Abstract
Pupil and gaze measurements have been shown to contain a surprising amount of information related to visual and cognitive processes. For instance, changes in pupil size have been associated with perceptual switches (Einhäuser et al., 2008), target detection (Privitera et al., 2010) or cognitive load (Kahneman & Beatty, 1966) and it has been suggested that microsaccades could reflect the orienting of attention (Laubrock et al., 2010). Here we consider a multivariate pattern analysis to study the diagnosticity of various pupil and gaze measurements available from binocular eye-tracking recordings to predict the perceptual state of human participants and the occurrence of switches over time during a binocular rivalry task. Our paradigm involves the presentation of a brighter stimulus in one eye and a non-fusable less bright stimulus in the other eye, inducing an alternation between two percepts.
In agreement with previous results (Fahle et al., ECVP 2010), we find that a perceptual switch to the darker stimulus results in significant pupil dilatation, starting significantly before manual report. On the other hand, a perceptual switch to the brighter stimulus results in an opposite, but similar amplitude, pupil constriction thus dismissing a motor artifact explanation (Hupé et al., 2008) and clearly vouching for an association between rivalry switches and pupil size changes. We further demonstrate that it is possible to reliably decode perceptual switches on a single-trial basis. Using a feature selection method, we investigate the predictive power of various oculometric and pupil measurements. Overall our results confirm the link between changes in pupil size and changes in perceptual states and bear important consequences for future studies of binocular rivalry.
This research was funded in part by a DARPA grant to TS (DARPA-BAA-09-31).