DISPERSION IN ARTIFICIAL DOUBLE REFRACTION. 



295 



noticed, it was discovers 1. not from any such curve, but from the actual circle 

 readings of the spectroscope, which usually increased by steadily increasing differ- 

 ences ; in this glass, just after the readings corresponding to X 5000, the difference 



decreased instead of increasing. Keai li ll^s Were taken several times \\itli '_'le:it care. 



and the effect was confirmed in each case. This demonstrated that the cause was 

 not accidental. 



4500 5000 



//V T{A/TH-MTRS. 



1st order olw. 



Ilnd 



5500 



6000 



Tension obs. 

 Pressure 



6500 6700 



Fig. 8. Glasses 3453. Diagram of residuals. 



Clearly a wrong determination of a reference line is out of the question ; this could 

 not cause an irregularity in the differences of circle reading. 



With regard to (2) and (3), the weights and the divisions of the circle were tested 

 with great care and found correct. 



As to personality, the jump of 40 tenth-metres between X 5000 and X 4800 would 

 require an error of 6' in locating the centre of the band, and a change of personality 

 of this amount, in a fairly bright region of the spectrum, is unthinkable. Besides, if 

 the effect is due to such a change, it should appear in all the glasses, which is not the 



c ise. 



Under these conditions it oeoma safe to assert that between wave-lengths 4500 and 



