S»2 



Life and Tjetters of Francis Galton 



"It ia now the pArt of thoco who have to fix the Rcales of marks to determine the weight 

 to be given nwpectively to relative rank and to absolute performance in examinations of each 

 different kind of service." (p. 32; B. A. R. p. 477.) 



A point may be noted here before we consider Galton's further contri- 

 butions to the marking problem. In the paper just discussed Galton illustratt-s 

 by isogmms his method of exhibiting graphically the correlation of the rank 

 for one character with the deviate for a second associated character. He 

 undertook, however, the construction of a diagram of isograms for the case 

 in which rank of one character was correlated with rank of a second. For all 

 three cases — rank with rank, rank with variate, and variate with variate — 

 Galton constructed isograms or contour lines. There is little doubt that rank 

 with rank was the first way in which he approached correlation. In 1 875 (sec 

 our p. 1 87) Galton was dealing with the inheritance of size in sweet-pea seeds, 

 but before he obtained his data for sweet-peas, he appears to have tried what 

 he could do with much smaller seed, apparently that of cress. The correlation 

 of the seed of mother and daughter plants was dealt with by the metho<l of 

 his memoir of 1875, and that is, I think, the probable date of his first crud( 

 correlation table, which he obtained from five groupings of each size of seed ; 

 the isograms are represented by ink lines on the sheet of glass covering 

 the little compartments which contain the ranked seeds of the daughter- 

 plants. These isograms have been smudged and almost obliterated by tin 

 wear and tear of fifty years, but can still be traced. The facts (a) that he 





k.i *iJ br (C»:V. 



% \ 



iv^,V.^^; ^'^ 





..^'■11 



s^tM, s 



OfjK 



W 



v^c 



42 



+/" 











Jiiillii 



M 



Oalton's first illustration of Correlation, circa 1875. From the GalUmiana. 



uses the word averfuje when he later used median, (b) that he divides not 

 into the f>ercentile8 of his later work, but into quartile and double-quartile' 



' Oalton himwlf says one lto«lfth from the end of the range, but if the reader looks at 

 Oalton'ii table on p. 42 of his paper on "HtatiHtics by Intercoinparison" (see our p. 338), he will 

 find 82 in 1000 given for two quartile units, which Oalton takes app^oxim^tely as jij. 



