Fig 1.
a) Gradual continuous performance task. Each block started and ended with 7 scrambled images. The participant was asked to respond to city scenes but not to mountain scenes. Each image continuously morphed into the next, with an 800-ms interval between 100% coherence levels. b) Behavioral results. Data are smoothed for display only. All measures showed a significant linear increase with time-on-task, p-values are listed in the lower right corner of each panel. Error bars represent the SEM.
Fig 2.
Trial-averaged modeled (red) and empirical (black) stimulus-related pupil dilations. The vertical dashed line represents event (stimulus or response) onset. Error bars represent the SEM.
Fig 3.
Baseline diameter and derivative.
a) Time-on-task effect for baseline diameter, and b) for diameter derivative. p-values are listed in the top left corner of each panel. The derivative is shown as variance-normalized but without the mean removed. Values below the horizontal dotted line indicate that on average the pupil is constricting, whereas values above the line indicate that the pupil is dilating. c) The relationship between pupil diameter and its derivative. Baseline pupil diameter plotted as a function of the derivative. Diameter is smallest when the pupil is dilating the fastest. USD: Units standard deviation.
Fig 4.
The relationship between pupil diameter and behavior.
Regression coefficients are shown per pupil measure and behavioral measure. USD: Units standard deviation. Error bars represent the SEM. *: p < 0.05; **: p < 0.01; ***: p < 0.001.
Fig 5.
Relationship between pupillary measures and behavior, before (a,b) and after (c,d) regressing out time-on-task.
Pupil data were z-scored within participants and blocks, aggregated across participants, and then divided up into 30 bins, and the behavioral data were sorted according to pupil diameter. Large positive values on the Y-axis indicate relatively poor behavioral performance. The initially linear relationship between baseline diameter and behavior becomes U-shaped after controlling for time-on-task, whereas the relationship between the derivative of baseline diameter and behavior remains linear after controlling for time-on-task. Straight lines are least squares regression lines, curved lines are fitted 2nd-order polynomials.
Fig 6.
The relationship between pupil diameter and behavior, after statistically controlling for time-on-task.
Regression coefficients per pupil measure and behavioral measure with time-on-task included as a variable of non-interest. Error bars represent the SEM. *: p < 0.05; **: p < 0.01.