Fig 1.
Recording epochs for pupillary responses.
Note that on Loss outcomes, there is no ensuing audiovisual feedback and therefore T2 cannot be modelled separately from T1. On the other outcomes, if the duration of audiovisual feedback was less than 2.5 seconds, the T1 and T2 epochs would overlap; see S1 Table for average durations.
Fig 2.
Change in pupil diameter by outcome type in a representative participant and the overall group.
(A) Pupil diameter trace for a representative participant through Baseline, and at the start (T1), and end (T2) of audiovisual feedback. Lines represent the different outcome types. Loss trials have no T2. (B) Mean pupil diameter for all participants. Lines represent the mean of different outcome types. Shaded areas represent the standard error between subjects for that point in time. (C) Predicted pupillary response to different slot machine outcomes, as percentage change from Baseline, minus the response for loss outcomes. Asterisks represent the effect of the outcome relative to the reference category, loss trials. Bars represent the 95% confidence interval. LDW = Loss-Disguised-as-a-Win, * p < .05, ** p < .01, *** p < .001.
Table 1.
Event-related change in pupil diameter as a percentage of baseline by trial number, prior credits, outcome type, and outcome phase.
Fig 3.
Illustrations depicting the sequence of on-screen events on free-spin bonus features and winning outcomes, for the two slot machine models used in the present study.
These illustrations capture the approximate colour schemes and image dimensions of the key game elements. Researchers requiring actual screenshots should contact the corresponding author. Losses-disguised-as-wins follow similar in-game processes to the winning outcomes shown, and loss outcomes are similar to wins during the reel spin but do not contain any audiovisual feedback. Note the contrasting luminance properties for Buffalo Spirit (as a light screen background) and Ice Empress (with a darker screen background).
Fig 4.
Buffalo spirit slot machine screen luminance by outcome phase.
Average screen luminance (L) component from CIELAB color space time-locked to event onset. L range = 0–255, Audiovisual Feedback Time ranged from 0–140,430 ms; a narrower range limited at 30,000 ms is presented for easier inspection and comparisons to T1. Ribbons represent one standard deviation above and below the mean of 7284 loss (79.2%), 909 win (9.9%), 942 loss-disguised-as-win (10.2%), and 62 bonus (0.7%) outcomes at a given point in Time. Baseline and Reel Spin outcomes do not align due to discarded frames in the latter.
Table 2.
Averaged slot machine screen luminance, full epochs.
Table 3.
Buffalo spirit averaged slot machine screen luminance, measured epochs.
Fig 5.
Ice Empress slot machine screen luminance by outcome phase.
Average screen luminance (L) component from CIELAB color space. L range = 0–255, Audiovisual Feedback Time ranged from 0–176,220 ms; a narrower range limited at 30,000 ms is presented for easier inspection. Ribbons represent one standard deviation above and below the mean of 3060 loss (82.0%), 406 win (10.9%), 229 loss-disguised-as-win (6.1%), and 36 bonus outcomes (1.0%) at a given point in Time.
Fig 6.
Scrambled images of slot machine outcomes.
L = objective luminance, cd/m2 = photometer measured brightness rating in situ ambient lighting.
Fig 7.
In the top row, lines indicate the mean pupil diameter from 200 to 2500 ms on Trial B for each participant (n = 12). In the bottom row, lines indicate the mean pupil diameter from 200 to 2500 ms on Trial B aggregated across participants. Pupil diameter is estimated as pixels as observed in the eye on the camera image, and is therefore not corrected for perspective.
Table 4.
Difference score medians, IQRs, Wilcoxon signed-rank test p-values and effect sizes.