Although prevention of bicycle–motorized-vehicle-crashes will result in significant improve-ments in bicycle mortality rates, little is known about the actual safety critical events leading up to these crashes. Recently, the understanding of critical events has been advanced by the use of Naturalistic driving techniques. This technique gathers data about driving in ‘natural conditions’ by means of the instrumentation of a large fleet of cars with sensors and cameras. To date, these studies have mainly focussed on car-car interactions rather than on interactions with vulnerable road users. This paper discusses the potential of Naturalistic Driving for un-derstanding critical interactions with vulnerable road users. To this end it presents some of the results from additional analysis of the data from previous Naturalistic Driving studies. In addi-tion, the paper presents the objectives and design of the recently started large scale Natural-istic Driving study UDRIVE in which a large fleet of motorized vehicles will be equipped with cameras and sensors, resulting in data on 470 years of driving. In contrast to previous and on-going Naturalistic Driving studies, UDRIVE is – amongst other issues- specifically designed to address interactions with cyclists and pedestrians.