Abstract We used differences in soil carbon δ13C values between forested sites and grasslands dominated by the C4 grass Schizachyrium scoparium (little bluestem) to detect the presence of former grasslands in the historical landscape of the coastal sand plain of Martha's Vineyard, Massachusetts, U.S.A. Soil δ13C was measured at (1) sites with long-term forest or grassland vegetation and (2) sites with known histories where forest vegetation invaded grassland and where forest converted to grassland. The δ13C of soil under long-term grassland was –24.1‰ at 0 to 2 cm depth and –23.4‰ at 2 to 10 cm and was enriched by 3.4‰ and 2.8‰ compared with soil under long-term forest. In forests that invaded grasslands dominated by S. scoparium, soil δ13C decreased as C derived from trees replaced C from S. scoparium. This decline occurred faster in surface soils and in the light soil organic matter fraction than in the mineral soil. In forests that converted to grasslands, soil δ13C increased and the rate of increase was similar in surface and mineral soil and in the different soil organic matter fractions. Rates of change indicated that soil δ13C could be used to detect changes in vegetation involving the presence or absence of S. scoparium during the last 150 years. Application of this model to a potential grassland restoration site on Martha's Vineyard where the landscape history was not known indicated that the site was previously unoccupied by S. scoparium during this time. The δ13C of surface mineral soil can be useful for detecting the presence of historic S. scoparium grasslands but only in the period well after European settlement of these coastal sand plain landscapes.