Although gas chromatography-pyrolysis-isotope ratio mass spectrometry (GC-Py-IRMS) has allowed us to make online compound-specific delta18O measurements for about the last ten years, this technique has hardly been applied. We tested different pyrolysis reactor designs using standards (vanillin, ethylvanillin, a fatty acid methyl ester and alkanes) in order to optimize the GC-Py-IRMS delta18O measurements. The method was then applied to methylboronic acid (MBA) sugar derivatives (pentoses, 6-deoxyhexoses and hexoses). Plant- and microbial-derived monosaccharides were extracted hydrolytically from litter and topsoils before derivatization. The measured delta18O values of samples and co-analyzed reference material were first drift-corrected by use of regularly discharged pulses of CO reference gas. Secondly, they were corrected for the amount dependence of the delta18O values. Thirdly, the delta18O values were calibrated using the reference material (principle of 'Identical Treatment'), and, finally, a correction was applied by taking the hydrolytically introduced and water-exchangeable oxygen atoms into account. Our results suggest that the delta18O values of plant-derived monosaccharides in litter reflect the climatic conditions of the last year, whereas delta18O values of the respective topsoils reflect the averaged climate signal of the last decades or even centuries. This demonstrates the high potential of the method for palaeoclimate reconstructions.