Online mouse tracking as a measure of attention in videos, using a mouse-contingent bi-resolution display

Abstract

Data on human visual attention is increasingly collected online, but there are limited tools available to study attention to video stimuli in online experiments. Webcam-based eye tracking is improving, but it faces issues with precision and attrition that prevent its adoption by many researchers. Here I detail an alternative mouse-based paradigm that can be used to measure attention to videos online. This method uses a blurred display and a high-resolution window centered on the user’s computer mouse location. As the user moves their mouse to view different screen content, their mouse movements are recorded, providing an approximation of eye movements and the attended screen location. To validate this Mouse-Contingent Bi-Resolution Display (MCBRD) paradigm, mouse movements collected from online participants watching twenty-seven videos were compared to eye movements from the DIEM dataset. Display settings of window size and blur level were manipulated to identify the settings that resulted in mouse movements most similar to eye movements. This validation study found differences in speed between mouse and eye movements, but similarities in attended regions of interest, especially when the MCBRD screen was blurred with the highest tested Gaussian blur sigma of 0.45 degrees of visual angle. These results suggest that the MCBRD paradigm can be used to measure what regions viewers find salient, interesting, or visually informative in online videos.