Detecting changes between real-world objects using spatiochromatic filters

Abstract

Our ability to detect change between real-world scenes is rich in information about how we represent the visual world. Of particular recent interest has been the information also available from change detection failures, cases in which observers are seemingly "blind" to the occurrence of a change (Rensink, 2002;. Whereas correct change detection judgments imply a successful representation and comparison of the objects undergoing change, systematic failures to detect change may indicate weaknesses or holes in a representational structure.In the present study, I use behavioral and imageprocessing techniques to explore the representational constraints revealed by a change detection task. I will first discuss three forms of representational limitation likely to affect change detection behavior, loosely characterized as memory-, attention-, and similarity-based constraints. I will argue that renewed attention should be paid especially to the visual similarity relationships between the objects undergoing change. Second, I will explore the theoretical implications that these various forms of constraint have for capacity-limited models of change detection, arguing that higher level processes or parameters should be assumed only after similarity-based constraints have been found inadequate for explaining the observed data. Third, I will report a behavioral experiment demonstrating a similarity-based constraint, one involving an effect of orientation and categorical similarity on change detection. Last, I will outline an imagebased representational substrate capable of capturing these effects of visual similarity on real-world change detection performance. This f inal contribution of the study is particularly important in that a simple technique is described that can be used to quantify the visual similarity relationships between pre-and postchange patterns regardless of their featural complexity. Moreover, this technique is framed in terms of a computational model. Despite the prominent use of realistic stimuli in the change detection literature, no computationally explicit model has been demonstrated to work for real-world changes. The present work is intended to help bridge this gap between change detection theory and the stimuli commonly used to investigate this behavior. Sources of Representational Constraint in a Change Detection TaskThe representational weaknesses revealed by change detection failure are described most often in terms of memory and attentional constraints. Memory-based detection failures will result if the objects undergoing change are not represented in an enduring form (Grimes, 1996;O'Regan, 1992), or if a coherent representation of the prechange object does not exist at the time of the postchange comparison (Rensink, 2000a). Memory might also constrain performance if the number of objects in the prechange scene exceeds a capacity limit imposed by working memory (but see Rensink, 2000b), or if the build-up I thank George McConkie and Gary Wolverton for the stimulus presentation environment used in...

Eye movements serialize memory for objects in scenes

“…Rather than an extremely sparse representation (O'Regan, 1992), change detection might therefore be served by a fallible but relatively dense representation extending over the last seven objects fixated during viewing, and possibly many more. In this sense, our memory-constrained view is consistent with recent explanations of change detection suggesting that information exists for many objects in a scene, but that this information is impoverished and not sufficient to affect performance on every memory task (Angelone, Levin, & Simons, 2003;Simons, Chabris, Schnur, & Levin, 2002;Zelinsky, 2003). If observers fail to fixate both the pre-and postchange objects during scene viewing, even prerecency memory would be unavailable to the task, and the probability of detection should approach chance (Henderson & Hollingworth, 1999b;Hollingworth, Schrock, & Henderson, 2001; but see Zelinsky, 2001).…”

Memory for sound, with an ear toward hearing in complex auditory scenes

“…2, performance improved when the object that changed was more acoustically distinct from the sound it replaced (cf. Zelinsky, 2003). But the acoustic manipulation had no effect on object-encoding performance, even though it resulted in more spectral differences within one of the scenes.…”

Effect of Decision Load on Whole-Display Superiority in Change Detection

“…In many studies, researchers demonstrated that even when participants were unable to detect a change, both explicit measures (i.e., recognition memory task) (Mitroff et al, 2004;Yeh & Yang, 2008) and implicit measures (i.e., perceptual identification task) (Silverman & Mack, 2006;) revealed that the pre-change representation was preserved in memory. This evidence suggests that CB can occur because of retrieval and comparison failure (Hollingworth, 2003;Mitroff et al, 2004; or decision difficulty (Wilken & Ma, 2004;Yeh & Yang, 2008;Zelinsky, 2003).…”