Carbon (C) exchange of a ∼200-year-old eastern hemlock (Tsuga canadensis L.) forest in central Massachusetts, USA, was estimated from mid-October 2000 through October 2001 based on eddy covariance measurements and statistical modeling from microclimatic data. Measurements were made in 68% of the hours during the year of study, with > 50% coverage in all months except December and August. Data were filtered by wind direction and atmospheric turbulence to remove invalid measurements. Analysis of filtered data showed that photosynthetically active radiation (PAR) was significant in predicting C exchange, except during the winter. Daily minimum air temperature affected C exchange in autumn and winter, whereas time of day, water vapor pressure deficit and air temperature had significant effects on C storage in spring, summer and fall. Most C storage in the stand occurred in April through July and in October 2001, with maximum rates in April and May. Persistent cold weather prevented C storage in December through March. In early spring 2001, C uptake was sensitive to nocturnal frost: daily minimum air temperatures below 0 °C reduced C fixation, and minima below –5 °C caused its virtual cessation. Soil temperature was a poor predictor of C balance during this period. In August, high soil and air temperatures (averaging 16.7 and 21.1 °C, respectively) drove high ecosystem respiration, which approximately balanced C uptake. These patterns show potential for stimulated C storage in hemlock forests in a warmer climate with fewer spring and autumn frosts, but reduced C storage during warmer summers. Estimated annual C storage was 3.0 Mg ha−1, which is higher than for younger coniferous and deciduous forests during earlier years in the northeastern USA. Long-term data are needed to determine if the estimated high C storage in this hemlock forest is a result of interannual climate variation or an effect of forest composition.