850 BELL SYSTEM TECHNICAL JOURNAL 



transmit correctly one of these test sentences over an ideal system 

 to the time required over the system under test. With the crew used 

 in making these tests, and with an ideal transmission system, it 

 required an average time of 5.2 seconds after the speaker started to 

 pronounce the sentence before the observers grasped the idea. It 

 will be seen from the curve that for systems having approximately 

 20 per cent articulation, the time required is twice as great. Fig. 11 

 shows that one out of every four of the sentences is mistaken for this 

 value of articulation. If it is assumed that an observer asks that 

 only sentences which he fails to understand be repeated, it can be 

 shown that this time ratio is equal to the discrete sentence intelli- 

 gibility." 



It is evident from Figs. 11 and 12, that the observed time ratio 

 is appreciably less than the discrete sentence intelligibility. This 

 difference may be taken to indicate that an observer not only asks 

 that sentences which he fails to understand be repeated, but also that 

 sentences about which he is uncertain be repeated. In other words, 

 the time element reflects both factors, the understandability and the 

 uncertainty. 



As has been previously mentioned, tests have been made with 

 various types of English word lists. Because of the manner in which 

 the words were selected, and also due to uncertain psychological 

 factors entering into the tests when such words are used repeatedly, 

 it is difficult to compare the results so obtained with syllable articula- 

 tion results. 



However, it was found that if a definite rule were followed In 

 selecting words from a newspaper, consistent results could be obtained 

 with lists containing 500 or more words per list. The method of 

 selection was to take the first word from every third line of a newspaper 

 column. In this selection all proper names and the following six 

 most frequent words of English were excluded, the, of, and, to a, in. 

 When a word was hyphenated from the previous line, the whole 

 word was used. Each of eight callers called a list of 66 words to 

 four observers in the manner of an ordinary standard articulation 

 test. Tests were made with the carbon transmitter circuit and the 

 six circuits indicated in Fig. 3. The data were analyzed to give the 

 discrete word intelligibilities for the one, two, three, four, and five- 

 syllable words occurring in the lists, as well as for the lists as a whole. 

 The lists on the average contained 46.3 per cent one-syllable, 29 per 

 cent two-syllable, 16.8 per cent three-syllable, 6.4 per cent four- 



11 "^ Theoretical Study of Articulation and Intelligibility of a Telephone Circuit," 

 John Collard, Electrical Communicatton, 7, page 168, January, 1929. 



