Using CRISP to model global characteristics of fixation durations in scene viewing and reading with a common mechanism

Abstract

Fixation durations vary when we read text or inspect a natural scene. Past studies suggest that this variability is controlled by the visual input available within the current fixation. The present study directly compared the control of fixation durations in reading and scene viewing in a common experimental paradigm, and attempted to account for the control of these durations within a common modelling framework using the CRISP architecture (Nuthmann, Smith, Engbert, & Henderson, 2010). In the experimental paradigm, a stimulus onset delay paradigm was used. A visual mask was presented at the beginning of critical fixations, which delayed the onset of the text or scene, and the length of the delay was varied. Irrespective of task, two populations of fixation durations were observed. One population of fixations was under the direct control of the current stimulus, increasing in duration as delay increased. A second population of fixation durations was relatively constant across delay. Additional task-specific quantitative differences in the adjustment of fixation durations were found. The pattern of mixed control of fixation durations obtained for scene viewing has been previously simulated with the CRISP model of fixation durations. In the present work, the model's generality was tested by applying its architecture to the text reading data, with task-specific influences realized by different parameter settings. The results of the numerical simulations suggest that global characteristics of fixation durations in scene viewing and reading can be explained by a common mechanism.

Keywords:

• Computational modelling
• Eye movements
• Fixation durations
• Reading
• Saccadic inhibition
• Scene perception

Acknowledgements

Data collection was supported by a grant from the Economic and Social Research Council (ESRC) of the UK to JMH (RES-062-23-1092). We thank Tim J. Smith for research assistance and discussion.

Notes

¹When modelling the data from the scene onset delay paradigm, we assumed a unidirectional adjustment of fixation duration: Difficulties in processing the scene during the delay period inhibit (i.e., delay) saccade initiation, leading to longer fixation durations (Nuthmann et al., 2010). In contrast, the toy simulations in Figure 2 describe a bidirectional (speed up and slow down) adjustment of fixation durations. Whether one-way or two-way effects occur in scene viewing is an open empirical question.

²Note that these parameters were fit by eye; no advanced fitting procedure was implemented.