462 B. BAY LANKESTER. 



the comparison slit, by which, for a given wave length, an 

 apparently equal luminosity in the absorbed and in the com- 

 parison slits was given. The apparatus allows one easily to 

 read off a slit-width as small as 0*0005 mm. Every measure- 

 ment was five times repeated. The number set down on the 

 records is in each case the mean. Their probable error is in 

 general barely more than 1 per cent, of the measured value ; 

 only in the outer red and violet, and with very powerful 

 absorption, is the error greater on account of the diminished 

 intensity of light. The course of the absorption curve (as 

 drawn on the charts) will then in all essentials faithfully re- 

 present the fact. 



An objective test for the criticism of the trustworthiness of 

 the measurement results is afforded by the comparison of the 

 numbers arrived at when two different thicknesses of the 

 absorbing layer are used. If for any wave length the intensity 



of the original light is weakened to per cent, by passage 



through a layer of the thickness 1, then for the same wave 

 length the intensity {=. i) after passage through a layer of the 



thickness n is given by the formula i^ = — - . Accordingly, if 



we have measured the course of the absorption for a known 

 thickness of layer, we can reckon it also for every other possible 

 thickness of layer. I have carried out the calculation for neutral 

 Bonellin and neutral Chsetopterin, and inserted the calculated 

 values in thick type in the tables. The agreement between 

 calculation and observation is amply sufficient, especially when 

 we consider that important alterations in the light intensity 

 value must be brought about by minute changes in the position 

 and breadth of the spectrum strip, the average light intensity 

 of which is being determined, in those parts of the spectrum 

 where there are sharp alterations in the absorption. The 

 greatest care was given on this account to the exact position and 

 borders of the spectrum strip, and every time it was carefully 

 determined whether the wave length scale was exactly in its 

 proper position. The breadth of the spectrum strips, separately 



