Once more, it's the most commercial time of the year, and every one is coming out with their Holiday Book recommendations. While the NY Times and the Washington Post put out the most well known best books list (and I'm peeved, but not surprised, that Thomas Pynchon and Richard Powers didn't make these top 10 lists), Discover Magazine has put out a list of the best 25 science books of all time.
The blog for the National Book Critics Circle notes that this list "is more about scientific umph than readability." Frankly, I don't even think that's true - I really can't see what bizarre criteria would generate a list that includes Newton's Principia (not that readable), Lovelock's Gaia (not much scientific umph), and Watson's Double Helix (very readable, but not a scientific milestone like the Principia). "Scientific umph" is frankly a bad criterion for a list of great scientific books, because major accomplishments in science are generally not first presented in book form (with a few obvious and older exceptions - Newton's Principia, Darwin's Origin. Sheer literary value is not a good criterion either, because you can have well-written 'science' books where the science itself is not really that substantial, but the writing is so good they're worth reading anyway. (Some people are going to hate me for this, but I find Lewis Thomas's books to be in this category. Personally, I don't find them as insightful about science as the books on my list below, but I know others disagree.)
A list of great books should have both scientific umph, and literary value, but in this sense: they should be compelling, well-written books that convey something deep about science - how it works, how its practitioners think, or about a specific concept. They should make a serious contribution to how we (scientists and the public) think about some aspect of science.
So I've put together my (currently partial) list of what I think are the best science books (in English), books which should be readable (though not necessarily easily readable - understanding science does take effort) by people who aren't professionals in the subject dealt with:
The Making of the Atomic Bomb, Richard Rhodes (1986) (and the 'sequel', Dark Sun, about the hydrogen bomb). This is the most amazing scientific history I've every read. Rhodes writes in absolutely stunning prose, and gives the best current historical account of the development of nuclear physics. Along with the elements of scientific history and biography, Rhodes also develops other themes, like the relationship of science to war, the state, and secrecy. Science writing doesn't get any better than this.
The Character of Physical Law, Richard Feynman (1965). I have fantasies about some day teaching an Intro to Science course using these incredible lectures about scientific reasoning. Feynman was a master lecturer (keep in mind that essentially all of Feynman's books are edited transcriptions of lectures). So while the prose, in a literary sense, may not be Pulitzer material, as lectures, Feynman's language is superb. This book provides very deep insight into scientific thinking, with (naturally) a large slant towards physics.
On the Origin of Species, Charles Dawrin (1859) (Also, The Descent of Man, which I think is a better second Darwin book to read than the Voyage of the Beagle.) One of the few books of the last 200 years to be in itself a major scientific contribution. The Origin is still popular because it is one of the most readable original scientific contributions out there. Darwin was apologetic about his writing style, but hey, in general those Victorians were pretty damn literate, so it's a good read. Actually, Darwin's style is quite good - he usually avoids the pathologically florid style which other scientific writing of that era often succumbed to. This book is substantial because it lays out a major scientific discovery and is an excellent example of scientific reasoning.
Feynman Lectures on Physics, Feynman, Leighton, and Sands (1964) (In principle, this should be generally readable, since it begins with Freshman physics; in reality, it's very tough. Lay readers can do almost as well with the subset of the lectures contained in 6 Easy Pieces and 6 Not So Easy Pieces.) Why am I giving this a separate entry, when I've already included The Character of Physical law? These lectures, while providing similar insight into general scientific reasoning, give an excellent presentation of specific scientific content - of what we have actually learned through scientific reasoning. If you're a person whose willing to plow through dense books by people like Kant, Wittgenstein, or James Joyce, the same effort with the lectures on physics will be well worth the payoff. In terms of language, again keep in mind that these are edited lectures - from this perspective, the language is very well crafted.
The Double Helix, James Waton (1968) This is a well-written account of one of the major discoveries in biology.
The Selfish Gene, Richard Dawkins (1976) This is sort of a popular book, sort of an original contribution, but like Darwin's Origin, a well-written exposition of one way of reasoning using evolutionary concepts.
What is Life, Erwin Schroedinger (1944) This is somewhat dated, and some of the ideas presented have been rejected. However, this book single handedly pushed many physicists into biology, some of whom turned out be major figures in the development of molecular biology. This book is valuable for more than just historical interest though - Schroedinger raises unsolved issues about the Origins of Life.
Genius, James Gleick (1992) (Chaos is another great Gleick book that I think is at the same level.) This is my favorite scientific biography of all time. Gleick, like Rhodes, is a superb writer. I almost didn't put this book on this list though, because I don't think Gleick effectively conveys the content of Feynman's scientific work. (In Gleick's defense, quantum electrodynamics is incredibly abstract and tough to convey.) But, as a work about what makes a scientist tick, and how a great scientist develops, this book ranks among the best.
Goedel, Escher, Bach, Douglas Hofstadter (1979) A great big, substantial meditation on math, music, and logic, composed in vivid, well-crafted language.
A Brief History of Time, Stephen Hawking (1988) A classic discussion of some of the big questions facing physics.
I'm going to stop the list here (out of sheer laziness, and to leave room for other great authors I haven't read yet), but obviously there are many more that can be included. There are a bunch of other books that are some of my personal favorites, but don't, for one reason or another, quite have the timeless/universal quality of the books above:
The Road To Reality, Guns, Germs, and Steel, Ahead of the Curve, The Demon-Haunted World, Human Natures, The Triumph and Tragedy of J. Robert Oppenheimer, The Creationists (this is a really well-written book about people trying to cope with their faith and the progress of modern biology), This is Biology, Boltzmann's Atom, Silent Spring... - this list can go on for awhile.
My favorite science book published in 2006 was Nicolas Wade's Before the Dawn, about human evolution. Wade is a skilled writer who can effectively convey scientific content in accessible language. On the downside, I don't think this book does enough to distinguish hypotheses and speculation from widely accepted scientific conclusions, always a tough issue when you're describing science at the frontier of a given field.
There's a lot of good stuff to read out there.
UPDATE 12-4-06: I forgot to include one major book that should have been on Discover's list: The Eight Day of Creation, by Horace Freeland Judson (1979) This is one of the most insightful histories of science. It deals with the molecular biology revolution, and is full of excellent material about how the founding molecular biologists chose their questions, thought about them, and largely solved them. There is no better book on molecular biology.