>>12174766replication crisis is a direct result of "publish or perish", aka "papers are currency"
The generation of data at lightning speed these days means you have to constantly publish to keep up with your competitors when it comes to grant writing/awards. In addition, the need to publish a first author paper as a grad student means a LOT of paper are just being pushed out there; its entirely possible to do 5 years of good science, and you just happen to find out a hypothesis is mostly negative. So data gets pressure-massaged until something "significant" comes out, it gets pushed to a mid-tier journal, and the grad student graduates.
Most studies published are bunk. In my field, there are only a handful that making any lasting impact, as the rest just don't stand up or aren't that interesting.
The only way this changes is if the way publishing works changes. Something has to be reworked so "publish or perish" no longer exists; some stop-gaps have been introduced recently which could help alleviate the problem a little (micro-publications, for example, and of course biorxiv), but honestly I think if we switched to "publish experiments as they are finished", aka "build a paper publicly" it would help turn the focus away from whole-papers and to the individual experiments, allowing more visibility to problems before they are all cemented together.
The other thing; methodology. Go to 100 labs and ask them to do a standard western, and you'll get 100 slightly different/tweaked protocols. Some will be very different. The methods sections of all their papers, however, will read the exact same: "All samples were run by SDS-PAGE and blotted"
Personally I wish the methods was literally a collection of protocols from lab notebooks used to a T. I tried replicating a result once, couldn't get it, asked the lab for a protocol- holy fuck it was maybe 20% similar to what they wrote in the paper methods. Fuck that noise.
