Editing Consilience (section)

==Description==
Consilience requires the use of independent methods of measurement, meaning that the methods have few shared characteristics. That is, the mechanism by which the measurement is made is different; each method is dependent on an unrelated natural phenomenon. For example, the accuracy of laser range-finding measurements is based on the scientific understanding of [[laser]]s, while satellite pictures and [[metre-stick]]s (or yardsticks) rely on different phenomena. Because the methods are independent, when one of several methods is in error, it is very unlikely to be in error in the ''same way'' as any of the other methods, and a difference between the measurements will be observed.<ref group=note>Note that this is not the same as performing the same measurement several times. While repetition does provide evidence because it shows that the measurement is being performed consistently, it would not be consilience and would be more vulnerable to error.</ref> If the scientific understanding of the properties of lasers was inaccurate, then the laser measurement would be inaccurate but the others would not.

As a result, when several different methods agree, this is strong evidence that ''none'' of the methods are in error and the conclusion is correct. This is because of a greatly reduced likelihood of errors: for a consensus estimate from multiple measurements to be wrong, the errors would have to be similar for all samples and all methods of measurement, which is extremely unlikely. [[Random errors]] will tend to cancel out as more measurements are made, due to [[regression to the mean]]; [[systematic errors]] will be detected by differences between the measurements and will also tend to cancel out since the direction of the error will still be random. This is how [[Scientific theory|scientific theories]] reach high confidence—over time, they build up a large degree of evidence which converges on the same conclusion.<ref group=note>Statistically, if three different tests are each 90% reliable when they give a positive result, a positive result from all three tests would be 99.9% reliable; five such tests would be 99.999% reliable, and so forth. This requires the tests to be [[statistically independent]], analogous to the requirement for independence in the methods of measurement.</ref>

When results from different strong methods do appear to conflict, this is treated as a serious problem to be reconciled. For example, in the 19th century, the [[Sun]] appeared to be no more than 20 million years old, but the Earth appeared to be no less than 300&nbsp;million years (resolved by the discovery of [[nuclear fusion]] and [[radioactivity]], and the theory of [[quantum mechanics]]);<ref>{{Cite web |last=Bahcall |first=John N. |date=29 June 2000 |title=How the sun shines |url=https://www.nobelprize.org/prizes/themes/how-the-sun-shines |access-date=2023-01-23 |website=NobelPrize.org |language=en-US}}</ref> or current attempts to resolve theoretical differences between quantum mechanics and [[general relativity]].<ref name=Weinberg>
{{cite book|last=Weinberg|first= S|year=1993|title=Dreams of a Final Theory: The Scientist's Search for the Ultimate Laws of Nature|publisher=Vintage Books, New York}}</ref>