Soil respiration is an important component of the annual carbon balance of forests, but few studies have addressed interannual variation in soil respiration. The objectives of this study were to investigate the seasonal and interannual variation in soil respiration, temperature, precipitation, and soil water content in two New England forest soils and to develop and evaluate empirical models for predicting variations in soil respiration using temperature and soil moisture content. We have been measuring soil respiration, using dynamic chambers in well-drained upland sites and poorly drained wetland sites since 1995 at the Harvard Forest, Massachusetts, and since 1996 at the Howland Forest, Maine. The upland sites had consistently greater rates of respiration than wetlands. Prolonged drought periods in 1995, 1998, and 1999 at the Harvard Forest resulted in decreased soil respiration rates in the uplands, particularly once soil moisture contents decreased below about −150 kPa. In contrast, wetland respiration increased upon drying. The interannual variation in soil respiration at the Harvard Forest, 0.23 kg C m−2 yr−1, exceeds the interannual variation in net ecosystem exchange (NEE), 0.14 kg C m−2 yr−1 previously measured for this forest, indicating that interannual variation in soil respiration can have an important influence on NEE. Interannual variation was lower at the Howland Forest, and the effects of low soil moisture content on respiration rates were more subtle. The onset of spring was variable among years at both forests, owing to variation in both temperature and precipitation, and contributed to 33–59% of the annual variability in total carbon release. At the upland sites, parameterization of empirical regression models for respiration as a function of soil temperature was inconsistent among years, indicating an important effect of interannual variation in soil water content. The negative residuals of the Harvard Forest temperature regression model were best explained by drought conditions (soil matric potentials ≤−150 kPa). This function was only applicable during severe drought and did not account for less severe dry periods that also reduced soil moisture and soil respiration. An empirical regression model for the wetlands as a function of temperature was significantly improved with the addition of a soil moisture function, which increased respiration rates under dry conditions and decreased it under wet conditions. Climatic changes resulting in drier conditions will likely decrease soil respiration rates in uplands and increase soil respiration in wetlands.