Neural oscillations are thought to underlie coupling of spatially remote neurons and gating of information within the human sensorimotor system. Here we tested the hypothesis that different unimanual motor tasks are specifically associated with distinct patterns of oscillatory coupling in human sensorimotor cortical areas. In 13 healthy, right-handed subjects, we recorded task-induced neural activity with 122-channel electroencephalography (EEG) while subjects performed fast self-paced extension-flexion movements with the right index finger and an isometric contraction of the right forearm. Task-related modulations of inter-regional coupling within a core motor network comprising the left primary motor cortex (M1), lateral premotor cortex (lPM) and supplementary motor area (SMA) were then modeled using dynamic causal modeling (DCM). A network model postulating coupling both within and across frequencies best captured observed spectral responses according to Bayesian model selection. DCM revealed dominant coupling within the ß-band (13-30 Hz) between M1 and SMA during isometric contraction of the forearm, whereas fast repetitive finger movements were characterized by strong coupling within the ¿-band (31-48 Hz) and between the ¿- (4-7 Hz) and the ¿-band. This coupling pattern was mainly expressed in connections from lPM to SMA and from lPM to M1. We infer that human manual motor control involves task-specific modulation of inter-regional oscillatory coupling both within and across distinct frequency bands. The results highlight the potential of DCM to characterize context-specific changes in coupling within functional brain networks.