The binary decision making process requires evidence accumulation and integration over time in order to choose one alterative over another. However, most environments are dynamic, causing newer accumulated evidence to differ from previous evidence. Such a change can be surprising and produce an arousal response. Surprising events that activate the arousal system often involve activity in the LC brainstem. This brainstem activity is correlated with a pupil dilation response. While the arousal system can cause a change in pupil diameter, other influences involved in the pupil light reflex can affect pupil size, such as luminance levels.

Using pupillometry as an indirect measure of activation of the arousal system, I created a regression model to determine whether the stability of an environment would affect the arousal state of a monkey. 

A monkey performed a dots-reversal random-dot motion task to simulate the decision making process. Throughout every trial of this task, a camera was focused on the monkey's pupil in order to determine its diameter at every point in time. In this task, a monkey fixates on a screen where during the first epoch of each trial a collection of moving dots appear. Some of the dots move in incoherent directions while others move coherently together left or right. At every change point, the dots reverse directions. For half of the trials there are a low frequency of change points (2), while the other half have a high frequency of change points (6). The environment with the low frequency of change points is more stable than that of high frequency. In the final epoch, the monkey is presented with one more potential change point where the dots may or may not switch directions, and the monkey is prompted to decide which direction the dots move in the final epoch.

In order to determine whether the evoked change in pupil size was significantly different between these two testing environments, I filtered the data using a low-pass filter and z-scoring across all trials. I ran a sliding window linear regression model that could account for several factors that could have an effect on pupil size, such as light levels, baseline pupil size, whether the final change point switched directions or not, and switch frequency. After running the regression and calculating the values of each parameter at every point in time, it could be concluded that the influence of switch frequency on evoked change in pupil size was non-zero. As a result, there is likely a difference between the evoked change in pupil size between environments of more versus less stability. It can be concluded that context stability may affect the arousal-driven pupil dilation response. Future directions may include performing a similar experiment on an auditory instead of visual task in order to eliminate the constriction effect of luminance levels from the testing screen itself, as well as possibly directly measuring activity from the LC instead of using pupil size as an indirect measure.