1319 
Group Bi/p Di/p, 
| I 1.47 | 2.00 
Il 1.47 2.30 
| 
| lll 1.22 1.76 
| 
| IV 1.62 2.50 
| Vv 1.69 1.23 
| 
the comparison to £. In that case he would have found for 
2 3619.54 (group I) a better agreement between B, and D, than 
between B, and D,. By the comparatively small number of obser- 
vations 5, however, this investigation would not have brought much 
new light. 
If for a moment we compare stil the results for the shift per 
atm. from the observations at 10 atm. overpressure in the spark- 
spectrum with those in the are-spectrum (viz. B, with D,), we see 
that generally the shifts in the arc-spectrum are larger than those 
in the spark-spectrum, though also here are exceptions found 
e.g. for 43783.67 in group 1, 23775.74 in group H, 23972.32 in 
group III, 23769.62 in group IV and 3973.75 in group V. 
(Must we consider it as a coincidence that these lines 4 3769.62 
43775.74, 23783.67) and (4 3972.32 and 43973.75) lie near each 
other in two groups? Is it possible that in the observations on the 
pressure-effect a mutual influence can play a part here, as is some- 
times found in observations on the Zppmar-effect >). 
The ratio between the mean values of B, and D, is different for 
the different groups, as is evident from the following table: 
Group Di/g, 
I 1.68 
I 2.05 
HI 1.26 
| IV 0.62 
V 1.20 
Only in group IV B, > D,, but perbaps this may be aseribed 
to the small number of observations that have been made in this group. 
Let us now proceed to the investigation, in how far there exists 
