Contaminant transport through fractured crystalline rock is simulated with the use of two fracture models. Development of the models is based on detailed field measurement and statistical analysis of fracture data, interpretation of aerial photograph fracture traces and aperture data reported in literature. Fractures observed on aerial photographs are ranked by the ease with which they can be identified on successively higher altitude photographs and provide the basis for development of the fracture network. Fracture data from the field investigation are utilized for calculation of rock block hydraulic conductivities and porosities. Fractures recorded during the field investigation were obtained from a subjective (biased) survey designed primarily to provide data on fractures which appear to occur as sets in outcrop exposures. Analysis of outcrop and aerial photograph data indicate that fractures are not random and can be separated into two broad sets. These sets are oriented roughly orthogonal to one another and are identified as the north- and east-trending sets. Although the fracture sets are oriented as such, the style of fracturing, as observed in the aerial photographs, is not systematic and appears to consist of a number of localized and over-lapping orthogonal sets. Analysis of biased orientation distributions suggest fracture orientations approximate normal distributions. Statistical analysis of fracture spacing and length show log-normal distributions. Transport of a non-reactive contaminant, in plan view, is simulated using the finite element code FEMWATER1-FEMWASTE1. Two fracture models were developed. Within both models fractures are treated as discrete fracture zones and inter-rock block areas as porous media equivalents. The two models differ primarily in the scale of observation and in the node/element design. The areally larger model simulates transport within 66 interconnected fractures and is designed utilizing 574 nodes and 587 elements. The smaller model represents 8 fractures in the northwest corner of this model and is designed utilizing 656 nodes and 617 elements. Analysis of velocity data show that flow in both models is principally occurring within the fractures and that the direction of flow within rock blocks is influenced by the closest adjacent fractures. Contaminant transport simulations show that the geometry of the fracture network defines the overall shape of the contaminant distribution and that use of dispersivities leads to increased longitudinal spreading within both the rock blocks and fractures. Under the defined boundary conditions, longitudinal dispersion within both fracture models occurs principally in response to the north-trending fracture set and lateral spreading as a result of intersecting fractures and abrupt fracture terminations. Porous media equivalent simulations show the smaller model does not encompass a large enough area to be simulated as a porous media equivalent. The larger model can be simulated as a porous media equivalent using longitudinal dispersivities of 2.0 m and 5.0 m within the northern and southern portions of the model, respectively. Associated lateral dispersivities are 0.10 and 0.25 m, respectively. Comparison of contaminant distributions shows that the two models cannot be quantitatively compared with one another because of the effects of boundary conditions and minor changes in fracture network design. Based on the shape of the concentration contours, it is shown that the areally smaller model is more consistent with fracture hydraulic conductivity data.