This paper reviews models that emphasize the cognitive components of driving behavior. Studies of individual differences have sought predictors of accident histories. Typically low correlations and reliance on post hoc explanations reflect theoretical deficiencies and problems with the use of accident measures. Motivational models emphasize transient, situation-specific factors rather than stable, individual predictors. However, neither testable hypotheses nor suitable methods have been developed to study situational factors and motives that influence driving. More recent models have incorporated a hierarchical control structure, which assumes concurrent activity at strategic, maneuvering, and operational levels of control. At the same time, automaticity has emerged as a central construct in cognitive psychology. All activities are assumed to combine fast, automatic components with slower, more deliberate, controlled processing. It is argued that identifying the situational factors that increase drivers' uncertainty and thus trigger a shift in attention from automatic to controlled processing will help integrate concepts of automaticity and motivational models. Finally, recent theorizing has suggested that errors associated with the inherent variability of human behavior may be more important to roadway crash causation than systematic errors, which are attributable to the known limits of the human information-processing system. Drivers' abilities to recover from errors may also be important to crash causation. It is concluded that the hierarchical control structure and theories of automaticity and errors provide the potential tools for defining alternative criterion measures, such as safety margins, and developing testable theories of driving behavior and crash causation. Two examples of models that integrate information-processing mechanisms within a motivational framework are described.