APPLICATION OF ARTICULATION TESTING 349 



The results obtained by a single caller-observer pair, even when 

 well-trained, are far from constant in successive tests on the same 

 circuit. These fluctuations may be ascribed in part to the smallness 

 of the sample of speech contained in one list, in part to the inability 

 of a caller to repeat speech sounds uniformly and in part to variations 

 in the concentration exercised by the observer. When a large crew is 

 used the effects of such fluctuations among the individual pairs tend to 

 neutralize each other more than with a small crew, so that the addi- 

 tional time consumed by using eight callers is offset by the fact that 

 with fewer callers more tests must be made to reach the same precision. 

 At the same time the use of eight callers provides a more representative 

 result. However, even with eight callers and four observers the effects 

 of the individual fluctuations are still of such importance that it is the 

 regular practice to repeat each test in a program at least once. 



Interleaved Tests on Pairs of Circuits 

 In addition to repetition of the tests on each circuit, a further control 

 over the fluctuations in crew skill is provided by the practice of testing 

 circuits not singly, but in pairs. That is, instead of completing a test 

 on one circuit before proceeding to another, two circuits are tested 

 almost simultaneously by interleaving the two tests. The first caller 

 reads a list of testing syllables on one circuit and follows it immediately 

 by reading a list on the comparison circuit. The second caller then 

 reads a list on the comparison circuit and follows it immediately by a 

 list on the first circuit. This procedure is followed by the other six 

 callers. Thus when the eighth caller has ended his second list, two 

 single tests have been made. Such a pair of interleaved tests on two 

 circuits is called, for convenience, a double test. By suitably arranging 

 the schedule for the callers and observers the effects of erratic fluctua- 

 tions can be neutralized to a large degree. 



The testing equipment and the circuit elements are arranged so as to 

 facilitate this method of carrying out the tests. A single master key, 

 controlling a switching system, is provided so that the change from one 

 circuit to the comparison circuit may be made almost instantaneously. 

 The two circuits may be different throughout, even to the magnitudes 

 of noise under which they are tested, and the intensity used by the 

 caller; they may differ in two or three elements, such as transmitters, 

 types of set, and length of trunk; or the difference may be in a single 

 element, such as trunk length, or in one of the operating conditions, 

 such as the magnitude of line noise. The data recording apparatus 

 types out with the analyzed data whether the list has been called on 

 "Condition A" or "Condition B." 



