This was posted a few day ago. See http://lesswrong.com/r/discussion/lw/3jx/the_decline_effect_and_the_scientific_method_link/ .
The follow-up on Wired
An interesting article on how statistical randomness, unconscious confirmation bias, publication bias, and lack of sufficient replication can affect reported data and therefore scientific conclusions. I'm curious as to what LessWrongians think about it.