A geostatistical reservoir study provides an ensemble of possible reservoir models, with a variability reflecting the uncertainty of the geology and fluid-flow properties. A production test links the static geostatistical model with dynamic fluid-flow data and thus provides possible means of validation and selection. We used a highly efficient 3D single-phase simulator to simulate pressure transients without compromising the fine-grid resolution typic al for geostatistical models. This simulator was applied to Boolean and Gaussian geostatistical models that represent reservoirs consisting of heterogeneous fluviodeltaic deposits. Simulated buildup and interference tests were analyzed and related to 3D permeability and connectivity patterns. Effective well-test permeability was compared with 3D averages obtained from different volumes within the geostatistical model, The cases studied show that combining geostatistical models and well tests can reduce uncertainty with respect to geometric connections and the permeability distribution. However, the study also confirms that well tests have a limited capability to assess lateral continuity precisely and uniquely. Simulation of an interference test has the potential to screen the geostatistical model for high-permeability connections between wells. This pragmatic integration of pressure transients and detailed heterogeneous reservoir models through forward simulation provides a simple means to evaluate these models. It allows testing a model. based on small-scale (core-plug) permeabilities and qualitative geologic information vs. routine field measurements representing large-scale hydraulic behavior of a reservoir.