Current models of soil organic carbon (SOC) include a passive pool representing refractory soil organic matter (RSOM) with turnover times of hundreds to thousands of years. These models suggest that, as total soil C is depleted, it becomes proportionally enriched in RSOM. The objectives of our study were to quantify clay-associated organic matter resistant to peroxide treatment in soils presumed to have differing proportions of RSOM, hypothesizing that peroxide-resistant C in the clay fraction belongs to RSOM, and that its proportion will increase with total C depletion. Clay fractions (<2 micro m) from three soils from a cultivation sequence, differing in the duration of cultivation, one long-term cultivated soil and one long-term bare fallow soil corresponded to samples increasingly depleted in total organic C. Samples were suspended in 30% hydrogen peroxide and treated until no changes in C concentration were observed. Total C in the clay-peroxide suspensions decreased exponentially and displayed kinetics corresponding to labile, intermediate and peroxide-resistant pools. Carbon isotope analyses showed an enrichment of 13C in samples after peroxide treatment, compared with before, that decreased from 8 per mil in forest samples to 0 per mil in long-term bare fallow. The proportion of peroxide-resistant C did not differ between soils and represented 12% of initial clay-associated organic C. No proportional increase with soil C depletion was observed and when expressed on a whole-soil basis, the results underestimated proposed values for the RSOM pool, suggesting that peroxide treatment may not be appropriate for the estimation of the RSOM pool.