Resistance coefficients have been presented from measurements taken in compound and simple channel shapes in both laboratory and field conditions. Laboratory data were measured using the UK Flood Channel Facility (FCF), a large-scale experimental compound channel. Boundary roughness conditions in the FCF included smooth main channel with both smooth and rod-roughened flood plains, and a sand main channel bed with both smooth and rod-roughened flood plains. The flood plain roughness consisted of 25 mm surface-penetrating dowel rods. Data were also obtained for simple channels for comparison by confirming flow to the main channel. The field data were obtained from a local river reach which has been reconstructed to form a symmetrical compound cross-section. The field and laboratory data enabled calculation of roughness coefficients for compound sections at a range of depths. In addition, velocity measurements enabled calculation of discharges and hence roughness coefficients for main channel and flood plains separately. Smooth simple channel resistance coefficients were found to conform to known relationships, while mobile bed simple channel roughness coefficients increased with depth to a maximum, responding to increasing dune size. Surface-penetrating roughness produced continuously increasing roughness coefficient with depth. Natural inbank data exhibit more scatter and followed a decreasing trend with depth, as relative roughness decreased. Roughness coefficients for the compound sections exhibited complex behaviour and were strongly influenced by the surface-penetrating flood plain rod roughness. Smooth main channel roughness coefficients were strongly influenced by flood plain roughness, while the introduction of a sand bed rendered the main channel data largely independent of flood plain roughness. River data display generally larger values than the laboratory channel, but trends are similar to the smooth boundary data. Differences are due to the presence of inwardly sloping flood plains on the river cross-section.