Neural dynamics of perceptual competition

Abstract

This research aims to understand the neural dynamics and mechanisms underlying perceptual bistability. In perceptual rivalry, ambiguous sensory information leads to dynamic changes in the perceptual interpretation of fixed stimuli. This phenomenon occurs when participants receive sensory stimuli that support two or more distinct interpretations; this results in spontaneous alternations between possible perceptual interpretations. Perceptual rivalry has been widely studied across different sensory modalities including vision, audition, and to a limited extent, in the tactile domain. Common features of perceptual rivalry across various ambiguous visual and auditory paradigms characterise the randomness of switching times and their dependence on input strength manipulations (Levelt's propositions).

Binocular rivalry occurs when the two eyes are presented with incompatible stimuli and perception alternates between these two stimuli. This phenomenon has been investigated in two types of experiments: 1) Traditional experiments where the stimulus is fixed, 2) Eye-swap experiments in which stimulus periodically swaps between eyes many times per second~\citep{logothetis1996rivalling}. In spite of the rapid swapping between eyes, perception can be stable for many seconds with specific stimulus parameter configurations. Wilson introduced a two-stage, hierarchical model to explain both types of experiments~\citep{wilson2003computationala}. Wilson's model and other rivalry models have been only studied with bifurcation analysis for fixed inputs and different dynamical behaviours that can occur with periodic forcing have yet to be explored. Here I report 1) a more complete description of the complex dynamics in the unforced Wilson model, 2) a bifurcation analysis with periodic forcing. Previously, bifurcation analysis of the Wilson model with fixed inputs has revealed three main types of dynamical behaviours: Winner-take-all (WTA), Rivalry oscillations (RIV), Simultaneous activity (SIM). The results presented here reveal richer dynamics including mixed-mode oscillations (MMOs) and period-doubling cascade which corresponds to low amplitude WTA (LAWTA) oscillations. On the other hand, studying rivalry models with numerical continuation shows that periodic forcing with high frequency (e.g. 18 Hz, known as flicker) modulates the three main types of behaviours that occur with fixed inputs by the forcing frequency (WTA-Mod, RIV-Mod, SIM-Mod). However, the dynamical behaviour will be different with low frequency periodic forcing (around 1.5Hz, so-called swap), and in addition to WTA-Mod and SIM-Mod, cycle skipping and multi-cycle skipping behaviour exist, which can also lead to chaotic dynamics. This research provides a framework for either assessing binocular rivalry models for consistency checks against empirical results, or for better understanding neural dynamics and the mechanisms necessary to implement a minimal binocular rivalry model.\\

At present, it remains unclear whether the general characteristics of perceptual rivalry are preserved with tactile stimuli. I introduce a simple tactile stimulus capable of generating perceptual rivalry and explore whether general features of perceptual rivalry from other modalities extend to the tactile domain. In these experiments, vibrotactile stimuli consisted of anti-phase sequences of high and low intensity high-frequency pulses, each followed by a silent interval, delivered to the right and left index fingers. Participants perceived the stimulus as either one simultaneous pattern of vibration on both hands (SIM), or patterns of vibration that jumped from one hand to the other hand, giving a sensation of apparent movement (AM). For long stimulus presentations, perception switches back and forth between these two percepts. Furthermore, my results confirm that Levelt's proposition II extends to tactile bistability, and that the stochastic characteristics of irregular perceptual alternations agree with non-tactile modalities. An analysis of correlations between subsequent perceptual phases reveals a significant positive correlation at lag 1 (as found in visual bistability), and a negative correlation for lag 2 (in contrast with visual bistability).

In this study, a mathematical model of tactile rivalry is developed that focuses on accurately reproducing the dynamics of the perceptual alternations. The model of tactile rivalry presented here consists of two processing stages; first stage for producing perceptual alternations; and a second stage for encoding the percept types (SIM and AM). Putative neural populations of the first stage could be located early in the somatosensory pathway at brainstem nuclei, and the neural populations of the second stage could be located within area 3b of the primary somatosensory cortex, based on excitatory and lagged inhibitory components of their receptive fields. The powerful combination of bifurcation analysis along with optimisation tools have been used to tune certain features of the model, resulting in a good qualitative and quantitative match to my experimental data. As well as capturing the dynamical characteristics specific to the perceptual interpretations in tactile rivalry, the model presented here is able to produce the general characteristics of perceptual rivalry including Levelt's proposition, short-tailed skewness of reversal time distributions and a scaling property of this distribution's first three moments.