350 BELL SYSTEM TECHNICAL JOURNAL 



When a single circuit is to be investigated it is usual to make the 

 comparison circuit a standard reference circuit, the characteristics of 

 which are well known, and to which the data would naturally be re- 

 ferred. In a series of tests it is customary to choose one of the test 

 conditions of the series as the circuit with which the other test condi- 

 tions are compared. In a large program, which may include many 

 series of tests, a circuit condition from each series can be chosen as 

 temporary reference conditions to which the data from the various 

 series are referred. These circuits may then be related to each other 

 and also to a standard reference circuit by direct comparisons, provid- 

 ing base points for whatever residual corrections are needed to reduce 

 the results of the whole program to a common basis of crew skill; just 

 as in a triangulation survey of land certain base points are especially 

 well determined for use in the final adjustment of the whole survey. 



The substantial advantages of interleaving the tests on pairs of 

 circuits are indicated by an analysis of the results from more than 100 

 tests, representing from 2 to 5 comparisons on 42 different sets of 

 conditions. This analysis showed a reduction from 1.3 to 0.8 per cent 

 sound articulation in the average deviation of the differences between 

 two conditions when they were tested simultaneously as compared 

 with the average deviation of the differences obtained in an equal 

 amount of time from non-simultaneous comparisons. Since the aver- 

 age deviation is inversely proportional to the square root of the amount 

 of data obtained, it would have required from two to three times as 

 much testing to attain the same precision with non-simultaneous 

 testing. 



Caller's Control Circuit 



One of the principal variables in articulation testing is the intensity 

 with which the caller speaks. With practice a caller can learn to pre- 

 serve a moderately steady intensity throughout a list, but it is unlikely 

 to be the same intensity from day to day, and it is fairly certain to differ 

 from the intensities of the other seven callers. Some sort of control is 

 necessary. 



Control and measurement of the caller's intensity by means of 

 measuring instruments located in the circuit under test have obvious 

 disadvantages. A primary difficulty is that of specifying and com- 

 paring such measurements on circuits having characteristics vary- 

 ing with frequency in different ways. When, under such circum- 

 stances, are two intensities to be called equal? There is the secondary 

 difficulty that with carbon button transmitters variations in the repro- 

 duction efficiency of the button occur, and it is desirable to be able to 

 follow these independently of variations in the speaker's intensity. 



