The highly nonlinear and complex relationship between rainfall and runoff due to temporal and spatial variability of watershed characteristics, heterogeneity in precipitation as well as several other factors, make the streamflow forecasting a challenging task. For years numerous techniques have been suggested and employed for streamflow forecasting among which Artificial Neural Network has attracted considerable attention of scientists and engineers owing to its capability to learn hidden patterns from historical data and predict the future accordingly. In this study, we present the applicability of a Self Organizing Radial Basis (SORB) artificial neural network to daily streamflow forecasting. SORB utilizes Gaussian Radial Basis functions (RBFs) in conjunction with the Self Organizing Feature Map (SOFM) used in data classification. The performance and merit of the algorithm in comparison to two other architectures, Multilayer Feedforward Network (MFN) and Self Organizing Linear Output map (SOLO) are demonstrated. To take the best advantage of short data set and generalization of the model to achieve the optimal performance in training and testing, cross validation is employed. It is shown that cross validation results to a more stable and reliable parameter set and streamflow forecasting.