Technical Program

Paper Detail

Paper: PS-2A.72
Session: Poster Session 2A
Location: H Lichthof
Session Time: Sunday, September 15, 17:15 - 20:15
Presentation Time:Sunday, September 15, 17:15 - 20:15
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Bayesian parameter estimation for the SWIFT model of eye-movement control during reading
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
Authors: Stefan A. Seelig, Maximilian M. Rabe, Noa Malem-Shinitski, Sebastian Reich, Ralf Engbert, University of Potsdam, Germany
Abstract: Dynamical models are increasingly contributing to the development of cognitive theory. Here we discuss an example for eye-movement control during reading. The SWIFT model (Engbert et al., 2005) is a stochastic dynamical system that predicts spatial fixation positions in a given text as well as fixation durations. We exploit the sequential nature of the likelihood for dynamical models. The likelihood function is a combination of spatial and temporal likelihood. While the spatial part is a pseudo-marginal likelihood, the temporal likelihood is obtained by numerical approximation. We use a fully Bayesian framework for parameter inference using an adaptive Markov Chain Monte Carlo (MCMC) procedure. As a result, we obtain model parameter estimates and credibility intervals on the level of individual readers. Interindividual parameter variations capture key features of the behavioral variability of eye movements observed in reading experiments.