THE BLACK BOX 6/18 



(2) Carbon, hydrogen and oxygen are all practically tasteless, 

 yet the particular compound "sugar" has a characteristic taste 

 possessed by none of them. 



(3) The twenty (or so) amino-acids in a bacterium have none of 

 them the property of being "self-reproducing", yet the whole, with 

 some other substances, has this property. 



If these examples are compared in detail with the processes of 

 study and coupling of Black Boxes, it is soon seen that the examples 

 postulate much less knowledge of their parts than is postulated of 

 the Black Boxes. Thus the prediction in regard to ammonia and 

 hydrogen chloride is based on no more knowledge of each substance 

 than that it is a gas. Similarly, of the twenty amino-acids all that 

 is asked is "is it self-reproducing?" Were each amino-acid treated 

 as a Black Box the examination would be far more searching. The 

 input to a molecule is the set of electrical and mechanical forces, in 

 all distributions and combinations, that can affect it; and its output 

 is the set of all states, electrical and mechanical, that it can be in. 

 Were this complete knowledge available, then the method of S.4/8 

 shows how the behaviour of many coupled amino-acids could be 

 predicted; and among the predicted behaviours would be that of 

 self-reproduction of the whole. 



It will be seen that prediction of the whole's behaviour can be 

 based on complete or on incomplete knowledge of the parts. If 

 the knowledge is complete, then the case is that of the Black Box 

 whose canonical representation is known, the inputs or circumstances 

 being all those that may be given by the other Boxes to which it is 

 to be coupled. When the knowledge of the parts is so complete, 

 the prediction can also be complete, and no extra properties can 

 emerge. 



Often, however, the knowledge is not, for whatever reason, 

 complete. Then the prediction has to be undertaken on incomplete 

 knowledge, and may prove mistaken. Sometimes all that is known 

 of the parts is that every one has a certain characteristic. There 

 may be no better way of predicting than to use simple extrapolation 

 ■ — to predict that the whole will have it. Sometimes this proves 

 justified ; thus, if a whole is of three parts, each of pure copper, then 

 we shall be correct if we predict that the whole is of pure copper. 

 But often the method fails, and a new property can, if we please, 

 be said to "emerge". 



It does in fact very commonly happen that when the system be- 

 comes large, so that the range of size from part to whole is very 

 large, the properties of the whole are very different from those of 



111 



