Consistent Eye Movement Patterns in Static and Dynamic Face Recognition: A Hidden Markov Model Study

Background/Objectives: Eye movements provide important insights into face processing. Hidden Markov models of eye movements (EMHMMs) are a relatively new approach that identifies common patterns across observers, moving beyond region-of-interest analyses. Prior EMHMM studies with static faces have typically revealed two strategies: a central "holistic" style and a feature-based "analytical" style. However, it is unknown whether such patterns extend to dynamic faces, which more closely reflect real-world viewing. This study is the first to apply EMHMMs to dynamic face recognition. Methods: Participants completed a face learning task in which half of the identities were presented as static images and half as dynamic videos. Eye movements were analysed using EMHMMs during both learning and recognition phases. Results: Two consistent patterns emerged across conditions: Central-focused and Eye-focused. A small subgroup displayed a third, central-plus-right-eye pattern when learning static faces. Eye movement patterns were largely stable across static and dynamic conditions, with only 16-27% of participants switching between them. Patterns were generally unrelated to recognition accuracy; however, participants adopting Eye-focused patterns during static learning performed better on static recognition. Conclusions: EMHMM-identified patterns generalise from static to dynamic faces, indicating strong stability in face-viewing behaviour across stimulus types. This finding contrasts with previous region-of-interest analyses suggesting greater differences between static and dynamic faces. By extending EMHMMs to dynamic faces, this study underscores the value of diverse analytical approaches for capturing eye movement behaviour and advancing understanding in face processing.