Cude and other skeptics have recently brought up the undiscovered error hypothesis. Here is a response to that by Melich and Rothwell:
Undiscovered error hypothesis Some skeptics claim that there might be a yet-undiscovered error in the experiments. [As Beaudette wrote] “if the measurements are incorrect, then an avid pursuit of the ‘science’ must in due course explicitly and particularly reveal that incorrectness.” More to the point, the claim that there might be an undiscovered error is not falsifiable, and it applies to every experiment ever performed. There might be an undiscovered error in experiments confirming Newton’s or Boyle’s laws, but these experiments have been done so many times that the likelihood they are wrong is vanishingly small. Furthermore, skeptics have had 20 years to expose an experimental artifact, but they have failed to do so. A reasonable time limit to find errors must be set, or results from decades or centuries ago will remain in limbo, forever disputed, and progress will ground to a halt. The calorimeters used by cold fusion researchers were developed in the late 18th and early 19th century. A skeptic who asserts that scientists cannot measure multiple watts of heat with confidence is, in effect, rejecting most textbook chemistry and physics from the last 130 years. As a practical matter, there is no possibility that techniques such as calorimetry, x-ray film autoradiography or mass spectroscopy are fundamentally flawed. It must be emphasized that although cold fusion results are surprising, the techniques are conventional and instruments are used within their design specifications. Cold fusion does not require heroic measurement techniques. Heat and tritium are not usually measured close to the limits of detection, although they have been in some cases, and helium and transmutations have been. It has been argued that even though the instruments work, the researchers may be making mistakes and using the instrument incorrectly. No doubt some of them are, but most are experienced scientists at major labs. The effect has been confirmed at 180 major laboratories [Storms, Table 1]. If an experiment could be as widely replicated as this could be mistaken, the experimental method itself would not work.