232 Scientific Intelligence. 
Twenty measurements, in all, were made, the range of tem- 
perature varying from 5° C. to 16° C. From these the follow- 
ing result was obtained : 
6b = 5'°66 + 013, 
The probable error is larger than would be desirable, and a 
good deal of irregularity in the results is to be attributed to the 
fact that we cannot be quite sure that the thermometer gave 
the true temperature of the grating, and also to slight shiftings 
in the plane of the grating, which doubtless sometimes 
occurred. Indeed, it is hoped and expected to secure a more 
accurate value by making a very exact determination of the 
position of the line in the winter, and then again in the summer 
when the difference in temperature may amount to as much as 
15° C., and in this we shall be almost entirely free from both of 
the principal sources of error. — Ps 
From the above value the coefficient of expansion is— 
€ = 0000202, 
with a probable error of a little more than two per cent of the 
whole. As spectrum observations are generally made within a 
small range of temperature, this value is, perhaps, sufficiently 
accurate for their correction. Iam unable to refer at present 
to any determinations of the coefficient of expansion of alloys 
similar to this which contains eight parts of tin to seventeen 
parts of copper. Most authorities agree fairly as to the co- 
efficient for copper, but as to tin there is a considerable range— 
from ° 17 to 0000248, It is difficult, therefore, to deter- 
mine what the coefficient for this alloy ought to be, but I ven 
ture the opinion that the above result is more likely to be 
too high than too low. 
SCIENTIFIC INTELLIGENCE. 
I. CHEMISTRY AND Paysics. 
density two-thirds of its normal value, Crarrs and MEIER 
density decreased regu 
one-half of the normal density. The numbers obtained in th 
two investigations differed considerably, however, caper = 
1050°, at which point Meyer found 5°83 and Crafts and er 
‘1. More recently Troost obtained 5-7 at 1250°; from whic 
Meyer concluded that Dumas’ method shuld give the same re 
