July i, 1S97] 



NATURE 



211 



My attention has since been directed to perfecting a mechanism 

 for obtaining a record of the vibrations. 



The instrument now before you, which I shall call a phono- 

 ip-aph- recorder, traces out, on a large scale, the curves of the 

 indentations on the wax cylinder corresponding to each vibration 

 of sound, and it does so in a way that seems to be highly 

 satisfactory.^ 



Since the apparatus was brought to its present condition, I 

 have been able to record the vibrations of the tones of several 

 instruments, and also the tones of the human voice, both in 

 singing and in speech. Illustrations of these I now show 

 you (Fig. 3 . 



First, with reference to speech, I wish to point out that when 

 the record of a word is examined it is found to consist of a long 

 series of waves, the number of which depends (i) on the pitch 

 of the vowel constituents in the word, and (2) on the duration of 

 the whole word or of its syllables individually. There is not for 

 each word a definite wave form, but a vast series of waves, and, 

 even although the greatest care be taken, it is impossible to 

 btain two records for the same word precisely the same in 

 . haracter. A word is built up of a succession of sounds, all 

 usually of a musical character. Each of these sounds, if taken 

 individually, is represented on the phonograph-record by a 



waves ends and where another begins. For example, in the 

 word Con-stan-ti-nopk, the predominant sounds are those of 

 o-a-i-o-ill, and the' variation in pitch is observable to the ear if, 

 in speaking X}c\& word, we allow the sound of the syllables to be 

 prolonged. If we look at the record of the word, we find these 

 variations in pitch indicated by the rate of the waves, or, as the 

 eye may catch this more easily, by the greater or less length of 

 wave, according to the pitch of the sound. The consonantal 

 sounds of the word are breaks, as it were, in the stream of air 

 issuing from the oral cavity, and these breaks (I am not dis- 

 cussing the mechanism at present) produce sounds that have also 

 often the character of vowel sounds. Thus, at the beginning of 

 'Constantinople," we have, as will be observed on pronouncing 

 the syllables very slowly, the sound iikkd. This sound is repre- 

 sented in the record by a series of waves. Then follow the waves 

 of the vowel 0. Next we have the sound ;;;/ (sound through the 

 nose), also represented by a series of waves. Next the hissing 

 sound is, which has first something in it of the vowel e or i, and 

 then the iss-s. This sound also is shown Ijy a series of waves. 

 Then there is ta, which has a double series of waves — (i) those 

 for it or t, and the next for a. This passes into the prolonged 

 vowel a, this into in, this followed by ti passing into the vowel 

 i, then another in, then a long 0, then a sound like op, and, 



Part of word Constantinople ; a vowel. 



Piccolo. 



Cornet. 



Sound of explosion o. a quick-firing gun 

 on board the Benboiv. 



Noise in a boiler maker's shop. 

 Flute. 



Flute. 

 Military band 



Fig. 3. — Specimens of curves obtained by phonograph recorder. Four inches = i/i2oth sec. The curves read from left to right. The lower curve of the 

 flute is continuous with the upper one, to show change of pitch. These curves are given to show the great variety of wave form. They are not 

 sufficiently accurate for analysis. To obtain curves suitable for analysis, the greatest care must be taken. 



greatei- or less number of waves or vibrations, according to the 

 pitch of the sound and its duration. The pitch, of course, will 

 depend on the number of vibrations per second, or per hundredth 

 of a second, according to the standard we take, but the number 

 of the waves counted depends on the duration of the sound. As 

 it is almost impossible to utter the same sound twice over in 

 exactly the same fraction of a second, or in the same interval of 

 time, the number of waves counted varies much in different 

 records. The rate per unit of time determines the pitch, the 

 number the duration of the sound. In a word, these successive 

 sounds blend into each other, and, in many records, the passage 

 from one pitch to another can be distinctly seen. The speech 

 sounds of a man vary in pitch from 100 to 150 vibrations per 

 second, and the song sounds of a man from 80 to 400 vibrations 

 per second. The sounds that build up a word are chiefly those 

 of the vowels. These give a series of waves representing a varia- 

 tionin pitch according to the character of the vowel sound. In 

 the record of a spoken word the pitch is constantly moving up 

 and down, so the waves are seen in the record to change in 

 length. It is also very difficult to notice where one series of 

 1 For a diagram of the apparatus, see Proc. Roy. Soc. Ed., December 7, 



NO. 1444, VOL. 56] 



lastly, the sound ill, a sort of double- vowel sound. As so many 

 of these sounds have the character of vowels, it is impossible, 

 by an inspection of the record, to say where one set of waves 

 begins and another ends. There are no such breaks corre- 

 sponding to the consonants ; the vibrations of the consonants 

 glide on as smoothly as those of the vowels. The number of 

 waves producing a word is sometimes enormous. In " Con- 

 stantinople " there may be 50c, or 600, or 800 vibrations. A 

 record of the words " Royal Society of Edinburgh," spoken 

 with the slowness of ordinary speech, showed over 3000 vibra- 

 tions, and I am not sure if they were all counted. This brief 

 illustration gives one an insight into nature's method of pro- 

 ducing speech sounds, and it shows clearly that we can never 

 hope to reach such records in the sense of identifying the curve 

 by an inspection of the vibrations. The details are too minute 

 to be of service to us, and we must again fall back on the power 

 the ear posses.ses of identifying the sounds, and on the use of 

 conventional signs or symbols, such as letters of the alphabet, 

 vowel symbols, consonant symbols, or the symbols of 

 Chinese, which are monosyllabic roots often meaning very dif- 

 ferent things, according to the inflection of tone, the variations in 

 pitch being used in that language to convey shades of meaning. 



