324 BELL SYSTEM TECHNICAL JOURNAL 



relative occurrence of a number of the sounds materially, although the 

 general range of percentages covered is changed very little. Omission 

 of the 1,500 least common words has a negligible effect. When the 

 words are weighted by the number of conversations in which they 

 occurred, out of 500, instead of by their total occurrence, the effect 

 on the distribution of sounds is surprisingly small, considering the 

 radical change in method. 



While the analysis into speech sounds for purposes connected with 

 the design of telephone circuits was the real goal of this study, it is 

 hoped that the information concerning both words and sounds will be 

 of service also to those working in the fields of phonetics and philology. 



