Turbidites were produced experimentally by releasing suspensions of spherical beads from a lock into a 6 m long horizontal channel. Flow properties were varied by changing initial flow geometry, and sediment size, sorting, and concentration in the initial suspension. Initial flow geometry was varied using two lengths of lock (0.29 and 0.59 m) and four depths (0.15, 0.2, 0.3 and 0.4 m). To vary the properties of the initial suspension, three nearly uniform sizes (2.0, 2.7, and 3.6 phi) of glass beads and one (2.5 phi) of plastic beads (specific gravity 1.52) were used, and all experiments were run at concentrations of 20% and 40% by volume. The effect of poor sorting was investigated by mixing the coarse and fine glass beads to produce a poorly sorted sediment with a median grain size equal to that of the intermediate-size glass beads. Bed thickness was almost uniform near the source except for beds deposited from poorly sorted sediment suspensions, which tended to be wedge-shaped. High concentration produced more uniform bed thickness, and (for equal volume of sediment) shorter, thicker beds. Increases in lock height and length increased bed thickness almost equally. If all other variables were held constant, bed thickness was directly proportional to grain size. The dimensionless ratio of the square of bed thickness divided by the volume of suspension (per unit width) was found to be directly proportional to the dimensionless ratio of settling velocity divided by the speed of the head of the current. The 'constant' in the regression equation appears to depend mainly on the sediment concentration. The experimentally determined equation gives plausible values for turbidity current velocities when extrapolated to scales over a million times larger than those of the experiments.