1902.] affecting the Transmission of Electric Waves. 255 



any way detracting from their value on account of the main object in 

 view being at the time of a different nature. 



One of the great difficulties in accurately comparing results, separated 

 by even the shortest intervals of time in wireless telegraphy, is, un- 

 doubtedly, in having to depend on an imperfect electrical contact for 

 the comparison, as even the best coherers, as they are now generally 

 termed, cannot be depended upon to give absolutely similar consecu- 

 tive indications, under similar conditions of excitation ; and in the 

 filings coherer, for instance, which I have chiefly used, the probability is 

 infinitesimally small that the filings will rearrange themselves identi- 

 cally, in any two successive signals, taking into account the shaking 

 to which they are subjected immediately following the reception of an 

 excitation ; though the practical results obtained with those of good 

 manufacture might almost lead one to an opposite opinion, owing to 

 the accuracy with which signals are generally recorded. Careful 

 observation over an extended period has led me to the conclusion 

 that the error arising from this cause in the experiments described 

 does not exceed 5 per cent., i.e., in any one hundred consecutive signs 

 in a series of signals, ninety-five of them will be accurately recorded 

 at the distance where the excitation is nearing the limits at which 

 it will affect the coherer. At shorter distances, with moderately 

 strong excitations, the error is much less, probably less than 1 per 

 cent. Experiments conducted under favourable circumstances, with 

 the object of obtaining the percentage of accuracy, have frequently 

 confirmed this. The method used to carry this out was to select a 

 long signal whose details were known, and count the errors in the 

 number of signs composing the signal. 



Though a possible error of 5 per cent, in any experiment may, in 

 systematic comparative trials, or quantitative analysis, appear large, 

 this error may be considered to be eliminated when the mean of a 

 considerable number of excitations is taken, and the results judged 

 by the general effect, which was whether or no excitations were received 

 at a given distance for a considerable period of time. For instance : 

 the experiments, as a rule, consisted in one station sending signals 

 consecutively for (say) 2 minutes, at the rate of 100 signs a minute. 

 If these signals were received correctly, or nearly so, the receiving 

 station was considered to be within the limits at which signals could 

 be received under those circumstances ; if only 50 to 90 per cent, of the 

 signs were received, it was considered at the extreme limit of signalling 

 distance ; if 20 to 50 per cent, were received, it was considered to be just 

 outside the limits ; if less than 20 per cent., or, as frequently occurred, 

 no signs were received, it was considered to be beyond the limits of 

 signalling. In the two latter cases, careful inquiry and examination of 

 the records of the transmitting station were always made to prove 

 that the signals had actually been sent at that time. At the con- 



