Understanding global variation in evapotranspiration (ET) is critical for accuracy of climate models, predictions used in water resources management, and assessment of land use change impacts on the water balance of ecosystems—yet we lack unifying principles to predict when transpiration (T) varies with land use. Plant T is a dynamic and often dominant component of ET, and is affected by a variety of processes controlled by land use changes superimposed onto edaphic conditions. We propose the following three principles that determine whether T will vary with changes in vegetation: variation will result if energy balance partitioning has been altered, if deeper or shallower active rooting depth has changed the amount of soil moisture accessible to plants, or if temporary changes in water use add up over longer time scales. Clearly these concepts are not new; however, they are often overlooked in favor of blanket assumptions that large changes in vegetation inevitably alter T. Not so. Our suggested framework incorporates both edaphic and plant traits that determine whether T will vary or not in response to altered land cover conditions. We suggest that this simple set of principles unifies results of wide-ranging studies of T following land use change and can explain underlying causes of T variation. These principles are illustrated through case studies from four different environments: Pacific Northwest forest, Texas mesquite brushland, Texas savanna and woodland, and Middle Rio Grande riparian forest. The proposed principles seem broadly applicable, should be further evaluated, and are directly relevant for land management. Copyright © 2011 John Wiley & Sons, Ltd.