( 716 ) 



which has now become a chissic in aslroiiomy, was based on obser- 

 vations of' satellites, on perturbatioJis in the motion of comets, and 

 of the planets Themis, Polyhymnia and Saturn, it seems to me 

 advisable to retain of these only the determinations from the three 

 planets. Of the older observations of the satellites the uncertainty of 

 the scale-value (which is increased threefold in the mass of the planet) 

 is such that their weight, compared with the modern observations, 

 and with the determinations from the perturbations of planets, is 

 absolutely negligible. Nkwcomb has also, for this same reason, assigned 

 a very small weight to these observations of the satellites. 



The use of observations of comets seems to me very dangerous. 

 It is very uncertain, if not improbable, that the observed centre of 

 light should retain the same relative position with respect to the centre 

 of gravity throughout one apparition of (he comet, and a fortiori 

 in diiferent apparitions. Newcomb also points out that the results based 

 on observations of comets are unreliable for this reason. Nevertheless 

 he assigns a large weight to the determination by vOxN Haerdtl from 

 Winnecke's comet, on the ground that the normal places of this 

 comet are so well represented by von Haerdtl's results. It appears 

 to me that this good representation does not diminish the stringency 

 of the argument stated above, and in my opinion it is advisable to 

 i-eject also this determination, together with those from other comets. 



There remain the determinations from the three planets, which I 

 adopt with the same weights assigned to them by Newcomb, and the 

 modern observations of satellites, which were only made, or at least 

 reduced, after Newcomb's discussion was published. For the^e latter 

 the scale-value is determined in an entirely satisfactory manner by 

 simultaneous observations of standard stars. Nevertheless I have 

 assigned to these observations a relatively smaller weight than to 

 the determinations from the planets, to allow for the possibility 

 of small systematic errors in transferring the scale- value from the 

 distance of the standard stars to the mutual distances of the satellites. 



In my reduction of Gill's observations of i89J 1 have included 

 in the probable error of ^}^ the effect of the uncertainty of the stan- 

 dard stars used for the determination of the scale-value. The probable 

 errors stated by Oookson do not include this uncertainty. The distances 

 of the stars used by Cookson are not so accurately known as of the 

 stars used in 1891. 1 have for these reasons assigned a smaller weight 

 to Cookson's two determinations than to (jills. The several determi- 

 nations and their probable errors and adopted weights are given in 

 Table YI. 



