ABSTRACT
A core question underlying neurobiological and computational models of behavior is how individuals learn environmental statistics and use them for making predictions. Treatment of this issue largely relies on reactive paradigms, where inferences about predictive processes are derived by modeling responses to stimuli that vary in likelihood. Here we deployed a novel proactive oculomotor metric to determine how input statistics impact anticipatory behavior, decoupled from stimulus-response. We implemented transition constraints between target locations, and quantified a subtle fixation bias (FB) discernible while individuals fixated a screen center awaiting target presentation. We show that FB is informative with respect the input statistics, reflects learning at different temporal scales, predicts saccade latencies on a trial level, and can be linked to fundamental oculomotor metrics. We also present an extension of this approach to a more complex paradigm. Our work demonstrates how learning impacts strictly predictive processes and presents a novel direction for studying learning and prediction.
Footnotes
↵* giuseppe.notaro{at}unitn.it
3 The four Beta values were. β1:t(20) = 6.64, p < .001, d = 1.45; β2:t(20) = 5.33, p < .001, d = 1.16; β3:t(20) = 3.17, p = .01, d = 0.69; β4:t(20) = 4.41, p = .001, d = 0.96.