Detecting cognitive interactions through eye movement transitions

Abstract

Many eye tracking studies are designed to reveal the co-activation of representations in interactive cognitive systems, such as lexical candidates in the human language system. Such co-activation is presumed to occur within participants on a trial-level. However, traditional analyses mostly use the viewing tendency of participants over trials (e.g., average fixation proportions to visual referents), rather than individual fixation patterns within trials (e.g., consecutive fixations across visual referents). Instead, we argue that assessing temporal dependencies of eye movements between relevant referents is better suited for detecting co-activation in an interactive system, compared to other oft-used methods that may falsely accept or reject interaction hypotheses. We demonstrate how to analyze eye movement transitions with a multilevel markov modeling approach using a relevant experimental example (bilingual co-activation in a visual world paradigm), and discuss the practical applications and theoretical implications when analyzing transitions in any type of eye tracking data.