A novel framework is presented for the analysis of ecophysiological field measurements and modelling. The hypothesis ‘leaves minimise the summed unit costs of transpiration and carboxylation’ predicts leaf‐internal/ambient CO2 ratios (ci/ca) and slopes of maximum carboxylation rate (Vcmax) or leaf nitrogen (Narea) vs. stomatal conductance. Analysis of data on woody species from contrasting climates (cold‐hot, dry‐wet) yielded steeper slopes and lower mean ci/ca ratios at the dry or cold sites than at the wet or hot sites. High atmospheric vapour pressure deficit implies low ci/ca in dry climates. High water viscosity (more costly transport) and low photorespiration (less costly photosynthesis) imply low ci/ca in cold climates. Observed site‐mean ci/ca shifts are predicted quantitatively for temperature contrasts (by photorespiration plus viscosity effects) and approximately for aridity contrasts. The theory explains the dependency of ci/ca ratios on temperature and vapour pressure deficit, and observed relationships of leaf δ13C and Narea to aridity.