Visualization of large data sets is a demanding task, especially if different users are interested in different aspects of the same data set. Today?s visualization techniques often do not distinguish different aspects and thus visualize all aspects of the data, which leads to overcrowded and cluttered representations. Moreover, users are provided with information irrelevant to them. In this paper we describe an event-based approach for visualizing large data sets. The basic idea is to let users describe interesting aspects of the data by means of events and then adapt the visual representation of the data on occurrence of events. We present a formal description of events, assuming that the data are given in relational form. Furthermore, a model for event-based visualization is developed. A brief overview of an early realization of the model along with first visual results is part of this paper.