EFFECT OF PRESSURE UPON ARC SPECTRA. 145 
on the steeper side and less easily distinguishable by the eye. (See accompanying 
small diagram.) 
The magnitude of this phenomenon is of an order which only interferes materially 
with the measurements of the lines of Group I. 
Though roughly divisible into three groups, it is by no means certain that the 
shifts of all members are the same ; indeed, the converse is indicated by the displace¬ 
ment curves (Diagram II.), in which there is a tendency for the lines to maintain 
the same relative displacements throughout the range. The mean values of the shifts 
agree, however, to an extent which cannot be disregarded, in assigning to the ratios 
of the three groups the values 1:2:4, which Humphreys states holds good for the 
spectra of other elements. 
13. The Relation between the Pressure and the Displacement. 
In August, 1906, the measurement of Set A was completed, but as the displacement 
curves then showed a departure from a linear relation between the displacement and 
the pressure at 15, 20, and 25 atmospheres (where the values were roughly double 
those required by the readings at other pressures for a linear connection to hold good, 
Diagrams I., III., V.), it was deemed advisable to re-measure the plates taken at those 
pressures. The second measurements agreed well with those originally obtained, but 
it was then found possible to measure a few of the strongest lines on another plate, 
Dl8, taken at 25 atmospheres, which had previously been rejected as being too 
nnder-exposed. The fourteen readings obtained from it indicated that the linear 
relation would be satisfied at that pressure if they were substituted for the readings 
given hy photograph Dl9. 
On both the photographs at 25 atmospheres the comparison spectrum was in the 
central strip, and as the precautions that had been taken (p. 129) seemed to preclude 
the possibility of an undiscovered fictitious shift, it was decided to investigate the 
phenomenon carefully, and to repeat the whole series of photographs. 
The photographs of the second set, B, give results more consistent with a linear 
connection betw r een the pressure and the displacement (Diagrams II., IV., VI.), except 
m the case of Group III., where the displacements at 20 atmospheres are again too 
high, though the readings given by Groups I. and II. agree well with those obtained 
at higher pressures in making the function a linear one. 
At 5 atmospheres most of the lines of Groups I., Set A, as well as most of the lines 
VOL. CCVIII.—A. 
U 
