the Laws of Chromatic Dispersion, 259 



whence »=l-9, log €„=0 1316156, and tf n = 0*003955. The 

 differences between the calculated and observed indices are 



B+ C- D-f E+ F- G- H+ 



0-000080 0000122 0-000012 0000081 0000118 0000006 0000086 



and with the corrected indices the extrusions are 



b x — ex— dx-\- ex-\- fx-\- gx— hx— 



0000154 0000051 0000107 0000163 0-000137 0000013 0-000189. 



For distilled water, temp. 15° C, their indices are 



giving n=l-87, log e n = 0*1215388, and ^=0*004023, or to 

 equalize errors, a n = 0*004018. The differences between the 

 calculated and observed indices are 



B+ C- D-f- E- F- G- H+ 



[0-000005 0-000030 0000016 0000036 0-000046 0000054 0000048; 



and with the corrected indices the extrusions are 



bx— Cx— dx-\- ex-{- fx-\- gx— hx— 



0-000148 0000049 0-000102 0000158 0000131 0000012 0000182. 



In these three last cases, the differences between the calculated 

 and observed indices all he within the limits of probable error, 

 and with the corrected indices the extrusions are all regular. 



The results deduced from Messrs. Dale and Gladstone's ob- 

 servations on water are of the first class, and agree closely with 

 those deduced from Fraunhofer's two sets of observations. The 

 large discrepancies presented by Powell's observations on this 

 medium are thus shown to be entirely caused by experimental 

 errors. 



These results all tend to establish the perfect universality of 

 the laws developed in the author's former paper. 



For the six media above discussed, Messrs. Dale and Gladstone 

 have given the index of the fixed line A. Fraunhofer left no 

 record of the normal wave-length corresponding to this line. By 

 comparing together Messrs. Dale and Gladstone's indices, while 

 keeping in view the likelihood that this normal bears some 

 definite relation to the others, the conclusion has been reached 

 that its most probable value in reference to B as unity is 

 1-121019, log 0049613. This makes the value of the fraction 

 A-rE = 1-466', which is =66 (t — v), thus establishing a definite 

 relation between this and the other normals. 



This value of A being assumed, we have the following discre- 

 pancies between the calculated and observed indices of this line 

 for the six media : — 



