There is a recent podcast in The Weekly Weinersmith that talks about the problems related to Scientific Reproducibility, and what some people are trying to do about it at the "Reproducibility Initiative".
In theory, reproducibility is one of the foundations of Science. The scientific method rests on the idea that you are looking for repeatable patterns: Actions that can be done over and over again, giving very predictable results. In practice, however, making sure that your research is reproducible, or trying to reproduce research made by other scientists requires a significant effort. And this effort is not rewarded by the current academic structure of our universities and funding institutions. The result is that there is a LOT of research out there that is not, in fact reproducible. And thus, from a scientific point of view, can't really be regarded as fact.
But the people in the podcast can explain this much better than I do, so please listen to them!