In recent years, the modeling of reaction systems has tended to become more complex. Especially in systems biology researchers are faced with network models growing in size, complexity and scope that become hard to understand and navigate. The hierarchical and modular decomposition of such models (which results in modular models) and their consequent visualization can help to manage and understand them. In particular, visual representations allow researchers to access, integrate and analyze complex biochemical information, and thus support the process of decision making.This thesis exploits existing and contributes novel visualization and exploration techniques that can be used to support the modeling process of intracellular signaling systems. As stated in the following, several aspects are discussed.In close cooperation with modelers, formalized visual representations for dynamic and logical models have been speci ed. For each modeling approach a symbol set and speci c visualization rules are provided. They are applied to the modular and hierarchical structures of complex signaling models and aimed at visualizing them in a biologically intuitive manner, but also at maintaining their systematic approach.For convenient use, not only a visual representation of the model is crucial; a graphical network needs to be combined with exploration techniques. Hence, a zoomable visual interface for modular models has been developed to support highly interactive tasks such as navigation through complex structures and exploratory analysis. Hereby, interactive techniques, semantic zooming as well as content-aware navigation are used and combined in order to integrate focus and contextual information, and e ciently navigate and explore the modular model.Although formalized representations are necessary, they are not always the best way to graphically depict knowledge. In many cases, visual representations have to be tailored to a speci c modeling task, a given dataset or individual preferences. Thus, the concept of 'Visual Scenarios' has been developed. Applying di erent visual scenarios, the modeler is able to modify the visual representation of elements of modular models in order to highlight or suppress them depending on context, structure or search criteria.The drawing of modular models can be covered by graph layout strategies. However, general purpose layout algorithms do not consider the speci c structure of modular models. Hence, two layout strategies have been proposed. The rst approach is based on the spring-embedder paradigm and incorporates di erent domain-speci c constraints. Results suggest that it is able to produce comprehensible visualizations for modular models of signaling systems. In contrast, the second approach provides a way to generate comprehensible, at visual representations.