Effect of Electrical Convection* 195 



pended by silk threads. From the two poles of the Voss 

 machine ran two wires to two terminals of the reversing 

 switch S. From the other two terminals ran two wires to the 

 reverser. The brushes of the reverser could be arranged so 

 that at any instant the two disks could be charged either 

 alike or oppositely, as desired. 



If the disks are set rotating in the same direction, the 

 commutator and Yoss machine also set rotating so as to charge 

 the disks at anv instant with the same sign, then on throwing 

 the galvanometer in series with the coil, the spot of light on 

 the galvanometer-scale should move off a distance D from its 

 position on open circuit. If the order of charging with respect 

 to the commutation is reversed, as can be done by reversing 

 the high potential switch S, then the spot of light should 

 move off a distance D on the other side of its position on open 

 circuit. Hence, on reversing the high potential switch S, 

 keeping everything else unchanged, the spot of light should 

 change its position by an amount 2D. This was the deflexion 

 always observed. This deflexion varied from 30 to 80 mm., 

 depending upon the conditions. Its direction always reversed 

 with the direction of rotation of the disks, and was always 

 such as should be expected from Ampere's rule. This latter 

 was determined by comparison with the deflexion produced 

 by a conduction-current, reversed in the same manner as 

 the charge on the disks, flowing in the test-coil T on the disk 

 apparatus. 



Under the conditions first employed, the deflexion 2D 

 expected was therefore 20 mm. A deflexion, apparently of 

 this order of magnitude, was first unmistakably observed as 

 early as December 22. At this time, however, the accidental 

 deflexions, due to extraneous effects not then eliminated, were 

 CDnsiderably greater than this, so that it was impossible to 

 make quantitative measurements. However, there was un- 

 mistakably a deflexion obeying the reversing of the switch, 

 and also reversing with the direction of rotation of the disks. 

 On January the lGth and 17th, the conditions were so far 

 improved that a series of quantitative measurements was 

 attempted. The mean of a number of determinations of 2D, 

 with the disks running positively, was 18*4, with the disks 

 running negatively 25*0. The direction in the two cases 

 reversed, and was as should be expected from Ampere's rule. 



The time from the middle of January to the middle of 

 March was spent in making various improvements in different 

 parts of the apparatus. There was never any difficulty in 

 observing the effect sought, though it was onlv gradually 



2 



