The flow through fractured rock is of great importance for many civil engineering projects, for example, when considering the safe storage of nuclear waste or the stability and profitability of dams, tunnels and slopes. How the fluid actually flows through a fractured rockmass is still a matter of vivid discussion within the hydrogeologic community. Various approaches to calculate the flow through a fractured rockmass exist. Which approach is best depends on the rockmass properties and the size of the rockmass under investigation. This study presents a new model which assumes that groundwater flow will take place preferably at the intersection of fractures. Study of the scientific literature showed that flow through intersections of fractures exists and can play a major role in flow through a fractured rockmass. Fieldwork has been conducted at outcrops close to Granada, Spain. At these outcrops fossilized flow paths have been observed at the intersections of fractures. During this fieldwork, fracture data have been gathered, which could be used to reconstruct the fracture geometry. As part of this research a mathematical model has been developed to simulate flow along the intersections of fractures, neglecting flow within the fractures or through the rock matrix. To this end a computer program, CPA, developed for a different purpose, has been modified and completed in order to generate a stochastic tubular network of fracture intersections and calculate flow rates through this network. The new version of the CPA code has been applied to a number of problems to test the correctness of the model and investigate the effects of the different input parameters. The application part of this study can be roughly divided into: network generation, sensitivity analysis, and the comparison with a model, Joint-OKY, that assumes flow to take place through the fractures. The field data gathered during fieldwork in Spain have been used to confirm the geometric modelling capabilities of the CPA program. A comparison between the network model generated by CPA and the observations in the field showed good agreement. The use of an eigenvec- tor approach to represent fracture orientation distribution has proven to be a good and simple method. To better understand the influence of various input parameters in the CPA model, sensitivity analyses were performed on the models generated by the CPA code. The following parameters have been investigated: fracture density, size of the model area, fracture size and anisotropy of conductivity. Finally a comparison between the CPA code and the Joint-OKY code, which assumes flow to take place within the plane of the fractures, has been made. This comparison showed that a model assuming flow within the fractures is more conductive. Further studies are needed to better include the transport of contaminants in the current CPA program.