Open main menu

Page:Popular Science Monthly Volume 66.djvu/383

This page has been proofread, but needs to be validated.



To the Editor: In his quotation from Augustus De Morgan in reference to the application of the law of averages to the detection of authorship and in his remarks thereon (see the December number of this journal) Dr. Raymond Pearl is, unconsciously of course, guilty of the exact fault which he, by implication at least, attributes to others; namely, ignorance of the work of previous writers upon the same subject. As his note seems to invite others to share with him in that ignorance, it may be desirable to explain, once more, that it was this very suggestion of De Morgan's that started the investigation of more than twenty years ago; that in the presentation of! the first results to the American Association for the Advancement of Science, De Morgan was fully credited with the idea; in the publication of these results in Science (about 1885) indebtedness to De Morgan was distinctly acknowledged; also in subsequent publications and papers, the latest being an article in a recent number of this journal (August, 1904);—all of which can easily be seen and known by anybody who cares enough about the subject to look it up. In the first, and, as far as I know, the only thorough test of De Morgan's idea, made more than twenty years ago, it was found to be difficult, perhaps impossible, to discriminate among authors by means of simple 'Average word lengths,' as suggested by De Morgan. The scheme for the graphic display of the variations in the average frequency of occurrence of words of different lengths was then devised, proving to be a vastly more powerful means of revealing peculiarities of composition. This is the only feature of the work which has been claimed as original, and the results of an exhaustive application of it were published in 1901, confirming me in my confidence in the truth of the general principle stated by De Morgan, though not in the value of the specific application of it suggested by him.

I believe that the scheme of analysis by 'Characteristic Curves,' devised a quarter of a century ago, has been 'rediscovered' one or twice since; but it should never be overlooked that the germ of the thing was in a brief remark that I found in that now, but never-ought-to-be-out-of-print book, the 'Budget of Paradoxes.' The 'Memoir' from which the letter is quoted by Dr. Pearl did not appear until many years later. As I now remember it, the original suggestion was much less elaborated than in the letter, which I think I have never seen before.

I do not understand why Dr. Pearl calls this the 'Sherman Principle.'

T. C. M.
Rome, Italy,
December 17, 1904.