Microbial decomposition of soil organic matter (SOM) releases around 98 Pg of C (as CO2) to the atmosphere annually. Quantifying CO2 emissions from SOM is necessary to monitor and manage them but is complicated by proximate respiration of CO2 from plant roots, and by the influence of roots on SOM decomposition rate. Differences in the natural abundance of 13C in root and SOM-derived respiration (of < 10 ‰ in most temperate ecosystems) can be used to apportion their contributions to soil-surface CO2 efflux. However, this is challenging because all three δ13CO2 measurements are susceptible to significant sampling errors, which this study set out to identify and resolve, as follows. Respired CO2 sampled from excised roots is 13C-depleted by 1.8 ‰ (± 0.47) compared to intact roots due to the contribution of CO2 from root wounds. Root-respired δ13CO2 is more reliably measured using chambers around live, intact roots. These chambers also permit detection of diurnal changes in root-respired δ13CO2. Soil disturbance during sampling and root removal changes the carbon substrates available to microbes and this is reflected in a rapid (1-2 hours) decrease in δ13C of respiration of c. 4 ‰. This change can be regressed to estimate the δ13CO2 of microbial respiration from undisturbed soil. Techniques for measuring soil-surface efflux δ13CO2 induce method-specific biases of as much as 5 ‰, as measured in intact mesocosm soil and when simulated using a numerical diffusion model. Discrepancies between measurements and model predictions may be due to complexities of gas transport not currently accommodated in diffusion models, namely, near-surface advection and non-uniform soil diffusivity. Using improved techniques, this study used natural abundance 13C partitioning to assess priming effects, identify distinct environmental drivers of root-respired and SOM-derived CO2 fluxes, and detect differences in soil carbon cycling between tree species, possibly attributable to mycorrhizal type.