Abstract
We have been studying the task facing observers who need to detect a reduction in the distance to an object (e.g. closing during driving). Ignoring stereo cues, which in any event may not be operative for viewing objects beyond 10 m distance, there are a number of monocular cues that might be used in the task. Linear parameters of a target object (height, width), its area, change in the spatial scale of texture, distinctness of features, and relation to fixed objects are some that might play a role. Experiments to study this applied problem can look at performance, one cue at a time, and then look at combined cues to gain a better picture. Our initial experiments of this type revealed that observers do better with width than with height for rectangular objects, but that area was little better than width by itself (Toyofuku, Cohn, and Nguyen 2002 http://www.journalofvision.org/2/7/676/ ).
In the present study we have tried to see which cues were employed. We reasoned that cues that were ignored would lead to a neutral classification image, while those that were employed would reveal themselves, although rules of combination would not be discernible. We measured the classification image for a transient increase in the size of a rectangular frame. Three normal observers were presented a detection task in which the signal to be detected was a brief increase in overall size of the frame at a fixed 1 M distance. 10,000 trial images were calculated. Each observer's classification image was unique. In common, all observers required width information. Two used the location of the upper edge of the frame. One used the entire frame. One of the observers used false frame indicators at the original location as a negative.