High-resolution neuroimaging and multi-electrode physiological recordings provide us with the opportunity to understand brain functions from simultaneously measured signals. Brain functions arise from the networks of interacting neural systems and these interactions are directional in the sense of signal transfer. Extracting such directional information (or effective connectivity) from measured brain data facilitates a better understanding of the neural computation involved in achieving a brain function. We have recently developed new model-free Granger causality techniques that address the question of directionality of neural interactions from measured signals. These new techniques, which are based on widely used Fourier and wavelet spectral methods, not only overcome some of the existing problems in parametric techniques, but also open up new applications in neuroscience. In this seminar, the usefulness of the nonparametric Granger causality techniques will be illustrated with applications to local field potentials recorded from monkeys and to functional magnetic resonance imaging time series from humans, both in the context of sensorimotor tasks.