Survey of Northeast Greenland. 185 
It might seem a very likely a priori supposition that the in- 
fluence of the time of day on the co-efficient of refraction must be 
very small in high arctic regions, particularly during summer and 
winter, because the day period in the temperature practically dis- 
appears at these times, and only plays any part at all during the 
spring and autumn. In the case in question it would consequently 
have been natural to undertake researches of the variation of the 
co-efficient of refraction for two periods, according as the sun is 
circumpolar or not, but our material does not permit of such a 
manner of proceeding, because the single times of day and night in 
this manner would be covered by far too small a number of obser- 
vations. I have therefore come to the conclusion that it would be 
most correct to treat the whole material in a lump. In order to 
eliminate the disturbing influence of the seasons as much as possible 
I have corrected all the single values of 4k for the special increase 
due to the season. Ås an example I shall mention that the first named 
value in the table below has been arrived at by taking the value of 
ДЕ on August 24th, at 3215 from the table on р. 156 and adding to 
this 0.004 (cf. the summary of the variation of the co-efficient of re- 
fraction according to the season, р. 179). The values of 4k given in 
the table below are expressed in units of the third decimal place. 
In grouping the observations according to the time of day I 
have aimed at getting, within each group, such a large number of 
observations, that one might reasonably hope that the casual character 
of the material would not influence the means of the groups to such 
an extent as to be fatal to the sense. The unsystematic manner in 
which the observations have been undertaken, has quite naturally 
brought about the result that these observations group themselves 
more closely round certain hours within the twenty-four than round 
others. The intervals between the means of the single groups, there- 
fore, become unequal; for instance it has not been possible to avoid 
a very large interval near midnight. 
The levellings at the sea horizon were undertaken on the same 
days as the other levellings, but either before or after the latter. As 
it was of great importance to me that the groups formed in the case 
of Horizon II should come to contain single values of the same date 
as the corresponding groups for the other lines of vision, a seeming 
arbitrariness has been introduced at a few points; in that the value 
of Мау 31st at 11250 has been reckoned with the group before noon, 
whereas the value of October 9th 11450 has been reckoned with the 
group immediately after noon; in the same manner the observations 
made at 8P40 on August 27th and September 11th have been given 
in two different groups. 
