Effect of Electrical Convection. 203 



and Voss machine were then started, and a test made to see 

 that there was no effect when the disks were at rest. The 

 disks were then set in rotation, and when they had obtained 

 a constant speed, a second observer noted the speed of each, 

 and also of the commutator, while the first observer measured 

 the potential of the disks, with the switch S first thrown 

 north, and then south. The deflexion produced by the con- 

 vection-current was then determined by the first observer as 

 follows : — With the switch S thrown north, the galvanometer 

 was thrown in circuit, and a reading taken. The switch was 

 then thrown south, a second reading taken, then north again, 

 and a third reading taken. The difference between the second 

 and the mean of the first and third readings was taken as the 

 value of 2D. The switch was then thrown south, then north, 

 then south, a reading taken each time. Again the difference 

 between the second and the mean of the first and third was 

 taken. In this way variations in the zero were largely 

 eliminated. The above operation was repeated twice. The 

 speeds and potentials were then determined, a similar set of 

 readings taken, and the speeds and potentials once more 

 determined. This usually required the greater part of one of 

 the forty minute periods during which the galvanometer 

 could be used. During the next period the direction of 

 rotation of the disks was reversed, and a similar set of read- 

 ings made. It is hardly necessary to state that with appar- 

 atus as complicated as that used, accidents were frequent, and 

 often a whole night was spent without obtaining a satisfactory 

 set of readings. 



There are two ways of employing the disk-apparatus to 

 put in evidence the effect sought. The two disks may be 

 rotated in the same direction, and, at any instant, charged 

 alike ; or, they may be rotated in opposite directions, and, 

 at any instant, charged oppositely. Both methods have been 

 employed, but the majority of readings were taken with the 

 disks charged oppositely and rotating in opposite directions 

 as the needle of the galvanometer was much more steady 

 under these conditions. 



Table III. gives an example of a complete set of readings, 

 taken under these latter conditions. The first four columns 

 give the readings for the sensibility, the next three the read- 

 ings for the potentials, the next eight the readings for the 

 deflexion, and the last three the speeds. S is the sensibilitv 

 of the galvanometer, M the mass on the scale-pan of the 

 electrometer, 8 and B' the distance apart of the electrometer- 

 plates, less *26 cm., the zero-correction of the scale of the 

 electrometer, when the disks are charged positively and 



