Wave-length of Light. 353 



were firmly secured at both ends to one and the same casting, 

 which in turn was firmly bedded in a brick pier. In addi- 

 tion the size of the apparatus was such that a variation of 

 even 1'' in the angle was quite improbable. The angle, 

 measured in the ordinary way with a collimating eye-piece, 

 could be determined to 1" of arc, exclusive of errors of gra- 

 duation in the circle. At first there appeared to be distinct 

 variations in the angle as determined at the beginning of 

 each series of observations, reaching sometimes more than 

 10". It soon appeared, however, that when the same part of 

 the circle was used the angle between the telescopes was 

 sensibly the same, and the apparent variations were then 

 traced to a periodic error in the divided circle, which by the 

 method of repetition was completely eliminated from the 

 measurements of angles of deviation, and only appeared in 

 the determinations of 6. This error was finally eliminated by 

 measuring 6 in various portions of the circle. 



The method of determining </> was as follows : — The in- 

 strument being adjusted by the ordinary methods, a suitable 

 line was selected for measurement, and then the angle 6 was 

 slightly increased or diminished until, by measurement of a 

 double deflection, n<f) was found to be very close indeed to 

 360°. Then a double deflection was carefully measured, and 

 if time permitted several times repeated, an observer always 

 being at the eye-piece to see that the line did not move 

 from the cross hairs while the micrometers were being read. 

 Then, clamping the main circle, the grating-holder was turned 

 through 2(f> until the line was very closely upon the cross 

 hairs, any slight readjustments made necessary by this dis- 

 turbance of the instrument were made, and the process was 

 repeated. In this way the initial line of the circle was finally 

 reached and a value of n(f> obtained which depended only on 

 the algebraical sum of the micrometer-readings, always a 

 small quantity. 



The determination of the temperature, a very difficult and 

 uncertain matter in the case of glass gratings, is here com- 

 paratively simple. A sensitive thermometer (Baudin 6156) 

 was kept in contact with the grating, its bulb being carefully 

 shielded by cotton. The construction of the spectrometer 

 made it impracticable to shield the grating effectively from 

 radiation from the observer's body ; but the thermometer 

 apparently proved effective in giving the real temperature, 

 since no discrepancies in the results could be traced to thermal 

 causes. The thermometer-readings were made to O- 05, and 

 the temperature of observation rarely varied more than two 

 or three degrees from 20° C. 



