RATIONALE: In stable isotope mass spectrometry, isotope values are normalized to internationally recognized reference scales using certified reference materials and working standards. Numerous techniques exist for performing this normalization, but these methodologies need to be experimentally assessed to compare their impact on reproducibility of isotope results. METHODS: We tested normalization methods by the number of standards used, their matrix, their isotope range, and whether normalization required extrapolating beyond the isotope range. Using 8 certified reference materials and 5 working standards on a ThermoFinnigan Delta-V IRMS and Elementar VisION IRMS for nitrogen and carbon isotope composition via solid combustion with an elemental analyzer, we computed every possible isotope normalization (n=6272). Additionally, we assessed how sample matrix impacted linearity effects on both instruments. RESULTS: Normalizations composed of three or four reference materials had better performance than one-point and two-point methods, especially when the normalization was matrix-mixed or extrapolated, and normalizations with an isotope range greater than 15‰ were more accurate under these conditions. Normalizations that were matrix-matched and were not extrapolated exhibited the highest accuracy. Linearity effects were found to exceed instrument precision by two orders of magnitude irrespective of sample matrix and were not predicted by reference gas diagnostics. CONCLUSIONS: To maximize interlaboratory comparability of isotope results, operators of EAIRMS systems should use at least 3 calibration standards to construct their normalizations, select standards with a large isotope range to avoid extrapolation, and match the matrix of their standards to their samples to the best extent possible.