Abstract An experimental agricultural spraying machine for investigating the effects of applying spatially variable herbicide doses is described. Design features of the machine which affect spatial resolution and steady-state accuracy of herbicide dose are examined. Experiments have shown that the steady-state dose can be controlled to within ±5% of the set value over the system range. When using the injection metering systems to control dose, the sprayer will typically travel 11 m before the effect of a change in set dose rate is seen at the nozzles. Using solenoid valves, four levels of dose rate can be applied within 0·5 m of a command. The finite time taken for injection metering systems to respond to changes in water flow rate as sections of the boom are switched on and off, can cause dynamic errors in applied dose.