This paper describes a newly developed microcomputer-based drowsiness warning system, which detects changes in the driver's alertness through his steering behavior. In developing this system, we first quantified several levels of alertness based on such physiological factors as brain activity and blinking. Tests were then conducted in which drivers fell into different degrees of drowsiness.Using the quantified alertness levels, we defined the “drowsy driver” and found unique steering patterns that could not be seen in normal driving. These patterns were entered into the memory unit of the microcomputer. When the sensor built into the steering wheel detects a drowsy steering pattern, the microcomputer recognizes the driver's drowsiness and activates a buzzer to warn the driver.In this paper, the process of determining the alertness levels is explained, along with the steering characteristics of the drowsy driver.