EOF/PC analysis is applied to low-pass filtered deseasonalized CO2 growth rates from a global observational network, and two statistically-significant modes of interannual variability are identified. The spatial structure of the 1st mode is characterized by an interhemispheric gradient, while the 2nd mode is characterized by a land-ocean dipole in the Northern Hemisphere midlatitudes. The gravest PC reflects ENSO-like variability, although the phase relationship to ENSO appears to change around 1990. However, the leading PC exhibits a statistically-significant, stationary phase relative to an index of the Pacific Decadal Oscillation (PDO) throughout the sampling period. The next-to-leading PC, on the other hand, shows little relation to any of the climate indices examined here. These relationships intimate that, while ENSO may play a substantial role in interannual CO2 growth rate variability, the ENSO-growth rate relationship is not stationary, and additional modes of interannual variability may significantly influence year-to-year changes in CO2.