422 Dr. C. V. Burton on the Sun's Motion 



of available observations, is about 330; and e, the " probable "" 

 discrepancy between the observed and calculated time of an 

 eclipse, is estimated by Prof. Sampson at about 4*5 seconds. 

 The numerical values thus derived from (10) are 



p.e. in a = 44*0 kilom. per second -\ 

 „ ,,6 = 237 „ „ L . . (11) 



„ „ c= 10,000 „ „ J 



14. The determination of the velocity-component c (per- 

 pendicular to the plane of the ecliptic) is so badly conditioned 

 that the investigation can hardly be considered to afford 

 any light on that point. It may be admitted as probable 

 that the velocity of the solar system through the aether is 

 very far below 10,000 kilom. per second ; otherwise we 

 should have to suppose (for example) that practically all stars 

 whose radial velocities have been measured are moving- 

 through the aether with velocities of thousands of kilometres 

 per second. This certainly appears unlikely, though perhaps 

 the possibility ought not to be too lightly denied. An in- 

 direct argument against a very high velocity of the solar 

 system may be derived from other considerations. For we 

 know that the velocity components in the plane of the ecliptic 

 must be relatively moderate, otherwise there would be a 

 marked anomaly, having a period of about 12 years, in the- 

 observed times of the eclipses of Jupiter's satellites*, and 

 a priori it is unlikely that the velocity of the solar system 

 should be nearly in some arbitrarily assigned direction, such 

 as that perpendicular to the plane of the ecliptic. 



15. On the other hand, if we begin by admitting complete 

 ignorance regarding the velocity-components to be deter- 

 mined — save only our assumption that they are small com- 

 pared with the velocity of light — then 10,000 kilom. per 

 second will represent approximately the probable error in 

 the value of the component c ; and in general the determi- 

 nation of the component of velocity resolved in any direction 

 in space will be affected by a large probable error arising 

 from the uncertainty in the value of c. On this view, the 

 exact choice of coordinate axes becomes of importance, the 

 relatively large probable error in b given by (11) being due 

 to the influence of the error in c. So long as the axes 

 remain as specified in § 12, this nnduly large error in b can 

 only be avoided (if at all) by an excessive arbitrary weighting 

 of the observations, so designed as to secure the vanishing 



* It may be hoped that the investigation now proposed will indicate- 

 these components more definitely, or at least a superior limit to them. 



