Detecting changes in visual scenes during saccades: Replicating and extending John Grimes's experiments

In the mid-1990’s, John Grimes’s work on how saccades influence visual perception revealed unexpected results: subjects failed to detect retrospectively obvious scene changes that occurred during saccades, despite having fixated before and after on the changed stimulus (Grimes, 1996). Since his original work, many studies have demonstrated other ways of inducing change-blindness, supporting the general finding that fixations upon changed stimuli can fail to yield detection of those changes. However, to date, no attempt has been made to replicate the original work by Grimes. In this investigation, we replicated, refined, and extended the original Grimes paradigm in two laboratories (at the University of Florida and Lingnan University) to evaluate the relationship between saccades and visual perception. In our task, the first of several planned experiments, 21 participants at each institution (42 total) viewed a newly-developed set of 100 change-blindness images. Participants were given instructions that their primary task was to remember as many details as possible, but they needed to report whenever they noticed that something changed in the image. For half of the images, one object in the image changed during a pseudo-randomly determined saccade. For the other half of the images, nothing changed. Results revealed that the “miss” rate for subjects across both sites was significantly above zero. Follow-up analyses showed that the size of the image redraw area, the amplitude of the saccade, the redraw time during the saccade, and the distance between gaze position and change center were all factors that influenced change-detection performance. Together, our results (1) replicate Grimes’s work, and (2) reveal important eye-movement and image-based factors that contribute to change-detection failures. We will discuss how these results can inform current theories of visual awareness (e.g., Higher-Order Theory and Integrated Information Theory), and how this paradigm can incorporate metacognitive measures to provide additional insights.