This paper presents an empirical approach for estimating driver's cognitive workload using driving performance, especially lateral control ability through readily available sensors such as lane position and steering wheel angle. To develop a real-time approach for detecting cognitive distraction, radial basis probabilistic neural networks (RBPNN) were applied. Data for training and testing the RBPNN models were collected in a simulator experiment in which fifteen participants drove through a highway and were asked to complete auditory recall tasks. The best performing model could detect cognitive workload at the accuracy rate of 73.3%. The results demonstrated that the standard deviation of lane position and steering wheel reversal rate can be used to detect driver's cognitive distraction in real time.