Sadly it is the very reasonable assumption that is wrong: important findings often do not get replicated. Usually the setup and execution (even working from gre

Sadly it is the very reasonable assumption that is wrong: important findings often do not get replicated. Usually the setup and execution (even working from great notes) is long and difficult, and often involves something specialized like trans-genetic mice (engineered for experimenting on this specific thing). And the benefits for being the lab that successfully repeats an experiment are approaching 0, and because there are so many thing that could go wrong a failure to replicate is not often a slam-dunk repudiation. So there is little incentive for people to do this. This should probably be fixed, maybe with some right-of-passage for Post-Docs being assigned to replicate important works, with explicit funding from the NIH (or someone) earmarked for this as a sort of training budget.