Telephonic Intelligibility. 155 



The consonants may be divided into three groups, as indicated 

 by the heavier lines in the table, which are sufficiently distinct so 

 that no consonant in one group is mistaken for any consonant 

 in the other groups as frequently as 5 per cent, of the time. 



The totals in the last column o£ Table I. show that the 

 records were not evenly divided among the twenty syllables. 

 Thus hue was recorded only 27 per cent, as many times as 

 it was called, while tea was recorded nearly twice as often 

 as called. The scattering of the records increases somewhat 

 with the total number of records, but not proportionately. 



In general each consonant was much more frequently 

 mistaken for some one consonant than for any other. 

 Table II. gives the correct records and the leading mistakes 

 for transmission through the air, and in Table III. the same 

 records are given for telephonic transmission. The last 

 columns of these tables show that 96 per cent, of the records 

 were correct for transmission through the air, and 59 per cent, 

 correct for telephonic transmission. The leading mistake 

 made for air transmission occurred in 3 per cent, of the cases 

 (total correct and incorrect records), while for telephonic 

 transmission the leading mistake amounted to 26 per cent, 

 of the cases. The mistakes made through the air were also 

 made over the telephone, but additional mistakes also occurred, 

 and often these are the more frequent ones. The most common 

 mistake made over the telephone was in recording me when 

 knee was called. This occurred in 67 per cent, of the cases 

 when knee was called. Knee is thus natural lv the syllable 

 least often recorded correctly, occurring in only 24 per cent, 

 of the cases. Lee was recorded correctly in OS per cent, of 

 the cases, a record excelling the average for direct transmission 

 through the air. 



The records tabulated in the above tables were all made 

 by Mr. W. L. Richards and Mr. E. C. Molina. They dif- 

 fered considerably in the distribution of records, but pre- 

 sumably the averages are about normal as far as the personal 

 element goes. With the best possible articulation the intel- 

 ligibility would undoubtedly be considerably higher. 



Our tests indicate that the greater part of the distortion 

 occurs in the subscriber's set. and that a short length of line 

 may actually improve intelligibility as compared with zero 

 line. As the length of line reaches the commercial limit the 

 intelligibility gradually falls off. The tests indicate, as would 

 be expected, that changes in the telephone apparatus and the 

 line affect the different syllables quite differently. 



It will be seen that the syllables fee and see are confused 

 with each other in telephonic transmission. They were 

 interchanged in 27 per cent, of the cases, and recorded cor- 

 rectly in 58 per cent, of the cases, which makes f and s an 

 average pair of consonants, as the twenty consonants were 



