The Michelson–Gale–Pearson experiment was a large-scale version of the Michelson-Morley Experiment and the Sagnac-Interferometer which attempted to measure the Sagnac Effect due to the rotation of the earth. It has been claimed that this experiment provided evidence for the earth's rotation.
“ The outcome of the MGP experiment was ambiguous, though maybe no more ambiguous than the small persistent positive shift observed in MM experiments. Composed of 269 separate tests with readings that varied from -0.04 to +0.55 of a fringe, and a mean at +0.26 fringes, the MGP experiment could be interpreted to yield a positive result of ≈ 0.3 km/s - therefore near the speed of the earth's rotation, but the result was of borderline significance. It could be said that the experiment was inconclusive because it adduced neither proof that there was a shift in the phase of the light beams, nor that there wasn't one. ”
Essentially the tests saw wild results. There was almost no change to light's velocity in one test, and then a lot of change in another test. It is perplexing that the rotation of the earth would start and stop when tested at different times. Only through the statistics was it claimed that the experiment saw the rotation of the earth. As stated above, the inconsistent results were ambiguous in nature and could offer no evidence of the shift in the phase of the light beams.
Importance of Consistency
In attempt of correlating such results with the rotation of the earth the mean is discussed, but what of the average or the median? What of the fact that one of the extremes is near zero? Which is the baseline that is being modified?
Consistency in empirical experimental investigation is of prime importance to science integrity. If one were to conclude from such an experiment that the earth is rotating, but also imagines that a mechanism is modifying the results from their own favored ideal to create the inconsistent results seen, the conclusion is fallacious. That imagined mechanism which modified the results could equally be creating those results. One sees that consistency is required for any valid test of a phenomena. In sciences which are not desperate to prove something, experiments with inconsistent results are often rejected altogether for that very reason.