314 



BELL SYSTEM TECHNICAL JOURNAL 



digits position for most of the sounds and as to order of magnitude for 

 the infrequent sounds. 



The effects of restricting the word Ust in various ways are shown in 

 Figure 3. The first hne shows graphically the relative occurrence per 



^ 



RELATIVE OCCURENCE OF THE INITIAL CONSONANTS 

 IN TELEPHONE CONVERSATION 



P T K F TH' S SH CH M N NG L R W Y H Compound 

 B D G V TH" Z ZH J 



Fig. 3 — The relative occurrence of initial consonants — effects of restricting the 

 word list. 



Line I — Relative occurrence of initial consonants for all words. 



Line II — Differences resulting from omission of minor parts of speech (118 words). 

 Line III — Differences resulting from omission of the 100 commonest words. 

 Line IV- — Differences resulting from omission of the 1,500 least common words. 



hundred for initial consonants as in Table VIII. If the minor parts of 

 speech are excluded before the analysis, which eliminates only 118 

 different words, but nearly half the total words, the resulting changes 

 are shown on the second line. Notable decreases occur for "th," "w" 

 and "y," which may largely be traced to the omission of "that," 

 "they," "this," etc.; "will," "with," "would," etc.; and "you," 

 respectively. These elisions enhance the relative contributions from 

 "get," "see" and "know." When the 100 most common words are 

 omitted there are also large changes, as shown by the third line. Since 

 50 of the 100 most common words are of the minor parts of speech the 

 similarity of this line to the second is not surprising. The omission, 

 on the other hand, of the 1,500 least common words, namely, those 

 omitted from the vocabulary of Table III, changes the distribution by 

 negligible amounts as shown in the fourth line. Since, then, the 737 



