GENERAL DISCUSSION OF SPECTROSCOPIC RESULTS. 
461 
From the accordance between the two sets of measures for each spectrum the 
accidental errors of the mean positions may be estimated at about ’01 millim. The 
mean error is less than this for the best defined lines, but greater for the broad or 
diffuse lines. This error corresponds to an error in wave-lengtli of T6 tenth-metre at 
. 5000, decreasing to '07 at 4000, and '04 at the end of the spectra at 3500. 
It does not, of course, follow that the wave-lengths in the tables can be relied on 
within these limits, except for isolated lines of which the measures are unaffected by 
any disturbing causes, such as faint companion lines or shadings. But this degree of 
accuracy seems actually to have been attained in a large proportion of the iron and 
titanium and other well-identified lines (see Table I., p. 478). 
The mean values for the hydrogen lines, given separately in Table II., agree very 
closely indeed with the computed values. Thus in nineteen lines, in a total of 
twenty-eight, the differences do not exceed 'UI tenth-metre, and in four lines only the 
differences reach T tenth-metre, three of these being the lines y, S and e, which are 
difficult to bisect on account of their great width. This result will, perhaps, l)est 
indicate the general accuracy of the wave-length work. 
No corrections of any kind have been applied to the results, and it may be well to 
emphasise here the fact that no corrections are needed for apparent displacements 
due to the different altitudes to which the various gases ascend in the chromosphere. 
In making the measures, the settings were made at the position of maximum density 
in the case of the broad over-exposed hydrogen lines, the finer lines being simply 
bisected without reference to the apparent edge of the moon’s limb. 
It is probable that the positions of maximum density of the images of the stronger 
lines correspond to radiations coming from a region within 2" of the photosphere, 
\vhilst in the fainter ultra-violet liydrogen series the mdiations are almost confined 
to the flash spectrum layer, the emission fi'om the upper chromosphere being almost 
insensilfie for these lines. 
Assuming that all the settings were made on arcs radiated from a region within 
2 " of the photosphere, this being the approximate limit to which tlie reversing 
layer extends, no ajDpreciable error will be made by bisecting the images; for a 
difference of 1" of arc in the positions of the various gases above the moon’s limb 
would make an apparent shift on the plate at 188 centims. focus of '0091 millim., a 
quantity about equal to the accidental errors of measurement. 
On the other hand, if the settings were made on the inner edge of the arcs, that is, 
at the apparent limb of the moon, a considerable error would be inti'oduced, depending 
on the intensities of lines, which spread inwards as well as outwards as a result of 
irradiation. 
It is particularly noticeable that a large number of the arcs of the flash spectrum in 
these prismatic camera photographs are narrow lines sharply defined on both sides, 
there is no diffuseness on the outer side, as might be expected were the arcs true 
images of the strata producing them. They are in reality, as I have previously 
