A two-systems account of predictive processing
Max Planck Institute for Psycholinguistics
I will present a mechanistic two-systems account of prediction. I will argue that predictive processing can be broadly divided into passive and active mechanisms influencing pre-activation of upcoming input. Passive mechanisms include associations (at all levels from phonetic/phonological to event structures), resting state level dynamics, structure repetition, regularity detection, and sharpening/fine-tuning of perceptual/cognitive structures and procedures. Active mechanisms include reasoning (logical inference, ‘linguistic discourse processing’, etc.), simulation (‘conscious mental imagery’), and turn-taking. Passive mechanisms can in principle operate independently but feed into active mechanisms. Prediction thus is a dynamic interplay of (evolutionary old and new) passive and active procedures.