INCESSANT TRANSMISSION 9/19 



Other sciences need not follow suit. In biology especially "noise" 

 will seldom refer to this particular source; more commonly, the 

 "noise" in one system will be due to some other macroscopic system 

 from which the system under study cannot be completely isolated. 



Should the two (or more) messages be completely and simul- 

 taneously recoverable, by de-coding of the output, the concept of 

 noise is of little use. Chiefly it is wanted when the two messages 

 (one wanted, one unwanted) interact with some mutual destruction, 

 making the coding not fully reversible. To see this occur let us go 

 back to the fundamental processes. The irreversibility must mean 

 that the variety is not sustained (S.8/6), and that distinct elements at 

 the inputs-are represented at the output by one element. Consider 

 the case in which the input is a vector with two components, 



the first having possible values of ^, B or C 

 second ,, ,, ,, ,, E, F or G. 



,, .JV,WVHV» ,, 



Suppose the output is a variable that can take values 1,2, . . ., 9, 

 and that the coding was 



I 



If now the input message were the sequence BACBACAABB, 

 while the "noise" gave simultaneously the sequence G F F E E E G 

 F G E, then the output would be 



1, 4, 7, 2, 6, 3, 2, 4, 1, 2 



and the de-coding could give, for the first component, only the 

 approximation 



B, A, C, A or B, A, C, A or B, A, B, A or B. 



Thus the original message to this input has been "corrupted" by 

 "noise" at the other input. 



In this example the channel is quite capable of carrying the message 

 without ambiguity if the noise is suppressed by the second input 

 being held constant, at E say. For then the coding is one-one: 



.ABC 



^623 

 and reversible. 



It will be noticed that the interaction occurred because only eight 

 of the nine possible output states were used. By this permanent 

 restriction, the capacity of the channel was reduced. 



187 



