556 BELL SYSTEM TECHNICAL JOURNAL 



having negligible distortion, it is noticed that the distorted speech is 

 accompanied by certain audible effects which appear to be extraneous 

 to the speech and transient in character. As discussed above phase 

 changes in the component frequencies of steady state waves cause 

 little if any change in the character of the sound. This would indicate 

 that the so-called audible effects of phase distortion arise in the transi- 

 tion periods, i.e., in the period between the approximate steady state 

 of one speech sound and that of the succeeding sound, and are due to 

 the spreading out effect of phase distortion. Data on the amount of 

 distortion that will cause just noticeable effects will be discussed in a 

 later paragraph. 



Before discussing the various factors affecting the recognizability, it 

 Is of interest to consider the importance as obtained from articulation 

 tests, ^ of different portions of the duration intervals of speech sounds, 

 and also the importance of different portions of the speech frequency 

 range. 



Fig. 3 shows the effects upon sound articulation of limiting the 

 transmitted frequency range by means of high or low pass filters 

 in a system having neglible distortion in other respects. Although the 

 curves do not intersect at 50 per cent nor do the articulation values of 

 complementary filters add up to 100 per cent, they may be used with 

 qualifications, to measure the contribution or importance to articulation 

 of a portion of the speech frequency range. Thus the slope vs. cut-off 

 frequency for the low pass filter curve gives a measure of the contribu- 

 tion to articulation of a frequency range A/ when contiguously added 

 to the range 275 to/. 



Articulation tests that have been made with voice operated relays 

 give an indication of the importance to articulation of portions of the 

 duration intervals of speech sounds. In the tests, syllables of the 

 consonant-vowel-consonant type were spoken at intervals of about 3 

 seconds. A circuit having a relay adjusted so as to break contact 

 almost simultaneously with the beginning of a syllable, was used. The 

 contacts of a second relay formed a short circuit across the receiver. 

 The operation of the first relay caused the second relay to break contact 



■* Articulation Testing Alethods, H. Fletcher and J. C. Steinberg, Bell Sys. 

 Tech. Jour., Oct. 1929. In an articulation test lists of syllables are spoken into the 

 transmitter of a system having phase distortion and observers at the receiving end 

 write down the sounds which they hear. The observed lists are compared with the 

 spoken lists and the errors determined. The percentage of the total number of spoken 

 syllables that are correctly observed is called the syllable articulation. A syllable is 

 considered to be incorrectly observed if one or more of the fundamental speech sounds 

 which it contains are mistaken. The percentage of the total number of spoken 

 speech sounds which are correctly observed is called the sound articulation. When 

 attention is directed toward a specific sound such as "e," the term individual sound 

 articulation is used. It is the percentage of the number of times that "c" was spoken 

 that it was observed correctly. 



