This paper is the culmination of previous work to determine if steering behavior could be used to unobtrusively detect driver fatigue. The driving performance of 17 sleep-deprived heavy-truck drivers was monitored on a closed track. Functions in the time, frequency, and phase domains were developed to quantify changes in steering wheel input. The steering-based weighting functions which correlated most strongly with independent measures of driver fatigue and drowsiness were used to develop a simple algorithm. The algorithm predicted fatigue for all 17 volunteer drivers before the end of their test. The algorithm identified 12 drivers before a lane breach occurred, and only two drivers were not captured until a lane breach greater than 15 cm occurred. These data and the algorithm demonstrate the potential for a steering-based fatigue detection algorithm.