Accountability in Research, 2000

What should a scientist do with an anomalous experimental result (one which flies in the face of prevailing theories), which is statistically significant (at the five sigma + level), yet which cannot be repeated at will? This is the problem which plagued us early on in our non-plasma or "cold" fusion experiments which began in spring 1986 at Brigham Young University, 2.5 years before we heard of the ostensibly-related work of Drs. Pons and Fleischmann. It is a question which haunts us still. I invite the reader to seriously consider the question posed above: what would you do with such data? If you walk away from an anomalous result, you could miss something important. Indeed, is it scientifically honest to ignore such data? One can argue that scientific instruments often show "glitches," and this is probably just one of these. Quite possibly. But what if the "glitch" is not only statistically significant, but also has interesting characteristics, such as the correct energy (2.5 MeV) for neutrons from deuteron-deuteron fusion, and the expected signal-width? [1] What if other data runs which are individually significant only at the 2-sigma level also contribute this peak? [1]