Slatixtical Inve4stujutiuiut 375 



and the idea of her tickled the fancy of Mr Punch, who in the issue that 

 followed the publication of Galton's paper thus ajKwtrojihises her: 



The Squeeaa of »&. 

 Maidfii of tlif ini),'lity ihuhoIoh, HuhImdcIh in' it sadly Hlut<-il, 



Tliert' ix'oordi'd, you would be Havu Ix^oii known thrir wivi« to whack, 



Fniiious in nil niimly tusHles, You, unless you're ovpr-nit^-d. 



And its very clt-nr to n>e, Could give such endearments back. 



That if in tli(< dim hereafter Yours the task to try correction, 



Any husband should play tricks, Till your husband and your "chicks," 



You would with derisive laughter, Had a lively recollection 



Give a "S«iueeze of 8(5." Of your "Squeeze of 86." 



Puneh, April 16, 1884. 



(Jalton's second table exhibits the full advantages of his method of per- 

 centiles, but presents also its disitdvuntages, as it does not allow us to deduce 

 the usual t'reijuency constants with any reasonable accuracy. At the same 

 time it is easily understood by the non-statistical, who can readily determine 

 from it their rank for any character. Tims suppose a man 'oetween 23 and 

 26 years of age to have a vital or breathing capacity of 190 cubic inches, he 

 I sees at once from the table that he is surpassed by 70 "/, of men (199) and 

 himself surpiusses 20°/, (187). A simple rule of three sum then indicates 

 to him that he lies ^ or \ of the way from 80 to 7^ or at 77*5. In other 

 words he would rank between the 77th and 78th men in a population of 100 

 young adults similar to those who visited the anthropometric laboratory. 



The ages and numbers in the table' on the following page possibly require 

 some justification. All Galton says is that he had 



"groups of appropriate cases cxtractwl from the duplicate records by Mr J. Henry Young of 

 the General Registry Office. I did not care to have the records exhausted, but requested him 

 to take as many as seemed in each case to be sufficient to give a trustworthy result for these 

 and certain other purposes to which I desired to apply them. The precise number was deter- 

 mined by accidentfil matters of detail that in no way implied selection of the measurements." 

 (p. 278.) 



Now-a-days we should consider it needful to keep the probable errors in 

 view, which are occasionally somewhat large for small series treated by the 

 method of percentiles. Galton deduced his percentiles from the fretjuency 

 distributions by summing, plotting, drawing a curved line through the plotted 

 points and then interpolating for the actual percentiles by graphiail inter- 

 polation^ 



"the (Mipulation of England hardly contains enough material to form even a few regiments of 

 efficient Amazons." 



' This table with the description of its method of preparation was also published in Nature, 

 January 8, 188;) (Vol. .xxxi, pp. 223-5), a month or two before its appearance in the Journal 

 of the A iithroix^oyical Iiistitule. 



' The method with more detail was discussed by Galton under the title "The Application 

 of a Graphic Methotl to Fallible Measures" in a f>aper communicated to the Jubilee Congress 

 of the Royal Statistical Society and published in the JubUee Voltiine of the Slatuticnl iSocieti/, 

 1885, pp. 262-5. The method consists in dmwing an ogive curve from the data and interpo- 

 lating for the ptMventiles. The siimo end could l>e reached by integrating the frequency 

 histogram and dividing its final onlinatc into equal parts corresponding to the percentiles. This 

 may, of course, be rapidly done by the integraph. 



