The role of the magnetic (B-) field in the star-formation process is highly debated. How important is the magnetic field in the presence of gravity and turbulence? Conclusive observational results have been scarce due to both limited available data and due to the generally challenging measurements requiring very sensitive observations. We present observational results illustrating the role of the B-field on three representative scales in the star-formation process: (1) the filamentary-scale infrared dark cloud G34. Here, the local B-field correlates with the local velocity gradients on the largest scales. Based on a benchmark analysis that quantifies the various energy components we argue that a different relative importance between B-field, turbulence, and gravity leads to different observed fragmentation types on a next smaller scale; (2) the molecular-cloud-scale in the high-mass star-forming region W51. Here, increasingly higher resolutions from the SMA to ALMA resolve new B-field sub-structures. In particular, we see zones of symmetrically converging B-field lines, cometary-shaped satellite cores, local collapse features, and a possible new phenomenon of magnetic channelling; (3) the protostellar-source-scale in B335. Here, we provide evidence for magnetic braking from detailed high-resolution observations of the kinematics of neutral and ionized gas tracers. Besides these textbook cases on three different scales, we further present statistical resultsfrom a 50-source sample of molecular clouds that reveal generic B-field features and a systematically locally varying importance of the B-field versus gravity.