To cope with crises we need evidence and experts - not asses

Hugh Pennington is that all too rare being, a top-grade scientist who writes like a dream, analyses coolly yet thinks dramatically, and is alive to the unfolding human activity behind scientific crises. This praise will not come as a surprise to anyone who has heard him lecture or read one of his long, thoughtful pieces in the London Review of Books on the various food crises or who knows his passionate belief that we humans can learn and are not condemned to repeat history. He is, in short, a progressive - someone who believes in the potential of medicine and scientists to improve, not harm; who knows that, with encouragement, the state and companies can learn; and who promotes the improvement of public policy, yet is not afraid to point out when laws, inspectorates and political processes have been asses.

This book is an important expansion of that work, and it deserves to be widely read in all those disciplines and courses that now look at food as subject matter - whether in terms of risk management, food technology or science policy. In a world where risk is now a central notion within public policy, When Food Kills ought to be required reading for anyone appointed as "adviser" to frontline politicians. As further inducement, I should stress that one of its greatest strengths is its readability. Just when the narrative delves into complex issues and when the tired reader might start to skim, the author will give a personal thought, or deviate to tell a story about someone from the past or from his own family, or from his own experience (chairing the inquiry into the appalling outbreak of E. coli at Wishaw, Scotland, in 1996). The result is a gripping account of how risks are conceived, treated and ignored.

I suspect strongly that his motivation for writing this book is to have the space to put down all his thoughts rather than just the headlines. I found myself thinking of the late W. G. Sebald's The Rings of Saturn - part travelogue, part history, part rumination, part acute observation; both mirror and beacon. Unlike Sebald's work, When Food Kills is a book of, in and about science; a guided walk across difficult terrain in science, politics, philosophy and culture.

It is an erudite text yet not a textbook. Bibliographic sources are listed at the end, and key texts are summarised in the main text. Words are chosen carefully; Pennington tells us what he wants us to note. A purple section recounts Ludwig Fleck's brilliant 1935 discourse on the making of scientific "facts" - "discoveries are not made in a vacuum. They are deeply influenced by the training, experiences, attitudes and assumptions of researchers". This is a book with a deep set of messages. Royal Society inquiry chairs, please note.

Pennington's tale is complex because his subject matter ranges widely.

Although a medical man and emeritus professor of bacteriology at Aberdeen University, his interests are broad. The book covers disasters not just in food but in railways, water, mental asylums, oil rigs, mines and nuclear technology. It is this broad perspective that gives it value. Pennington is not frightened to draw conclusions or to side with arguments and individuals. His account of the BSE saga points out the role played by scientific jealousies, discusses why brilliant science was ignored to suit policy frameworks and explores why German rather than, say, British science dominated early work on transmissible spongiform encephalopathies (TSEs).

Through his examples of risk discovery and attempts at management, his big thesis begins to emerge: that science is framed by cultures and that what can lift scientists from the repetitive is a readiness to take risks, think dangerously and listen to evidence rather than preconceptions. But this requires a framework that sees awkward questioning as valuable.

The book is a many-layered discourse, in part on how science develops, in part on how waters can be muddied by political exigency (pressure to deliver analyses, results, sops to public opinion) and in part an appeal to scientists to get real about how even the most apparently abstract work carries ideological potential. Pennington is stringent about mismanagement and missed opportunities. He calls for teamwork and cross-disciplinary pooling. "Epidemiology is an essential tool in the investigation of outbreaks or infection. But it is insufficient on its own. It is always much better if it is used alongside biological information drawn from biochemistry, immunology, and genetics, just as crime investigations need evidence not only from witnesses but from fingerprints, tyre marks, and DNA as well." Why, he asks about the BSE saga, were experts on TSEs not among the Department of Health's team? Why did the state's apparatus back certain people and world-views? Things could have been so different, Pennington suggests, if people such as Alan Dickinson, head of the Neuropathogenesis Unit group at Edinburgh, had been asked to test the then dominant model of BSE's aetiology. Science fails the public interest, Pennington suggests, not just by what it does but when its principles - fearless questioning and rigorous pursuit of evidence - are not followed. Instead, so often, a dominant theory emerges that is acceptable, threatening but not too threatening and, above all, manageable.

When Food Kills is highly critical of institutionalised inspectorates. I had to re-read various passages to ensure that I read it right. Too often, the book argues, inspectorates fail to do their job; this happened at Wishaw, but not just there. Personal relations, class sympathies, being "one of us" intervene. The book concludes with a caustic but wonderful account of public inquiries. These ought to be the formal mechanism for all parties - experts, professionals, the state, citizens, industry - to learn lessons and correct practices. Too often, argues Pennington, inquiries are long, tortuous and add to, rather than dispel, policy fog. Importance is lost, awkward minutiae are carefully managed in the "story" presented at the launch. Only later or with forensic energy does the full story emerge, from, say, key issues buried in appendices in volume 94. Good governance requires such details, managing disaster science requires a preparedness to face complexity. Even before today's instant media, the real facts could belie the headlines.

This is an account by someone who has been on the inside as well as the outside of policy circles. Pennington favours quick inquiries that meet the 1994 assessment by Louis Blom-Cooper QC: an inquiry can be useful to assuage horror, to allay fears of whitewash, to go beyond individual issues to the general, to give opportunities for people to make representations and to make recommendations. As yet, Pennington has not matched Blom-Cooper's 11 inquiries but his own recommendations are worth following: aim for openness, inclusivity, evidence-based thinking, recognition that politicians hold preconceptions, and speed. I vote for that.

Tim Lang is professor of food policy, City University and co-author of Food Wars .