456 Messrs. Rowland and Hutchinson on the 



diminish the deflexion, but did not bring the needle back to 

 zero as the glasses did. On looking at the figure (Plate IX. 

 fig. 1) it will be seen that there is a brass plug (I) closing the 

 bottom of the tube in which the needle is placed. The rapid 

 rotation of the disks caused a very appreciable exhaustion at 

 the centre, and consequently a steady stream of air was sucked 

 down the tube through the open mouthpiece, and out through 

 the imperfect connexion of the plug. Air-currents were not 

 at first suspected, as the deflexion was so very steady. The 

 brass plate used was smaller than the glass, and hence did not 

 completely shield the tube. 



After the brass back-pieces (H H) had been taken out, and 

 a hard rubber substituted, it was found that with one direction 

 of rotation the needle was extremely unsteady ; it would I'un 

 up the scale for several centimetres, stop suddenly, &c. — 

 evidently a forced vibration. This was traced to air-currents 

 also. Now, the air blew into the open mouth of the cone. 

 The apparatus had been run for some months with this open, 

 and not the slightest irregularity had been seen. But the 

 hard rubber pieces were very much larger than the brass ones 

 which were removed ; they filled up the lower space to a 

 greater extent, and deflected the air upwards more than 

 before, causing the unsteadiness. With the opposite rotation 

 the air was thrown down instead of up, and consequently 

 did not aftect the needle. 



The first systematic observations were made in January 

 1889, with the disks charged and plates earthed. The de- 

 flexion on reversing was got without difficulty, and it was in 

 the direction to be expected ; that is, with positive electrifica- 

 tion, the effect was equivalent to a current in the direction of 

 motion of the disk. A number of series were taken in the 

 next two months ; they agreed among themselves well 

 enough, but did not follow the law assumed. The deviation 

 can best be explained in this way : — The equations above show 



N D N 

 that for a fixed position of the disks A cc o- — — oc If then, N 

 ^ pep 



and ^ being constant, the condenser plates are moved up to 



the disk, step by step, thus varying e, and D be changed at 



the same time so as to keep Tf/e Qc cr, a constant, the deflexions 



should be constant. 



Such was not found to be the case ; the deflexions were 



directly proportioned to e instead of being constant : that 



is, with greater difference of potential, the deflexions were 



greater, although the surface-density remained constant. 



Finally this was found to be due to a charge on the back 



surface of the gold coating. The end of the axle comes 



