We use output from a hindcast simulation (1958–2007) of an ocean biogeochemical and ecological model to inform an observational strategy for detection of a weakening Southern Ocean CO2 sink from surface ocean pCO2 data. Particular emphasis isplaced on resolving disparate conclusions about the Southern Ocean CO2 sink that have been drawn from surface ocean pCO2 observation studies in the past. We find that long-term trends in ΔpCO2 (pCO - pCO) can be used as a proxy for changes in the strength of the CO2 sink, but must be interpreted with caution, as they are calculated from small differences in the oceanic and atmospheric pCO2 trends. Large interannual, decadal, and multi-decadal variability in ΔpCO2 persists throughout the simulation, suggesting that one must consider a range of start and end years for trend analysis before drawing conclusions about changes in the CO2 sink. Winter-mean CO2 flux trends are statistically indistinguishable from annual-mean trends, arguing for inclusion of all available pCO data in future analyses of the CO2 sink. The weakening of the CO2 sink emerges during the observed period of our simulation (1981–2007) in the subpolar seasonally stratified biome (4∘C < average climatological temperature < 9∘C); the weakening is most evident during periods with positive trends in the Southern Annular Mode. With perfect temporal and spatial coverage, 13 years of pCO data would be required to detect a weakening CO2 sink in this biome. Given available data, it is not yet possible to detect a weakening of the Southern Ocean CO2 sink with much certainty, due to imperfect data coverage and high variability in Southern Ocean surface pCO2.