A letter of Stokes to Sir John Lubbock, printed in the _Scientific Correspondence of Sir George Gabriel Stokes_, states his recollection of the matter, and gives Thomson the credit of having inferred the method of spectrum analysis, a method to which Stokes himself makes no claim. He says, "I know, I think, what Sir William Thomson was alluding to. I knew well, what was generally known, and is mentioned by Herschel in his treatise on Light, that the bright D seen in flames is specially produced when a salt of soda is introduced. I connected it in my own mind with the presence of sodium, and I suppose others did so too. The coincidence in position of the bright and dark D is too striking to allow us to regard it as fortuitous. In conversation with Thomson I explained the connection of the dark and bright line by the analogy of a set of piano strings tuned to the same note, which, if struck, would give out that note, and also would be ready to sound it, to take it up, in fact, if it were sounded in air. This would imply absorption of the aërial vibrations, as otherwise there would be a creation of energy. Accordingly I accounted for the presence of the dark D in the solar spectrum by supposing that there was sodium in the atmosphere, capable of absorbing light of that particular refrangibility. He asked me if there were any other instances of such coincidences of bright and dark lines, and I said I thought there was one mentioned by Brewster. He was much struck with this, and jumped to the conclusion that to find out what substances were in the stars we must compare the positions of the dark lines seen in their spectra with the spectra of metals, etc.... "I should have said that I thought Thomson was going too fast ahead, for my notion at the time was that, though a few of the dark lines might be traced to elementary substances, sodium for one, probably potassium for another, yet the great bulk of them were probably due to compound vapours, which, like peroxide of nitrogen and some other known compound gases, have the character of selective absorption." It will be remembered that the experimental establishment of the method of spectrum analysis was published towards the end of 1859 by Bunsen and Kirchhoff, to whom, therefore, the full credit of discoverers must be given.