Abstract
The paper introduces the use of onion graphs as a focus+context technique for visualizing large UML class models. The focus area, which can be manually or automatically derived, is visualized using the standard UML notations. The remainder of the model is abstracted (context) and presented at varying levels of detail in onion notation. A selective aggregation technique for achieving the abstractions is presented. Finally, the technique is demonstrated by examples on two subsystems of an open source project.