THE SOUi\'l>S DF Sl'liliCIl 



615 



sjH\ikcrs, the "r" characteristics are best displayed toward tiie end 

 of the record, though there is no sharp traiisilioii point. In the sta- 

 tistical stutly of these sounds the data were taken from the latter 

 pt)rtions of the records; init in the mechanical anaKsis it was thought 

 best to use the whole recortl. Now abstracting and condensing the 

 data obtained in these two ways we have (ignoring; fnndnmcnial 

 tones) the following table of frequencies: 



r {fir and er) 



These may be compared with Paget's results (from the second 

 memoir, in which r is classified as a consonant sound) taking one of 

 his general results from a mass of experimental data: 



r (Paget: reference 9a, 9b p. 154) 



"Throat or back resonance". . 400-700 cycles 



"Middle resonance" 1 149-1824 cycles 



" Front resonance" 1824-2169 cycles 



(all varying with the associated vowel) 



The italicized values in the first table above indicate correspondences 

 with Paget's data, and we conclude that these roughly define the r 

 sound, in terms of the steady-state theory. 



Before taking leave of the vowel diagram, we should note not only 

 the location of the resonant ranges but also their extent, and their 

 relative separation from other resonant ranges in order to arrive at 

 essential characteristics of the vowel sound. In other words, the 

 individual vowel quality depends not only on a certain characteristic 

 region of resonance but on the relative pitches in case there is more 

 than one region of resonance. This eflfect is clearly shown to some 

 degree in every group save one (VII:r) in Fig. 13. It will be noted 

 that for the characteristic maxima of energy in the spectrum of a given 

 sound, the peaks in the curve for female voices tend to occur at a 



