Three short essays on what science is and what it can and can't do, by someone who was not only a very distinguished scientist but also a remarkably talented writer. And probably the sort of book that can be read and appreciated equally by working scientists, philosophers of science, and complete lay-people. Always assuming that they aren't the sort to be easily scared by references that leap back and forth between Coleridge, Shelley, Dr Johnson, seventeenth century playwrights and philosophers, and experimental work in immunology...

"An essay on scians" considers a whole raft of assertions about what science is and how it is perceived, often responding to them in unexpectedly playful ways — for example, he suggests that one of the joys of science is that essentially anyone can do it, and as a career-path it is, like sport in developing countries, a great way for ambitious young people from modest backgrounds to widen their horizons.

"Can scientific discovery be premeditated?" argues against the fashionable idea that scientists should be commissioned by funding bodies to answer specific (useful) questions: he points to numerous examples where someone researching in one field has made a discovery that turns out to be useful in quite a different area (e.g. x-rays). But he doesn't want us to call this "luck" — scientists go into such situations with their eyes open, and have trained themselves to see possible connections and crossovers.

The title-essay "The limits of science"looks into the consequences of the idea that there are certain types of question that are not susceptible to scientific examination — the famous "why are we here?" type of question. Medawar rejects the approach that these should be dismissed as not being valid questions: obviously they are questions some of us have a real need to ask. But he doesn't accept that this means we should put forward myth, metaphysics or religion as a more valid (or even equally valid) way of answering such questions. As long as they do not provide answers that can be empirically tested, he's not buying it. (But he does accept that metaphysics, in particular, can help to suggest ways of approaching difficult questions that scientists can learn from.)

"Limits" also provokes him to ask whether there is a built-in limit to the potential of science to answer questions that are susceptible to scientific investigation, just as there seem to be hard limits to things like population growth or the maximum size of aircraft. Is there ever going to be "too much knowledge" for scientists to keep an overview and do useful research into new things? He doesn't think so. The notion that there was ever a time when "one person could know everything" is silly, people have always specialised and worked in teams, and they continue to do so. Perceptively (given that he was writing in 1984, at the end of a long career), he also points out that computer databases have eliminated the need for an individual researcher to carry any but the most relevant technical knowledge around in their head.

Lovely writing, clear thinking, and only a hundred pages long. What's not to like? ( )