Abstract Two distinct modes of the East Asian winter monsoon (EAWM) have been identified, and they correspond to real and imaginary parts of the leading mode of the EAWM, respectively. Analyses of these modes used the National Centers for Environment Prediction (NCEP) and National Center for Atmospheric Research (NCAR) monthly mean reanalysis datasets for the period 1968–2003, as well as the Southern Oscillation index (SOI), North Atlantic Oscillation index, and eastern equatorial Pacific sea surface temperature (SST) data. Results were obtained by resolving a complex Hermite matrix derived from 850-hPa anomalous wind fields, and determining the resulting modes’ associations with several climate variables. The first distinct mode (M1) is characterized by an anomalous meridional wind pattern over East Asia and the western North Pacific. Mode M1 is closely related to several features of the atmospheric circulation, including the Siberian high, East Asian trough, East Asian upper-tropospheric jet, and local Hadley circulation over East Asia. Thus, M1 reflects the traditional EAWM pattern revealed in previous studies. The second distinct EAWM mode (M2), which was not identified previously, displays dominant zonal wind anomalies over the same area. Mode M2 exhibits a closer relation than M1 to sea level pressure anomalies over the northwestern Pacific southeast of Japan and with the SOI and equatorial eastern Pacific SST. Unlike M1, M2 does not show coherent relationships with the Siberian high, East Asian trough, and East Asian upper-tropospheric jet. Since atmospheric circulation anomalies relevant to M2 exhibit a quasi-barotropic structure, its existence cannot simply be attributed to differential land–sea heating. El Niño events tend to occur in the negative phase of M1 and the positive phase of M2, both corresponding to a weakened EAWM. The Arctic Oscillation does not appear to impact the EAWM on interannual time scales. Although the spatial patterns for the two modes are very different, the two distinct modes are complementary, with the leading EAWM mode being a linear combination of the two. The results herein therefore demonstrate that a single EAWM index may be inappropriate for investigating and predicting the EAWM.