Abstract
Purpose: :
Using artificial field loss generated by a gaze-contingent display, we studied how visual field loss affected visual search performance and to what degree search training might improve performance.
Methods: :
The subject searched for an "O" target among "C" distracters. The orientations of the C’s were randomized. The stroke width of the "C" and "O" was 7.9 arcmin. Search items were positioned on a 6x6 virtual grid subtending 20x20 deg. Set sizes of 1, 8 or 32 were used. Each search session had 36 target-present and 18 target-absent trials. A 500 Hz eye tracker was used to create an artificial scotoma on normal subjects by either presenting only a square area of the search stimulus centered at the fovea (tunnel vision, TV) or masking all stimulus in the same square area and exposing the rest (central scotoma, CS) at 100 Hz. The square area was either 6 or 10 deg in size. Up to 16 search sessions for each field loss were collected from each subject. Search performance without artificial field loss (free viewing) was also measured.
Results: :
When searching with an artificial CS, accuracy depended on set size. Accuracy was the same as free viewing when set size was 1 and did not improve with practice. Accuracy was initially impaired when set size was 8, but improved to free viewing level through practice. When set size was 32, accuracy was acceptable with the 6 deg CS, but was near guessing level with the 10 deg CS, and did not improve with practice. Using larger, 13 arcmin search items improved search accuracy with the 10 deg CS. An artificial TV had no effect on accuracy. All field loss initially lengthened reaction time (RT). More severe field loss (larger CS and smaller TV) resulted in slower search. For the small (6 deg) CS, RT approached free viewing level at all set sizes after practice. For the large CS and for the large and small TV, practice shortened RT, but search performance after many sessions of training was still much slower than free viewing. The number of fixations made during search correlated highly with RT under all conditions. Practice reduced the number of fixations. Oculo-motor parameters such as fixation duration, saccade velocity and amplitude during search were not affected by the presence of field defect and were not improved through practice. These results held for both target-present and target-absent search trials.
Conclusions: :
Human subjects with field loss could learn to search faster and more accurately, mainly through reducing the number of fixations during a search trial. The degree of improvement achieved through training depends on the type and severity of field loss, and may be restricted by factors such as crowding.
Keywords: low vision • visual search • eye movements