Much of the research in the Visual Cognition Lab focuses on a phenomenon known as "Change Blindness." Under normal viewing conditions, changes to a scene produce a motion signal that is readily detected. However, when a change coincides with another event that disrupts the motion signal, observers are often blind to surprisingly large changes. Recently, a number of laboratories have revealed change blindness for changes occuring during saccades, flashed blank screens, blinks, motion picture cuts or pans, and even during real-world occlusion events. In addition to studies illustrating the magnitude of change blindness in motion pictures and the real world, current research in the Visual Cognition Lab examines a number of related issues.

      • Click here for an overview of change blindness quoted from Simons (2000)
      • For other information about change detection, visit the change detection database

Preserved representations
       One fundamental question underlying work on change detection is whether the existence of change blindness indicates the absence of a representation of the changed object or feature. In order to successfully detect a change, observers must represent the initial state of the changed object. However, the converse does not logically follow: even if observers fail to detect a change, they may still have a representation of the changed feature (See Simons, 2000 in Visual Cognition for a discussion of this issue). These projects investigate the nature of our representations in the face of change blindness. Do we retain an implicit representation even if we lack an explicit one? Do we have an explicit representation that for some reason does not lead to successful detection?
       In the lab, we use eye tracking to determine if observers look longer or more often than expected at changed objects even when they do not report the change. Because observers are often unaware of their eye movements, fixations can sometimes indicate change detection without awareness. Similarly, we have used analyses of response times to explore implicit change detection. In another project, we use techniques derived from the literature on eyewitness recognition (including lineups and questionnaires) to determine if observers show above-chance recognition for changed objects even when they cannot report the changes. Our studies of preserved representations use both intentional and incidental change detection tasks with stimuli including simple arrays of objects, photographs of natural scenes, motion pictures, and real-world events.

Category coding
       Another line of research examines the role of categorization in scene representation. Studies of scene perception suggest that we can rapidly identify the gist of a photograph (e.g., see work by Potter, Intraub, and Biederman). Research on the effects of scene context on object recognition has produced mixed results. Some early studies indicated that object recognition is facilitated by a consistent scene context, but more recent studies by Hollingworth & Henderson suggest that the perception of scene context and object recognition are relatively independent processes. Our research asks whether observers rapidly abstract the category and meaning of objects and scenes. Findings of change blindness for intentional change detection tasks (e.g., flicker) often lead to the conclusion that successful change detection requires the effortful abstraction of features and a comparison of such abstract representations. In other words, we lack a visual representation, so successful detection must depend on an abstract representation.
       These projects are designed to examine the nature of such abstractions. One line of studies examines how rapidly we can detect an inconsistency in a photograph -- that is, how rapidly can we detect that an individual object is inconsistent with the broader scene context. Another line of studies examines the spontaneous coding of categorical information about people. In a real-world change detection study, Simons & Levin (1998) found that observers were less likely to detect a change to a person if that person was a member of a different social group from the subject. In an ongoing series of studies, we are using video change detection tasks to explore the detection of changes to the race or sex of an actor as well as to explore the impact of embedding a common superordinate category (e.g., changing the race if both actors are dressed as construction workers). Another project examines the link between categorical perception of shapes (circle or square) and colors and change detection. If people spontaneously categorize objects, they should be more likely to detect a change that crosses a category boundary than an equally large change that does not.

Mechanisms of detection
       Although recent studies of change blindness have demonstrated that surprisingly large changes are not detected, relatively few studies have examined the mechanisms underlying change detection. To what extent do motion signals eventually allow detection in the flicker task? Are different types of change more or less salient? What role does the size of the change play in detection? This project uses standard psychophysical techniques to explore the relationship between change magnitude, rate of change, set size, and timing parameters in change detection tasks. The ultimate goal is to gain a better appreciation for the role of motion in change detection and to explore the interaction between the magnitude of the change signal and the need for attention. Experiments use simple arrays of precisely controlled objects as well as natural scenes.