252 CARNEGIE INSTITUTION 



atlas, its components will be given at once by the known compo- 

 nents of the apparatus curve, thus avoiding a long and expensive 

 analysis. 3. Curves of pure speech sounds can be directly engraved 

 on gramophone plates without any intervention of the voice itself. 

 This apparatus, made in Munich, after a former rough model, is 

 nearly completed. 



In studying the curves another phenomenon was observed. A 

 spoken vowel changes its character from beginning to end ; in the 

 hundreds of vowels inspected no two waves of a vowel were ever 

 exactly alike. Neighboring waves were similar and the change oc- 

 curred more or less gradually, according to the nature of the vowel. 

 Except in long vowels the ear cannot detect even the existence of 

 such a change, and all phoneticians have considered the short vowels 

 as constant things. The ear gets only an impression of the general 

 effect and what the vowel really sounded like at beginning, mid- 

 dle, and end remains unknown. I find that the ear classifies as the 

 same short vowels a number of tj'pes having in reality no acoustic 

 resemblances. I also find that a suggestion from spelling or from 

 another word will cause the ear to hear as different several vowels 

 that are acoustically the same, and as the same several vowels that 

 are acoustically different. The phonetic spellings in the dictionaries 

 are certainly to a considerable degree erroneous in respect to the 

 short vowels. The study of short vowels should be extended to 

 include many hundreds of cases, and a machine should be con- 

 structed that will make it possible for the ear to hear each stage of 

 the vowel separately. In respect to the long vowels a somewhat 

 similar condition was found ; for example, the usual statement that 

 long vowels, such as ifd' in " see, ' ' are diphthongized does not always, 

 or even generally, hold good in ordinary American conversation. 



To elucidate the curves of vibration from the vocal cavities it was 

 necessary to make some study of the cavities themselves — for exam- 

 ple, as formed by the position of the tongue. One method of study- 

 ing the position of the tongue is that of employing an artificial palate 

 with a chalked surface ; the chalk is removed where the tongue 

 touches. This method was devised in 1887 by the American physi- 

 cian Kingsley, who registered about a dozen contacts. For French 

 the study has been systematically carried out by many hundreds of 

 registrations by Rousselot (Paris) and his pupils ; for German a few 

 results have been obtained ; for English nothing has been done. I 

 have accumulated and prepared for publication many hundreds of 

 registrations of the American sounds in their varieties, using not 



