The perceptual prediction paradox : do actions exert a distinct influence on perception?

Perception of expected action outcomes has been thought for decades to be attenuated or ‘cancelled’. Underlying mechanisms are proposed to highlight informative unexpected events that signal the need to take new courses of action and update our models of the world. These accounts are thought to explain why we cannot tickle ourselves and unusual aspects of action in clinical populations. However, theories outside of action purport that predicted events are perceptually enhanced, allowing us to generate largely accurate representations of our noisy sensory world. This thesis presents a series of experiments investigating whether action prediction mechanisms necessarily generate attenuation, or whether they can enhance perception – as typically observed in domains outside of action. Chapter 1 outlines these theoretical accounts and evaluates the range of supporting literature across domains. Chapters 2 and 3 examine how action predictions influence tactile perception. Chapter 2 uses a force-judgement paradigm that typically demonstrates attenuation and controls for some potential confounds. Chapter 3 explicitly dissociates the predictive contribution of these effects with manipulations of statistical probability from broader sensory cognition domains. Finding in these experiments that action predictions can enhance, rather than attenuate tactile perception, Chapter 4 sought to investigate the neural mechanisms underlying visuomotor prediction in Primary Visual Cortex using functional magnetic resonance imaging (fMRI). Univariate and multivariate analyses demonstrated that expected outcomes are represented with superior fidelity than unexpected outcomes. Chapter 5 used high-resolution (7T) fMRI to investigate whether a key assumption of predictive coding theories could reconcile the opposing neural prediction effects reported in the literature. Chapter 6 summarises the findings of this thesis that sensorimotor predictions formed on the basis of statistical learning do not appear to exhibit a distinct influence on perception from other forms of prediction, and are instead likely generated by comparable underlying mechanisms