Abstract
Humans cannot discriminate eye of origin of visual input, when the informative cues such as feeling-in-the-eye are made uninformative (Ono and Barbeito, 1985), nor can they improve discrimination by one week of daily practice with feedback (Blake and Cormack, 1979). We ask whether improvement is possible by more extensive training (Zhaoping and Xiao 2015). The two authors practiced this task with feedback for multiple weeks, with occasional holiday breaks. Typically, each trial presented the observer a brief (200 ms) dichoptic test stimulus containing an array of monocular items. The observer had to report the ocular origin of the central target item, which was distinctively shaped from the other, task-irrelevant, non-target items. The ocular origin and the luminance of all the monocular items were randomly and independently assigned, and the non-target items in the two eyes were bars tilted in opposite directions from vertical. Binocular dots between monocular items were used to anchor vergence. This test stimulus was preceeded by binocular fixation stimulus and followed by a binocular mask. A random third of the trials were control trials containing no target item, prompting for a distinct response. In initial days, the target was a vertical bar. After 14-20 training days, the task performance (for non-control trials) rose consistently above the chance level to contain correct responses in 70-80% of the trials. One observer noticed, on her 31st training day, that the apparent tilt of the target was informative since it correlated with its ocular origin, perhaps caused by her astigmatism. Modifying the stimulus design to remove each confirmed or suspected informative cue often immediately dropped performance, and further training on the modified design often raised performance, revealing additional informative cues, which were then further removed. Eventually, performance could not be improved beyond 55% correct by more than two weeks of training.
Meeting abstract presented at VSS 2016