Does the atmosphere drive the oceans or do the oceans drive the atmosphere? That was one of the questions asked at the 2000 Chapman conference (American Geophysical Union) on the North Atlantic Oscillation. At such a meeting this question could appear to be targeted to the theoretical or numerical modelling community. As an oceans community we often see potential evidence for the longer-term influence of the oceans on the global climate system. Our understanding of ocean thermohaline circulation, with its sinks and sources, is linked to the concepts of oceanatmosphere interactions and feedback systems. But what does the presence of feedback in a physical system really imply? A linear system does not have feedback. So a system must be non-linear to be able to produce, or receive, a feedback signal. Quantifying natural systems involves considering both inherent and measurement noise. So now a discussion of non-linear physical systems, with noise, sounds as though we should be passing our question back to the engineering world of signal processing. Or does it?