To test the hypothesis that in oligotrophic areas of the ocean respiration exceeds production, a 12-month study was undertaken of in vitro-determined net oxygen production and consumption in the top 150 m of the water column at the extreme oligotrophic site, Station ALOHA, in the North Pacific subtropical gyre. Throughout the year the water column was observed to be in metabolic deficit, the calculated cumulative shortfall being 9±1.7 mol O2 m−2 a−1 (approximately 100 g C m−2 a−1), an amount equivalent to 40% of measured production (annual estimated rates of production and consumption were, respectively, 22 and 31 mol O2 m−2 a−1). We consider three possible explanations for the observed deficit: (1) the in vitro oxygen rate measurements, in themselves, are fundamentally flawed and should be discounted, (2) the observations are correct and the observed deficit is a true account of the balance of oxygen (and organic carbon) at Station ALOHA, or (3) the observations are correct as they stand, but need not be interpreted as organic carbon imbalance for that ecosystem. We find no error unique to the oxygen rate measurements themselves. We find also no evidence that the associated organic carbon deficit can be sustained over the long-term by internal organic reserves or by external subsidy. Accordingly we accept the geochemical findings that calculated in situ oxygen flux requires the euphotic zone of the water column at this site to be slightly (circa 2 mol C m−2 a−1) autotrophic, in contrast to the simple analysis of our observations which gives a net heterotrophic water column. We discuss a number of processes that may give rise to the observed discrepancy. In part it may derive from the difficulty of reproducing the variations in the light field experienced by an algal cell due to vertical advection. It may also derive from the intermittency of production. This latter effect would manifest itself in the following manner. Because of its universal distribution in the food web, respiration has greater integrating properties than photosynthesis and so will give a more accurate estimate of the long-term mean in studies with coarse sampling frequencies. If the system is undersampled, then short bursts of photosynthesis are prone to be missed from the integration of the production term but will be seen in the consumption term: hence the apparent deficit. The corollary of this line of reasoning is that, in undersampled systems, respiration has the potential to give a more accurate measurement of integrated system production than photosynthesis.