This is a good review of the book titled "On Fact and Fraud: Cautionary Tales from the Front Lines of Science" by David Goodstein. It is a good review because Laura Greene, who herself is an acclaimed physicist, gave a few examples highlight from the book that represents the various categories highlighted by the book's author. In the process, Greene conveys a little bit of an idea on how science works and why the idea of reproducibility is utterly crucial in science.
Then, in 1986, Georg Bednorz and Alex Müller reported that they had measured a Tc of about 40 K in a LaBaCuO compound. Their discovery was first made known to most of the community at a Materials Research Society meeting in Boston, where two other independent and eminent scientists, Ching-Wu (Paul) Chu from Houston and Koichi Kitazawa from Tokyo, reported similar findings. Many of those present were convinced enough to repeat the experiments. Within weeks, the results were being reproduced in dozens of laboratories worldwide. The following January, a new compound with Tc ~ 90 K was announced; it too was widely reproduced within a month. Today, any new claim of HTS is met with well-equipped and capable laboratory scientists all over the world, so if the claim is not reproduced broadly and quickly, then it is not taken seriously. We have learned from this field that the most important diagnostic for determining scientific fact is reproducibility.
I think this concept has been severely underemphasized. In 'soft science' areas such as economics, social science, etc., this concept is almost non-existent. In science, it is what distinguish between a valid observation versus a fleeting "accident". It is why Fleishmann and Pons' "cold fusion" didn't make it. It is also what distinguishes anecdotal evidence from scientific evidence. This is an idea that needs a lot more coverage to be conveyed to the public.