I 



June 21, 1888] 



NATURE 



175 



amely, the head length, head breadth, foot length, and 

 iddle-finger length of the left foot and hand respectively, 

 ach of these is classified according as it is large, 

 medium, or small. There are thus three, and only three, 

 divisions of head lengths, each of which is subdivided into 

 three divisions of head breadth ; again, each of these is 

 further subdivided into three of foot length, and these 

 again into three of middle-finger length ; thus the num- 

 ber of primary classes is equal to three multiplied into 

 itself four times — that is to say, their number is eighty- 

 one, and a separate pigeon-hole is assigned to each. All 

 the exact measures and other notes on each criminal are 

 written on the same card, and this card is stored in its 

 appropriate pigeon-hole. The contents of each pigeon- 

 hole are themselves sub-sorted on the same principle of 

 three-fold classification in respect to other measures. 

 This process can, of course, be extended indefinitely, but 

 how far it admits of being carried on advantageously is 

 another question. The fault of all hard-and-fast lines of 

 classification, when variability is continuous, is the doubt 

 where to find values that are near the limits between two 

 adjacent classes. Let us take the case of stature, for 



C-l C C-2 



e-i E e-2 



B 



Fig. 4. 



illustration of what must occur in every case, represent- 

 ing its distribution by what I have called a " scheme," as 

 shown in Fig. 4. 



Here the statures of any large group of persons are 

 represented by lines of proportionate length. The lines 

 are arranged side by side at equal distances apart on a 

 base, A B, of convenient length. A curve drawn through 

 their tops gives the upper boundary of the scheme ; the 

 lines themselves are then wiped out, having served their 

 purpose. If the base A B be divided into three equal 

 parts, and perpendiculars, c D, E f, be erected at the 

 divisions between them, reaching from the base up to the 

 curve, then the lengths of those perpendiculars are pro- 

 portionate to the limiting values between the small and 

 the medium group, and the medium and the large group, 

 respectively. The difference between these perpendiculars 

 in the case of stature is about 2*3 inches. In other 

 words, the shortest and tallest men in the medium class 

 differonly by that amount. We have next to consider how 

 much ought reasonably to be allowed for error of measure- 

 ment. Considering that a man differs in height by a full 

 third of an inch between the time. of getting up in the morn- 

 ing and lying down at night ; considering also that 



measures are recorded to the nearest tenth of an inch 

 at the closest, also the many uncertainties connected 

 with the measurement of stature, it would be rash not to 

 allow for a possible error of at least ± half an inch. 

 Prolong C D, and note the points upon it at the distance 

 of half an inch above and below D ; draw horizontal lines 

 from those points to meet the curve at d.i, d.z, and from 

 the points of intersection drop perpendiculars reaching 

 the base at c.\,c.i. A similar figure is drawn at F. Then 

 the ratio borne by the uncertain entries to the whole num- 

 ber of entries is as c x c 2 + e x c. to ab. This, as seen by the 

 diagram, is a very serious proportion. There is a dilemma 

 which those who adopt hard and-fast lines of classifica- 

 tion cannot avoid : either the fringe of uncertainty is 

 dangerously wide, or else the delicacy with which 

 measures are made is not turned to anything like its full 

 account. If the delicacy is small, the fringe of uncer- 

 tainty must be very wide ; if the delicacy is great, the 

 fringe will be narrow ; but then the other advantages of 

 possessing delicate observations are wasted through em- 

 ploying only a few classes. The bodily measurements 

 are so dependent on one another that we cannot afford 

 to neglect small distinctions. Thus long feet and long 

 middle-fingers usually go together. We therefore want 

 to know whether the long feet in some particular person 

 are accompanied by particularly long, or moderately long, 

 or relatively short fingers, though the fingers may in the 

 two last cases be long as compared with those of the 

 general population, and will be treated as long in M. 

 Bertillon's system of classes. Certainly his eighty- one 

 combinations are far from being equally probable. The 

 more numerous the measures the greater would be their 

 interdependence, and the more unequal would be the 

 distribution of cases among the various possible combina- 

 tions of large, small, and medium values. No attempt 

 has yet been made to estimate the degree of their inter- 

 dependence. I am therefore having the above measure- 

 ments (with slight necessary variation) recorded at my 

 anthropometric laboratory for the purpose of doing so. 

 This laboraiory, I may add, is now open to public use 

 under reasonable restrictions. It is entered from the 

 Science Collections in the Western Galleries at South 

 Kensington. 



Mechanical Selector. — Feeling the advantage of possess- 

 ing a method of classification that did not proceed upon 

 hard-and-fast lines, I contrived an apparatus that is quite 

 independent of them, and which I call a mechanical 

 selector. Its object is to find which set out of a standard 

 collection of many sets of measures, resembles any one 

 given set within specified degrees of unlikeness. No one 

 measure in any of the sets selected by the instrument can 

 differ from the corresponding measure in the given set, by 

 more than a specified value. The apparatus is very 

 simple, it applies to sets of measures of every description, 

 and ought to act on a large scale with great rapidity, and 

 as well as it does on a small one, testing several hundred 

 sets by each movement. It relieves the eye and brain 

 from the intolerable strain of tediously comparing a set 

 of many measures with each successive set of a large 

 series, in doing which a mental allowance has to be made 

 for a plus or minus deviation of a specified amount in 

 every entry. It is not my business to look after prisoners, 

 and I do not fully know what need may really exist for new 

 methods of quickly identifying suspected persons. If 

 there be any real need, I should think that this apparatus, 

 which is contrived for other purposes, might, after 

 obvious modifications, supply it. 



The apparatus consists of a large number of strips of 

 card or metal, c\, ci (Fig. 5), say 8 or 9 inches long, and 

 having a common axis, A, passing through all their smaller 

 ends. A tilting-frame, T, which turns on the same axis, has a 

 front cross-bar, F, on which the tips of the larger ends of all 

 the cards rest whenever the machine is left alone. In 

 this condition a counterpoise at the other end of T suffices 



