471 



aity, U* probability that any one line of the spectrum of a gas will coincide 

 with oa* of these a lines is 



ud when p large compared with n, this becomes nearly -. If there are 

 r lines in the spectrum of the gas, the probability that each and every one 

 hall coincide with a line in the solar spectrum is approximately . Hence, 



in the case of a gas whose spectrum contains several lines, we have to com- 

 pare the results of two hypotheses. If a large amount of the gas exists in 

 the sun, we have the strongest reason for expecting to find all the r lines in 

 the solar spectrum. If it does not exist, the probability that r lines out of 

 tin* n observed lines shall coincide with the lines of the gas is exceedingly 

 small. If, then, we find all the r lines in their proper places in the solar 

 apectrum, we have very strong grounds for believing that the gas exists in the 

 sun. The probability that the gas exists in the sun is greatly strengthened 

 if the character of the lines as to relative intensity and breadth is found to 

 correspond in the two spectra. 



The absence of one or more lines of the gas in the solar spectrum tends 

 of course to weaken the probability, but the amount to be deducted from the 

 probability must depend on what we know of the variation in the relative 

 intensity of the lines when the temperature and the pressure of the gas are 

 made to vary. 



Coincidences observed, in the case of several terrestrial substances, with 

 several systems of lines in the spectra of the heavenly bodies, tend to increase 

 the evidence for the doctrine that terrestrial substances exist in the heavenly 

 bodies, while the discovery of particular lines in a celestial spectrum which do 

 not coincide with any line in a terrestrial spectrum does not much weaken 

 the general argument, but rather indicates either that a substance exists in the 

 heavenly body not yet detected by chemists on earth, or that the temperature 

 of the heavenly body is such that some substance, undecomposable by our 

 methods, is there split up into components unknown to us in their separate 

 state. 



We are thus led to believe that in widely-separated parts of the visible 

 universe molecules exist of various kinds, the molecules of each kind having 



