Transient physiological markers of arousal may help the brain update internal mental representations when the environment changes. The investigators found that electroencephalographic (EEG) activity and pupil dilation tracked inferred latent state transitions, were associated with reduced perceptual bias, and predicted context-dependent adjustments in learning, findings the researchers said support the latent states theory of arousal.
Researchers tested the theory in 63 healthy adults who completed a color prediction and reproduction task while undergoing simultaneous EEG and pupillometry. The task was designed to distinguish two forms of environmental change: changepoints, in which underlying conditions shifted and remained stable, and oddball events, in which surprising observations were transient and followed by a return to the previous state. Behavioral data were available for all participants, with usable pupil data from 60 participants and EEG data from 57 participants. The preregistered study evaluated five predictions using pupil dilation and the P300 event-related potential as indirect markers of arousal.
Both pupil dilation and EEG responses were found to increase during trials with higher model-derived probabilities of latent state transitions. Among the P300 components, the early frontal P3a reflected latent state transitions and their predicted behavioral consequences, whereas the later P3b component did not show the same relationship. Other event-related EEG signals were also associated with state-transition probability.
Stronger physiological evidence of latent state transitions was also shown to be associated with reduced perceptual bias. Participants relied less on prior expectations when surprising observations suggested that the underlying state of the environment had changed. This association was observed with both EEG- and pupil-derived measures across changepoint and oddball conditions, and combining the two physiological measures strengthened the relationship.
Learning responses differed according to environmental context. Researchers found that stronger physiological state-transition signals were associated with greater learning following changepoints, when updating beliefs was adaptive, but with reduced learning following oddball events, when isolated surprising observations were less informative about future outcomes. The combined EEG and pupil measure showed the strongest relationship with these bidirectional learning adjustments, matching the study's preregistered predictions.
Additional analyses supported the remaining preregistered hypotheses. Researchers found that pupil dilation aligned to the prediction phase was greater following oddball than changepoint events, consistent with an internally generated transition back to the previous latent state following a transient anomaly. Participants with stronger delayed pupil responses also demonstrated larger differences in learning between the two task conditions. In addition, participants whose arousal signals more strongly predicted reductions in perceptual bias also showed stronger arousal-related modulation of learning, supporting the researchers' proposed common mechanism linking both behaviors.
The study was conducted in healthy adults performing a controlled laboratory task, which may limit the generalizability of the findings to more complex real-world settings or clinical populations. The investigators also noted that pupil dilation and EEG are indirect, nonspecific physiological markers and do not directly measure activity of the locus coeruleus–norepinephrine system proposed by the latent states theory. The researchers acknowledged that other neuromodulatory systems may contribute to the observed physiological responses and said future studies using direct neural recordings or causal manipulation will be needed to test the proposed mechanism.
"Our results support the overarching idea that the brain optimizes learning and perception through dynamic transitions of internal latent state representations," wrote lead study author Tiantian Li, of Brown University, and colleagues. "Transient markers for arousal, including pupil area and the P3a, reflect these transitions and relate to learning and bias behaviours accordingly."
Disclosures: The authors reported no competing interests.
Source: Nature Human Behaviour
