608 BELL SYSTEM TECIIXICAL JOVRNAL 



record, and a maximum and minimum figure determined for the 

 frequency \vherc\er possible. The behavior of the amplitude of the 

 high frequency component during the cycle was noted, and a rough 

 estimate made of its magnitude. Practically all the vowel records 

 show frequencies above 2500 cycles and the amplitudes in some cases 

 are large. In only two records out of 104 was the high fre(iuenc\' 

 component too small in amplitude to gi\'e a frequency determination. 

 These high frequency components may or may not be characteristic 

 of the given sound; this question is more fully dealt with later. 



To complete the examination of each record its duration was noted, 

 and this time was divided into three intervals: (1) a building up period 

 in which the oscillations rise from zero to an amplitude which shows 

 all the components clearly; (2) a middle period in which the general 

 ampilitude remains nearly constant, but in which some \'ariations in 

 the amplitudes and phases of the com]jonent frequencies usually take 

 place; and (3) a period of decay in which the components disappear 

 and the oscillation gradually loses its characteristic wave form. 



The procedure may be illustrated by its application to the first 

 record for which the following data were recorded: 



Plate No. 1, 00 as in pool. Speaker MA. (Male). 



Time to build up, .05 sec; Middle period, .20 sec; Period of decay, 

 .06 sec; Total Duration .31 sec. 



Fundamental: 102 at start, rises to 108 in middle, rises to 120 at 

 end. Pitch Variation normal. (See explanation below). 



Low Frequency Characteristic: 400 at start, 430 at middle, 440 

 at end. Amplitude greater than that of fundamental. Ai^proxi- 

 mately, a fourth harmonic of fundamental, but amplitude 

 \'ariation (luring the r\(le suggests a transient. 



High Frequency Component : Minimum, 3300 cycles. Maximum, 

 :<(')()() cycles. Noticeable tiiroughout; amplitude \-ariation sug- 

 gests a transient. 



Xo other frequencies. 



'i'his routine was ajiplied to each of the 104 \-owel records and a 

 general summary made of the results, gi\'ing approximate values of the 

 vowel cliaracteristics which forecasted the more accurate results ob- 

 tained later from the mechanical harmonic analysis. 



The simpli'st phenomena to summarize are the general character- 

 istics of the individual speakers. These are based on I lie mean ])er- 



