Abstract
Purpose:
In a prior driving simulator study (Bowers et al. 2014), inadequate head scanning was associated with failing to detect intersection pedestrians by drivers with hemianopia; however, eye position was not tracked so it was unknown how close gaze came to the pedestrians. Here we report pilot data addressing the relationship between gaze scanning and detection.
Methods:
Twenty six people with complete homonymous hemianopia (HH) and without neglect drove three city routes in a simulator. Detection rates were determined for life-size stationary pedestrians at five T-intersections. Pedestrians appeared just before a turn on either the far left or right of the intersection (at the edges of the clear-sight triangle), or the center left or right on the opposite side of the intersection. Gaze behaviors (6-camera SmartEye tracker) were analyzed from the time the pedestrian appeared to the time of detection (horn press) or the time of exit from the intersection, when not detected. Misses were categorized as: no scan in the direction of the pedestrian; scan with inadequate magnitude; or scan with adequate magnitude but did not see (looked-but-failed-to-see; LBFTS).
Results:
Miss rates were high for far pedestrians on both the blind (79%) and seeing sides (65%, p = 0.07). These misses were a result of no scan (blind side 36% and seeing side 24%), inadequate scans (57% and 67%, respectively), and LBFTS (7% and 10%, respectively). For far-left pedestrians (at 85°), inadequate scans brought gaze to a median of 24° and 43° from the pedestrian for left and right HH drivers, respectively (p = 0.06). Miss rates were lower (< 30%) at center left on a left turn, and all were LBFTS events. However, the right HH group had higher miss rates (64%) at center right on a left turn than the left HH group (14%, p = 0.005). These right HH misses were LBFTS (43%), inadequate scans (43%), and no scan (14%).
Conclusions:
Gaze scan deficits, in particular not scanning sufficiently far, were the main reason for detection failures in both the blind and seeing hemifields. The data also revealed detection failures with the characteristics of LBFTS, especially in more central areas of the intersection, which in some cases appeared similar to change blindness events. A follow-up study will evaluate the extent to which gaze scanning deficits impair detection of moving-vehicle hazards and whether HH increases the incidence of LBFTS relative to normal controls.