Abstract
Vehicle control is possible because the human nervous system is capable of producing complex sensorimotor actions. Drivers must monitor errors and initiate steering corrections of the correct magnitude and timing to maintain safe lane positions. The perceptual mechanisms determining how a driver processes visual information and initiates corrections remains unclear. The perceptual-motor action literature suggests two potential alternative mechanisms for responding to errors: (i) perceptual evidence (error) satisficing fixed constant thresholds (Threshold), or (ii) the integration of perceptual evidence over time (Accumulator). To distinguish between these mechanisms three experiments were conducted using computer generated steering correction paradigms. In the first two experiments, drivers (N=20) steered towards an intermittently appearing ‘road-line’ that varied in position and orientation with respect to the driver’s starting position and trajectory respectively. In the third experiment (N=50), silent automation failures were induced at varying severities. Drivers had to disengage automation and initiate steering to avoid exceeding lane boundaries. Threshold and Accumulator accounts predicted different steering patterns responding to these errors: a Threshold account predicted a fixed absolute error response across conditions regardless of the rate of error development, whereas an Accumulator account predicted that drivers would respond to larger absolute errors when the error signal developed at faster rates. Results in all three experiments show that drivers responded faster, responded to larger quantities of error, and steered with greater magnitude, as the rate of error signal development increased. These findings are in line with an Accumulator account, thus we propose that models of steering and silent failure takeovers should integrate perceptual evidence over time to capture human perceptual performance.