376 



BELL SYSTEM TECHNICAL JOURNAL 



unity, we may alweiys expect a certain percentage of observations to 

 fall outside the limits even though the system of causes be constant. 

 In other words, the acceptance of this assumption gives us a right to 

 believe that there is an objective state of control within limits but 

 in itself it does not furnish the practical criterion for determining when 

 variations in quality, such as given in Fig. 3, should be left to chance. 

 Furthermore, we may say that mathematical statistics as such does 

 not give us the desired criterion. What does this situation mean in 

 plain every day engineering English? Simply this: such criteria, if 

 they exist, cannot be shown to exist by any theorizing alone, no matter 

 how well equipped the theorist is in respect to probability or statistical 

 theory. We see in this situation the long recognized dividing line 



0.5 



AVERAGE 



AVERAGE 



AVERAGE 



1923 - 1924 



1925 



YEAR 



1926 (9 months) 



Fig. 6 — Evidence of improvement in quality with approach to control. 



between theory and practice. The available statistical machinery 

 referred to by the magazine Nature is, as we might expect, not an 

 end in itself but merely a means to an end. In other words, the fact 

 that the criterion which we happen to use has a fine ancestry of high- 

 brow statistical theorems does not justify its use. Such justification 

 must come from empirical evidence that it works. As the practical 

 engineer might say, the proof of the pudding is in the eating. Let us 

 therefore look for the proof. 



3. Evidence that Criteria Exist for Detecting Assignable Causes 

 A . Fig. 6 shows the results of one of the first large scale experiments to 

 determine whether or not indications given by such a criterion applied to 

 quality measured in terms of fraction defective were justified by experi- 



