
APRIL 29, 1915 | 
NATURE 
247 

rial and to protect the user from carelessness or fraud. 
But it has also acted as a stimulus to manufacturers 
to standardise and improve their products. Once it 
is known how good a quality can be produced, all 
manufacturers strive to reach it, and the average 
quality is raised. The rapid development of cement 
and of the higher qualities of steel has depended on 
the determination of their superiority by accurate tests 
in the properties required for special services. 
With the introduction of mild steel, the quality of 
which ranges between wide limits, the need of 
systematic testing became urgent. Now there are 
testing laboratories in the works of most railway com- 
panies and in steel works. Also the practice of test- 
ing has been adopted in the case of many other mate- 
rials. With the extension of testing has come the 
need of standardising tests themselves, a work now 
largely accomplished by the International Association 
and the British Standards Committee. But with the 
increasing stringency of specifications, perhaps more 
attention should be given to the calibration of the 
instruments used in testing. In physical investiga- 
tions great attention is given to the determination of 
the errors of instruments and methods, and perhaps 
in material testing this has received too little attention. 
I do not suppose there are important errors in the 
indications of large testing machines, though a com- 
parison of results in different laboratories would be 
interesting. Errors in subsidiary apparatus are prob- 
ably more frequent. It is possible with a standard 
bar and a good extensometer, used by the same prac- 
tised observer to measure the agreement or disagree- 
ment of different testing machines. I have found also 
the use of copper cylinders subjected to crushing to be 
a very convenient means of checking load indication$ 
of machines. 
On the Continent and in the United States there 
are public laboratories, supported by the State, where 
anyone can have tests made, at moderate fixed rates, 
by a trained scientific staff and with great accuracy. 
Such institutions assist industry if, on one hand, they 
meet industrial requirements, and, on the other, are 
unbiassed by private interests. They have the advan- 
tage over private laboratories that they are under the 
direction of experts of exceptional experience and 
reputation, and are able to pursue investigation, con- 
tinued often for long periods, beyond immediate re- 
quirements into fruitful by-paths. The view taken in 
Germany is that the work accomplished is partly of 
advantage to the manufacturer who calls for assist- 
ance, and so far should be paid for by him, and 
partly is of advantage to the nation in advancing 
science, which is the justification for a State subsidy. 
Hence it is incumbent on the public institution to 
publish results as freely as is possible without injur- 
ing private interests. 
At Grosslichterfelde the annual income from fees 
for testing amounts to 20,000l., and the annual State 
subsidy is 12,0001. On the staff there are 230 persons, 
of whom seventy-five have had university training and 
thirty-eight technical high-school training. Since 
1880 the work accomplished has steadily and regularly 
increased. 
The Bureau of Standards at Washington is on a 
still larger scale, and does very similar work. It 
receives an annual subsidy from the Federal Govern- 
ment of 100,000]. The equipment is extraordinarily 
complete. For instance, it has a 6o0o0-ton testing 
machine at Washington, and a _ tooo-ton testing 
machine at Pittsburgh, both of the highest sensitive- 
ness and accuracy. Such machines do not exist in 
this country. Happily, we have now a similar insti- 
tution in the National Physical Laboratory at Ted- 
dington. Its functions are somewhat restricted hy the 
NO. 2374, VOL. 95| 

characteristic English jealousy of State action, which 
I think is diminishing. It would be difficult to over- 
estimate the service it has done in the solution of 
various mechanical, electrical, and physical problems 
which were unlikely to be attacked by private persons. 
Some of the researches started by this institution have 
been carried out there, and valuable papers contributed 
to our Proceedings. The superiority of English aero- 
planes has been demonstrated, and their services for 
scouting and for directing artillery fire have been 
invaluable. When the Government required difficult 
experimental investigations to be carried out on aero- 
planes, it found at Teddington a staff and an organisa- 
tion already in existence and suited to the purpose. 
It is a condition of commercial testing that the 
results should be available in a short time and at 
small expense. Hence ordinary tests are of a some- 
what arbitrary character, and do not completely 
imitate the conditions of actual service. We test 
specimens of steel to destruction and measure ductility 
by plastic deformation, though in use the stresses do 
not exceed the elastic limit and the deformation is 
elastic. We test cement in tension and use it in com- 
pression. There is need for constant criticism of 
methods of testing and for the invention of new tests, 
such as tests for hardness and brittleness, although 
new tests must be cautiously introduced. 
One small change could be easily made, and in 
some cases would be interesting. When a set of tests 
on a material has been made, the mean of the results 
is taken as the best value of the property measured. 
In physical investigations generally a further step is 
taken by calculating what is termed the ‘ probable 
error,’ which broadly is a measure of the trustworthi- 
ness of the results. Calculation of the probable error 
is troublesome, but Martens has pointed out that the 
‘“mean error’’ is nearly as accurate, and good enough 
for practical purposes. The mean error is the sum 
of the deviations of the individual results from the 
mean, irrespective of sign, divided by the number of 
observations. The mean error is conveniently ex- 
pressed as a percentage of the mean value. If the 
mean error is large, either the method of testing, the 
uniformity of the material, or the preparation of the 
test specimens is at fault. Generally the source of the 
error can be inferred. A large mean error in tests 
of steel would indicate want of uniform quality in the 
material, in cement—probably faults in preparing the 
test briquettes. Of two supplies of a material, that 
with the smallest mean error is preferable. 
There is another kind of testing, likely in the future 
to be of increasing importance—that is, the measure- 
ment of strains in membeis of completed structures 
in order to determine the stresses to which they are 
actually subjected in service. Some years ago, after 
the loss of the Viper, measurements of the strains in 
the skin of a torpedo-boat slung in various ways were 
made at the request of the Admiralty. The object was 
to obtain information on the stresses in a structure 
supported on waves. Strain measurements on_ the 
members of bridges during the passage of trains have 
been made in Holland, in France, by Mr. La Touche 
and Mr. Sales in India, and by Prof. Turneaure in 
the United States. These observations throw light on 
two points—the trustworthiness of theoretical calcula- 
tions of the stresses and the magnitude of the stresses 
due to dynamical actions which cannot be calculated. 
Recently Mr. I. E. Howard, engineer-physicist at the 
Bureau of Standards, has initiated a very extensive 
investigation which is to extend to large bridges, the 
Panama loclk-gates, and steel-framed buildings. Some 
results on the stresses in the shell of a simple cylin- 
drical tubular boiler have been published. A cylindrical 
boiler shell is a very simple structure, and the straining 
