The quality of soil organic matter (OM) depends on its distribution among labile and recalcitrant pools and the quality of each pool considered. OM quality is assumed to decrease as decomposition proceeds, but to verify (his assumption it is necessary to define quality in operative terms. Here we study the change in OM quality during decomposition of mixtures of four plant materials (Medicago sativa whole ground plants, and ground litter of Eucalyptus globulus, Quercus ilex and Pinus halepensis) with a mineral red earth, incubated at different depths (5, 20, and 40 cm) for 2 years. OM quality was evaluated from acid hydrolysis, considering three pools: (a) Labile Pool 1, obtained by hydrolysis with 5 N H2SO4 at 105 degreesC for 30 min; (b) Labile Pool II, obtained by hydrolysis with 26 N H2SO4 at room temperature overnight, then with 2 N H2SO4 at 105 degreesC for 3 h; and (c) Recalcitrant Pool, the unhydrolyzed residue. In agreement with previously published results, the recalcitrant C/total OC (RIC), and recalcitrant N/total N (RIN) ratios are regarded as indicators of global OC and N quality. In addition, in Labile Pools I and 11, the ratio carbohydrate C/polyphenol C is used as indicator of OC quality. The main findings obtained by applying this approach can be summarized as follows: (1) In undecomposed plant materials, initial RIC ranged from 25% to 60% (Medicago and Pinus mixtures, at extreme values). Throughout decomposition, RIC values increased strongly (Medicago mixtures), slightly (Eucalyptus), or were roughly maintained (Quercus and Pinus), suggesting that strong decreases in OC quality occur only for easily decomposable plant materials. (2) Initial RIN values were between 15% and 30%, i.e., much lower than RIC ones. In contrast with the behaviour of RIC, the RIN values strongly increased in all cases, or, in other words, N quality clearly decreased for all plant materials, owing not only to a lower mineralization of the recalcitrant N, but also to a net incorporation of N to this pool. The amount of incorporated N is significantly related to the initial lignin content of the incubated plant material. Such incorporation seems to occur during wet periods; in contrast, its relationship with temperature was hardly detectable. No similar phenomenon was detected for recalcitrant C. (3) The C-13-CPMAS-NMR spectra of the recalcitrant pool showed prominent peaks in the 0-45 ppm region, which corresponds to the alkyl C and accounts for up to 50% of the total unhydrolyzable C in Quercus mixtures. In contrast, the aromatic zone, 110-160 ppm, was poorly apparent. These features were maintained more or less intact during the 2 years of field incubation, and suggest that lipidic polymers represent a substantial part of the recalcitrant pool. (4) Throughout the decomposition process, the ratio Labile Pool II/Labile Pools I + II decreased for carbohydrates, and increased for phenolic compounds. The use of these ratios is suggested to evaluate the degree of decomposition of plant residues. In the Labile Pool I the ratio carbohydrate C/polyphenol C remained the same, whereas for Labile Pool II this ratio decreased strongly, suggesting that the changes in quality may be restricted to a single pool. (5) Samples incubated in upper horizons (5-cm depth) were subjected to a much drier pedoclimate than those incubated at deep layers (20 and 40 cm), resulting in a slower mineralization of both the labile and the recalcitrant pools of C and N. Nevertheless, on a mineralized OC basis, most indicators of quality did not differ statistically between depths. Hence, the drought in the upper horizon retarded the decomposition, but dit not result in a different biochemical evolution. Because of its simplicity, chemical fractionation into three pools is a useful approach to characterize biochemical changes in C and N quality during plant residue decomposition. (C) 2002 Elsevier Science B.V. All rights reserved.