144 I. P. Косн. 
104”. If applying the factor of reduction determined above а 
the result would be ф — 76°46'13".7 which shows that the factor of 
reduction is in this case too great. As the observations of the sun 
are carried out in temperatures which are 10°—15° higher than the 
corresponding temperatures in which the star observations have taken 
place, this might in itself also seem quite reasonable. The material 
is, however, too slight to admit of anything being based upon it in 
this respect. 
If we consider the determinations of latitude at the great zenith 
distances from 71° to 86°, one must at once be struck by the fact 
that while the lower culminations give a fairly correct value of the 
latitude, the upper culminations give a value which is far too small; 
so in this case the refraction computed must have been too small. 
It is reasonable to assume that there are local disturbances which 
make themselves felt in the atmosphere. All the observations of 
the upper culmination with large zenith distances fall within the 
period Sept. 8th to October 12th i.e. the period of the year when 
mirages over the fjord and the sea are everyday phenomena (as to this 
see chapt. VI), and the sights at the upper culmination happen to 
traverse the fjord, where the disturbances in the lower strata of air 
must be supposed to be greater. 
The fact that the eight groups of observations at the lower cul- 
mination with zenith distances between 83° and 86° give such a 
good value of the latitude is due to fortuitous circumstances, as will 
easily appear from the desultory nature of the single values. There 
is, however, probably some connection between the fact that the 
error is rather evenly distributed on both sides of the real value, 
and the fact that these observations are made in the height of sum- 
mer, when the atmosphere is not influenced in any considerable 
degree either by the melting of snow or the freezing of the water. 
Through their determinations of time by means of the sun BORGEN 
and COPELAND found support for their supposition that the Besslian 
refraction was too great. It was, therefore, quite natural for us to 
direct our attention to this point. With the object of computing the 
correction of the mean time chronometer 57310, we made in Sep- 
tember 1907 a series of zenith distance observations of the sun near 
the prime vertical; as a further test the clock correction was a few 
times determined by transits on the vertical af Polaris. Each group 
of zenith distance observations consisted of twelve, three sets of ob- 
servations to the upper limb and three sets to the lower. Within 
the same group the observations to the same limb generally cor- 
responded very well, whereas the results of the upper and lower 
