Photographic Researches and Portraiture 331
ferences of keenness of vision had to be pooled. Galton considers that mutual mistakability may occur under any one or more of the following conditions, which he thinks should be noted alongside the index
`°aa. The portraits are apparently exact copies or reductions on different scales.
a. They appear to be portraits of the same person at about the same age, though differing in pose and dress.
b. They would be mistaken for portraits of the same person, even though they differ in sex and considerably in age, if the hair had been cut and dyed alike, and the dress arranged in the same way.
c. As above, if much disguised, as for theatrical impersonations.
b-c. Applies to cases intermediate between b and c.
p. Their resemblance is partial only, being confined to specified features.
The applications of the process are numerous, as must always be the case when a hitherto vague perception is brought within the grip of numerical precision. To myself it has the special interest of enabling the departure of individual features from a standard type to be expressed numerically. The departure may be from a composite of their race, or from a particular individual. The shortcomings of a pedigree animal from a highly distinguished ancestor could be measured in this way. Many other examples might be given." (p. 563.)
As in his profile work Galton used a very large number of pairs of photographs of relatives to test his index of mistakability upon. He asked in the newspapers for photographs of families, and they appear to have been rained down upon him; some material was suitable, some quite unsuitable ! It seems to me that to get reliable measures of resemblance special photographs should be taken-full face and profile, the hair being screened under a tight fitting elastic cap. Further if bearded individuals are to be one of the "comparates," then the comparison must further be made with the chin and lips screened; the eye is very apt to be misled in its judgments by extraneous characters such as hair and pose.
A manuscript typed and prepared for press in February 1906, entitled "The Measurement of Visual Resemblance," seems never to have been published. It adopts a somewhat different index to that finally chosen by Galton in October of the same year. He begins by saying that visual resemblance between any two objects may be measured in units whose value is strictly defined.
"Resemblance is independent of actual magnitude and has therefore to be expressed in angular units. It is curious that no popular terms exist to express them in the language of any civilised country, for not only would they be useful, but the diameter of the sun when paled by an intervening screen affords an excellent and practically constant standard for rough measurement. It would often be well to indicate objects in a distant landscape by describing them as so many sun-breadths to the right or left of some conspicuous feature, or to speak of a mountain seen from a specified place as towering so many sun-breadths in height, or as bulking so many [square] sun-breadths in area. But as sun-breadths are not terms in popular use, and as they are not the best unit for the purpose of this memoir, I will employ another that is. The sun's diameter may be taken as subtending an angle of 31.0 minutes of a degree. I will employ for my unit the diameter of an imaginary mock sun that subtends 34.4 minutes, and is therefore wider than that of the real sun in the proportion of 10 to 9. Its merit lies in the fact that the tangent as also the arc of 34.4 minutes differ insensibly from 0.01 ; in other words the angular unit is that which is subtended by 1 measure of any kind, at the distance of 100 measures of the same kind. I will call the arc subtended by this angle. at any specified distance a ' sol'."