[jakeduke]

Some loosely connected thoughts on this from a grad student in biomedical engineering:

A huge part of the problem that I didn't see addressed in the OP is that negative results don't get you into Science, Nature, PNAS, etc. There is such a push for positive findings with statistical significance that results like "the treatment resulted in a small but insignificant decrease in x" aren't usually very well received, no matter how sound the explanation of the result. It also sucks real hard to work for 2+ years on a study to realize at the very end that your main hypothesis didn't pan out, so I'm sure there's a fair amount of "how can I look at my data so I can find some significance" going on.

All this being said, when I read a paper, I can (usually) tell fairly quickly if the study is well designed and if the science is good or not. So if someone's hypothesis is supported by one figure with some questionably significant data, 5 other figures of loosely related stuff and discussion sections that do lots of hedging, I'm not going to be terribly inclined to take much from the paper. So I'm not quite sure how awful this is for science in general, because the people who are reading the stuff generally have decent BS detectors in their respective fields.