435 



of declination does not vanish in passing from the northern to the 

 southern magnetic hemisphere, but only changes signs at the equi- 

 noxes. I think every physicist will agree with you, that no thermic 

 hypothesis will be able to explain this annual variation, but as you 

 say (Toronto, ii. p. xix), it is ' obviously connected with, and de- 

 pendent on, the earth's position in its orbit relatively to the sun, 

 around which it revolves, as the diurnal variation is connected with 

 the rotation of the earth on its axis.' But you have given no hint 

 how this different position in its orbit can affect the magnetic condi- 

 tion of the earth, except so far as you suppose that the excentricity 

 of the orbit is the reason that the total magnetic force is about y^-a 

 greater at the perihelion than at the aphelion (page xciii) ; but even 

 granted that this variation is the effect of the excentricity, it cannot 

 be the cause of the annual variation of the declination, as this is of 

 contrary signs in the two semiannual periods. 



" I have thought that this annual variation might possibly be ex- 

 plained by the following considerations, which I (although with 

 extreme diffidence) shall venture to lay before you. 



" As the recent magnetic observations have proved without doubt 

 the direct magnetic action of the sun, or that the sun itself is a 

 magnet, the sun must accordingly have magnetic polarity or mag- 

 netic poles. Now in our ignorance of the causes of the magnetic 

 condition of the heavenly bodies, I think it reasonable to connect it 

 in some way with their rotation on their axes, and to assume that 

 generally their magnetic axis will nearly coincide with the axis of 

 rotation ; at all events, if these do not coincide, but include a small 

 acute angle, the sum of the magnetic influences on a distant mag- 

 netic body during a whole rotation, will be nearly the same as if the 

 magnetic poles were placed in the axis of rotation. If we suppose 

 a magnet E revolving about another S, the magnetic axis of E re- 

 maining parallel with itself, but not parallel with the axis of S (as 

 the earth around the sun), then the magnetic induction of S on E 

 will depend on the relative position of both magnetic axes. More- 

 over, if we only regard the mean of the magnetic induction of the sun 

 on the earth during several rotations of both on their axes, we may 

 approximately assume that the magnetic axis of both coincides with 

 the axis of rotation, and compare their relative position during a 

 whole revolution or annual period, with the magnetic variation in 



