( 696 ) 



This filial result may therefore he assumed to represent the residual 

 influence of the personal errors on the p.m. 



For the value n^ of this mean T find 



7, = — O.-'OOO 4 



In deriving this result the hours with many nebulae did not get 

 any greater w^eight than the hours with only a few objects. Owing 

 to this cause the final weight is found to be only 0.4 of what it 

 would have been had the distribution been uniform. 



We shall get a result of appreciably greater weight if in the first 

 place we combine by twos the hours lying symmetrically in respect 

 to the apex. In these mean values the parallactic motion is already 

 eliminated ; we may therefore further combine the twelve partial 

 results having regard to their individual weights. 



In this way I find 



Ja = -\- O.-^OOOG. 



It thus appears that Mönnichmeyer has succeeded remarkably well 

 in getting rid of the influence of the personal errors. 



As mentioned just now these errors appear still further diminished 

 in the result for the parallactic motion. 



There thus seems to be ample reason for neglecting any further 

 consideration of them. In order to enable the reader to get at once 

 a pretty good insight in the accuracy really obtained, I have divided 

 the whole of the material not only into the three classes [of 

 Mönnichmeyer, but I have subdivided each of them into a certain 

 number of sections, each of about the same weight. 



I thus got the following summary. (See p. 697). 



The values of t have been included in the table merely in order 

 to show that in them too no traces of any personal error are visible. 



In order to get the yearly parallaxes tt, 1 have divided the secular 



parallaxes - by 4.20 ; this number being, according to Campbei,l's 



Q 

 determination, the number of solar distances covered by the solar 



system in a year in its motion through space. 



The probable errors were derived in the hypothesis that the com- 

 ponent V is wholly due to errors of observation. 



If we compute the probable error of one of our 13 results from 

 their internal agreement we get 0."023. This number ditfers very 

 little from the values directly found. Here again we have an indication 

 that systematic errors must be small. 



The last row of numbers contains the simple aA-erages of the 13 

 individual results. 



