SECONDARY STANDARDS. 
943 
The differences between two or more series of comparisons of the same weights, 
though small, are larger than the probable error of each series would lead us to 
expect. Of the errors which affect the results of weighing, some partake too much 
of the nature of constant errors to be fairly estimated by the method of least squares. 
Of this kind is the error due to small differences of temperature of the weights. 
Whenever it was practicable, the weights to be compared were left in the balance- 
case during the night previous to the day on which they were compared. This pre- 
caution, however, was in some measure defeated, when a single weight was compared 
with the sum of several others ; for the latter would be in advance of the former in 
following the changes of temperature during the time occupied by the comparisons. 
The effect of temperature on the apparent weight of any object appears to be due to 
currents of air ascending or descending, according as the weight is hotter or colder 
than the air in the balance-case. A brass kilogramme that had been left for several 
hours in the balance-case where the temperature was 5°’2 C., appeared to be about 
5 milligrammes lighter after it had been heated up to 16°‘4 C. The hygroscopic 
matter contained in some of the auxiliary weights, from which it was difficult to free 
them entirely by digestion in boiling water, may also have introduced a small error 
in one direction. In order to diminish, as much as possible, any inaccuracy result- 
ing from this cause, all the more important weighings into which these weights 
entered, were made within the narrowest practicable limits of time. Many observa- 
tions that could not be brought within such limits of time as were considered satis- 
factory, were rejected, the chance of a larger irregular error belonging to a small 
number of comparisons being considered less injurious to accuracy than the error in 
one direction to be apprehended in a larger number, extending over a considerable 
interval of time. That part of the weighing which depended upon the performance 
of the balances was most satisfactory. When the large balance, constructed by 
Mr. Barrow, was loaded with a pound in each pan, the probable error of a single 
comparison, by Gauss’s method, was 0‘00056 grain, or less than one-12 millionth of 
the weight in either pan ; with a kilogramme in each pan, the probable error of a 
single comparison, by Borda’s method, was 0'00]62 grain, or less than one-9 millionth 
part of the weight in either pan; by Gauss’s method it was 0‘00112 grain, or one- 
14 millionth of the weight in either pan. 
Legalization of the new Standards. 
Legal authority has been given to the new Standard lb. and its four copies in pla- 
tinum by an Act of Parliament, entitled ‘‘An Act for legalizing and preserving the 
restored Standards of Weights and Measures.” The most important of those pro- 
visions of the Act which relate exclusively to the Standards of Weight, are contained 
in the following extracts : — 
“Whereas by an Act of the Fifth Year of the Reign of King George the Fourth, 
Chapter Seventy-four, ... it was enacted ... that from and after the First Day of May 
