It is common that open-channel flows are turbulent and the channel-bed surfaces are rough. The presence of roughness elements on the channel-bed has significant influences on the flow characteristics in the bed vicinity and further above. Some earlier researchers have created two-dimensional roughness in laboratory channels by placing rectangular ribs on the bed and three-dimensional roughness by gluing sediment grains of irregular shape on the bed. Elements in the form of cubes are arguably more suitable to quantify the influences. However, laboratory experiments for detailed measurements of flow velocity and turbulence over cubical elements are rarely available. In this thesis, laboratory experiments of turbulent flow over acrylic cubes in a rectangular channel were carried out. Cubes with sides of 1.905 cm (3/4 inches) are uniformly mounted as roughness elements on the horizontal bed. The approach flow depth ranges from 8.8 to 9.1 cm or 4.62 to 4.78 times the roughness height. The experiments cover three conditions of cube spacing: being 2, 4, and 6 times the cube side dimensions; these relative spacing conditions represent the d-type, intermedium-type, and k-type roughness, respectively. Using an acoustic Doppler velocimeter (ADV), detailed measurements of three-dimensional flow velocities and turbulence quantities were made at locations above and around the cubes. At each location, the measurements lasted 60 s, at a sampling frequency of 120 Hz. The positions of the ADV probe were controlled accurately ( 1 mm) by a step motor. The velocity distributions are shown to be much more complex than the boundary-layer velocity distributions widely used in the literature. We categorized vertical profiles of the longitudinal velocity component into six distinct types (types 1 – 6) and vertical profiles of turbulent shear stress into five types (types 1 – 5). Based on measurements of the longitudinal velocity component and turbulent shear stress, we calculated velocity distributions for the d-type roughness by using the logarithmic law. There are large discrepancies between the calculation results and measurements. Moreover, we obtained contours of the secondary flow intensity in various planes. The distributions of the secondary flow are symmetrical with respect to the channel centreline at some cross sections but asymmetrical at other cross sections. The k-type roughness is shown to produce stronger secondary flow than the d-type and intermediate type of roughness. The experimental results are useful for model validations.