Stable carbon isotopes can be used to partition the net ecosystem–atmosphere exchange (NEE) of carbon dioxide (CO2) into its photosynthetic and respiratory components, but the method has not been generally adopted due to instrumental and theoretical limitations. Here, motivated by recently improved instrumentation, we extend the theory of isotopic flux partitioning to include photorespiration, foliar daytime ‘dark’ respiration, and other refinements, arriving at a general yet practical formulation from which all previous formulations can be derived as simplifying approximations. We use a full growing season of isotopic eddy covariance flux data from a temperate deciduous forest to demonstrate the method, quantify its uncertainties, and determine biases associated with previously published formulations. We find that when δ13C of CO2 is acquired with high precision (0.02‰ RMSE for 100s integration times), the statistical uncertainty in the partitioned fluxes is comparable to that in NEE itself—i.e., as good as practicably possible. Assessable systematic uncertainty is ±17% of gross ecosystem production (GEP), due mostly to uncertainty in the isotopic fractionation by carboxylation. Additional, currently unquantifiable systematic uncertainty is associated with treating the canopy as a single “big leaf”. Both sources of systematic uncertainty could be greatly reduced by feasible supporting leaf-level measurements. Our extended theory corrects systematic biases in previous isotopic approaches, including overestimation (by 13%) of GEP due to the omission of photorespiration. The partitioning determines the isotopic signature of photosynthesis, which we find to vary seasonally between −24 and −28‰ such that the isotopic disequilibrium between ecosystem carbon input and output remains stable at approximately −0.5‰ through most of the growing season. The key advantage of isotopic partitioning over standard, regression-based partitioning is that it enables controls on the ecosystem-scale photosynthetic and respiratory fluxes to emerge from observations, without having to assume functional relations to environmental drivers a priori. As an example, we show how isotopic partitioning reveals certain large variations in daytime NEE to be caused by shifts in the flux tower sampling footprint between regions of high and low respiratory flux—a finding unobtainable by standard partitioning. For this reason, isotopic partitioning can be more precise than standard partitioning for quantifying environmental controls on NEE.