Abstract Last-century climate change has led to variable increases of the intrinsic water-use efficiency (W i; the ratio of net CO2 assimilation to stomatal conductance for water vapor) of trees and C3 grassland ecosystems, but the causes of the variability are not well understood. Here, we address putative drivers underlying variable W i responses in a wide range of grassland communities. W i was estimated from carbon isotope discrimination in archived herbage samples from 16 contrasting fertilizer treatments in the Park Grass Experiment, Rothamsted, England, for the 1915 to 1929 and 1995 to 2009 periods. Changes in W i were analyzed in relation to nitrogen input, soil pH, species richness, and functional group composition. Treatments included liming as well as phosphorus and potassium additions with or without ammonium or nitrate fertilizer applications at three levels. W i increased between 11% and 25% (P < 0.001) in the different treatments between the two periods. None of the fertilizers had a direct effect on the change of W i (ƊW i). However, soil pH (P < 0.05), species richness (P < 0.01), and percentage grass content (P < 0.01) were significantly related to ƊW i. Grass-dominated, species-poor plots on acidic soils showed the largest ƊW i (+14.7 μmol mol−1). The ƊW i response of these acidic plots was probably related to drought effects resulting from aluminum toxicity on root growth. Our results from the Park Grass Experiment show that W i in grassland communities consistently increased over a wide range of nutrient inputs, soil pH, and plant community compositions during the last century.