Abstract: Contributions of C3 and C4 plants to respiration of C3-C4 ecosystems can be estimated on the basis of their contrasting 13C discrimination. But accurate partitioning requires accurate measurements of the isotope signature of whole system respiratory CO2 (δR), and of its members (δ3 and δ4). Unfortunately, experimental determination of representative δ3 and δ4 values is virtually impossible in nature, generating a need for proxies (surrogates) of δ3 and δ4 values (e.g., the δ of leaf biomass). However, recent evidence indicates that there may be systematic differences among the δ of respiratory and biomass components. Thus, partitioning may be biased depending on the proxy. We tested a wide range of biomass- and respiration-based δ proxies for the partitioning of respiration of mixed Lolium perenne (C3) - Paspalum dilatatum (C4) stands growing at two temperatures inside large 13CO2/12CO2 gas exchange chambers. Proxy-based partitioning was compared with results of reference methods, including (i) the δ of whole plant respiratory CO2 (δ3 and δ4) or (ii) respiration rate of intact C3 and C4 plants. Results of the reference methods agreed near perfectly. Conversely, some proxies yielded erroneous partitioning results. Partitioning based on either the δ of shoot or root respiratory CO2 produced the worst bias, because shoot respiratory CO2 was enriched in 13C by several ‰ and root respiratory CO2 was depleted by several ‰ relative to whole plant respiratory CO2. Use of whole plant or whole shoot biomass δ gave satisfactory partitioning results under the constant conditions of the experiments, but their use in natural settings is cautioned if environmental conditions are variable and the time scales of respiration partitioning differ strongly from the residence time of C in biomass. Other biomass-based proxies with faster turnover (e.g., leaf growth zones) may be more useful in changing conditions.