Investigating the Underlying Mechanisms of the Motion Silencing Effect

Abstract

Detecting changes in our visual environment is important for many tasks, but the Motion Silencing Effect demonstrates that we cannot detect changes reliably when objects are moving quickly. This thesis investigates the underlying mechanisms of this visual illusion using behavioural and neural methodologies. After a detailed background introduction (Chapter One), we test if motion silencing applies to dynamic orientation changes, a novel feature for silencing (Chapter Two). Using a change detection task, we demonstrate that silencing occurs for this feature, marked by a decrease in change detection accuracy with increasing motion of the display. We then investigate if attention can modulate silencing, by using cues to direct spatial attention to a specific target (Chapter Three). Cueing attention with a valid cue improved performance in comparison to an invalid or neutral cue, suggesting that attention can lessen motion silencing. These behavioural results highlight that silencing can be modulated by a higher-level cognitive process.
Building on these behavioural results, we use Magnetoencephalography (MEG) to explore the underlying neural processing time course of the silencing illusion (Chapter Four). Using multivariate pattern analysis (MVPA), we provide evidence for being able to decode i) the global motion signal from 135 ms after stimulus onset, ii) the presence or absence of feature change from 640 ms after stimulus onset and iii) whether a trial was a hit or miss (silenced) trial from 810 ms after stimulus onset. Finally in the General Discussion (Chapter Five), we integrate these results with feedforward and feedback neuroanatomical models of perception, arguing that feedback signals from attentional regions and area MT alter how feature change signals are processed in early visual cortex, which affects the processing of these signals in areas further upstream. Therefore, besides a low-level receptive field bottleneck, silencing is likely generated by higher-level feedback.

Date of Award	1 Oct 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Chris Kent (Supervisor) & Iain D Gilchrist (Supervisor)