&12 BEl.I. SYSTEM TECIIXIC.IL JOURNAL 



In the sounds of tin.' first class the high frequency component is usually 

 small in amplitude, more subject to individual bias in its frequency, 

 and may or ma\- not build up in amplitude as early as the low frequency 

 characteristic. In the sounds of the second class the high frequency 

 characteristic is usualh' prominent from the start and builds up very 

 rapidK-; while there is less \ariation in its frequency with the individual 

 speaker. In sounds of the first class there is no decided suggestion 

 of a transient in the high frequency (23 out of 40 records, (iroups 

 I to V inclusive) while in sounds of the second class the transient effect 

 is pronounced {3(1 out of 40 records. Groups VIII, IX, XI, XII, XIII). 



With these considerations in mind there is presented in table IV 

 a summary of the data obtained from this preliminary examination of 

 the vowel records. The mean duration time, and its subdi\isions, 

 are shown in the second column for each pure vowel sound, with mean 

 duration only for the sounds ar (Group VII) and er (Group X). The 

 fundamental and characteristic frequencies of each sound are shown in 

 the 3 columns headed "Mean Fundamental," "Mean Low Character- 

 istic" and "Mean High Characteristic Frequency" respectively. Each 

 mean is taken from four records. The two columns headed "Scattered 

 Low" and "Scattered High Frequencies" contain mean \alues of 

 additional components, occurring in one or more records, in certain 

 frequenc\- ranges, the number of records in which such components 

 are noted being shown in parentheses following the mean. The table 

 illustrates and emjihasizcs many points which ha\e been brought out 

 in the preceding discussion, particularly the closeness with which the 

 high frequency characteristics are defined in the \f)wels of the second 

 or "doubly-resonant" class. 



The table however gives no quantitative statement of the energy 

 distribution among the different frequencies and it is necessary now to 

 refer to the results of a harmonic analysis of these records which has 

 been made and published' from which the diagram of Fig. 13 is taken. 

 The machine method for analysing these wave-forms has been described 

 by Mr. Sacia in detail elsewhere;- it suffices here to note mereh' the 

 essentials in the treatment of the data. 



For the dynamical study, the whole record from start to finish was 

 taken as the unit for analysis, and the data obtained are therefore 

 the average characteristics of the sounds throughout their duration. 

 In the form of an endless belt each of these records was passed repeated- 

 ly through the analysing machine. A single record is of course 



'"Dynamical .Sludv of the Vowel' Soumls." Hill S\stein rechiiic.il Journal, 

 111, No. 2, .^pril, l'>24. 



' (". K. Siiria: " I'lioloinechanUal Wave .Xnalyzcr .Applied to iiiliarrnonic Analysis;" 

 Jour. Opt. Soc. .Am. ami Ki\ . of .Sii. Inst., 9, Oct., 1924, |). ■iSl. 



