ULTRASTABILITY IN THE LIVING ORGANISM 9/4 



attempt that it causes the pulse-rate to increase. He will almost 

 certainly ask himself 4 What has changed ? ' 



Such facts provide valid evidence that some variable has 

 changed value. I need not elaborate the logic for no experi- 

 menter would question it. What has been sometimes overlooked 

 though, is that we are also entitled to draw the deduction that 

 the variable, being as it is an effective factor towards the system, 

 must, throughout the previous day, have remained constant ; 

 for otherwise the reactions observed during the day could not 

 have been regular. For the same reason, it must also have been 

 constant throughout the next morning. And further, the two 

 constant values cannot have been equal, for then the hearts' 

 behaviours would not have been changed. Assembling these 

 inferences, we deduce that the variable must have behaved as 

 a step-function. Exactly the same argument, applied to the 

 changes of behaviour shown by Jennings' Stentor, leads to the 

 deduction that within the organism there must have been vari- 

 ables behaving as step-functions. 



Is there any escape from this conclusion ? It rests primarily 

 on the simple thesis that a determinate system does not, if started 

 from identical states, do one thing on one day and something 

 else on another day. There seems to be no escape if we assume 

 that the systems we are discussing are determinate. Suppose, 

 then, that we abandon the assumption of determinism and allow 

 indeterminism of atomic type to affect heart, Stentor, or brain 

 to an observable extent. This would allow us to explain the 

 ' causeless ' overnight change ; but then we would be unable to 

 explain the regularity throughout the previous day and the next 

 morning. It seems there is no escape that way. Again, we 

 could, with a little ingenuity, construct a hypothesis that the 

 pharmacologist's experiment was affected by a small group of 

 variables, whose joint action produced the observed result but 

 not one of which was a step-function ; and it might be claimed 

 that the theorem had been shown -false. But this is really no 

 exception, for we are not concerned with what variables ' are ' 

 but with how they behave, and in particular with how they 

 behave towards the system in question. If a group of variables 

 behaves towards the system as a step-function, then it is a step- 

 function ; for the ' step-function ' is defined primarily as a form 

 of behaviour, not as a thing. 



Ill 



