834 



BELL SYSTEM TECHNICAL JOURNAL 



or 2,200 consonant errors were selected at random from the articulation 

 data and the number of these 2,200 errors that occurred after each 

 vowel sound ascertained. Similarly, the number of vowel errors, 

 out of a total of 1,100 errors, that occurred after each of the consonant 

 sounds, was determined. 



Probability studies indicate that the distribution of these errors 

 as shown in Table IX is of the same order as that to be expected on 



TABLE IX 



Distribution of Vowel and Consonant Errors 



the basis that the distribution of errors is due entirely to chance. 

 Since, from the way the lists were constructed, the occurrence distri- 

 bution is due to chance, it is evident that the errors in recognition 

 of the sounds do not depend upon the particular sounds that they 

 follow. Although the analysis was not made, it would be expected 

 that a similar situation obtains for initial consonant and vowel errors. 

 These data may be interpreted to mean that the consonant articulation, 

 the vowel articulation, the sound articulation, or the syllable articu- 

 lation, is approximately independent of the particular sound combi- 

 nations, when a wide variety of combinations are used. The results 

 obtained with these lists, therefore, are as representative of speech 

 as the results that would be obtained with lists employing particular 

 sound combinations in proportion to their frequencies of occurrence 



