o-i 



?j 



0-2 



0-2 



)) 



0-3 



0-3 



j> 



0-4 



0-1 



J, 



0-5 



0-5 



)) 



0-6 



0-6 



JJ 



0-7 



07 



JJ 



0-8 



0-8 



JJ 



0-9 



0-9 



55 



1-0 



6 Coincidences between the Lines of different Spectra. 



Between O'O and 01 Xth metre, 8 coincidences. 



jj J JJ 



JJ ' JJ 



„ 2 



jj 8 jj 



j, ^ 



;; 1 JJ 



„ o 



J? -*■ JJ 

 JJ -*• JJ 



Here again two points are very much out; but an over- 

 estimation of i\)th of a Xth metre (a quite possible mistake) 

 in reading off Angstrom's scale at three different places would 

 replace these two points in the curve; we may therefore fairly 

 consider that the method affords support to Griinwald's 

 hypotheses. 



A curious point in connexion with figs. 7 and 8 lies in the 

 fact that in both the first experimental ordinate, instead of 

 being the greatest, is smaller than the second. Is it possible 

 that this may indicate a systematic error in Grriinwald's cal- 

 culations? The probability of this seems increased by an 

 examination of the errors with regard to sign ; for, in the 

 comparison of the hydrogen and water spectra, the average 

 positive error (obtained by dividing the sum of all the errors 

 in which the predicted exceeds the observed value by the 



number of such errors) is — — = 0*6 Xth metre ; while the 



17*2 

 average negative error is " = 0'5 Xth metre. Since a 

 oo 



constant arithmetical error is highly improbable, the only 

 explanation of this seems to be a small systematic difference 

 between Hasselberg's scale of measurement for the hydrogen- 

 spectrum (from which Griinwald's water lines are obtained 

 by halving the wave-lengths of the hydrogen-lines) and 

 Angstrom's scale. As it would seem to be very difficult 

 to detect such a difference of scale in any other way, this 

 example adds another to the purposes to which tins method 

 may be applied. 



