828 BELL SYSTEM TECHNICAL JOURNAL 



{(-) Carbon Transmitter Circuit (see Hg. 7). — ^The average values for 

 five transmitters are L = 93 per cent, 5" = 81 per cent. In these 

 tests the sensation level of the received speech was 75 db, and the 

 calling level as measured by a volume indicator bridged across the 

 line side of the input repeating coil was — 12.5 db. 



{d) Master Reference System Plus Filters. — System set for a sen- 

 sation level of 70 db. 



3,750" Low Pass Filter L = 96.7% 6" = 91.0% 



750" High " " L = 96.7% 5=91.0% 



1,500" Low " " L = 77.7% 5=49.5% 



1,500' High " " L = 91.0% 5=76.0% 



The foregoing discussion has been concerned with methods of 

 correcting the articulation results obtained by a given crew at different 

 times to an arbitrary stage of practice or experience. To do this it is 

 necessary to calibrate the crew for types of distortion that are similar 

 to those of the systems for which the corrections are desired. The 

 method has been described in detail because there are times when it 

 is necessary to make such corrections. However, it has been our 

 experience that such practice effects become negligible with a crew 

 that has been set up in accordance with the methods previously 

 described, when the crew's experience with types of distortion is 

 diversified and when unusual circuits are tested successively until no 

 further increase in articulation with practice occurs. 



These methods may also be used to correlate the articulation data 

 of various crews and various techniques, provided that the only 

 essential difference between the crews and techniques is in the demand 

 or burden that is placed upon the observer. This means that the 

 crews must have similar vocal characteristics and similar hearing 

 abilities, and that the testing lists must have similar speech sound 

 content. It has been found, for example, that a crew of women 

 callers obtain a considerably higher articulation than men callers on a 

 circuit which eliminates all frequencies below 1,500 cycles and vice 

 versa on a circuit which eliminates all frequencies above 1,500 cycles. 

 It is obvious, therefore, that the methods described above could not 

 be used to correlate the two crews for such circuits. Similarly, the 

 methods could not be used for comparing two crews, if the hearing 

 level of one is 10 db below the other, or to compare two techniques, 

 one of which is made up entirely of vowel sounds and the other entirely 

 of consonant vSounds. As shown in Fig. 4-b, data have been obtained 

 with various crews on various circuits which can be correlated very 



