I'm about halfway through Guns, Germs & Steel and something's nagging me about the author's style. I just realised what's bothering me: he (intentionally, I think) blurs the distinction between hypothesis-generation and experimental validation. Specifically, his argument is largely based on anecdotal evidence and Gedankenexperiments. He'll present a series of known archealogical and anthropological facts and then make a few fairly reasonable logical processions to "prove" his explanation for some phenomenon. To be fair he frequently points out that his views aren't proven beyond doubt, but always with a wink and a nod that says, "But my explanation just makes sense doesn't it? So we all know I'm right."
Really the entire book is about the first phase of science: generating hypotheses. We take some facts we know and combine them logically to think up a plausible explanation. The next step, however, is the most crucial: we set out to prove our hypothesis wrong! We try to think of some way to definitively prove that our guess is wrong and thus refine our guess. Absent other evidence, the path of the sun through the sky lends itself to the logical explanation that it's movign around the earth.
Most of his arguments are plausible, but it irks me that he doesn't seem to be able to offer much convincing proof that they're actually true. I think I'm especially bothered because he does use actual data to demonstrate some of his founding principles and then drifts into the realm of conjecture without really acknowledging the difference.