culture of dialogue in research practice, e.g., by creating an open exchange-platform, and 3) to support young academics to become established and independent researchers, e.g., by preparing joint grant applications.

• DFG Programme: Scientific Networks