Purchase this article with an account.
Ruobing Xia; A unified network model for microsaccade and macrosaccade generation. Journal of Vision 2015;15(12):213. doi: https://doi.org/10.1167/15.12.213.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Featured by small amplitude and involuntariness, microsaccade shows similar profiles with normal saccade (macrosaccade) otherwise. For instance, despite the stochastic nature of microsaccade generation, its frequency and direction can be regulated by covert attention. Recent studies suggested that microsaccade generation might share the same circuit with macrosaccade. However, whether and how these two motor types are processed remains unclear. With this question, we built a continuous attractor network model to simulate both microsaccade and macrosaccade generation. In this network, neurons with various spatial preferences (from -90° to 90°) were organized in a one-dimensional axis, whose activity controlled the timing and target of saccades using a threshold mechanism. The recurrent connection pattern contained two components: a homogeneous pattern where connection weights decrease monotonically with the distance between neighboring neurons, which produced continuous attractor properties in the network; and a clustered connection pattern between neurons tuned to foveal regions, which built a point attractor so as to provide a tendency of keeping fixation. Additionally, a global inhibition mechanism was used to maintain stability and to generate competition between potential saccadic targets. In the simulation, each neuron received a small Poisson-noise input, and a peripheral visual input was present to a group of neurons to mimic the cueing effect in a covert attention task. We found that, while large visual input induced macrosaccades directly, microsaccades could be generated from noisy foveal activities. Interestingly, a small visual input would not lead to macrosaccades, but might bias the foveal activity indirectly due to the continuous attractor properties, and thus influence the frequency and direction of microsaccades. The simulation successfully replicated the signature of microsaccade distribution in covert attention experiments, indicating that this model could be used as a potential solution for explaining the attentional effect on microsaccade and unifying the microsaccade and macrosaccade generation.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only