Later, after radioactivity had been proven to be a significant source of the Earth's internal heat, he did privately admit that he might have been in error.
What is especially telling about this whole story is the conclusion of the absolute truth of the conclusion based on premises that are weak, or at least not adequately demonstrated.
Of course there seem to me to be fairly reasonable explanations for this observation which may allow for more slowly forming granitic rocks.
Potassium - Argon and Argon - Argon dating are based on the current understanding that radioactive Potassium-40 decays to the stable form, Argon-40 with a half-life of approximately 1.25 billion years.
There are some circumstances that can affect this rate such as magnetic fluctuations etc...
Based on these assumptions he at first suggested an age of the Earth of between 100 Ma and 500 Ma.
This estimate was actually reduced over his lifetime to between 20 Ma and 40 Ma and eventually to less than 10 Ma. Perry, in particular, a noted physicists and former assistant to Kelvin, showed that cooling calculations using different but equally likely assumptions and data resulted in ages for the Earth of as much as 29 Ga.
Chamberlain (1899) pointed out that Kelvin's calculations were only as good as the assumptions on which they were based.
"The fascinating impressiveness of rigorous mathematical analyses, with its atmosphere of precision and elegance, should not blind us to the defects of the premises that condition the whole process.He believed this even though he did admit that some heat might be generated by the tidal forces or by chemical action.However, on the whole, he thought that these sources were not adequate to account for anything more than a small faction of the heat lost by the Earth.Other factors and basic assumptions must also be considered.Of course, Kelvin formed his estimates of the age of the Sun without the knowledge of fusion as the true energy source of the Sun.Of course, the detected variation is no more than 0.2% of the published rates, but this paper is still quite interesting since such a correlation was never suspected before.