The aim of this study is to estimate the long-term behaviour of trace metals, in two soils differently impacted by past mining. Topsoils from two 1 km(2) zones in the forested Morvan massif (France) were sampled to assess the spatial distribution of Cd, Cu, Pb and Zn. The first zone had been contaminated by historical mining. As expected, it exhibits higher trace-metal levels and greater spatial heterogeneity than the second non-contaminated zone, supposed to represent the local background. One soil profile from each zone was investigated in detail to estimate metal behaviour, and hence, bioavailability. Kinetic extractions were performed using EDTA on three samples: the A horizon from both soil profiles and the B horizon from the contaminated soil. For all three samples, kinetic extractions can be modelled by two first-order reactions. Similar kinetic behaviour was observed for all metals, but more metal was extracted from the contaminated A horizon than from the B horizon. More surprising is the general predominance of the residual fraction over the "labile" and "less labile" pools. Past anthropogenic inputs may have percolated over time through the soil profiles because of acidic pH conditions. Stable organo-metallic complexes may also have been formed over time, reducing metal availability. These processes are not mutually exclusive. After kinetic extraction, the lead isotopic compositions of the samples exhibited different signatures, related to contamination history and intrinsic soil parameters. However, no variation in lead signature was observed during the extraction experiment, demonstrating that the "labile" and "less labile" lead pools do not differ in terms of origin. Even if trace metals resulting from past mining and metallurgy persist in soils long after these activities have ceased, kinetic extractions suggest that metals, at least for these particular forest soils, do not represent a threat for biota.