Abstract
The ability of observers to detect the presence of more than one change at a time was tested using a sequence of displays that always contained two simple items (e.g. rectangles) with constant locations. Each feature that was examined (orientation, color, and size) had two possible values for each item: A or A′ for item 1, and B or B′ for item 2. On half the trials the display sequence was such that only one change occurred at a time (e.g. AB, A′ B, A′B′, AB′); on the other half, it was such that a pair of changes occurred on every other transition (e.g., AB, A′B′, A′B, AB′). The task of the observer was to determine whether the display sequence in a given trial contained any change pairs.
Brief (100 ms) blank fields were presented between displays so that changes could not be detected via transients, but instead required focal attention (Rensink et al, 1997). Since only two items were present in any display, and since attention can be given to 3–5 items at a time, this technique allowed a determination of the extent to which change can be perceived at any moment, even when all items are attended.
Detection accuracy was measured as a function of display duration. For displays shorter than 200 ms, detection of change pairs was always poor. This is consistent with the temporal resolution needed to perceive simultaneous feature pairs (Holcombe & Cavanagh, 2001). For longer displays, an interesting pattern emerged: whereas performance improved considerably with display duration for changes made to different kinds of feature (e.g., color and orientation) or to the orientation of separated items, it remained poor for other kinds of changes, even for display durations exceeding 600 ms. This therefore appears to be a new form of induced blindness, in which—at least for some kinds of changes—observers cannot see more than one change at a time.
Supported by NSERC Canada and Nissan Motor Co.