Work

Perceptual Exploitation of Statistical Structure in Multisensory Signals

Public Deposited

Downloadable Content

Download PDF

Efficient and accurate processing of internally- and externally-generated information is enhanced by the presence of multisensory signals that can provide redundant information about percepts or events. However, efficient usage of multisensory signals requires implicit perceptual knowledge of the potential or likely relationships between signals encoded within each sensory modality. If the system has implicit or situationally-inferred knowledge about likely relationships between signals, it can exploit this knowledge to encode relevant information more accurately, more efficiently, or with greater sensitivity through crossmodal facilitation. Alternatively, knowledge of signal correlations can be used to attenuate processing of task-irrelevant signals that are predicted by other correlated signals. Therefore, learned or situationally-inferred perceptual knowledge about the statistical structure of multisensory signals plays a critical role in producing the perceptual and cognitive benefits of multisensory stimulation. Here, I identify two mechanisms used by the perceptual system to optimize perceptual and cognitive processing in accordance with the statistical structure of multisensory signals. First, I demonstrate that the perceptual system uses reliable correlations between mouth shapes and oral resonances to facilitate the detection of auditory speech signals. This correlation is likely learned in infancy and reinforced throughout the lifetime to enhance speech perception under adverse conditions. Second, I demonstrate that rapidly inferred relationships between self-generated actions and their sensory results are used to efficiently allocate attention away from expected distractions during cognition. Because the stimuli that were linked to actions in this experiment would never be naturally linked in the same manner, this result is likely driven by situationally-inferred knowledge of statistical relationships between signals. Together, these results indicate that implicit perceptual knowledge of the statistical structure of multisensory signals, learned over the course of either a few seconds or multiple years, can be used to facilitate both perception and cognition in multiple ways depending on the task at hand.

Last modified
  • 10/26/2018
Creator
DOI
Subject
Keyword
Date created
Resource type
Rights statement

Relationships

Items