TY - CONF
T1 - Reconstructing pupillary dynamics during free-viewing of movies: the roles of pupil light and orienting responses
AU - Cai, Yuqing
AU - Strauch, Christoph
AU - Naber, Marnix
PY - 2023
Y1 - 2023
N2 - Pupil size dynamically adapts to changes in low-level visual features, as well as cognitive factors. When cognitive factors are manipulated in pupillometric research, low-level visual features are usually strictly controlled and the fixation position is required to be constant. This severely limits the range of options for experimental designs. Instead of controlling for low-level features, the current study attempts to model and predict pupillary dynamics based on complex changes in low-level features. Any unexplained variance can then be attributed only to higher-level factors. Forty healthy participants free-viewed a collection of 60-second movie clips while gaze position and pupil size were recorded. Visual features, namely luminance changes and color changes, were extracted across the movie frames. Following the idea of linear time-invariant systems, visual feature changes were convolved with pupil response functions (PuRFs) for light and orienting processes separately. To find the model that best fitted the actual pupil size recordings, we systematically varied the peak latency, width, and amplitude of the PuRFs. The fitted models demonstrated that pupil responses predicted by light matched the real pupil size changes. The median proportion of explained variance across data from all movie clips (n = 453) was approximately 30%. In addition, this proportion significantly improved to 34% after including transient pupil orienting responses to changes in color space. In conclusion, these results illustrate that our model of the pupil light and orienting response can explain a substantial proportion of variance of pupil size changes during unconstrained viewing of complex visual stimuli. Extensions of the current model could be used to produce baseline pupil traces that allow researchers to (1) control for confounds of low-level features in experimental designs, (2) discover which low-level visual aspects drive pupillary dynamics, and (3) investigate the effects of higher-order factors such as attention in isolation of confounding factors.
AB - Pupil size dynamically adapts to changes in low-level visual features, as well as cognitive factors. When cognitive factors are manipulated in pupillometric research, low-level visual features are usually strictly controlled and the fixation position is required to be constant. This severely limits the range of options for experimental designs. Instead of controlling for low-level features, the current study attempts to model and predict pupillary dynamics based on complex changes in low-level features. Any unexplained variance can then be attributed only to higher-level factors. Forty healthy participants free-viewed a collection of 60-second movie clips while gaze position and pupil size were recorded. Visual features, namely luminance changes and color changes, were extracted across the movie frames. Following the idea of linear time-invariant systems, visual feature changes were convolved with pupil response functions (PuRFs) for light and orienting processes separately. To find the model that best fitted the actual pupil size recordings, we systematically varied the peak latency, width, and amplitude of the PuRFs. The fitted models demonstrated that pupil responses predicted by light matched the real pupil size changes. The median proportion of explained variance across data from all movie clips (n = 453) was approximately 30%. In addition, this proportion significantly improved to 34% after including transient pupil orienting responses to changes in color space. In conclusion, these results illustrate that our model of the pupil light and orienting response can explain a substantial proportion of variance of pupil size changes during unconstrained viewing of complex visual stimuli. Extensions of the current model could be used to produce baseline pupil traces that allow researchers to (1) control for confounds of low-level features in experimental designs, (2) discover which low-level visual aspects drive pupillary dynamics, and (3) investigate the effects of higher-order factors such as attention in isolation of confounding factors.
U2 - 10.1167/jov.23.9.4813
DO - 10.1167/jov.23.9.4813
M3 - Abstract
SP - 4813
EP - 4813
ER -