A 150-year-long record of intrinsic water-use efficiency (Wi) was derived from community-level carbon isotope discrimination (13Δ) in the herbage of the unfertilized, unlimed control treatment (plot 3) of the Park Grass Experiment at Rothamsted (England) between 1857 and 2007. 13Δ during spring growth (first cut harvested in June) averaged 21.0‰ (±0.5‰ SD) and has not shown a long-term trend (P=0.5) since 1857. 13Δ of summer/autumn growth (second cut harvested between September and November) increased from 21.3‰ to 22.0‰ (P < 0.001) between 1875 and 2007. Wi during spring growth has therefore increased by 33% since the beginning of the experiment, and Wi of summer/autumn growth has increased by 18%. The variation in 13Δ was mainly related to weather conditions. Plant available soil water explained 51% and 40% of the variation in spring growth 13Δ and summer/autumn growth 13Δ, respectively. In the 1857–2007 period yields have not increased, suggesting that community-level photosynthesis has not increased either. Therefore, the increased Wi probably resulted from a decreased stomatal conductance. Vapour pressure deficit (VPD) during spring growth (March–June) has not changed since 1915, meaning that instantaneous water-use efficiency (Wt) in spring time has increased and transpiration has probably decreased, provided that leaf temperature followed air temperature. Conversely, VPD in the months between the first and second cut has increased by 0.07 kPa since 1915, offsetting the effect of increased Wi on Wt during summer and early autumn. Our results suggest that vegetation has adjusted physiologically to elevated CO2 by decreasing stomatal conductance in this nutrient-limited grassland.