"Why Most Published Research Findings Are False"

This paper by John P A Ioannidis came out in 2005, but it's only recently come to my attention. The sobering conclusion of the paper is given in the title, and it has received a fair bit of attention. My first take is that this depressing statement is probably much more true for patient-based studies than for hard-core experimental or "wet bench" science. A more digestible summary of the paper can be found here.

But all is not lost. Some feel that Ioannidis' paper is itself false! Check it out here and here. Ioannidis responds here.

As pointed out by others, perhaps the real take home message is that a simple reliance on p values, especially the enshrined "0.05" cut-off, can be hugely misleading. Remember: "There are three kinds of lies: lies, damned lies, and statistics".