About the Group

We make thousands of eye movements (saccades) every day. Simply directing eye gaze to a new location affects our attention and requires the visual system to update objects from ever changing retinal coordinates to stable (spatiotopic) world coordinates. This is termed the problem of visual stability and is still unsolved (Cavanagh, 2011). Many theories attempt to explain these processes as local (Wurtz, 2008) or global (Deubel, 1998) mechanisms, but direct comparisons based on individual experiments often produce conflicting results. Computational models allow us to specify our best current theories with rigorous mathematical precision and test them against data from multiple experiments. These models can also behave as simulators of the cognitive processes in question and generate new hypothesis that can be tested experimentally.

Current models lay on a continuum between very low level neural models and high level symbolic models: the former more biological and the latter more cognitive. We will build on current models of low level visual salience (Itti & Koch, 2000; Gordienko & MacInnes, 2016), temporal diffusion of eye movement responses to spatial stimuli (Ratcliff, 2008; MacInnes, 2016) and cognitive models of top down attentional control (Griffiths, Kemp, & Tenenbaum, 2008; MacInnes, Hunt, Clarke, & Dodd, under review) to create an integrated model simulation that can predict established attentional phenomena such as facilitation effects, Inhibition of return (IOR), saccadic remapping and saccade selection.

We propose a solution that would allow for the easy interchange of model depth through a system of modular components. This will allow us to integrate the best combination of neural and cognitive stages in the large scale model to test key aspects of visual stability. Our model will be trained and tested on existing data from a variety of experiments, and finally validated on data from new eye tracking experiments.