Evapotranspiration is controlled by many climatologically-dependent factors and vegetation dynamics. Generally, climate is the major source of evapotranspiration variation, and the influence of vegetation is minimal. In the tropics, however, the climate is relatively consistent, but the vegetation dynamics are highly complex. Many species fill light-niches within the same vertical profile of leaf area index and they control stomatal water loss at different rates and in response to different stresses. Because of such complexity, modeling evapotranspiration in the tropics becomes rather difficult, and the measurement of it even more so. With the advent of the eddy covariance method, measurement of the land-atmosphere water flux has allowed for integration of these various evapotranspiring components across spatial and temporal scales that were effectively impossible to measure previously. But, evapotranspiration models, which traditionally were designed with temperate agricultural systems in mind, do not reflect tropical dynamics. Recently, a new ecophysiologically based, remote sensing driven model of evapotranspiration was validated across 16 eddy covariance sites globally (r2 = 0.90, RMSE = 16 mm/month). Although the model performed well globally, only 1 eddy covariance site was used from the tropics. Here, we test the model in comparison with a suite of traditional evapotranspiration models at 4 new eddy covariance sites in the Large-scale Biosphere-Atmosphere (LBA) experiment in the Amazon. We assess why and how evapotranspiration is controlled in the tropics, how the eddy covariance data respond to these controls, and how the evapotranspiration models perform in response to these controls. We scale-up to the Amazon basin as a whole using remote sensing data and discuss the implications of uncertainty from tropical evapotranspiration model estimates within larger hydrologic and climate change models.