Abstract
Many researchers are currently studying attention online, but have limited options for studying attention to dynamic video stimuli. Webcam-based eye tracking is improving, but is not yet accurate enough for many researchers’ needs. We offer a Mouse-Contingent Bi-Resolution Display (MCBRD) paradigm to measure attention to video in online experiments. In this paradigm, a circular high-resolution window is centered on the viewers’ mouse location, while the rest of the stimulus is presented at a lower resolution. The viewer can view different content by moving the high-resolution window with their mouse. Through this, mouse movements provide an approximation of eye movements and the attended location on the screen. This method, implemented in JavaScript, is one of the first to allow for video stimuli presented with a blurred periphery. By comparing these mouse movement results to eye movements collected from viewers watching 27 videos from the DIEM dataset, we aim to validate the MCBRD as a measure of attention, and determine the optimal blur and window radius settings to use when conducting mouse-contingent display studies with video stimuli. This validation tests window radii of 1.5, 3, and 4.5 degrees of visual angle, and Gaussian blur sigmas of 0.1, 0.3, and 0.5 degrees in a 3x3 within-subjects factorial design. A dynamic screen scaling method is used to present stimuli calibrated for visual angle relatively accurately to online participants. Preliminary analyses show mouse movements from this paradigm are similar to eye movements in terms of on-screen distance explored and areas of interest visited. Current analyses address the blur and window settings that produce mouse movements most similar to eye movements, measured with Normalized Scanpath Similarity and Dynamic Areas of Interest analyses. These validation results will help guide anyone wanting to use a mouse-contingent display to research attention to videos in online studies.