The current issue of Science News has an interesting article on the use and misuse of statistics in science. There is a good discussion on the general lack of understanding of the concept of statistical significance within the scientific community. The article also touches on the problem of testing multiple hypotheses simultaneously, and on the philosophical underpinnings of the classical statistical approach (p-values, confidence intervals). It concludes with a discussion of Bayesian approach as a potential remedy for some of these problems, a position which I personally think has some merit.