ARTICULATION TESTING METHODS 829 



well by means of the above curves. Fig. A-d gives data that were 

 obtained with other crews which show very poor correlation. In 

 neither case are the characteristics of the crew well enough known 

 to satisfactorily account for the observed differences. At the time 

 the work was done the significance of these factors was not so apparent, 

 so that they were not given the attention they now receive. During 

 the past two years a crew of 10 people has been used almost continu- 

 ously in testing work. During this time numerous changes in per- 

 sonnel have taken place and only five of the original members are 

 now on the crew. The data obtained during this time appear to be 

 strictly comparable. In some cases it is necessary to use the practice 

 curves. In other cases (circuits that are frequently tested), practically 

 identical results are obtained. For this reason, it is believed that if a 

 similar crew of 10 different people were to be selected as previously 

 described, comparable articulation results would be obtained. It 

 seems reasonable to expect that crews testing in various languages 

 should also obtain comparable results provided that the crews were 

 similar in the sense used here and that the lists were phonetically 

 similar. It seems desirable, therefore, to standardize on the factors 

 which affect the comparison of data, such as, the size and type of 

 crew, the type of list, and the type and number of reference circuits. 

 Best results are likely to be obtained when the crews do not differ 

 by amounts which correspond to values of x less than 0.7. Smaller 

 values indicate that the crews have not had sufficient testing experi- 

 ence, or have speech and hearing characteristics which are essentially 

 different, or that the phonetic content of the testing lists are appreci- 

 ably different. In the latter case the results may be correlated by 

 means of statistical relations that will be given in a later paragraph. 



Relation of Articulation to the Transference 

 OF Thought by Speech 



The foregoing paragraphs have been concerned with the practical 

 problems of setting up a suitable testing technique and correlating 

 the observed articulation results. The procedure that has been 

 discussed enables us to measure the percentages of the various speech 

 sounds which are correctly recognized when they are spoken in a 

 simple con-vow-con syllable. We desire at this point to consider the 

 broader significance of this measure. In other words, how is the 

 articulation result related to the transference of thought by means of 

 speech? This relationship involves many psychological factors which 

 are difficult to evaluate so it must not be expected that a compre- 

 hensive answer can be given here, but it is important to understand 



