ARTICULATION TESTING METHODS 833 



Table VIII gives the results of articulation tests that were made 

 with two such types of lists. In both cases the sounds were combined 

 at random into syllables of the con-vow-con type. The tests were 

 made on the auxiliary circuit of the master reference system. 



Realizing that the probable error in the articulation value given 

 for each sound is ± 5, there do not appear to be any outstanding 

 differences in the articulations of the various sounds with the two 

 types of list. The average articulations for the two lists differ by 

 less than the probable error. The test indicate, therefore, that lists 

 having uniform occurrence of sounds give the same individual sound 

 articulation values as lists having the frequencies of occurrence of the 

 sounds proportional to their frequencies of occurrence in speech. 

 At least this is true within the accuracy usually attained in making 

 such tests. The testing advantages of the former type of list have 

 already been pointed out. 



It is important to notice that the average sound or the average 

 syllable articulation may not be the same for the two types of lists. 

 even though the articulation for each sound is the same. The averages 

 shown in the table were obtained by assigning equal weights to the 

 articulation for each fundamental sound. If weights which are 

 proportional to frequency of occurrence of the sounds in speech be 

 assigned, the averages obtained will, in general, be slightly different. 

 For the particular circuit corresponding to the data of Table VIII, the 

 averages obtained in the two ways did not differ by more than the 

 observational error. Our data have shown that this is also true for a 

 large class of circuits ordinarily used in telephone work. However, 

 those transmission systems which have a specific effect upon certain 

 consonant or vowel sounds, for example, upon s which occurs 850 

 times in one list compared to 300 times in the other, would obviously 

 have different values for the sound articulation by using the two 

 methods of obtaining the average. 



In speech, certain combinations of sounds occur more frequently 

 than others. In other words, some consonants precede certain vowels 

 more frequently than they do other vowels, and similarly, some 

 consonants follow certain vowels more frequently than others. For 

 example, the combination "es" is used much more frequently than 

 the combination "us" (u as in foot). Since the testing lists are 

 made by random selection, the various con-vow and vow-con combi- 

 nations occur with uniform frequency. In order to determine how 

 this difference influences the interpretation of the sounds, articulation 

 data on various circuits were examined. Attention was focused first 

 on the final consonant sounds. One hundred errors for each consonant, 



