473 
The refraction was calculated by the tables given by ALBRECHT; 
the declinations were taken from the Nautical Almanac, i.e. from 
Newcoms’s fundamental catalogue. From all the observations of the 
same evening one zenith point was deduced which satisfied all of 
them in the best manner comparing the yr and gy, ie. so that 
PNL — PNR + ySK-— PSL=%; 
the mean results are, however, independent of this. On July 19 and 
July 20 the observations were incomplete, but they supplement 
each other; from both days together follows as zenith point 4840, 
while on July 22°¢, when another series of observations was made 
with the same position of the circle, 49"18 was found (see preced- 
ing tables). 
Where in the observations in each position of the circle the mean 
zenith distance of the north and the south stars was the same, the 
influence of systematic division errors and of flexure becomes eliminated 
in the half sum } (py + gs) for each position, and in that case it 
is certainly best to regard the mean of the 6 results as our final 
result. My second series of observations answered very perfectly to 
these requirements. In that series the same 8 stars were observed 
each evening and the mean zenith distance was for the northern 
stars 30°.3, for the southern ones 30°,5. (R.N. 33°, R.S. 33°, LN. 
28°, L.S. 28°). In the first series the zenith distances differ more; 
the mean differences zy— zs lie for the various positions of the 
circle between — 15° and + 138° and their mean value is + 3° 
(the mean values of z themselves were zy— 54°, zg = 50°.5, mean 
z= 52°). But even in this case the method mentioned above will 
probably give the best results possible. We thus get: 
dst Series 2nd Series 
— 5° 16' 58"46 = Fa Pe eee 
59.58 59.96 
59.44 59.66 
GY easy 58.62 
57.87 58.40 
61.59 58.63 
Mean — 5° 16'59"08+0"60 =O Ee Oat 
The mean errors added to the final results were deduced from 
the mutual agreement of the 6 partial results. Of the latter them- 
selves the mean errors are +1"48 and +066. Judged by this, 
the 2nd series proves to be much more accurate than the 1st. This 
is undoubtedly partly due to the fact that in the 2ed series twe 
