492 ANNUAL REPORT SMITHSONIAN INSTITUTION, 1924 



The "m" sound is nasal. The lips are pressed together and the 

 breath is released through the nose. The stop at the lips is broken 

 when the sound is terminated or when a succeeding vowel begins. 

 The method of producing the other stop consonants can readily be 

 followed from the table. 



The fricative consonants are characterized by the rushing sound 

 of the breath through a characteristic air outlet. We have voiced 

 and unvoiced consonants among the fricative sounds as well as 

 among the stop consonants. For example, the sound " f " is pro- 

 duced by forcing the breath through the air outlet between the 

 upper teeth and the lower lip. The sound " v " is formed in the 

 same way except with the accompaniment of vocal cord vibrations. 

 The method of producing the other fricative consonants is easily 

 seen by reference to Figure 1. 



The speech sounds thus produced in the course of conversation 

 are radiated from the speaker and transmitted through the air by 

 means of pressure waves. These air vibrations are very tiny and 

 exceedingly complicated. In physical analyses of speech it is 

 usually these pressure waves or their duplicates, converted into elec- 

 trical vibrations, which are studied. Many of the results here 

 described were obtained with a certain type of high-quality electrical 

 reproducing system or circuit as the basis of the experiments. This 

 system consists of a special form of telephone transmitter, a five- 

 stage vacuum-tube amplifier for magnifying the electric speech cur- 

 rents, and, to terminate the circuit, either a group of telephone re- 

 ceivers of special construction or an experimental type of recording 

 apparatus. The design and construction of this experimental sys- 

 tem is such that it is probably the most nearly perfect telephonic 

 reproducing apparatus so far built. Its quality is indistinguishable 

 from that of direct air transmission. 



In speaking a given letter sound, only the component frequencies 

 of the particular sound (i. e., a sort of " acoustic line spectrum ") 

 are being emitted. By impressing a steady sound on the reproduc- 

 ing system mentioned above and by rapidly inserting in succession 

 suitable sharply-resonant filters covering the range of interest, 

 harmonic analyses of the sustained tone may be made. Figure 2 

 shows the amplitude-frequency characteristics of some of the Eng- 

 lish vowels obtained in this way. While these results are typical, 

 it is to be noted that they represent the vowel sounds as pronounced 

 by one particular speaker. 



But diflPerent speech sounds have different components, and more- 

 over the same sound is frequently pronounced at different pitches, 

 since conversational speech has more or less melody to it. In the 

 aggregate speech may be taken to be represented by a band spec- 

 trum. Figure 3 represents the " acoustic spectrum " of English 



