124 
would take too much of your time. My object was to show 
that if the *• Quarterly Journal of Science” and the “ Chemi- 
cal News” are to represent scientific opinion with any degree 
of truth, they would do well to use a little discretion as to 
what they print. 
“ Remarks on Mr. Spence’s Experiments on the Effects of 
Cold on the Strength of Cast Iron,” by Joseph Baxendell, 
F.R.A.S. 
In concluding his paper read at the last Meeting of the 
Society, Mr Spence stated that “ he had so much confidence 
in the experiments then detailed, that he had no hesitation in 
giving it as an ascertained law, that a specimen of cast iron, 
having at 70° Fall. a given power of resistance to transverse 
strain will, on its temperature being reduced to zero, have that 
power increased by 3 per cent.” Now, in physical investiga- 
tions it is often very hazardous to rely too much on the 
simple means of sets of experiments or observations, however 
numerous, unless the theory of errors has been employed to 
test their value; and in the inquiry as to the effect of cold 
on iron, this remark applies with peculiar force. 
Mr. Carrick lias objected to Mr. Spence’s experiments 
that the differences between some of the breaking weights 
are very large; and also that the iron used was of an 
inferior quality; but the quality of the iron, unless it is 
actually very bad, is a matter of secondaiy importance, 
since its only effect will be to increase the range and 
diminish the average of the breaking weights; and with 
respect to the wide differences between some of the results, 
this is more than compensated for by the number of the ex- 
periments which is sufficiently great to afford the means of 
determining approximately the law of error to which they 
were subject, and thus of ascertaining whether the final 
results are entitled to the high degree of confidence which 
Mr. Spence has placed in them. When, however, I ran my 
eye over the columns in Mr. Spence’s table after the reading 
of his paper, it at once struck me that the differences of the 
individual breaking weights from the mean values in both 
