3i6 



The Journal of Heredity 



These are better than the correlation 

 for the two independent judgments 

 regarding the portraits in the Canadian 

 "Who's Who." There it was r = .29 

 approximateh'. If the same observer 

 is asked to classify the same material 

 a second time, or a third time after 

 several intervening days, these correla- 

 tions (between his own successive 

 judgments) will be high, even though 

 he has forgotten the individual faces. 



I found my own classifications cor- 

 related r = .62 between a first test and a 

 second test taken se\eral weeks after- 

 wards, and r=.60 between the second 

 tests and a third taken a few diiys 

 later. The first and second classifica- 

 tions made by Mr. A. E. Wiggam 

 gave an approximate correlation of 

 r = .75 and the second and third r = .81. 

 In time, anyone's two judgments would 

 give a correlation approaching perfec- 

 tion, or r=1.00, but that would come 

 about only as one grew acquainted 

 with the faces and remembered where 

 they were put before. 



All these correlations ha\e an im- 

 portance to the science of physiognomy, 

 not because of their accuracy (for 

 they are onl>' rough first approxima- 

 tions) but l)ecau.se the\' prove that 

 the method of visual judgments may be 

 sometimes successfully employed. 



Here the chief interest lies in the 

 fact that everyone of the ten observers 

 agreed in making the total number of 

 large or long nosed Canadian soldiers a 

 small number, compared to the sum of 

 other two groups. The figures below 

 show the opinions of the observers. 



Twenty-seven of these pictures are 

 here rej)roduced ( Kigs. H-lOj, they are 

 arranged in the order of the sizes of 

 their noses, according to the combined 

 and averaged opinitjns of the observ- 

 ers. Four pictures in the middle grade 

 have been omitted merel\- for conven- 

 ience in rejjresentation on the pages of 

 this journal, and the nine at the ex- 

 treme vn(\ of the small or short nosed 

 group have also been omitted so that 

 no one could possibK- taki- any per- 

 sonal offense at the utili/ation ol this 

 material. 



I,arj{f Short 



or Average or 



Long Small 



The totals 71-187-152 give, when 

 reckoned as percentages of the grand 

 total, 17.3—45.6—37.1. The inclu- 

 sion of the five repeated observations 

 does not alter the result even as much 

 as one per cent, and is therefore omit- 

 ted. 



The distribution from the group of 

 Montreal department store emplovees 

 was 5-20-10 or 3-30-2 with a total 

 8 — 50 — 12. This if in percents is 

 11.4—71.4—17.1, or in other words the 

 middle group is the largest, and the 

 small or short is next; just as was the 

 result from the coiubined tests on the 

 Canadian soldiers. The average of 

 these two sets of tests on "average 

 men" is 14 — 59 — ^27. 



Subse(|uentl\- some tests were made 

 on the photographs in Harvard "Class 



