120 PROCEEDINGS OP THE AMERICAN ACADEMY 



or endure. This cause also would tend to make the variation range 

 equally above and below the calculated value as the sparking distance 

 increased or diminished. 



Another explanation may be sought in the varying ohmic resistance 

 of the path of the spark, although this explanation is inadequate to 

 explain the whole efFect. In order to test it, a long (1.3 cm.) and 

 short (.4 cm.) spark were taken from a glass Leyden jar (see Fig- 

 ure 7). Not only could no appreciable difference between the two 

 plates be detected, but there was no variation in the time of successive 

 oscillations. 



In regard to the measurement of the negatives on the dividing 

 engine, the following points may be worthy of mention. At the time 

 the measurements were made, it was expected that the sparks from the 

 glass condenser would show the variations, and that the air condenser 

 would give the constant and theoretical period of oscillation. The 

 reverse of this appeared when the results of the measurements were 

 reduced. Moreovei", the measurements were made by a run of the 

 dividing engine from one end of the negative to the other ; so that if 

 an error was made in the setting of the cross-hair on the image of 

 one discharge, — for example, making the measurement of that oscil- 

 lation large, — a corresponding amount would be deducted from the 

 measurement of the next oscillation. The result of this would be, that 

 if the apparent variations were due to errors in measurement, the 

 periods of discharge would be alternately large and small, or at least 

 as often and as far below the theoretical value as above it. But this 

 is conspicuously not the case. 



A consideration of the curves which represent our results shows that 

 with quick oscillations which result from the employment of a small 

 air condenser, the air dielectric did not have time to recover completely, 

 in the time of one oscillation, from the strain to which it was subjected. 

 With the larger air condenser, the oscillations being slower, more time 

 was given for this recovery, and hence the periodicity which we have 

 discovered was not so marked. It seems, therefore, that not only 

 should an electrical resonator be turned for capacity and self-inductiou, 

 but also for a certain periodicity of strain in the dielectric. 



In the case of glass, we should not expect to obtain evidence of this 

 periodical recovery from a quick strain, since it is well known that the 

 recovery from strain is so slow that the discharge from a Leyden jar 

 is incomplete after a discharge lasting a second. The curve we give 

 for glass (Fig. 7) shows that this periodical recovery is too slow to 

 manifest itself during the time of quick oscillation. 



