Experiments: "convergence to a false truth" Long time ago, when animals still spoke, I had a course on ethics in research and one of the topics was the (involuntary, subconcious) elimination or ignoring of experimentally measured data if they were too far off compared to the currently accepted value, i.e. nobody likes to publish 'an outlier' result. The professor at the time named it "convergence to a false truth". 
Is this term still recognized as the reference for this behaviour? 
I was wondering if someone could point me to bibliographic references of reports on cases where the outlier in the end proved to be the correct value. 
 A: In the early to mid-1970's, particle accelerators were being used to look for evidence of new subatomic particles. Sketchy evidence of two which were later to be named the J/psi  and psi-prime began to show up in experiments in 1973 and a significant amount of time through 1974 was spent by the experimentalists involved to determine if the data points they had in hand were just outliers (statistical accidents caused by noise) or real evidence of a new particle. 
Key to the determination was multiple repetition of the same experiments on one accelerator, replication of those experiments on a completely different machine, and refinement of the experimental conditions, by which means the experimentalists could progressively zoom in on the outliers and study their properties. 
The outliers were not statistical flukes or noise: they were real and resulted in a fundamental breakthrough in particle physics which was named the November Revolution (see Riordan, The Hunting Of The Quark, chapters 12 and 13). 
