In order that tlie extent of these clilTerences may he seen T give below a table 

 of differenees in the wave-lengths of the principal lines, the numbers being 

 obtained by subtracting Angstrom's wave-length from that of each of the others, 

 the unit here and throughout this pa]ier being the generally accepted tenth- 

 metre. Angstrom's numbers are his tinal ''definitive values'' given in his memoir. 

 I am not able to refer directly to the original memoirs of cither of the others and 

 in thus quoting at second-hand some errors may be committed. Ditscheiner's 

 numbers are taken from a table given in the American Journal of Science for 

 April 1872 which was copied fnnn his paper by Professor Gibbs and gives the recom- 

 puted values. The values credited to van der ^Yilligen are from the same source 

 and those of Mascart are from Angstrom's memoir. The numbers under D belong 

 to Ditscheiner and tliose under V. W. and I\I. to van der Willigen and Mascart. 





r». 



V.W. 



M. 



r. 



7. 



4.22 



— 0.5 



c 



0.2 



3.47 



— 1.4 



D, 



2.27 



3.31 



— 0.83 



n>2 



1.98 



3.18 



— 1.12 



E 



b, 



LSI 



2.44 



— 1.69 



— 1.1 



b, 



1.84 



3.06 





F 



1.46 



3.26 



— 0.94 



G 



3.9.-) 



4.12 



-fO.30 



Of these residfs, tho.'.-e of Jlascart agree most nearly with Angstrom's tables, 

 but even here the difference amounts to a ma.ximimi of nearly twd units in the 

 case of E. It will be seen 1hal not only do these authorities differ from 

 Augstri'iin but they also differ fuim each other. In the measurement of wave- 

 length the most difficuU thing tu lU'tennine is the value of the grating space. 

 An error in its value will not, Imwevei', influence the ratio of the wave-lengths. 

 From the nature of the ditTerenccs e.\hibiteil in the abo\-e table it is clear that they 

 cannot be accounted for on the su])position of error in the grating space alone, 

 but that among them all. inc!\'ding, of course, those of Angström, there must 

 somewhere exist serious errors in angular measurement and doubtless errors in 

 space determinations beside. \\'hcn th( fe differences auKumt. as they do among 

 these observers, to several units, they are sufficient to influence materially much 

 spectroscopic work in which reference is made toak^olnte wave-length, according 

 as one or the other is accejited as correct. 



The University liaving rect ive<l from the makers, early in the year 1880, a 

 spectrometer of unusual power and excellence of construction and also a number 

 of diffraction gratings ruled by Mr. Chapman upon L. M. Rutherfurd's n.iachine, 

 of which one or two arc luiusually good, it appeared to me to be desirable to 

 make immediate use of it in the measurement of the wave-lengths of several of 

 the principal lines, it was not thought necessary to extend the work beyond a 



