Interannual variability of sea ice in the Bering Sea and its relationship to atmospheric variability is analyzed using a singular value decomposition (SVD) analysis of sea ice concentrations (SICs) and 1000 hPa wind speeds in winter and spring seasons. The statistically significant first and second SVD modes, explaining 76.3% and 17.6% in winter and 54.6% and 29.6% in spring of the squared covariance between the two fields, are identified for SICs both in the winter and spring seasons with 1 month leading wind speeds. The spatial structures show that the first (second) SVD mode explains the SIC variability in the northeastern (northwestern) Bering Sea, related to the local northwesterly (northerly) wind anomalies for the positive SIC anomalies both in the winter and spring seasons. A comparison of the first SVD modes between the winter and spring seasons suggests that the difference of dominant patterns of wind anomalies results in the difference of SIC anomaly distributions between two seasons. The relationship between sea ice and atmospheric circulation anomalies indicates that one mode of the leading two SVD modes in each season is related to large-scale atmospheric circulation associated with the Aleutian low and the other mode is related to relatively local atmospheric fluctuations related with pressure anomalies over Alaska. Furthermore, a slight difference of 700 hPa geopotential height anomalies results in the substantially different sea ice anomalies. These results suggest that in order to know the interannual sea ice variability in the Bering Sea, a better understanding of the wind anomalies over the Bering Sea are important.