558 



BELL SYSTEM TECHNICAL JOURNAL 



T — AG seconds. The data indicate that equal elements of time in the 

 duration intervals of the sounds are of approximately equal importance 

 to the average articulation of a group of sounds. It should be pointed 

 out that this might not be the case for individual speech sounds and 

 also for certain types of speech distortion. In the above tests a 

 carbon transmitter was used of a type which introduced attenuation 

 distortion. Tests have also been made with speech having negligible 

 attenuation distortion. Although they indicated a relation of the 



100 



90 



Z 80 

 O 



_l 70 

 O 



cr 60 

 < 



^ 50 



Z 



o 



^ 40 



o 

 o 



^30 



20 



10 



o 



* — . o 



0.02 0.04 0.06 0.08 0.10 0.12 0.14 



IMTERVAL OF TIME CLIPPED FROM INITIAL CONSONANT IN SECONDS 



Fig. 4 — Importance of an element of time to articulation. 



above type the data are somewhat questionable because of uncer- 

 tainties in the operating time of the relay. 



In the next series of articulation tests a nominal undistorted speech 

 frequency range of — 4500 cycles was divided into two parts by means 

 of filters and each part transmitted through a different channel. After 

 transmission the two parts were recombined. The phase characteristic 

 of each channel approximated a straight line over the greater part of 

 the frequency range. The slope of the characteristic of one channel 

 could be increased by various amounts over that of the other channel. 

 One channel thus introduced a definite time delay, in the sense used 

 here, with respect to the other channel, i.e., a delay given by the dif- 

 ference in the slopes of their phase characteristics. The observed 

 sound articulations plotted against time of delay are shown in Fig. 5. 

 The articulation values decrease with increasing delay and approach 



