The validity of nutrient use efficiency as a central concept in ecosystem ecology has recently been subject to challenge based upon arguments over autocorrelation of data, interpretation of graphical approaches, and appropriate statistical analyses. Much of the confusion on the measurement and interpretation of nutrient use efficiency results from the lack of a sound theoretical basis with which to examine experimental results. In this paper, we develop a theory of nutrient use efficiency based upon fundamental mass balance, present a graphical approach to appropriate testing of alternative hypotheses to avoid problems of autocorrelation in data, and suggest critical areas where experiments must be performed to distinguish among hypotheses. We show that nutrient use efficiency (production per unit nutrient uptake) must be distinguished from nutrient response efficiency (production per unit nutrient available). In contrast to the monotonic increase of nutrient use efficiency with decreasing nutrient availability originally proposed in the 1982 model of P.M. Vitousek, nutrient response efficiency is unimodal with maximum efficiency at intermediate levels of nutrient availability. However, nutrient use efficiency dynamics at low nutrient availability cannot yet be theoretically defined. We also show theoretically which plant traits control responses of ecosystem nutrient use or nutrient response efficiency along gradients of nutrient availability. Finally, we show how our model naturally leads to species replacement along nutrient availability gradients.