560 BELL SYSTEM TECHNICAL JOURNAL 



the sound articulation for a delay AT should be 



A = A. + K{A,- A,), (2) 



where i^ is a factor obtained from Eq. (1) for T = .17 seconds. 



This value of T is used because in these tests the articulation sylla- 

 bles were spoken as parts of introductory sentences, such as, "The 

 first syllable is ««/," etc. Oscillograph records for this manner of 

 speaking indicate an average duration time for vowel and consonant 

 sounds of an order of .16 to .18 seconds. This is somewhat less than the 

 duration time for sounds spoken in detached syllables, and somewhat 

 greater than the duration time for sounds spoken in connected speech 

 containing words of one or more syllables. For the latter case, the 

 average duration time is probably more of an order of .08 to .12 seconds. 



The solid lines shown in Fig. 5 were calculated from Eq. 2 which 

 involved only the time factor. The agreement between observed and 

 calculated results indicates that the two other factors were compara- 

 tively small in these cases. As regards the noise factor it should be 

 noted that the frequency range of the so-called noise and that of the 

 sound wave with which it interferes have no part in common. When 

 this is true, as previously pointed out, the interference from noise is 



small. 



The following tests were made with networks having curved phase 

 characteristics of the type shown in Fig. 2. One of these networks 

 was an all pass structure made up of two types of sections, a "B" 

 section having a critical frequency of 2000 cycles and an "A" section 

 having a critical frequency of 2500 cycles. By using difTerent numbers 

 of sections different amounts of delay distortion could be obtained. Fig. 

 6 shows the delay distortion for the conditions that were tested. The 

 attenuation characteristic was equalized up to 2500 cycles, and a 2400 

 cycle low pass filter having negligible phase distortion was associated 

 with the network. Fig. 7 shows the sound articulation values versus 

 the number of sections. A time factor Kj for an element A/ in the 

 frequency range to 2400 cycles, may be obtained from Eq. 1 by setting 

 Y — 0.17 seconds and AT equal to the delay distortion given in Fig. 6, 

 for the element A/. Delays in some frequency ranges will impair the 

 articulation much more than similar delays in other ranges because 

 some frequency ranges are of greater importance to articulation. 

 The importance of various frequency ranges is closely related to the 

 slopes of the curves in Fig. 3. To obtain an effective factor K, the 

 values of K; must be weighted in accordance with the importance of 

 the frequency ranges considered. This may be done approxi- 

 mately for the 2400 cycle range by averaging the values of Kf 



