516 MR JAMES RUSSELL ON THE SUPERPOSITION OF MECHANICAL VIBRATIONS 



On 14th and 15th July last I found this assumption to be absolutely correct. 

 Oscillations (produced in the usual way by means of two small Leyden jars fed by 

 induction coil, spark gap, and loosely coupled oscillation transformer) could either be 

 passed directly through a long iron wire, '041 cms. diameter, or through a coil of one layer 

 surrounding it. In both cases the effects of these oscillations, when superposed under 

 the A and also under the B conditions, were essentially the same as the effects of purely 

 mechanical vibrations upon magnetisation. Any of the corresponding curves of figs. I, 

 and II. (present paper) show the results obtained with sufficient accuracy. 



I should like, however, to point out an essential particular in which Dr Eccles'' 

 experimental methods differ from my own. He states that " the effects of the spark 

 are recorded as if the observations had been taken only on the ascending half of the 

 hysteresis curve ; the figures given being, in fact, the means of the measured effects at 

 points symmetrical with regard to the origin on the ascending and descending branches." 

 Consider, say, the 100 cycle of Table I. {Phil Mag., August 1906, p. 113). When the 

 field is +100 the effect of the spark is + 1 "1 8. But this result has been obtained by 

 taking the means at both cyclic extremes ; hence, when the field is — 100, the spark effect 

 is — 1*18. No readings are given in the third column when field is — 75, but when it 

 is reduced to — 50 the effect of the spark is + - 27. The experiments, therefore, do not 

 show how the transition from — 1*18 at the negative cyclic extreme (field —100) to 

 + 0*27 when the field is reduced to — 50 has taken place. This 100 cycle is typical of 

 the other three cycles given. 



Consequently, in the series of figures plotted from Table I. the curves measuring the 

 effect of the spark on the ascending half of the hysteresis curve begin in the second and 

 finish in the first quadrant, when in reality they ought to be represented by a 

 continuous curve beginning in the third and finishing in the first quadrant. 



Dr Eccles states that he found it possible " to get over and over again practically 

 the same magnetometer deflection for every spark, provided the effect of previous 

 oscillations was wiped out by taking the iron through a cycle.'" The fact that the 

 observations could be repeated does not prove that the effect of previous oscillations 

 had been wiped out. The cycle so obtained is not symmetrical about the origin, its 

 want of symmetry being determined by the "set" or permanent deformation given 

 to the magnetisation by the preceding oscillations. This is in accordance with the 

 known facts of magnetisation. Moreover, I found, previous to determining my own 

 experimental methods under the A conditions, that repeated reversals of the cyclic 

 field were not sufficient to wipe out the effect of immediately preceding oscillations, 

 and that to do so the magnetic metal must be demagnetised by decreasing reversals 

 after each superposition of an electric oscillation or of a mechanical vibration. 



One of the conclusions arrived at is as follows : — " It is evident, moreover, that the 

 magnitude of the effect at any point is closely connected with the slope of the 

 hysteresis curve." It is quite true that the maximum spark effect is closely connected 

 with that point of the loop where the slope is greatest, but the occurrence of a neutral 



