358 PROCEEDINGS OF THE AMERICAN ACADEMY. 



separate settings made by each of us agreed to within 45° on the head 

 of one of the screws. This means that the grating can be set by 

 plumb-line to within 3.3 minutes of arc. 



Opening the slit and hanging the bob so that the thread could be 

 seen through it, the various settings made by each of us agreed to 10° 

 on a divided head fitted to the tangent screw. This means by calcu- 

 lation 1.7 minutes of arc of rotation of the slit. 



On the other hand, using full length of slit, as in the previous case, 

 and appropriate width, about 1/1000 inch, various exposures of the arc 

 were taken on the same plate in the manner customary in making 

 focus plates, except that the camera box was left clamped and the slit 

 was oriented. Plates so taken showed no difference in the spectra 

 when the scale on the divided head of the tangent screw was rotated 

 90° clockwise or counter clockwise from the position of parallelism as 

 determined by plumb-line, making a change of 15.3 minutes in the 

 orientation of the slit — a change nine times as great as that in the 

 case of the plumb-line. However, the relative merits of the two 

 methods must not be taken as nine to one, but merely as about four to 

 one, for the plumb-line adjustment for the grating is only about one 

 half as accurate as that for the slit. 



The above facts make it extremely probable that the adjustment of 

 the slit in the previous investigation was good. And, further, if with 

 full length of slit no change in definition could be detected for a rota- 

 tion of 90°, it is all the more probable that with a slit of 5 mm. length, 

 as used in making regular exposures, the definition was the best 

 obtainable. 



(2) Further, as to shift as a function of the orientation of the slit, 

 series of plates were taken with the slit oriented approximately 1° and 

 0.5° of arc clockwise and counter clockwise, including a series at 

 parallelism ; or 360° and 180° counter clockwise, 0°, 180°, and 360° 

 clockwise on the divided head. If orientation introduce shift, the 

 shift-orientation curve should either show a point of inflection at zero 

 orientation or cross the displacement axis at that point. Table I, on 

 pages 356 and 357, is self-explanatory. The data given in the table 

 and the curves of Figure III show that for the two lines studied the 

 shift is not influenced by the orientation of the slit. 



The values of the shift obtained are, within the limits of error of 

 experiment, the same as those obtained in the previous investigation. 



The average deviation from the mean of two measurements (of the 

 shift of a line) on any one plate is 0.003 (Kent) and 0.004 (Avery) 

 t. m. for A 3900.68 ; and 0.002 (Kent) and 0.003 (Avery) t. m. for 

 X 3913.58. It will be noticed that the value of the shift given on the 



