Science is inherently conservative, based as it is on scepticism and the demand for evidence. We seek reasons for our beliefs, and this is why science claims a special status as a source for reliable knowledge. That's not to say, however, that scepticism and caution are cost-free.

On scientific matters with practical implications — climate change and what to do about it, for example — reticence may even diminish the influence of the scientific perspective. This is the view, at least, of NASA climate scientist James Hansen, the victim last year of Bush Administration efforts to prevent the flow of information from scientists to the public. In a recent paper, Hansen now suggests that the habitual reticence of scientists may be “inhibiting communication of a threat of potentially large sea-level rise”.

The problem, he argues, stems from strong nonlinearity in the physics, which could make polar ice sheets melt much faster than generally expected, raising sea levels next century by several metres (www.arxiv.org/physics/0703220). As white snow melts into darker wet ice, for example, it begins to absorb more solar energy, melting faster still. This nonlinear mechanism may already be at work in the ice sheets of Greenland and Antarctica, which satellite data show to be losing 150 km3 of ice each year, twice as fast as they were a few years ago.

Most public discussion of climate change remains largely dominated by linear thinking.

But the operative phrase here is “may be at work”. There are so many factors involved that no one can be absolutely sure. The sheer complexity of ice-sheet physics, or of any other part of the climate process, demands computational models able to integrate many factors, but these always seem open to legitimate criticism given the number of parameters they contain. The latest and biggest model may be 'the best', in some sense, but that doesn't mean it is any good. Hence, the reticence.

Even so, one can be cautious about specific mechanisms and outcomes, but far more vocal on general points that seem beyond dispute. Most public discussion of climate change — and many other matters of policy — remains largely dominated by linear thinking with its expectations of continuity and the legitimate extrapolation of trends. Yet we know that such expectations are unrealistic for nonlinear systems, which generically exhibit phase transitions and bifurcations. Talk of a catastrophic shutdown of the North Atlantic Conveyor, or of possible 'runaway' global warming, isn't irresponsible hysteria; it's plausible speculation that is consistent with everything we know about nonlinear systems. We'd all be better off, as the British theoretical ecologist Robert May once argued, “if more people realized that simple nonlinear systems do not necessarily possess simple dynamical properties.”

Human activity has not only significantly altered the concentration of atmospheric CO2 — from about 280 to 380 p.p.m. — but has done so at a rate that is historically unprecedented. Even without any climate modelling, this knowledge alone is cause for concern. What we shouldn't be reticent about are the inherent dangers of strongly disturbing a highly nonlinear system that we're not close to understanding, and on which our lives depend. We may not know the future, but we can have confidence that it won't unfold gradually and predictably. There will probably be plenty of surprises, driven by instabilities and positive feedbacks. Precaution would seem very well-advised.