Evapotranspiration (ET) vies with runoff as the primary mechanism for water loss from a watershed underlain by permafrost, yet past attempts to predict ET have proven to be less than completely successful in the Arctic. Imnavait Creek, a small 2.2 km2 watershed underlain by continuous permafrost has been studied for 4 years. Evapotranspiration on a watershed scale has been calculated from water balance studies. These results are compared with point measurements of pan evaporation and daily estimates of ET by the energy balance and Priestley-Taylor methods. Since it is difficult to determine the daily change in soil moisture, the energy balance approach appears to be the best method to determine daily ET. The water balance approach is the best method to determine total ET over the course of the summer because it is possible to delete the soil moisture term due to an insignificant change annually in this watershed. Priestley-Taylor gave adequate estimates of ET with only limited data. After a pan coefficient is determined, the evaporation pan functions well over extended time periods but is less accurate for shorter periods. Evapotranspiration is greatest in early summer, immediately following the spring snowmelt, during the period of maximum incoming radiation but not necessarily maximum air or soil temperatures. The cumulative potential evaporation is greater than the cumulative summer precipitation. The source of moisture for ET in early summer is from snowmelt or moisture stored in the active layer.