Agricultural practices are known to diminish soil organic carbon (SOC) stocks and alter carbon quality. We surveyed a diverse set of sites in heterogeneous landscapes to determine past land use histories and ages of agricultural abandonment in order to examine changes to and spatial controls on soil carbon pools. Soils were sampled using quantitatively excavated pits, to the extent of the rooting zone. Three regions (i. western New England (WNE); ii. southern Wisconsin; iii. northern Wisconsin) of northern hardwood forests with different patterns of agriculture abandonment, varying soil types and properties, and differing climates were examined. Carbon (C) and nitrogen (N) concentrations were measured and contents calculated using bulk-density relationships. In WNE, soils formerly used for agricultural practices accumulated soil organic carbon (SOC) at a rate of 0.33 Mg ha-1 y-1 for the first century of forest regeneration. Formerly plowed soils accumulated C in the organic (Oe + Oa), 0-10 cm, and deep mineral soil (> 20 cm), while formerly pastured or hayed soils accumulated C in the organic horizons and 10-20 cm portion of the mineral soil. Sites used for subsistence logging showed no accumulation trends. As expected, N accumulated with C, although the patterns of N accumulation were more varied. Physical fractionation of the top 20 cm of mineral soil (the maximum depth to which these soils were plowed) showed that the pool of C associated with soil minerals increased with stand age (0.04 Mg ha-1 y-1), but that modern agricultural soils possessed as much C in this fraction as the oldest forests. A two-month incubation of these soils demonstrated 48% more C was respired (as CO2) in modern agricultural fields than abandoned forests. Multivariate regression tree results demonstrated that the time since agricultural abandonment and climate were important determinants of SOC amounts within the western New England landscape. When those sites were compared with other northern hardwood forest soils from the Adirondack region of New York, the Green Mountains of Vermont, southern Wisconsin and northern Wisconsin, growing season degree-days (GSDD) was the best predictor of SOC totals (48% sums of squares explained), and the warmer regions heavily used for agricultural purposes were separated from the cooler montane forests. These results suggest that fundamental differences exist between the soils that are useful for agricultural purposes and the land that was abandoned or left undisturbed, and demonstrate the importance of regional soil carbon estimates.