L\ Bell — Absolute Wave-length of Light. 279 



made for nricrometric work and has a power of two hundred 

 and fifty diameters, while one division of the micrometer equals 

 0-28^. The average error of a single comparison between two 

 decimeters is rarely greater than 0*1//, while the temperature of 

 the observing vault can be kept for several days constant 

 within 0° "5 C. and during a day's observations usually remained 

 constant within half that amount. T^ he bars under comparison 

 were side by side, symmetrically placed with reference to the 

 illumination, and were at temperatures very near to 16 0- 67, at 

 which they were standard. 



The facts then concerning the speculum metal bars are 

 these : In about two and a half years S a 2 has shortened by 

 very nearly 1*0^ and S^ by a little over that amount. In 

 S a 2 this change has taken place exclusively in the last decimeter 

 and in S^ it has been confined to the first decimeter. 



The apparent slight increase in DmjS a 2 and Dm 2 S a 1 I do not 

 regard as beyond the effect of the experimental errors. The 

 changes in the lengths of the subdivisions of these standards 

 are very curious and some explanation may be offered by the 

 fact that the bars were cast in a nearly vertical position and an- 

 nealed in sawdust, a method hardly sufficient for a material so 

 strongly crystalline as speculum metal. I think, however, one 

 is justified in drawing the conclusion that speculum metal, so 

 tempting on account of its beautiful surface and the exquisite 

 sharpness of the graduations drawn upon it, is a material thor- 

 oughly unsuitable for standards of length by reason of its 

 tendency to change with time. I have thus entered into some- 

 what minute details in the case of these bars, because the whole 

 question of changes in standards of length is in a somewhat 

 unsettled state, and it seems desirable to put' on record this 

 case, which has been investigated with more than ordinary care 

 by both Prof. Rogers and myself, and in which the changes 

 found have taken place within a comparatively short time. 



It is quite well-known that in 1855 this question was raised 

 by Mr. Sheepshanks, then engaged in constructing the new 

 British standards. Discrepancies amounting sometimes to 2 

 or 3// appeared in his measurements, but after a considerable 

 amount of study, these differences appeared to be too irregular 

 to be fairly ascribable to actual changes. Slight variations of 

 temperature, especially when the standards compared were of 

 different materials, the lagging of the real temperatures of the 

 bars behind the thermometer indications, and particularly the ef- 

 fect of coarse and sometimes unsymmetrical defining lines, are 

 perhaps enough to account for the observation. 



The work, however, done on the U. S. bar " Bronze 11," 

 as reported in the report of the Coast Survey for 1877, seems 

 to show genuine change in that standard. 



