AUGUST 23, 1901.] 



SCIENCE. 



291 



which kingdoms — not with any one city — 

 London has to be compared. 



What is wanted in the faculty of engi- 

 neering is, therefore : 



1. Increase in staff of professors and in- 

 structors at existing centers — say £10,000 

 a year (£330,000). 



2. Extensions at existing centers in build- 

 ings and equipment to accommodate addi- 

 tional students — say £150,000. 



3. New centers — building, equipment, 

 and endowment of engineering departments 

 at— say seven at £50,000 (£350,000). 



4. ISTew subjects — provision for buildings, 

 equipment and endowment of centers for 

 marine engineering and naval architecture 

 (£100,000); civil and municipal engineer- 

 ing (£100,000) ; railway, dock and hydraulic 

 engineering, etc. (£100,000) ; electric trac- 

 tion (£100,000) ; architecture (£100,000), 

 etc. We may say, therefore, that the faculty 

 of engineering needs a capital sum of £1,- 

 330,000. 



SCIENTIFIC BOOKS. 

 Bihliotics, or the Study of Documents. By Per- 

 SIFOR Frazer. Third edition. Philadelphia, 

 J. B. Lippincott Co. 1901. Pp. xxiv+ 266. 

 The subject-matter of this book, not very 

 clearly suggested by the title, is the methods 

 used by the handwriting expert and by the 

 chemist, in the identification of writing and the 

 detection of forgery. A scientific man desirous 

 of getting some insight into these methods — 

 or into the best of them — will find this book 

 good reading. It is written with a scientific 

 bent. It considers handwriting much as a zo- 

 ologist considers animals. The determination 

 of the characteristics of a given handwriting 

 is like the determination of the characteristics 

 of a natural species. The older methods relied 

 on general impressions or on the description of 

 salient features ; the newer methods rely on 

 measurement of the details. A person's hand- 

 writing, like a natural species, is a variable 

 thing, and the exact study of it must deal in 

 averages and ranges of variation. " It will be 

 readily conceded that at least two factors ai'e 



present in the performance of an act which is 

 often repeated. One is the general similarity, 

 and the other is the variation in some details 

 which prevents any two acts or results from ever 

 being exactly identical. These are analogous to 

 the two important factors of the theory of evo- 

 lution, called, by Charles Darwin, hereditary 

 transmission and accidental variation. In order 

 to arrive at an ideal standard of similar recur- 

 rent actions, it is necessary to eliminate, as far 

 as possible, the accidental variations. The most 

 obvious way to do this is to take the average or 

 mean of the records of a number of such ac- 

 tions." The first method devised by the author 

 for arriving at the mean was the graphic 

 method of composite photography. This ap- 

 peals to the eye, and shows directly which pai'ts 

 of a signature are most uniform and which are 

 most variable. A disputed signature may bet- 

 ter be compared with a composite of several 

 genuine signatures, than with any one of them. 



The more exact method of averages, also de- 

 vised by the author, begins with actual measure 

 ment of details ; and it is the inconspicuous 

 details that are most characteristic. The salient 

 features can be changed more or less at the 

 writer's will, or imitated by another person. 

 But the little tricks of curvature and slant and 

 proportions, the minutiae of shading and align- 

 ment, as they are the expression of unconscious 

 habits, so they cannot voluntarily be laid aside, 

 and as they are undetected by the eye, so 

 they cannot be reproduced by a forger. The 

 most useful details for measurement are angles 

 and the ratios of different lengths ; these are 

 very inconspicious, yet fairly constant, not 

 changing with the size of the letters. 



If, then, a signature is disputed, several genu- 

 ine signatures are obtained, and a certain num- 

 ber of details are measured in all the specimens ; 

 the average measurements of the genuine signa- 

 ture are computed, and the measurements of 

 the disputed signature compared with them. 

 Close agreement throughout stamps the disputed 

 signature as genuine ; wide divergence as spu- 

 rious. Some divergence is of course to be ex- 

 pected, and in fact complete identity is evidence 

 of tracing. But just how much difference can 

 be allowed? How sure is the expert of his 

 decision ? It does not appear that the theory 



