Interesting article in New Scientist reporting some mathematical modeling work done by Mark Tanaka: under plausible assumptions, ineffective (quack) medical treatments spread more quickly than effective ones, because (a) most people eventually get better on their own (and then blame it on the treatment), but (b) people adopt treatments based primarily on what they see other people doing, rather than by looking at scientific evidence. Basically, the longer you stay sick, the more time other people have to imitate you. There are obvious analogies with software development practices... :-)