EEY. T. E. EOBINSON ON SPECTEA OF ELECTEIC LIOHr. 
■945 
the plain slit it recovered its power : the same was the case "with my prisms. That the 
sole cause of this was the narrowness of the slit, and not any peculiar action of the colli- 
mator, I verified by attaching to the stand of the theodolite a good object-glass of 8 feet 
focus, and placing it at that distance from the slit ( 3 - 0 ). This arrangement gave vision 
that was first-rate, and showed lines in an electric spectrum 14' beyond the last of those 
that were visible in the usual mode of observing. It follows from these facts, that the 
collimator’s focus should be as long as possible consistent with convenience ; and I will 
suggest that it should be a cassegrain instead of an achromatic. The equivalent focus 
of this is from six to seven times that of the large mirror, and I find by a rough trial 
that the image is very sharp. If the mirrors of A were glass silvered on Foucault’s 
plan, it would have as much light as the achromatic, and be free from all chromatic 
error. I have recently found that the brightness of faint lines is much improved by 
using a cylindric lens before the slit. The slit was generally one minute wide. 
3. The precision of the angles measured depends, on the determination of the index- 
correction, on the precision of the bisection, and on the reading of the circle. The zero 
was obtained at the beginning, and also often at the end of each set, by bisecting the 
slit when illuminated by nearly homogeneous light. At first this was done by inter- 
posing a slip of red glass ; latterly, in preference, by using the flame of a Bunsen burner 
in which chloride of sodium was present. The bisection is sometimes doubtful from 
the flicker of the discharges, more frequently from faintness of the lines, whether 
intrinsic or relative to the ground on which they are seen, and occasionally from want 
of sufficient light to see the point. The verniers read only to minutes ; but the half 
minute is easily estimated, and so 0''25 may be considered the uncertainty of reading. 
It is desirable to form some estimate of the probable amount of error due to the com- 
bined action of the three. This is done by observing twice over the same spectra, and 
comparing the differences of the observed lines ; from which may be obtained the pro- 
bable error, and the probabilities of given observed differences being errors of observa- 
tion, or evidences of the lines not being identical. Unless the graduation were much 
finer than it is, this process would not be of any real value ; and a much simpler one may 
serve. In this work 325 such differences were observed, of which number there were. 
Equal to 0' . . . . 
. . 86 . 
. 86 . 
Comp. 
. 239 
From 0' to +0''5 . . 
. . 131 
Not exceeding 0'’5 . 
. 217 . 
. 108 
From ±0'-5 to +l'-0 
. . 89 
Not exceeding 1'‘0 . 
. 306 . 
. 19 
From 4:1''0 to +1'’5 
. . 15 
Not above 1'‘5 . . 
. 321 . 
4 
Above l'*5 .... 
. . 4 
We may reason thus: if the difference of the places of a certain line in two spectra 
exceed 0'-5, either the lines are identical and the difference is error, or they are distinct; 
108 217 
the probability of the first happening and therefore that of the other or 
it is 2 to 1 they are not the same. If the difference exceed I'-O, it is 16 to 1, if 1''5, 80 
