106 I. P. Kocx. 
In this case the observation watch was apparently not able to 
stand the long exposure in the cold Observatory, but lost 05.4 as 
compared with 57396. That it was really 57203 which altered its 
rate, appears doubly sure from the comparisons with the mean time 
watch, in that the relative rate of the two watches can be regarded 
as established on the basis of the first-mentioned comparisons on p. 105. 
In the time which elapsed between 12'41™30s and 148467 00$ 
> 5 
— 205.1. However, it only gained 195.6 as compared with 57310; the 
difference is 08.5 which within the limit permissible corresponds with 
the above result. 
The examples quoted, which are quite ordinary, show that we 
may with considerable certainty read off an ordinary good pocket 
watch in tenths of a second; and further, the last example shows 
how careful one must be as regards watch comparisons, before and 
after an observation has taken place. 
57203, in comparison with 57310, should have gained 
Thread intervals. Before the commencement of the transit 
observations it was considered expedient to determine the thread 
intervals. As will appear from the example given below, which 
example is an extract of an observation group of twelve stars, we 
made use of stars with comparatively small declinations. The reason 
of this was that the thread intervals were determined during the 
latter part of August. The short nights were at that time still so 
light that stars of greater declination and of a suitable size were not 
visible. On the other hand the external conditions — light nights 
which made the use of artificial light superfluous; a temperature 
about zero — during the latter part of August and the beginning of 
September were particularly favourable. The meridian was pointed 
by directing the telescope for each position of the instrument against 
Cairn V, after which the instrument, by means of the graduation of 
the horizontal circle, was turned 1° 42’.3 to the right. 
In the case of each single star the value of the thread intervals 
was computed for a star of a declination of zero (9 — 0) according 
to the formula: 
BE 12.0080 
where F are the values corresponding to 9 — 0 whereas f are the 
thread intervals calculated in the column of differences in the ob- 
servation table. As the ultimate values of the F’s was used the mean 
of the single values (see p. 92). Of the thus determined F’s the 
thread intervals were again calculated for stars having a declination 
of 1°, 2°, 3°... as far as 40°, and from the table thus procured the 
