The ratio of phosphate and nitrate concentrations in the deep ocean matches closely the Redfield ratio required by phytoplankton growing in the surface ocean. Furthermore, the oxygen available from dissolution in ocean water is, on average, just sufficient for the respiration of the resulting organic matter. We review various feedback mechanisms that have been proposed to account for these remarkable correspondences and construct a model to test their effectiveness. The model's initial steady state is close to the Redfield ratios and stable against instantaneous changes in the sizes of the nitrate and phosphate reservoirs. When classic flux estimates are adopted, nitrate responds to perturbation in 1000--2000 years and phosphate in 40,000--60,000 years. However, recently increased estimates of the input and output fluxes of nitrate and phosphate suggest that they respond more rapidly to perturbation, nitrate in 500--1000 years and phosphate in 10,000--15,000 years. Nitrogen fixation tends to maintain nitrate close to Redfield ratio with phosphate, while denitrification tends to keep nitrate as the proximate limiting nutrient and tie it in Redfield ratio to dissolved oxygen. Under increases in phosphorus input to the ocean, the relative responsiveness of nitrogen fixation and denitrification determine whether nitrate remains close to Redfield ratio to phosphate or to oxygen. If nitrogen fixation is strongly limited (e.g., by lack of iron), increasing phosphorus input to the ocean can cause phosphate to deviate above Redfield ratio to nitrate. Hence nitrogen dynamics can control phosphate behavior and nitrate can potentially be the ultimate limiting nutrient over geologic periods of time. When nitrate and phosphate are coupled together by responsive nitrogen fixation, negative feedbacks on organic and calcium-bound phosphorus burial stabilize their concentrations. If anoxia suppresses organic phosphorus burial, the resulting feedbacks on phosphate (positive) and oxygen (negative) improve regulation toward the Redfield ratios. Variants of the model are forced with a global record of phosphorus accumulation in biogenic sediments as a proxy for changes in phosphate input to the ocean over the past 40 Myr. Nitrate is generally regulated close to Redfield ratio to phosphate, despite large changes in phosphorus input. If nitrogen fixation is strongly limited, then there is one interval (~15 Myr ago) when a very rapid increase in phosphate input forces phosphate above Redfield ratio to nitrate. Decreases in phosphorus input cause phosphate and nitrate to quickly deviate below Redfield ratio with oxygen, removing anoxia from the ocean, while increases in phosphorus input rapidly increase anoxia. Hence we conclude that there appears to be an element of chance in observing today's ocean on the edge of anoxia with nitrate, phosphate, and oxygen all close to the Redfield ratios. ¿ 2000 American Geophysical Union