404 ANNUAL REPORT SMITHSONIAN INSTITUTION, 19 63 



We have confined discussion thus far to examples of natural and 

 artificial helmsmanship — control under the guidance of information. 

 This is what the term "cybernetics" was coined to denote, in contra- 

 distinction to the older method of physical constraint typified by our 

 railway example. Unfortunately the popularity of the term has led 

 to its being used in some quarters to include situations of the opposite 

 sort — those where no element of information-guided selection is pres- 

 ent, but (as with the railway) the necessary corrective forces depend 

 wholly on the inertial reaction of the situation where they are required. 

 Thus a pendulum, or a ball at the bottom of a bowl, is described by 

 some writers as a cybernetic system under "feedback," because it 

 automatically suffers a force opposed to its displacement from equi- 

 librium. Even such lowly forms as the reaction of a floor to the 

 weight of an object placed on it cannot then be excluded. It is gi^ad- 

 ually being realized that such usage trivializes the term ; but for the 

 present at least the reader must expect to find it sometimes used in 

 confusing and even contradictory senses. 



Behind the confusion there is in fact an interesting ambiguity in 

 the cybernetic approach — in the very nature of control itself. The 

 cybernetician seeking to understand a complex system begins by trying 

 to discern the "pattern of subordination" — asking "which controls 

 what?" The difficulty is to define "control." Everyone would agree 

 that a watercock controls the flow of water into a cistern ; but if two 

 cisterns are connected together, so that the level in each affects the level 

 in the other, can we say that one controls the other? However we 

 decide, there is obviously an important difference between the two 

 cases. The second shows no more than a passive tendency to equi- 

 librium between action and reaction ; but the first includes an "active" 

 element (the cock plus water supply) whose control lever determines 

 the flow of water without itself suffering appreciable reaction. 



Suppose now that we link the cock lever to a ball floating on the 

 water, as in the familiar ball-cock. What now controls the level of 

 the water ? In one sense, we may still say "the cock" ; but in another 

 and more important sense, it is not now the cock per se, but its height 

 above tank bottom, which (other things being equal) governs the 

 final water level. If we had means of raising and lowering the whole 

 watercock-and-ball assembly, this would enable us to select the water 

 level at which the cock would turn itself off. 



What this example makes clear, I think, is the subtle and arbitrary 

 human element that underlies many cybernetic notions. Basically, by 

 saying that A controls B we mean that if ive could control A then we 

 could control B. In a strictly physical sense, divorced from the human 

 notion of "purpose," the notion of control tends to be ambiguous or 

 meaningless. The only objective physical distinction we can firmly 

 draw is between (1) devices, such as watercocks, steamvalves, transis- 



