A Bayesian approach to dynamical modeling of eye-movement control in reading of normal, mirrored, and scrambled texts

Abstract

In eye-movement control during reading, advanced process-oriented models have been developed to reproduce behavioral data. So far, model complexity and large numbers of model parameters prevented rigorous statistical inference and modeling of interindividual differences. Here we propose a Bayesian approach to both problems for one representative computational model of sentence reading (SWIFT; Engbert et al., 2005). We used experimental data from 36 subjects who read text in a normal and one of four manipulated text layouts (e.g., mirrored and scrambled letters). The SWIFT model was fitted to subjects and experimental conditions individually to investigate between-subject variability. Based on posterior distributions of model parameters, fixation probabilities and durations are reliably recovered from simulated data and reproduced for withheld empirical data, at both the experimental condition and subject levels. A subsequent statistical analysis of model parameters across reading conditions generates model-driven explanations for observable effects between conditions.

Predictive modeling of the influence of parafoveal information processing on eye guidance in reading

“…Finally, it is important to stress that the results presented in this work heavily rely on the success and quality of the parameter estimation. Parameter inference based on individual readers' experimental data might be a breakthrough for process-oriented modeling 20,22,23 , since interindividual differences are often comparable in size to the observed effects. Here we exploited the full potential of interindividual differences by running predictive simulations separately for each participants.…”

“…Another prerequisite for the investigation of quantitative predictions is a reliable framework for statistical inference. Recently, we implemented a fully Bayesian framework for parameter inference for the SWIFT model 22,23 , which permits parameter identification based on experimental data from single readers in a statistically rigorous way. Therefore, we implement our assumptions on the interaction of fixation duration with foveal and parafoveal processing in the SWIFT model to investigate the potential of various mechanisms in explaining the integration of foveal and parafoveal information during reading.…”

“…2). In the lastest version of SWIFT 22,23 , only the processing of the currently fixated word in the fovea affects the random timer through foveal inhibition. In this study, we investigate additional parafoveal inhibition from activation a n+1 (t) to the right of the fixated word n. In two different variants, parafoveal inhibition can act on the timer either immediately (τ = 0) or with temporal delay (τ > 0).…”

“…respectively) tested the preview validity effect. The model simulations of the boundary experiment provide a strong test of possible pathways from foveal and parafoveal information to gaze control within a well-defined mathematical model under statistically reliable procedures 20,22,23 .…”

“…Artificial fixation sequences were simulated for all subjects with parameter combinations specific to the subjects' estimation results. For each sentence a different set of parameters was randomly sampled from the posterior distribution of the respective participant 23 , and per sentence 10 sequences were generated. For the analysis of preview effects, the simulated sentences were processed in the same way, as the experimental data.…”