Investigating the Shared and Distinct Mechanisms Underpinning Perceptual Decision-Making and Metacognition

In order to successfully adjust to changes and learn in the sensory environment, humans must be able to detect whether or not decisions are based on reliable sources of information. Because we cannot always depend on the availability of objective feedback regarding the accuracy of our choices, computing a subjective sense of confidence to guide adaptive behavioural strategies is a necessity. Computational neuroscience has made substantial progress in parsing the basis of these metacognitive evaluations, which are reviewed in the introductory chapter of this thesis. Specifically, a set of mathematical models have been devised which hinge on the idea that the same sensory evidence accumulation processes that mediate perceptual decision-making might also govern the emergence of representations of confidence. Support for this prediction comes from invasive recording techniques in both non-human and human animals. However, many questions about exactly how these signals influence metacognitive judgements remain. For instance, it is unclear whether the decision variable is directly read-out to inform confidence, what the precise temporal dynamics of this evidence accumulation process are and what other information sources might influence the emergence of metacognitive representations. By adopting paradigms that allowed for the identification of these distinct neural signatures in non-invasive human brain recordings the current thesis aimed to elucidate the temporal dynamics as well as unique contributions of these different levels of processing along the sensorimotor hierarchy in the construction of confidence. The experiments presented in Chapter 2 aimed to address two gaps in current explorations of this relationship. Firstly, to extend on previous research by establishing that the sensory evidence accumulation process scales with choice confidence. This was confirmed in two experiments, showing that stronger build-up and higher amplitudes in this signal was associated with higher levels of confidence. This chapter further highlighted important methodological considerations for future work, by demonstrating that domain-general evidence accumulation does not simply continue to build until the choice reports and can be strategically terminated. Secondly, this study aimed to elucidate the role of motor level processes in the construction of confidence. Investigation of effector-specific mu/beta signals showed that, while contralateral motor preparation reached a fixed amplitude prior to response execution, higher activation for the unchosen alternative was associated with lower levels of confidence. These findings were complemented by an analysis of response conflict, which indicated theta band activity was similarly elevated on lower confidence trials. Additionally, the effects occur during a similar time window prior to the response, which provides a tentative indication that motor preparatory signals might indirectly influence metacognitive judgements through the expression of response conflict in theta power. Chapter 3 built on these findings and examined the temporal dynamics of post-decisional evidence accumulation across these different levels of decision formation and to what extent this process is contingent on external stimulus information. This was investigated through a paradigm where on some trials stimulus presentation continued after an initial perceptual decision. Exploration of domain-general evidence accumulation yielded two key insights. Firstly, post-decisional evidence accumulation occurred regardless of the presence or absence of stimulus presentations. Secondly, its temporal dynamics suggest this signal was comprised of two partially overlapping but qualitatively distinctive phases. Prior to the perceptual choice, sensory evidence accumulation mapped onto the choice alternatives, while the post-decisional evidence accumulation process more closely represented the accuracy of the first-order decision. An interpretation that was further supported by analysis of the differential characteristics of each stage in the emergence of confidence, changes of mind and first-order accuracy. Finally, an examination of effector-specific motor preparation during the post-decision window highlighted an important feature of these signals with respect to confidence. Specifically, a crossover between ipsilateral and contralateral signals was observed on change of mind trials, which reflects a resetting of the decision bound in favour of the final choice alternative. To our knowledge, this is the first demonstration of this behavioural phenomenon at the motor level, which could be further characterised in future research. A fundamental assumption of neuroscience is that behavioural paradigms tap into the same cognitive processes that govern daily-life decisions. The experiment in Chapter 4 aimed to contribute to the emerging literature exploring this assumption, capitalising on recent advancements in online data collection, and testing procedures. Specifically, behavioural data on a random dot motion task with confidence ratings was self-report measures of daily-life cognition and metacognition collected from a large representative sample, which builds on previous research that has largely focused on psychiatric symptom measures. Behavioural variables were decomposed into the mechanistic underpinnings of perceptual decision formation through computational models. The results from this study indicate that individual differences in multiple self-report measures were associated with metacognitive and perceptual decision formation variables. However, many of the expected relations between the parameters of computational models were not observed. While it is possible this reflects the limited ecological validity of the lab-based metrics and computational models, alternative explanations highlight the challenges of adopting these models in the study of inter-individual differences, which could be addressed in future research. Finally, Chapter 5 provides an overview of the key contributions of the research in this thesis as well as some avenues for future studies