The scale on which the service is carried on, is very large. It was begun in 1883, and by the end of 1887 no less than 60,000 sets of measures were in hand, but thus far only about one half of the persons arrested in Paris were measured, owing to the insufficiency of the staff. Arrangements were then made for its further extension. There are from 100 to 150 prisoners sentenced each day by the Courts of Law in Paris to more than a few days’ imprisonment, and every one of these is sent to the Dépôt for twenty-four hours. While there, they are now submitted to Bertillonage, a newly coined word that has already come into use. This is done in the forenoon, by three operators and three clerks; six officials in all. About half of the prisoners are old offenders, of whom a considerable proportion give their names correctly, as is rapidly verified by an alphabetically arranged catalogue of cards, each of which contains front and profile photographs, and measurements. The remainder are examined strictly; their bodily marks are recorded according to a terse system of a few letters, and they are variously measured. Each person occupies seven or eight minutes. They are then photographed. From sixty to seventy-five prisoners go through this complete process every forenoon. In the afternoon the officials are engaged in making numerous copies of each set of records, one of which is sent to Lyon, and another to Marseille, where there are similar establishments. They also classify the copies of records that are received from those towns and elsewhere in France, of which from seventy to one hundred arrive daily. Lastly, they search the Registers for duplicate sets of measures of those, whether in Paris or in the provinces, who were suspected of having given false names. The entire staff consists of ten persons. It is difficult to rightly interpret the figures given in the pamphlet (4) at pp. 22-24, as they appear to disagree, but as I understand them, 562 prisoners who gave false names in the year 1890 were recognised by Bertillonage, and only four other persons were otherwise discovered to have been convicted previously, who had escaped recognition by its means.

I had the pleasure of seeing the system in operation in Paris a few years ago, and was greatly impressed by the deftness of the measuring, and with the swiftness and success with which the assistants searched for the cards containing entries similar to the measures of the prisoner then under examination.

It is stated in the Signalements (p. 12) that the basis of the classification are the four measurements (1) Head-length, (2) Head-breadth, (3) Middle-finger-length, (4) Foot-length, their constancy during adult life nearly always [as stated] holding good. Each of these four elements severally is considered as belonging to one or other of three equally numerous classes—small, medium, and large; consequently there are 34 or 81 principal headings, under some one of which the card of each prisoner is in the first instance sorted. Each of these primary headings is successively subdivided, on the same general principle of a three-fold classification, according to other measures that are more or less subject to uncertainties, namely, the height, the span, the cubit, the length and breadth of the ear, and the height of the bust. The eye-colour alone is subjected to seven divisions. The general result is (pp. 19, 22) that a total of twelve measures are employed, of which eleven are classed on the three-fold principle, and one on the seven-fold, giving a final result of 311 × 7, or more than a million possible combinations. M. Bertillon considers it by no means necessary to stop here, but in his chapter (p. 22) on the “Infinite Extension of the Classification,” claims that the method may be indefinitely extended.

The success of the system is considered by many experts to be fully proved, notwithstanding many apparent objections, one of which is the difficulty due to transitional cases: a belief in its success has certainly obtained a firm hold upon the popular imagination in France. Its general acceptance elsewhere seems to have been delayed in part by a theoretical error in the published calculations of its efficiency: the measures of the limbs which are undoubtedly correlated being treated as independent, and in part by the absence of a sufficiently detailed account of the practical difficulties experienced in its employment. Thus in the Application pratique, p. 9: “We are embarrassed what to choose, the number of human measures which vary independently of each other being considerable.” In the Signalements, p. 19: “It has been shown” (by assuming this independent variability) “that by seven measurements, 60,000 photographs can be separated into batches of less than ten in each.” (By the way, even on that assumption, the result is somewhat exaggerated, the figures having been arrived at by successively taking the higher of the two nearest round values.) In short, the general tone of these two memoirs is one of enthusiastic belief in the method, based almost wholly, so far as is there shown, on questionable theoretic grounds of efficiency.

To learn how far correlation interferes with the regularity of distribution, causing more entries to be made under some index-heads than others, as was the case with finger prints, I have classified on the Bertillon system, 500 sets of measures taken at my laboratory. It was not practicable to take more than three of the four primary measures, namely, the head-length, its breadth, and the middle-finger-length. The other measure, that of foot-length, is not made at my laboratory, as it would require the shoes to be taken off, which is inconvenient since persons of all ranks and both sexes are measured there; but this matters little for the purpose immediately in view. It should, however, be noted that the head-length and head-breadth have especial importance, being only slightly correlated, either together or with any other dimension of the body. Many a small man has a head that is large in one or both directions, while a small man rarely has a large foot, finger, or cubit, and conversely with respect to large men.

The following set of five measures of each of the 500 persons were then tabulated: (1) head-length; (2) head-breadth; (3) span; (4) body-height, that is the height of the top of the head from the seat on which the person sits; (5) middle-finger-length. The measurements were to the nearest tenth of an inch, but in cases of doubt, half-tenths were recorded in (1), (2), and (5). With this moderate minuteness of measurement, it was impossible so to divide the measures as to give better results than the following, which show that the numbers in the three classes are not as equal as desirable. But they nevertheless enable us to arrive at an approximate idea of the irregular character of the distribution.

Table XVI.

Dimensions
measured.		Medium
measures in
inches and
tenths.		Nos. in the three classes respectively.
-
below.		0
medium.		+
above.		Total.
1. Head-length7·5to7·7101191208500
2. Head-breadth6·0"6·1173201126500
3. Span68·0"70·5137165198500
4. Body-height35·0"36·0139168193500
5. Middle-finger4·5"4·6180176144500

The distribution of the measures is shown in Table XVII.