An Electrophysiological Investigation into Temporal Factors in Human Perceptual Decision-Making

Abstract

At any given moment, the human brain receives a barrage of noisy sensory signals that convey important information about events taking place in the surrounding world. In order to function optimally in our environment we must use this sensory information to make perceptual decisions that guide our interactions with the external world. Convergent evidence from computational modelling and neurophysiology, which is reviewed in chapter one of this thesis, suggests that this process of perceptual decision-making is mediated by accumulating samples of noisy sensory information over time towards a decision threshold. Despite important advances that have been made in terms of understanding the fundamental mechanisms by which perceptual decisions are made, there are still many unanswered questions about how these mechanisms are affected by a wide range of factors. However, owing to recent breakthroughs in human electrophysiology, many of these questions can now be addressed by directly observing the decision-making process in the human brain through the measurement of freely evolving neural signatures of effector-selective and domain-general decision formation. By exploiting these newly characterised signals, the aim of this thesis was to examine the role of various temporal factors in perceptual decision formation. These include the impact of temporal uncertainty and prior experience of temporal structure on the onset timing of evidence accumulation, the role of urgency in perceptual decision-making under speed pressure, and mechanistic changes underpinning the effects of training on perceptual decision-making.

In the natural sensory world it is not always possible to predict precisely when a goal relevant sensory change will take place. Consequently, it can be challenging for the brain to determine when to start sampling evidence from a noisy sensory environment, particularly if the target sensory event has very low salience. Despite this, the majority of studies in the perceptual decision-making literature to date have focused on scenarios in which sensory events are both predictable and salient, leaving little ambiguity over when the brain should start accumulating evidence. Chapter two addressed this gap in the literature. It discusses crucial insights into the importance of accumulation onset showing that when the timing of a subtle goal relevant change in a stimulus feature is de-correlated across trials, observers are unable to accurately time the onset of evidence accumulation, much to the detriment of their performance. In addition, the findings of chapter two also suggest that accumulation onset timing can be adjusted strategically in order to optimise behaviour in spite of this overall performance limitation.

Chapter three built on the findings of chapter two by examining how the timing of accumulation onset is affected by recent experience of temporal structure under conditions of temporal uncertainty. Specifically, this chapter examined the sequential foreperiod effect, an example of intertrial dependency whereby the speed with which individuals respond to a sensory target is affected by the timing of that target on the previous trial. Such dependencies reflect our tendency to incorporate statistical information from the environment into our decisions. The findings of chapter three indicate that even when temporal structure is randomly varied across trials, observers nevertheless use recent temporal information to guide the onset of evidence accumulation giving rise to positive effects on performance under congruent conditions and negative effects on performance under incongruent conditions.

The aim of chapter four was to examine the hypothesis that the contingent negative variation (CNV), a frontocentral component of human EEG, indexes urgency in the context of perceptual decision-making. An urgency signal is one that grows in a time-dependent, evidence-independent manner reflecting the increasing impetus to response as an internal or external response deadline approaches. Such a signal is expected to increase statically in strength at baseline under conditions of speed emphasis, and to grow dynamically at an ever-increasing rate in anticipation of an impending deadline. In support of the hypothesis that the CNV represents such a signal, it exhibited precisely these characteristics.

Chapter five examined the impact of extended training on the mechanisms underpinning perceptual decision-making with results indicating that learning was mediated by two primary mechanisms. Firstly, there was a boost in the representation of sensory evidence and a corresponding increase in the rate of evidence accumulation. This mediated initial improvements in response time and led to fewer missed responses. Secondly, further improvement in accuracy was achieved by increasing the decision bound and thereby requiring more evidence to be accumulated. Such a change often entails slower responses but due to the increased rate of evidence accumulation, response times instead reached a plateau level and accuracy improvements alone were yielded.

Finally chapter six presents a general discussion of the main contributions of this thesis.