4 B. E. Moore 



it would escape detection at any rate. Again, any irregular con- 

 traction of the films in drying would cause appreciable error in 

 the measurement of very small quantities. Errors from such 

 sources are mitigated by taking several sets of plates. Two sets 

 of s- and ^-plates were at my command, but only one set of no- 

 field plates. When irregularities were present upon one set of 

 plates they were also present upon the other, but the difficulty 

 might be with the no-field plates with which they were compared. 

 For the purpose of detecting small variations, then, the method 

 is not altogether free from error. But if both s- and ^-components 

 are photographed together, i. e. without the intervening calcite, 

 neither does one then know whether the ^-component or the 

 ^-components are displaced in an unsymmetrical triplet ; and if 

 the ^-component happens to be a wide line and the .y-components 

 have small separations, the readings are then confusing and much 

 less accurate than the method which I have followed. Had I an- 

 ticipated the dissymmetry of lines at the time these photographs 

 were taken, I should have desired to have photographed the p- and 

 ^-components directly upon the same plate as the no-field lines, 

 by the long-established method of Rowland, viz., by taking two 

 exposures upon one plate, covering one-half of the plate for the 

 no-field exposure and the other half of the plate for either the 

 p- or the ^--exposure. Even then, it occurs to me, that the method 

 might not be accurate enough to test dissymmetries as small as 

 one might expect from the Voigt equation, as it also seems to 

 me that the method is not quite accurate enough to test quad- 

 ruplets and triplets for simple relationship between the magni- 

 tudes of the separations (e. g. Are these separations related to a 

 normal "a", and are they multiples of aliquot parts of "a"?) My 

 previously measured triplets would certainly not suggest the choice 

 of any value as a normal, and neither would the quadruplets in 

 the present investigation. 



The question, however, does arise, Is there adequate error to 

 allow one to collect them into groups whose separations are re- 

 lated to one another in a simple way ? I do not think the error so 

 large. However, it does seem to me what is wanted to answer 

 this interesting question and the dissymmetry question as well, 



92 



