A unique relation between water-surface elevation and water discharge usually does not exist for sand-channel streams. The relation is affected by changes in bed roughness and changes in bed elevation because of scour and fill. An artificial control on a sand-channel stream must control both the resistance to flow and the bed elevation in order to stabilize the relation between water-surface elevation and water discharge. The weir (control structure) in the Rio Grande conveyance channel near Bernardo, N. Mex., was designed on the basis of a model study and field data (Harris and Richardson, 1964). About 72 percent of the measurements used to define the base relation between water-surface elevation and water discharge falls within plus or minus 5 percent of the mean relation for the prototype. The stage-discharge relation is not affected by backwater for values of submergence less than 90 percent. There is no consistent relation between the ratio of measured discharge to rated discharge and submergence for values of submergence greater than 90 percent. The control does not restrict the channel capacity to less than the stated design capacity of 2,000 cubic feet per second. When the control is drowned out, or ineffective, the relation of water-surface elevation to water discharge is virtually the same as that prior to construction of the control for discharges greater than 1,500 cubic feet per second. When the control is not drowned out--that is, free-fall conditions exist--the water-surface elevation for a discharge of 2,000 cubic feet per second is greater than the minimum elevation, but is less than the maximum elevation that occurred at that discharge prior to construction. The model study was only partially successful in predicting the operating characteristics of the prototype. Some of the differences between prototype operation and model predictions may exist because the prototype was not built exactly as recommended on the basis of the model study. In general, the prototype has operated somewhat better than the model predicted.