Jon Turney resurrects the work of a neglected French thinker whose original ideas on the usefulness and limits of rationality in the life sciences are becoming increasingly relevant. Odd how intellectual fashion among English-speakers picks up some recent French writers, ignores others. Everyone knows about Foucault, or claims to, but few know anything of the man who accepted his Madness and Civilisation as a doctoral thesis.

The man was Georges Canguilhem, whose scholarly career began before and ended after Foucault's. He had all the right qualifications to be fashionable, starting his intellectual life as a fellow student of Sartre in the 1920s, becoming a hero of the resistance in the 1940s, and finally retiring as professor of philosophy at the Sorbonne in 1971. Yet by the time he died last autumn, in his nineties, few of his writings had made any impression outside France. A pity because, according to Nikolas Rose of Goldsmiths College, co-organiser of a recent conference in Canguilhem's honour in London, the profound issues raised by the contemporary technologising of life mean that we should get to know this neglected Frenchman better.

Canguilhem's life-long study was the way knowledge is established in the life sciences, and how this differs from the way we understand the development of physics and chemistry. One of his key precepts was that "the history of science is not only science's memory but epistemology's laboratory". The result was a body of work which explores both the conditions of rationality and its limits, which displays scientific truths as historically variable and plural, while nevertheless yielding privileged knowledge of the natural world.

So what of the claim for the contemporary relevance of Canguilhem's writing? At the conference, this came in three versions. One was argued by Richard Horton, editor of the Lancet, and one of the few outspoken critics of the current vogue for "evidence-based medicine". This term is shorthand for the idea that the randomised controlled trial is, as Horton put it, the "irreducible unit of truth" for clinical decision-making. This type of medicine has the rhetorical advantage that to oppose it is to appear to be in favour of medicine which is not based on evidence. But it is pernicious, in Horton's view, because it explicitly denies that any other kind of evidence apart from controlled trials has any value at all. It is, he suggested, an example of what Canguilhem called a scientific ideology, a concept developed in his Ideology and Rationality in the History of the Life Sciences. It means, Canguilhem said, "an epistemological concept with a polemical function", and Horton believes that evidence-based medicine fits the bill precisely. While trials are important, they mainly use crude measures of outcomes - they count deaths. And the bureaucratic and accounting enthusiasm for evidence-based medicine, boosted by commercial interests in selling "meta-analyses" to drug companies for advertising use, obscures the fact that doctors treat individual patients, not statistical populations.

³ÉÈËVRÊÓƵ

Horton, then, is looking to Canguilhem to support what is in some ways a traditional defence of clinical autonomy in the face of a reified set of decision-rules. The philosopher of science Ian Hacking, from the University of Toronto, uses a quite different strain in his work as an aid to reflection on a rather wider question: where our increasingly intertwined relationships with our information machines are taking us. There is a widespread feeling just now that we are set on a path which will lead us to become one with our technology - a race of cyborgs, part biological, part mechatronic.

Hacking drew mainly on a prescient lecture Canguilhem first delivered half a century ago on "Machine and Organism", which took issue with Descartes in a way that prefigures many of the questions which now preoccupy us. Descartes saw animals as machines, but humans as creatures which, while having machine-like bodies, were of a different order because of their faculties of judgement. Canguilhem, on the other hand, stressed the continuities at all levels.

³ÉÈËVRÊÓƵ

All organisms, in Canguilhem's view, seek to act on their environment using whatever devices they have at their disposal. For most organisms, these are necessarily biological - we should see organs as tools and machines. But his view is symmetrical: "Machines can be considered as organs of the human species", says Canguilhem. Or in Hacking's reading: "Rather than see technology as something which begins at our fingertips, we should see it as an extension of life itself". We need to consider this view further, he suggests, as we ponder whether what is happening in some of the most important parts of our culture now, as we develop new disciplines which treat biological objects as possible matter for (re)-engineering, is a kind of extension of what we have always done, or a real break.

The other area of the life sciences where "engineering" has been coupled with organisms in recent terminology is genetics, of course. And this is the third general area where Canguilhem's contemporary relevance is becoming apparent. Here, several speakers, including Paul Rabinow, harked back to his best-known work in English, The Normal and the Pathological. In it, he starts from his most important premise that the objects of knowledge of the life sciences are different from those of physics and chemistry, and that this implies that the kind of knowledge biomedical investigation produces is also different.

In particular, a medical practice which defines its task as restoration of normal biological function is problematic, because what is normal, abnormal or pathological can only be defined contextually, by observing shifting boundaries. The problem is partly that for a living organism the normal is never static, partly that of, as the philosopher of science Mary Tiles put it a few years ago, "the ambiguity in the concept of the normal between that which is usual and that which is as it ought to be". This is a familiar enough problem in the social sciences. Canguilhem places it at the heart of biology.

Originally, he was much concerned with physiological parameters, like blood pressure. But the movement of the life sciences means that we now have powerful new ways of displaying differences between individuals, most notably genetic differences. And it is far from clear how we will find the best language to speak of them in medical contexts. Suffice to say that when The Independent can report on a genetic "defect" which apparently prevents HIV infection taking hold in people who carry it, then we have some confusions still to clear up. And a writer whose emphasis is that "an anomaly is not an abnormality. Diversity does not signify sickness", may again be of some help.

³ÉÈËVRÊÓƵ

Jon Turney teaches science communication, University College London