I own something like 30 statistics and modeling books, accumulated over the years in classes and working on projects, and this is by far the best of them. I am a biologist with a decent background in statistics/modeling and a good knowledge of S-Plus (earned the hard way), but I am NOT a true statistician or mathematician. Crawley's descriptions are wonderfully lucid, written in ENGLISH rather than mathematical jargon, and his analyses are thoughtful and interesting. I learn as much from watching the way that he approaches a problem as I do from his explanations, and he gives you code so that you can try out and modify the examples. S-Plus and R are nearly identical, so although the book was written for S-Plus, it is equally useful in R. The book covers nearly every topic that a scientist could conceivably need for data analysis, to a degree of sophistication that will be adequate for 99% of its readers. For anything more, you are probably going to need to see a statistician anyway. The book's website includes 3 additional chapters (on gamma errors, additive models, and multivariate statistics).
I use this book in two ways: 1) as a valuable reference/cookbook for things I haven't tried, and 2) to remember, or to teach myself for the first time, how statistical tests work. In case after case, Crawley goes out of his way to show how and why statistical tests are calculated the way they are. S is ideally suited for this, since it makes the math painless and hides it behind nice graphical output, and lets you concentrate on understanding the concepts. If you are a student or professional who uses statistics and R or S-Plus, I can't recommend it highly enough, especially if you are someone who doesn't naturally think in mathematical symbols, or if you are more interested in learning how to do something and in understanding why it works, than in reading proofs or doing the underlying algebra.