394 BELL SYSTEM TECHNICAL JOURNAL 



one-hundredth of a second, the definition applies as well to consonant 

 sounds and in addition has the same practical significance as that of 

 the mean power of a vowel. 



Mean power is thus a variable function of time, starting from zero, 

 rising to a maximum and eventually falling to zero again as the sound 

 is being uttered.^ In studying an aggregate of speech sounds it is im- 

 practicable to have the final results in terms of these mean power 

 curves; the most important discriminant of such a curve of any sound 

 is its maximum ordinate, Pm- This value was used in the earlier study 

 and has been given the name "syllabic power" when used in connection 

 with the syllable as a whole. In the present case we shall abbreviate 

 by simply calling it the "mean power of the sound." Similarly, when 

 we are considering the consonant apart from the rest of the syllable we 

 select the maximum value of Pm for that consonant. 



Likewise, in considering the instantaneous power of a sound we 

 select the height of the greatest peak occurring therein and for con- 

 venience we call it the "peak power." 



All the averages hereinafter tabulated are the arithmetic averages 

 of such maximum ordinates and not the integrated averages. 



Normal and Conversational Values 

 We specify "normal" values as those derived from monosyllables 

 spoken disconnectedly without accent but also without being slighted; 

 while "conversational" values are derived from ordinary conversa- 

 tional speech. It does not follow that the arithmetic average of con- 

 versational values for a given sound should equal the average of the 

 normal value, for the reason that some sounds are slighted much more 

 frequently than others, as we shall see later. 



The Consonants and Semi-Vowels 



Of these sounds two independent sets of data are available: in- 

 stantaneous peak power and mean power. The former is summarized 

 in Table I. To explain the table in detail we take as an example the 

 consonant, "t" as in "tap." There being one observation upon each 

 of two speakers, the greatest observation showed 19 microwatts (peak) 

 from the lips of the one speaker while the other speaker reached a peak 

 of 13 microwatts, and the average of these two is 16. As in the paper 

 on Speech Power and Energy, the corresponding values of power in- 

 tensity in microwatts per square centimeter at the condenser trans- 

 mitter are given in the group at the right. Since the relating factor is 



'See "Speech Power and Energy," Fig. 1, page 628, for comparison of instantane- 

 ous and mean powers. 



