THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 313 



cerning the quality of a vowel which precedes "r." Likewise it was 

 found expedient to make a number of classifications of vowels in 

 unaccented positions. 



Since the figures of Table VIII are likely to find application as 

 weighting factors it is convenient to have them add exactly to 100 per 

 cent, consequently they are given to two places of decimals. An 

 estimate of the representativeness of these figures may be obtained 

 from the data presented in Table IX, which were worked out from 



TABLE IX 

 Comparison with Check Test 



observations mentioned before, conducted by a different observer at a 

 different time, but on the same set of toll circuits. Records were made 

 only of verbs, and for 250 instead of 500 conversations. The vocabu- 

 lary collected in the check test resembled that of the first observations 

 closely. Arranging the words in the order of occurrence, the first 17 

 words of the first observations are also the first 17 of the check test, 

 although the order is not repeated exactly. In the first observations 

 the first few words run: "get," "know," "see," "want," "go," "tell," 

 "think" and "say"; in the check test the order is: "get," "see," 

 "know," "want," "tell," "think," "go" and "say." Table IX shows 

 the analysis of the words as to the simple consonants, lumping initial 

 and final consonants together. Only one of the differences is greater 

 than 1 per cent and all but three are less than 0.5 per cent. One check 

 test is not sufficient for a final statement, but judging by these results 

 the observing method and the samples taken seem to justify considering 

 the figures of Table VIII as representative as far as the figures in the 



