Abstract
It has been argued that detecting a change between two images requires focused attention to the changed location (Rensink et al., 1997). We examine an alternative explanation based on the limitations of peripheral vision. If a change is discriminable in the periphery, it should be easily detected. Conversely, if the change is not discriminable in the periphery, it should be hard to detect. We evaluate this hypothesis using a `Spot-The-Difference' paradigm (Gur & Hilgard, 1975; Brunel & Ninio, 1997; Scott-Brown et al., 2000), where the goal is to identify differences between two images presented side-by-side. We gathered a set of 41 `Spot-The-Difference' puzzles from the web, 25 of which were derived from real-world photographs and the rest from illustrations. These puzzles, or image pairs, contained between 1 to 25 differences each. In Experiment 1, observers (n=6) viewed these image pairs freely, and for each pair, they identified differences by clicking on the corresponding image regions using a mouse. Observers were able to identify most differences (78%), with some differences being consistently harder to find than others. In Experiment 2, observers (n=7) viewed image pairs that were modified to contain either no differences or a single difference that was Easy, Medium, or Hard to find as established by Experiment 1. Observers were not allowed to move their eyes, and for each image pair, they indicated the presence or absence of a difference between two cued peripheral image regions using key presses. The accuracy at detecting a difference peripherally was significantly higher for the Easy differences (70%) than for the Medium (60%) or Hard (55%) differences (chance = 50%). These results support the hypothesis that performance on change detection tasks is at least partially constrained by low-level peripheral discriminability.
Meeting abstract presented at VSS 2014