(14) 
These errors are also represented in the lower diagram and con- 
nected by interrupted lines. Especially the last two positions before 
the periastron are now badly represented, a fact not to be wondered 
at, considering the large amount of the corrections of the elements. 
Although these positions have but the weights 0.6 and 0.3 I have 
yet proceeded to a second approximation. For the new 2nd members 
of the normal equations [ found : 
+0.32052 +0.66856 -+0.51950 -+0.20554 +0.30590 +0.37168. 
After a new solution of the normal equations «and 7 were again 
determined as above; the system of elements obtained is: 
| T— 1894.0900 i= 46° 19 
System IL { «= —71°.37069 St —= 4430.2 (1900.0) 
e = 0.5875 A= 212 BA 
The deviations left by this system in the normal positions are as 
follows. They have been connected by dotted lines in the diagram. 
1 : —0°.203 7: —0°.4383 13 : —0°.521 19 : —0°.300 
2: +-0 .209 8 : +0 .034 14 : —0 .205 20 : +0 .158 
8 : +0 .082 9: +0 .182 15 : —0 .218 21 : —0 .087 
4: —0.250 10: —0.032 16 : —0 .925 
5: —0.455 11: —0 161 17 : 40.778 
6 : —0.380 12: —0 .420 18 : +0 .098 
The outstanding errors are unimportant, but a certain regu- 
larity is unmistakable. The characteristic curvature in the original 
curve of errors before the periastron, is found back all but unchanged 
in the diagram of systems IP and II. The cause may be sought in 
a perturbation by a third (invisible) member of the system ; the suppo- 
sition however that not entirely eliminated personal errors have been 
at work seems to me more plausible. A third possibility remains : 
the not perfect accuracy of the coefficients of the equations of errors 
in the 2nd approximation might be the cause. Strictly speaking these 
ought to have been recalculated with the elements of system IP. 
7 
