XVIIL] VARIATION OF THE ATTACK. 253 



I always have reserved it, we have to consider whether the lines 

 are massed in certain regions of the spectrum. A rapid survey 

 was made with this object, the result showing that the same 

 thing did truly apply to these other regions with the dispersion 

 we had employed. 



Still full of the " solar line double " and " common impurity " 

 point of view, the attack was varied by an inquiry whether 

 there was any special character connected with the coincident 

 lines. In this way it was imagined that some light might be 

 thrown upon the question as to whether they were chance 

 coincidences. In short, the thing done was inquiring whether 

 these coincident lines varied their behaviour in some other 

 special manner from non-coincident lines taken at random. 



Supposing them to represent mere chance coincidences 

 " physical coincidences," as they have been called, or again, 

 lines so near together that our means cannot separate them 

 there is no reason why they should behave differently from 

 the other lines in a spectrum when the temperature is changed ; 

 while, as our hypothetical furnaces had taught us, if they be 

 truly common, they must vary with temperature. Further, they 

 must vary in such a way that other conditions being equal, they 

 shall become stronger when the temperature is increased, and 

 fainter when the temperature is reduced. 



Now what was the best mode of attacking this problem ? I 

 was unable to see a more expeditious one than that presented 

 to us by the sun's atmosphere. The following considerations 

 will show how we might hope for help in this quarter. 



Whatever be the chemical nature of this atmosphere, it 

 will certainly be hotter at bottom that is, nearer the photo- 

 sphere than higher up. Hence, if temperature plays any 

 part in moulding the conditions by which changes in the 

 resulting spectrum are brought about, the spectrum of the 

 atmosphere close to the photosphere will be different from that 

 of any higher region, and therefore from the general spectrum 



