L.Bell — Absolute Wave-length of Light. 363 



. Computing the similar quantities for the speculum metal 

 gratings III and IV, for grating III, 



5 = 0-0035 L 935 8 m,n 

 q> — 36° 0' 25 // -17 



6 = 6° 59' 5S"-56 

 I = 5133-89 



and for grating IV, 



s — 0'002534306 mm 

 <p = 35° 59' 59"-06 

 6 = 6° 58' 31" 

 I = 5914-37 



Reducing now these latter wave lengths to the corresponding 

 values of D 19 introducing the barometric corrections and com- 

 bining, the final results for that line are 



Grating W. L. 



I... 5896-18 



II 5896-23 



III 5896-15 



IV 5896-17 



Finally, then, the mean value of the absolute wave-length of 

 T> 1 in terms of the mean value assigned to S% is 



5896-1S 

 in air at 760 mm pressure and 20° C. temperature, or in vacuo, 



5897-90 

 It is no easy matter to form an estimate of the probable error of 

 this final result. So far as errors of observation go, the result 

 should be correct to within one part in half a million, but there 

 are so many complex sources of constant errors in this problem 

 that such a statement means little. My present result exceeds 

 the estimated probable error of my former result considerably, 

 though it falls within the limit set by Prof. Rowland and my- 

 self for the possible error and noted in his paper on " Relative 

 Wave-length " of the same date as my own. The cause of this 

 discrepancy is partly due to the varying temperature under 

 which the glass gratings were first measured, and partly to the 

 change in the value assigned to the standard of length.* 



Then too, the corrections applied to gratings II and III may 

 be slightly in error. Taking into account all these sources of 

 uncertainty it is my opinion that the above final result is not 

 likelv to be in error by an amount as great as one part in two 

 hundred thousand. 



2. * In terms of the length I originally assigned to S a 2 the wave-length of Di 

 would be 5896-14, while if the value deduced from the Berlin comparison were 

 taken it would be 5896-22. The wave-leugth quite certainly lies between these 

 values, but the proper weight to be given to the Berlin comparison relatively to 

 the others is rather uncertain. 

 Am. Jour. Sci— Third Series, Vol. XXXV, No. 209.— Mat, 1888. 

 22 



