444 



Prof. J. A. Ewing. 



[Feb. 10, 



Bat tlie actual amount of frictional error, in the first somewhat 

 roughly made instrument which has been set up, is so small as hardly 

 to affect the records to a sensible extent. To test it the following ex- 

 periment was made : — The whole apparatus, consisting of the plate /, 

 the clockwork to drive it, and the two levers, was put on a shaky 

 table, whose movements could be made to imitate those of an earth- 

 quake. The pointers were placed side by side instead of in their usual 

 positions, and the bob of one of them was held fast by a bracket from a 

 neighbouring wall. It was, therefore, perfectly steady during the 

 artificial earthquakes made by shaking the table. Two records were 

 then obtained upon the plate, by shaking the table transversely to the 

 levers : one was the true motion, known to be true because the bob of 

 that lever was held fixed during the test ; the other was the motion 

 as recorded by the other lever in the usual way. Both motions were 

 magnified in the same ratio of six to one. If the bob of the second 

 lever remained stationary (as it should have done except for friction), 

 the two records would have agreed exactly, and if not, then their 

 difference would show the extent and nature of the frictional error. 



Fig. 3. 



A x A 2 records produced by lever with fixed bob. 

 B x B 2 records produced by lever with free bob. 



The result was satisfactory beyond expectation. Figs. 3 and 4, 

 which are taken from a photograph of the plate, show the portions of 

 the pair of records produced by this method. The curves A x in fig. 3, 

 and A 3 in fig. 4, show the true motions as recorded by the lever whose 

 bob was held fixed from outside. B x and B 3 are the corresponding 

 records given by the other lever whose bob was free. It will be ob- 

 served that the agreement is very close. Recent observations have 

 shown that the parts of the curves in figs. 3 and 4, which are of rela- 

 tively small amplitude, resemble most closely the curves produced by 

 actual earthquakes. In these parts the divergence between the two 

 records is scarcely perceptible. 



The friction of the marking points on the plate gave some trouble 



