RATIONALE: We compared gas chromatography/isotope ratio mass spectrometry (GC/IRMS) and liquid chromatography/isotope ratio mass spectrometry (LC/IRMS) for the measurement of delta C-13 values in carbohydrates. Contrary to GC/IRMS, no derivatisation is needed for LC/IRMS analysis of carbohydrates. Hence, although LC/IRMS is expected to be more accurate and precise, no direct comparison has been reported. METHODS: GC/IRMS with the aldonitrile penta-acetate (ANPA) derivatisation method was compared with LC/IRMS without derivatisation. A large number of glucose standards and a variety of natural samples were analysed for five neutral carbohydrates at natural abundance as well as at C-13-enriched levels. Gas chromatography/chemical ionisation mass spectrometry (GC/CIMS) was applied to check for incomplete derivatisation of the carbohydrate, which would impair the accuracy of the GC/IRMS method. RESULTS: The LC/IRMS technique provided excellent precision (+/- 0.08 parts per thousand and +/- 3.1 parts per thousand at natural abundance and enrichment levels, respectively) for the glucose standards and this technique proved to be superior to GC/IRMS (+/- 0.62 parts per thousand and +/- 19.8 parts per thousand at natural abundance and enrichment levels, respectively). For GC/IRMS measurements the derivatisation correction and the conversion of carbohydrates into CO2 had a considerable effect on the measured delta C-13 values. However, we did not find any significant differences in the accuracy of the two techniques over the full range of natural delta C-13 abundances and C-13-labelled glucose. The difference in the performance of GC/IRMS and LC/IRMS diminished when the delta C-13 values were measured in natural samples, because the chromatographic performance and background correction became critical factors, particularly for LC/IRMS. The derivatisation of carbohydrates for the GC/IRMS method was complete. CONCLUSIONS: Although both LC/IRMS and GC/IRMS are reliable techniques for compound-specific stable carbon isotope analysis of carbohydrates (provided that derivatisation is complete and the calibration requirements are met), LC/IRMS is the technique of choice. The reasons for this are the improved precision, simpler sample preparation, and straightforward isotopic calibration.