550 HENEY A. ROWLAND 



Indeed, no process of angular measurement could approach the accuracy 

 of this one. 



Thus, using a line P to start with, I determine other groups of lines, 

 a', V, c', d', etc. From these again I find groups, some of which may 

 be the same as the first; then again from these, other groups. The 

 process can he continued further, hut we are apt to come hack to the 

 same lines again, and we are further limited by the visibility of the 

 lines. Thus the limit of great accuracy by eye observation in either 

 direction is practically 4200 and 7000; although in a dark room, especi- 

 ally in the first spectrum, one can see much further, even beyond the A 

 group, although it is difficult to set on the lines, and one is apt to mis- 

 take groups of lines for single lines. 4 When one uses a group as a 

 standard, and one or more of the group is an atmospheric line which 

 varies, the measures will of course vary also, unless the atmospheric 

 line is in the centre of the group. This is a very common source of 

 error, and has caused me. much trouble. In a grating with a very 

 bright second spectrum, I have, however, obtained the coincidence of A 

 with the region whose wave-length is about 5080, and have thus con- 

 firmed the value given in my preliminary table, which was obtained by 

 a very long interpolation passing from the first into the second spec- 

 trum. 



The accuracy of these primary standards can be estimated from the 

 equations given in Table VII. It is there seen that there is scarcely 

 any difference in the different measures as derived from different lines. 



It is to be specially noted that the wave-length of P and the lines 

 directly determined from it have no more weight than any of the 

 others. The table might just as well have been arranged with the D 

 line, or any other, first. The true way of discussing the results is to 

 form a series of linear equations, about twenty-six in all, and solve 

 them. This is the method I have used, although I have not discussed 

 them by the method of least squares. 5 



Some miscellaneous observations not included in the table allowed 

 me to add a few more line? to these primary standards. 



Having completed these primary standards, I then observed several 







4 la a very bright grating I have faintly seen, and even measured, lines down to 

 wave-length 8500. My assistant, Mr. L. E. Jewell, can see far into what is called 

 the ultra violet, even to wave-length 3500 or beyond. 



5 The calculations of this paper have involved about a million figures, of which I 

 have personally written more than half. Hence I am not anxious for more labor of 

 this kind. 



