March 



1910] 



NATURE 



127 



NUMERALISED PROFILES FOR CLASSIFICA- 

 TION AND RECOGNITION. 



\\JHE'S children or savages attempt to draw a 

 » » human profile, the result is usually a rude 

 figure that lays stress on five cardinal points. These 

 are the notch between the brow and the nose, the 

 tip of the nose, the notch between the nose and the 

 upper lip, the parting of the lips, and the tip of the 

 chin. Supposing these five points, B', N', U', U, and 

 C', to be located with fair precision, as will shortly 

 be shown to be feasible, then Fig. i is directly de- 

 ducible from them, together with the vertical and hori- 

 zontal axes, C'B', and C'X at right angles to C'B'. 

 The position of the five cardinal points varies in 

 different profiles much more than the probable error 

 of measurement. So though Fig. i is a mere skeleton, 

 which determines what may be called the set of the 

 features, and corresponds to the primary triangulation 

 of a country, otheeii^ints are to be derived from it, 

 and similarly utilised. Among these are the intersec- 

 tions with the outline by perpendiculars, drawn from 

 the middle or other specified division of the lines. 

 This skeleton serves as an excellent basis for the classi- 

 fication of profiles and for anthropological statistics. 



Peculiarities of profile, as a racial or family char- 

 acteristic, can be expressed numerically by an exten- 

 sion of this system in a way that promises to be 

 serviceable for eugenic records. It was, in fact, 

 largely with this object in view that I began the 



FIG.Z 



FIG. 3 



inquiry. The replacement in all scientific work by 

 numerical values, in the place of vague adjectives, 

 is a gain of first-class importance. There is no wa\' 

 known to me, other than this, by which likenesses 

 can be " lexiconised," that is, arranged as words in 

 a dictionary. A needed portrait may by its means be 

 discovered by a formula, as a spoken word is found 

 in a dictionary, by the letters that express its 

 sound. There are many simple purposes of ne\ys- 

 paper interest to which this same method might be 

 applied, but with more elaboration. 



The practice of cataloguing profiles may perhaps 

 become useful as a secondary means of identification 

 when the number of persons who may require to be 

 identified shall have become too large to be readily 

 dealt with by finget -prints alone. 



It will be shown (Fig. 5) that four telegraphic 

 "words" are sufficient to convey a very fair pro- 

 file likeness. The cost of sending an extra four words 

 by telegram to any part of the British Isles being only 

 twopence, and of a moderate amount over-seas, the 

 practice of telegraphing profiles of persons of current 

 interest, might become common. A refugee criminal 

 could easily be outstripped by his portrait, sufficiently 

 like to him to justify, in cpnnection with corrobora- 

 tive evidence, his being placed for a while under police 

 observation. The measures of profiles must, of 

 course, be reduced to uniformity. Thus, by utilising 

 ' two out of the five cardinal points to give direction 

 ! and scale, the mean positions of the remaining three 



NO. 2109, VOL. 83] 



points may be determined for any given race or 

 family, together with the frequency of deviations of 

 any given amount from those mean positions, and 

 such other deductions as can, be reached by the modern 

 methods of statistics. 



The corrected values are here described by the same 

 letters as the original ones, but without the dashes. 

 The standard scale that is used is such that BC, the 

 corrected value of B'C, shall be always 50 units in 

 length (see Fig. 5). The reduction is, of course, 

 effected by multiplying each measure in the portrait by 

 50 divided by its B'C. The number 50 is preferable to 

 100, which would probably first suggest itself, for a 

 variety of practical reasons, into which I need not now 

 enter. Two figures are assigned to each measure, so 

 the values o, i, 2, -.9, have to be written 00, 01, 02, 

 .09. The measures are recorded to the nearest 

 integer, there being no room for fractions, decimal 

 or other. A millimetre is a convenient unit for pur- 

 poses of drawing, more so than one-tenth of an inch ; 

 therefore, in reproducing the corrected measures, BC 

 becomes 50 millimetres, and the other measures are 

 altered in the same proportion. 



A thick beard interferes with determining L' and 

 U', but their positions can usually be inferred with a 

 useful degree of precision in moderately bearded faces. 



The accuracy with which the five cardinal points 

 can be located differs considerably. The most exact 

 determinations in an unbearded face are those of the 

 points C and N', and the direction of the line C'B'. 

 U' comes next in order of exactness, then B', and, 

 lastlv, L'. The distance between a line joining 

 C'X' and a parallel line tangential to U', can be fixed 

 with precision but is not used here. C and N' are 

 each defined by the intersection of two tangents, as 

 shown for X' in Fig. 3. 



It is well to examine these conditions more closely, as 

 they bear on the treatment of curvatures generally. A 

 knowledge of them permits rough and ready drawing, 

 in which the principal matters are attended to, the less 

 essential ones being more or less disregarded. One 

 of the tangents is parallel to C'B', which is treated as 

 vertical ; the other is inclined to the vertical at 45°. Con- 

 sequently, the curve of N' is contained in an obtuse angle 

 of i8o° — 45°= 135°. The tip of each prominence and the 

 bottom of each hollow is represented by one or other of 

 the three short circular arcs shown in Fig. 2, which are 

 suflRclentlv numerous for the purposes to which they are 

 here applied. The centres of all circles that touch both 

 the vertical and the diagonal will necessarily lie in the 

 line that bisects the obtuse angle between them : con- 

 sequently, N'O forms an angle with the vertical that is 

 equal to half 135", or 67^°. The tangent of this angle is 

 2-4142; therefore the position of the line of centres may 

 be found by laying off a point V in a vertical direction, at 

 10 units of length from O, and by drawing another line 

 from V horizontally to W, at a distance from it equal to 

 24- 14 of the same units as before. Then the line of centres 

 passes through O and W. It is easily shown (Fig. 3) that 

 the points of contact between the circle and the two 

 tangents are exactly 45° of arc apart. The length of the 

 chord of that angle is equal to about three-quarters of 

 its radius. The shortness of the chord, when the radius 

 is small, is well seen in Fig. 2, and must be borne in 

 mind ; it accounts for the scarcely noticeable differences 

 in the cur\'atures, and consequently fof the fewness of the 

 standard arcs that are necessary. The arc of 45° is shown 

 bv a heavy line in Fig. 3, where the circle has a radius 

 of 10 mm. There is often a colloquial confusion between 

 the obliquity of the planes between which an edge lies 

 and that of the sides of the edge itself. The former may 

 be very acute, and the angle of the edge would be equally 

 acute if the planes were prolonged until they met ; but 

 usually they do not meet, the edge itself being more or 

 less rounded. The acutely inclined faces of a knife mav 

 have a blunted edge, that fails to cut the skin without 

 much pressure, while a broken piece of glass, the fracture 



