Purchase this article with an account.
Braden Purcell, Jeremiah Cohen, Richard Heitz, Jeffrey Schall, Gordon Logan, Thomas Palmeri; Developing a neuromimetic accumulator model of perceptual decisions. Journal of Vision 2009;9(8):831. doi: 10.1167/9.8.831.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Neurons in the frontal eye field (FEF) contribute to decisions about where and when to shift attention and gaze. Visual neurons differentiate a task-relevant stimulus from distractor stimuli. Movement neurons trigger a saccadic eye movement when activity reaches a fixed threshold. A database of FEF visual and movement neuron activity from three macaque monkeys performing a singleton visual search guided the development of an accumulator model of perceptual decision-making. Previously, we used the activity of FEF visual neurons as input to stochastic accumulator models representing hypothetical movement neuron units (Purcell, B.A., R.P. Heitz, J.Y. Cohen, G.D. Logan, J.D. Schall & T.J. Palmeri (2008) Modeling interactions between visually-responsive and movement-related neurons in FEF during saccade visual search. Vision Sciences Society 8: 1080.). These simple models successfully accounted for response times during visual search, but lacked neural plausibility. In the present work, visual neuron activity drove more complex stochastic accumulator models. The models varied on whether hypothetical movement neuron units (1) accumulated visual input independently or competitively, (2) were subject to self-inhibition or (3) tonic inhibition that acts as a gate on visual input by suppressing the accumulation of activity below a certain level. Accumulator models with different architectures accounted for the distributions of response times. However, by comparing the dynamics of model activation to the activity of FEF movement neurons, we resolved this model mimicry. Independent or competitive accumulation of visual activity with tonic gating inhibition provided the best account for behavioral and neural data. The complexity necessary to account for both response times and the form of neural activity indicates that simple diffusion or race models of sensory decision processes may be inadequate.
This PDF is available to Subscribers Only