Assessments of terrestrial carbon fluxes require a thorough understanding of links between primary production, soil respiration and carbon loss through drainage. In this study, stem girdling was used to terminate autotrophic soil respiration including rhizosphere respiration and root exudation in a temperate Norway spruce stand. Rates of soil respiration and dissolved organic carbon (DOC) formation were measured in the second year after girdling, comparing an intact plant rhizosphere continuum with an exclusive decomposer system. The molecular and isotopic composition of DOC in the soil solution was analysed with a coupled Py-GC/MS-C-IRMS system to distinguish between the carbon sources of dissolved carbon. Pyrolysis products were grouped according to their precursor origins: polysaccharides, proteins or of mixed origin (mainly derivates of lignins and proteins). When dead roots became available for decomposition, rates of heterotrophic soil respiration in girdling plots peaked at 6.5 mu mol m(-2) s(-1), comparable to peak rates of total soil respiration (autotrophic and heterotrophic) in control plots, 6.1 mu mol m(-2) s(-1). A significant response of soil respiration to temperature was found in control plots only, showing that an unlimiting supply of organic substrates for microbial respiration may mask any temperature effects. The enhanced decomposition in girdled plots was further supported by the isotopic composition of DOC in soil solution; all three precursor groups became isotopically enriched as the growing season progressed (polysaccharides by 2.3 parts per thousand, proteins by 1.9 parts per thousand, mixed origin group by 2.2 parts per thousand). This indicates a trophic level shift due to incorporation of organic substrate into the microbial food chain. In the control plots' mixed origin fraction, the isotopic composition changed over time from a signature resembling that of lignin (-28.9 parts per thousand) to one similar of the protein fraction (-25.7 parts per thousand). Significant temporal changes of structural DOC composition occurred in the girdling plots only. These results suggest that changes in the microbial community and in decomposition rates occurred in both girdled and control plots in the following ways: (i) increased substrate availability (dead roots) gave rise to generally enhanced performance of the decomposer community in girdled plots, (ii) root-derived exudates probably contributed to enhanced decomposition of recalcitrant lignin in the control plots and (iii) the structural composition of DOC seemed to be more a result of decomposition than of plant root exudation in all plots. (C) 2007 Elsevier Ltd. All rights reserved.