76 



REPORT OF THE SECRETARY. 



This criterion lias been elsewhere mentioned, and indeed is sufficiently obvious to 

 need little explanation; for if among a large number of curves representing the same 

 spectral region a deflection asserts itself constantly at the same place, that deflec- 

 tion is evidently due to a constant cause and not to an accidental one. Now, from 

 the nature of the process, it is hardly possible to imagine any efficient constant 

 cause of such minute abrupt deflections as are in question, which is not a solar or 

 telluric cause, and it is these only we seek. Again, it will be seen that, provided 

 the deflections are of a finite order of minuteness, and that our mechanism can record 

 with such precision that the true or solar line falls at absolutely the same place on 

 each of even two plates, the chances are infinite against any exact coincidence of 

 lines which are not ''true." In practice, where absolute accuracy is unattainable, 

 we mast determine experimentally between what limits error customarily presents 

 itself, in the case of the true lines and in that of the false, and then ascertain what 

 probable error attaches to the result from the final comparison of any number. 



During the preceding year certain main deflections, somewhat less than one 

 hundred in number, were so established as "true;" but when within the past year 

 the work reached a state of progress warranting the more detailed investigation 

 of the spectrum alluded to, upon applying the criterion mentioned above to the 

 deflections representing the minute lines with which the work was now dealing, it 

 became evident that the number of very small casual deflections made it desirable 

 to subject the method of making the curves to a renewed examination, in which 

 the source of each class of minute error should be as far as possible discriminated. 

 This examination has shown that the site which the observatory occupies is exposed 

 to local tremors and magnetic disturbances in' a still greater degree than had been 

 anticipated. The appended table gives a rough classification of the sources of these 

 errors. 



Classification of errors affecting holographs. 



Kind of error. 



Causes. 



Remarks. 



(1) In the position of a 



ine.. . 



Variation of temperature of the 



Mostly due to temperature 







prism, inaccurate clockwork, etc. 



conditions, the direct result 

 of the present building. 



(2) The suppression of 



small 



Inaccuracy in photographic reduc- 





true lines. 





tion to line spectra. 





(3) The production of 



false 



Earth tremors, magnetic fluctuations, 



Due (1) to location of build- 



small lines. 





thermal currents in bolometer cir- 



ing; (2) to avoidable instru- 







cuit, change of potential in bat- 



mental defects. 







teries. 





From what has been said it will be seen that in order that the agreement 

 of a number of separate curves might furnish conclusive evidence of the reality of 

 a line, it became necessary to establish experimentally the amount of accidental 

 deflection due to local tremor wholly, and also the admissible amount of final error 

 inherent in the mechanical, optical, and photographic processes employed. The 

 following degree of accuracy has been judged possible as a final attainment with 

 the present means, and is that aimed at, although not yet reached. This aim is to 

 produce curves which, in a length of 60 centimeters, shall be free from any probable 

 error of a magnitude greater than one-tenth of a millimeter — a quantity whose 

 minuteness relative to the means employed may be understood when it is mentioned 

 that it corresponds to six-tenths of a second of arc, as measured on the circle of the 

 spectrometer. 



For the above reasons it will be seen that the improvement of the apparatus used, 

 as well as the investigation of sources of possible error, became the most important 

 work of the observatory before the publication of authoritative results. 



