There is considerable interest in modeling isoprene emissions from terrestrial vegetation, since these emissions exert a principal control over the oxidative capacity of the troposphere. We used a unique field experiment which employs a continuous gradient in CO2 concentration from 240 ppmv to 520 ppmv to demonstrate that isoprene emissions in Eucalyptus globulus were enhanced at the lowest CO2 concentration, which was similar to the estimated CO2 concentrations during the last Glacial Maximum, compared to 380 ppmv, the current CO2 concentration. Leaves of Liquidambar styraciflua did not show an increase in isoprene emission at the lowest CO2 concentration. However, isoprene emission rates from both species were lower for trees grown at 520 ppmv CO2 compared to trees grown at 380 ppmv CO2. When grown in environmentally-controlled chambers, trees of Populus deltoides and P. tremuloides exhibited a 30 - 40% reduction in isoprene emission rate when grown at 800 ppmv CO2, compared to 400 ppmv CO2. P. tremuloides exhibited a 33% reduction when grown at 1200 ppmv CO2, compared to 600 ppmv CO2. We used current models of leaf isoprene emission to demonstrate that significant errors occur if the CO2 inhibition of isoprene is not taken into account. In order to alleviate these errors, we present a new model of isoprene emission that describes its response to changes in atmospheric CO2 concentration. The model logic is based on assumed competition between cytosolic and chloroplastic processes for pyruvate, one of the principal substrates of isoprene biosynthesis.