ID# 2183:
"On the Anthropometric Laboratory at the late International Health Exhibition," by Francis Galton
Date:
1885
Pages: (1|2|3|4|5|6|7|8|9|10|11|12|13|14|15|16|17|18|19|20|21|22|23|24|25|26|27|28|29|30)
Source:
University College London, GP, 182

&quote;On the Anthropometric Laboratory at the late International Health Exhibition,&quote; by Francis Galton

at the late International Health Exhibition. 21 strikingly regular curve. (I shall speak further on about this one partial exception.) Per-centiles were then drawn to the curve, corresponding to abscissae that were respectively 5 per cent., 10 per cent., 20 per cent., &c., of the length of the base line. As the length of the base line was 275, these per-centiles stood at the graduations of 13.8[degrees symbol], 27.5[degrees symbol], 55.0[degrees symbol], &c. Their values, as read off on the sectional paper, are those which I have given in the Table. It will be understood after a little reflection that the 9th rank in a row of 10, the 90th rank in a row of 100, and the 900th rank in a row of 1000, are not identical, and that none of them are identical with the 90th per-centile. There must always be the difference of one half-place between the post which each person occupies in a row of [italics]n[end italics] individuals, numbered from 1 to [italics]n[end italics], and that of the corresponding graduation of the base on which he stands, and which bears the same nominal value, because the graduations are numbered from 0 to [italics]n[end italics] and begin at a point one half-place short of the first man, and end at one half-place beyond the last man. Consequently the graduations corresponding to the posts of the 9th, 90th, and 900th man in the above example, refer to the distance of those posts from the beginning at 0 of their several base lines, and those distances are related to the lengths of the base lines in the proportions of 8.5 : 10, 89.5 : 100, and of 899.5 : 1000, which when reckoned in pre-cents. of the several base lines are 85, 89.5, and 89.95 respectively. The larger the number of places in the series, the more insignificant does this half-place become. Moreover, the intrusion of each fresh observation into the series separates its neighbours by almost double that amount, and propogates a disturbance that reaches to either end, though it is diminished to almost nothing by the time it has arrived there. We may therefore ignore the existence of this theoretically troublesome half-place in our ordinary statistical work. There is a latent source of error that might affect such statistics as these, as well as many others that are drawn up in the usual way, which has not, so far as I know, been recognised, and which deserves attention. It is due to uncertainty as to the precise meaning of such headings as 30-, 31-, &c. If the measurements, no matter whether they were made carefully or carelessly, are read off from the instruments with great nicety, then a reading such as 30.99 would fall in the column 30-, and the mean of all the entries in such a column might fairly be referred to a mean value of 3.50. But if the instruments are roughly read, say to the nearest half inch, the reading of a real instrumental value of 30.99, and even that of a real value of 30.76, would both be entered in the column 31-. The column 30- would then contain measurements whose real instrumental values ranged between 29.75 and 30.75, and column 31- would contain those that ranged between 30.75 and 31.75; consequently, the means of all the entries in those columns [end]

Copyright 1999-2004: Cold Spring Harbor Laboratory; American Philosophical Society; Truman State University; Rockefeller Archive Center/Rockefeller University; University of Albany, State University of New York; National Park Service, Statue of Liberty National Monument; University College, London; International Center of Photography; Archiv zur Geschichte der Max-Planck-Gesellschaft, Berlin-Dahlem; and Special Collections, University of Tennessee, Knoxville.
The images and text in this Archive are solely for educational and scholarly uses. The materials may be used in digital or print form in reports, research, and other projects that are not offered for sale. Materials in this archive may not be used in digital or print form by organizations or commercial concerns, except with express permission.