The carbon dioxide emissions from cropland play important roles in the regional carbon budget. In this study, continuous measurements of the ecosystem respiration (RE) were obtained using the eddy covariance technique in a winter wheat-summer maize double cropping agroecosystem mainly between 2004 and 2012 in order to identify the among-year variations in RE and the related factors responsible. The annual RE, estimated by Lloyd and Taylor model, which was the most accurate, was 1866.4 ± 105.75 g C m-2 year-1 and it ranged from 1650.68 g C m-2 year-1 to 1945.57 g C m-2 year-1 during the eight years. The seasonal RE values were 867.98 ± 125.24 g C m-2 year-1 and 890.55 ± 131.34 g C m-2 year-1 for wheat and maize, respectively. Additionally, crop residue carbon ranged from 322.73 g C m-2 year-1 in 2012 and 453.49 g C m-2 year-1 in 2007. Correlation analysis indicated that the interannual variations in wheat and maize RE were correlated with the seasonal mean soil water content (W-Ws) and maximum leaf area index (W-LAImax) of wheat, and seasonal mean air temperature of maize (S-Ta), respectively. A rest method was attempted to investigate whether these relationships were occasional or inevitable. The rests of RE, i.e. the difference between simulated and observed RE values, were significantly influenced by LAI of wheat and hourly Ta of maize season but not by hourly Ws of maize season, indicating that the influence of W-LAImax and S-Ta on RE were inevitable outcomes and that of W-Ws on wheat RE was occasional. So we suggested that one should not confirm the controlling factors of interannual variations in carbon fluxes just from simple relationships, which may be statistical coincidences and do not correlated with biotical processes.