ABSTRAC ted behavi lanes. models can log data from the driver, the car, and the environment. The collected data were analysed and modelled with the help of a" knowledge discovery framework"(Georgeon, Mille, & Bellet, 2006). The first analysis of all lane changes caused by a slow leading vehicle focuses on the following indicators: glance to the left outside mirror, turn signal, and lane crossing. It is shown that the glance to the left outside mirror could serve as a predictor with a high potential to get information about the intention to change lanes in a very early stage. However, it is important to combine this predictor with additional information to avoid a high false alarm rate.