Abstract
Subjects are able to detect quickly animals and vehicles in previously unseen cluttered scenes presented peripherally even when their attention is distracted. They are, however, unable to discriminate rotated letters (T/L) and bisected color disks (Red/Green) in the same conditions (Li et al, PNAS 02). We explore this phenomenon further by variations of the original experiment.
The first experiment was designed to further probe the extent to which natural scene categorization is possible in the absence of attention. Subjects were instructed to respond whether there was an animal in one of the two natural images presented peripherally in random locations, while concurrently performing an attentionally demanding central task. We could measure no significant difference in performance between one and two images. This result strengthens the view that attention is not a critical resource in this task.
The second experiment was designed to verify whether T-L (or bisected color disks) discrimination was poor due to a lack of signal. The number of rotated letters (or bisected color disks) peripherally was increased to four instead of one. Even though there were more potential “features” for subjects to detect (e.g. four T junctions than just one), our subjects still failed to discriminate between T's and L's (or between Red/Green and Green/Red disks) when attention was distracted.
In a third experiment we explored the nature of the mechanisms that are critical for the T-L task. We reasoned that rotated T and L presented peripherally differ from images of animals in that they do not constitute an “object class”: we are trained to recognize upright letters in the central region of the visual field. While distracting attention we presented our subject with upright T-L discrimination tasks in either the periphery or in the center of the visual field. The performance of our subjects improved significantly.