Field studies of atmospheric CO2 effects on ecosystems usually include few levels of CO2 and a single soil type, making it difficult to ascertain the shape of responses to increasing CO2 or to generalize across soil types. The Lysimeter CO2 Gradient (LYCOG) chambers were constructed to maintain a linear gradient of atmospheric CO2 (~ 250 to 500 µL L-1) on grassland vegetation established on intact soil monoliths from three soil series. The chambers maintained linear daytime and nighttime CO2 gradients. Temperature variation within the chambers affected aboveground biomass and evapotranspiration, but the effects of temperature were small compared to the expected effects of CO2. Aboveground biomass on Austin soils was 40% less than on Bastrop and Houston soils. Biomass differences between soils resulted from variation in biomass of Sorghastrum nutans, Bouteloua curtipendula, Schizachyrium scoparium (C4 grasses) and Solidago canadensis (C3 forb), suggesting the CO2 sensitivity of these species may differ among soils. Evapotranspiration did not differ among the soils, but the CO2 sensitivity of leaf-level photosynthesis and water use efficiency in S. canadensis was greater on Houston and Bastrop than on Austin soils, while the CO2 sensitivity of soil CO2 efflux was greater on Bastrop soils than on Austin or Houston soils. The effects of soil type on CO2 sensitivity may be smaller for some processes that are tightly coupled to microclimate. LYCOG is useful for discerning the effects of soil type on the CO2 sensitivity of ecosystem function in grasslands.