Katrina A. Wolters¹, Veith Weilnhammer¹, Jefferson Ortega¹, Andrey Chetverikov², David Whitney¹; ¹University of California, Berkeley, ²University of Bergen

Cognitive modes are distinct patterns of conscious attention and mental engagement that shape how individuals perceive their external environment. Recent evidence suggests that human perception alternates between two distinct modes: an internal mode, where perception is biased toward previously seen stimuli, and an external mode, where perception is aligned more closely to the external world (Weilnhammer et al., 2023). Hidden Markov models have previously been used to identify these dynamic modes of perception during orientation judgment tasks, and have found that they can identify states of attractive, dynamic templates (i.e. serial dependence; Weilnhammer et al., 2024). In the current study, we aimed to investigate whether similar dynamic states can be identified while observers perform an ensemble judgment task. Additionally, we expanded upon previous studies by using a novel modeling approach by fitting hidden Markov models to metathetic (circular) response data. Observers completed 800 trials of an ensemble judgment task, where 36 orientations were presented on the screen for 1000ms, followed by a 500ms delay and a response period. During the response period, participants were instructed to reproduce the average orientation of the 36 presented lines. Interestingly, our analyses revealed contrasting patterns of perceptual biases. We found an internal mode-associated state that exhibited a serial dependence effect, in which observers’ ensemble judgments were attracted toward the previously presented stimulus. We also found an external mode-associated state exhibiting a repulsive bias away from the previously presented stimulus. These results indicate that distinct patterns of perceptual bias (either attractive or repulsive) can arise during cognitive modes.

Acknowledgements: Supported in part by the National Institute of Health (grant no. R01CA236793) to D.W and (grant no. 1F99NS141343) to J.O.