A mass spectrometric method to determine the absolute intramolecular (position-dependent) nitrogen isotope ratios of nitrous oxide (N2O) has been developed. It is based on the addition of different amounts of doubly labeled 15N2O to an N2O sample of the isotope ratio mass spectrometer reference gas, and subsequent measurement of the relative ion current ratios of species with mass 30, 31, 44, 45, and 46. All relevant quantities are measured by isotope ratio mass spectrometers, which means that the machines' inherent high precision of the order of 10(-5) can be fully exploited. External determination of dilution factors with generally lower precision is avoided. The method itself can be implemented within a day, but a calibration of the oxygen and average nitrogen isotope ratios relative to a primary isotopic reference material of known absolute isotopic composition has to be performed separately. The underlying theoretical framework is explored in depth. The effect of interferences due to 14N15N16O and 15N14N16O in the 15N2O sample and due to 15N2+ formation are fully accounted for in the calculation of the final position-dependent nitrogen isotope ratios. Considering all known statistical uncertainties of measured quantities and absolute isotope ratios of primary isotopic reference materials, we achieve an overall uncertainty of 0.9 per thousand (1 sigma). Using tropospheric N2O as common reference point for intercomparison purposes, we find a substantially higher relative enrichment of 15N at the central nitrogen atom over 15N at the terminal nitrogen atom than measured previously for tropospheric N2O based on a chemical conversion method: 46.3 +/- 1.4 per thousand as opposed to 18.7 +/- 2.2 per thousand. However, our method depends critically on the absolute isotope ratios of the primary isotopic reference materials air-N2 and VSMOW. If they are systematically wrong, our estimates will also necessarily be incorrect.