Fig 1.
Pavlovian probabilistic reversal learning task and independent pupil signals.
(A): Example trial. Participants continuously fixated a white dot at the center of the screen. After 500ms, the dot turned into a cue (either purple or yellow) that signaled for 1000ms a monetary outcome with 80% validity. After 3000ms, the monetary outcome (either reward or loss) was indicated by a sound. (B): Example run. The participant monitored the reward contingency across trials and reported a detected reversal in the reward contingency with a keypress. Yellow and purple lines indicate the reward probability of the cues, which reversed three times during this particular run. These reward contingency reversals constituted four reversal blocks: one block prior and 3 blocks after the experimental reversals. Dashed arrows indicate the moments in time the participant reported a detected reversal and highlight the correspondence between reversal detection and increases in tonic pupil size (see panel C). Grey bars below the cue reward probabilities indicate the latter half of trials within a reversal block that were used in the analyses of transient pupil responses. (C): Example of independently filtered tonic and transient pupil signals across a run.
Fig 2.
(A): Average hit rate and false alarm rate across subjects (B): Response time distribution of correctly detected reversals across subjects. (C): Relation between the hit rate and correct reversal detection time. (D): Relation between correct reversal detection time and detection time variability. (E): EBR distribution across subjects. (F): Blink density across a run of 8 minutes averaged across subjects. (G): Mutual information (MI) between experimental and reported reversals plotted as a function of time-shifts in the reported reversals. Across subjects, shifting the reported reversals 5 trials back towards the experimental reversal (at trial shift = 0) resulted in the highest MI estimate (dashed line). Time units on x-axis are trials. Solid lines reflect individual MI estimates, colored by individual EBR rank (low EBR = light gray, high EBR = dark gray). (H): Individual MI estimates projected on the average MI estimate were positively related to EBR, indicating that individuals who detected reversals more consistently across runs had relatively higher EBR (low EBR = light gray, high EBR = dark gray).
Fig 3.
Arousal-based pupil responses during value expectation and outcome evaluation show different correlation patterns with EBR.
(A): Reward and loss expectation elicited identical pupil responses. (B): Reward expectation correlated significantly with EBR, indicating that reward expectation elicits stronger pupil dilation in individuals with low compared to high EBR. (C): Violated reward and loss expectations resulted in unsigned, transient pupil dilation. No dilation differences were observed between U- and U+ responses, which was reflected in the ΔU response. (D): ΔU correlated significantly with EBR, indicating that the variability in the ΔU response related to individual differences in EBR (E & F): A median split on individual differences in EBR visualizes the correlation between ΔU and EBR, where pupil dilation response patterns to unpredicted outcomes reversed for individuals with low compared to high EBR. Horizontal significance designators indicate time points where p<0.05. Error bars are standard error of the mean. Statistics based on cluster-based (correlation) permutation tests, n = 1000.
Fig 4.
Tonic, report-locked pupil responses track reversal detection and correlate with EBR.
(A): Reversal detection was reflected in a tonic, report-locked response starting 27s. prior until 13s. post-event. No significant pupil response was observed at the time of the experimental reversal (both the report and experimental reversal are plotted at t = 0). A positive correlation between the report-locked response and EBR (dashed line) starting 21s. prior until 7s. post-event indicated that individuals with relatively high EBR show stronger reversal detection responses (black solid horizontal significance designator; correlation values correspond to Pearson r values on the right y-axis). (B): Individuals with relatively high EBR showed stronger reversal detection responses that started earlier in time and lasted longer. Error bars are standard error of the mean. Statistics based on cluster-based (correlation) permutation tests, n = 1000.
Fig 5.
Blink density patterns and their relation to individual differences in EBR.
(A): No deviations were observed in blink density patterns prior to the detection of a reversal. At the moment of the behavioral report, blink density significantly increased 2s. pre-event until 6s. post event (black horizontal significance designator). Individual differences in EBR did not correlate with blink density patterns during the interval of reversal detection (grey dotted line, corresponding to the Pearson r values on the right y-axis). (B): No differences in blink density pattern were observed between reward and loss expectation events (left panel) nor between unexpected reward and unexpected loss events (right panel). (C): Blink density patterns did not significantly correlate with individual differences in EBR during the expectation of reward and loss (left panel), nor during the experience of unexpected reward and unexpected loss (right panel). All statistics based on cluster-based (correlation) permutation tests, n = 1000.