This study employs the stochastic frontier production function approach to measuring the productive performance of canal irrigation in terms of farm-specific technical efficiency. The 1993 production data from samples of rice farmers from Khageri (government managed) and Pithuwa (farmer managed) irrigation systems in the Tarai of Nepal were analyzed using a stochastic production frontier, including a model for the technical inefficiency effects. Relative both to a common or pooled production frontier for the two systems and a separate or own frontier for each system, on average, the farmers from the farmer-managed system were found to be more efficient than those from the government system. Furthermore, the government system showed higher spatial variability in technical efficiency in rice production than the farmer-managed one, which suggests that irrigation management has a significant influence on production and hence on the distribution of productivity gains among the farmers. However, better performance of the farmer-managed irrigation system relative to the government system should be weighed against the higher transaction costs associated with irrigation management by farmers. The results indicate that substantial potential exists for increasing rice production by increasing the application of chemical fertilizer, the planting of seed, and the use of irrigation and by improving technical efficiency at the farm level. ¿ 2001 American Geophysical Union