The benefits of noise: Stochastic resonance across behaviour and cognition

Abstract

Noise is ubiquitous and generally regarded as an annoyance in various biological and non-biological systems. However, noise can also produce some beneficial effects as observed in the stochastic resonance (SR) phenomenon. In SR, the presence of optimal amounts of noise can enhance performance. In neuroscience, the importance of SR has been shown via both neural (i.e., noise in the brain) and external noise (i.e., artificially introduced noise) to enhance physiological and cognitive functions. Despite the known advantages of noise, the ecological benefits of SR are still not completely understood. For instance, it is not clear how naturally occurring SR due to neural noise can affect performance and behaviours in everyday tasks. Also, whilst the ecological benefits of external noise exist in the auditory, and tactile/motor domains, no studies have shown the ecological benefits of external noise in our most dominant sense, vision.
Therefore, the current thesis aims to investigate enhanced performance due to 1) neural noise, that is, showing naturally occurring SR, and 2) external noise, that is, showing artificial SR in ecologically relevant tasks. To highlight the benefits of neural noise, this thesis examined neurotypical individuals with varying levels of autistic traits and related the findings to autism spectrum disorder (ASD), a condition characterised by elevated levels of neural noise. It is hypothesised that the high levels of neural noise in ASD can lead to enhanced performance in a visual detection task when there is no external noise present (i.e., this indicates the presence of naturally occurring SR due to high neural noise). However, it is also hypothesised that the high neural noise in ASD will result in impaired performance on the same task in the presence of high external noise (i.e., cannot achieve SR artificially due to high neural noise).
To outline the ecological benefits of external noise in the visual domain, we developed a novel ecological experimental SR paradigm using augmented reality (AR) and tested its effects on healthy and on individuals diagnosed with age-related macular degeneration (AMD). We investigated AMD which represents a visual disorder that is the leading cause of blindness in Australia. It was hypothesised that optimal amounts of visual external noise introduced in this device will enhance visual acuity in the healthy and clinical group for both binocular and monocular vision.
This thesis includes seven chapters, starting with the introduction (Chapter I), followed by a literature review (Chapter II), four research papers (Chapters III-VI) and a discussion (Chapter VII). The research papers present four studies that address the aims and hypotheses of this thesis. Study 1 (Chapter III) and Study 2 (Chapter IV) focus on showing the benefits of neural noise (natural SR) through studying links to autism traits. In study 1, we compared the performance of higher and lower autistic trait individuals (representing high and low neural noise levels respectively) on a visual detection task under varying external visual noise levels (stimulus noise). In this study, we first presented a neurocomputational model that predicted that artificially increasing stimulus noise results in artificial SR for individuals with low neural noise (e.g., neurotypical), but not for those with higher neural noise (e.g., autistic, or neurotypical individuals with higher autistic traits). It also predicted that at zero or low stimulus noise, individuals with high neural noise outperform those with lower neural noise, indicating the presence of a naturally occurring SR for the high neural noise group. To further test the model, we performed two experiments, one in a laboratory setting (Experiment 1) and one online (Experiment 2). Both groups showed SR due to stimulus noise in Experiment 1, and this did not support our model strongly. In Experiment 2, significant stimulus SR was not found in any group, however, participants with higher autistic traits outperformed those with lower autistic traits at low stimulus noise levels, which is consistent with our modelling. Study 2 provides a review of the high neural noise and SR hypothesis for ASD. Here, we looked for studies and presented additional data that showed enhanced performance for ASD over the neurotypical group in the presence of low external noise. Additionally, we highlight how the neurocomputational model for ASD presented in Study 1 can explain common ASD phenotypes.
Study 3 (Chapter V) and Study 4 (Chapter VI) investigate the ecological benefits of external noise in the visual domain. In Study 3, we introduced a novel experimental setup using AR technology to introduce visual external noise to enhance binocular vision in healthy and AMD groups. As hypothesised, optimal amounts of external noise enhanced visual acuity in both the healthy and clinical groups. In Study 4, we conducted an exploratory study and tested the same setup on healthy and clinical (AMD) monocular vision. Here, we did not see any SR effect for monocular vision thus, not supporting our hypothesis but giving us new insights. More specifically, our study indicates that a binocular component may be necessary for SR to be produced, such that it allows for a signal and noise interaction beyond the retinas.
Overall, the findings of this thesis outline the ecological benefits of both naturally occurring SR (neural noise) and artificially induced SR (external noise). The implications of these findings lie in the development of potential novel assessment and diagnostic tools for ASD, including an increased understanding of ASD behaviours, as well as the development of a novel noise-based visual prosthesis device that could significantly improve the quality of life of AMD patients

Date of Award	2025
Original language	English
Supervisor	Jeroen van Boxtel (Supervisor), Faran SABETI (Supervisor) & Maxime PELLAND (Supervisor)