Front vowels in Parisian sociolects

A spectrographic study of front vowels in Parisian French has been carried out with the view of investigating differences between female sociolect upper-class and neutral pronunciation. The study was based on recordings from natural conversation. The formant data a r e normalized so a s to remove the influence of possible anatomical variations, e. g. average vocal t ract length. It i s found that the sociolect subjects aim at more extreme [i] and [ e] vowel targets than the neutral group. This feature of articulatory tenseness i s also apparent in adjacent dental and palatal consonants, which a r e produced more close and attain greater duration than in the neutral group. A non-uniform normalization technique, Fant (1975), i s adopted for converting the female data to the most probable equivalent male data thus allowing a comparison with other languages. Introduction In previous articles, Mettas ( 1970), Mettas, Fant, and St3lhammar (1971), Lacnderey and Mettas (1973), the phonetic qualities of some vowels ( A , /AN/ /m of two Parisian sociolects were investigated. The purpose of the present study i s to compare the front vowels [ i ] , [ e l , [ E ] , [ a J [ a ] of these two sociolects applying normalization techniques (Fant, 1975) to spe ctr ogr aphic data. . Subjects This study i s concerned with female subjects of age 1830. I t will be recalled that in Sociolect I subjects belong to two close social groups aristocracy and upper middleclass ("haute-bourgeoisie1') and a r e characterized by their long established social position in Paris . In Sociolect 2 (neutral pronunciation), speakers a r e middleclass Parisians (I'bour geoisiel'). Speech material The vowels analyzed were extracted from recorded spontaneous conversations of four speakers in Sociolect I and three speakers in Sociolect 2. As often a s possible we chose for analysis vowels having the same consonantal frame in both sociolects. * address: 42, Avenue en& Coty, P a r i s 75014, France. Mrs. Mettas has visited the department of Speech Communication several times, the las t time was in April 197 6. . ' \ . . . I STL -QPSR 23/ 1977 2. It was previously mentioned ( ~ e t t a s , 197 3) that, among the various realizations of the vowels in Sociolect 1, the [i] and [ e l often seem to have a more close sometimes a more open articulation from an auditory point of view than in the neutral pronunciation. Vowels analyzed in the present study belong to the f irst type. It was also mentioned that the [ E ] sometimes seems to tend towards [a], which i s not the case of the vowels analyzed here. It should also be observed that most of the [a] vowels of Sociolect i were sampled from unstressed words within a rhythmical group. In a stressed position [a] i s usually pronounced with a more posterior sound (Mettas, Fant, and Stillhammar, 197 1). In the neutral group, half of the f a ] sampled for analysis a r e stressed vowels, the other half being unstressed. The position of the [ a ] vowel did not affect the audible quality and the systematic differences in formant frequencies were found to be small. It must be added that the [ a ] vowels analyzed were in most cases the [ a ] of hesitation, which explains the long duration of most of them. Formant frequency measurements Spectrograms of the words were made by means of the Voiceprint spectrograph using the broadband pass filter of 300 Hz. The f irst three formants of the vowels to be compared were measured for each speaker in each group. To minimize variations related to coarticulation samples were taken a t locations where such contextual influences were judged to be minimum, i. e. where formants were closest to target values. Formant frequency values Fi, F2 , and F 3 of the vowels analyzed for each subject a r e tabulated in Table I-A-I together with the duration (D) of each vowel. The e r ror s in measuring formant frequencies from the broad-filter (300 Hz) a r e estimated to be larger than the ~ ~ / 4 given by Lindblom ( 1961) for male voices and by Karlsson (1975) for rather low frequency F fe0 males. When Fo becomes of the order of 300 Hz the Sonagraph broad filter shows harmonics instead of formants and the interpretation of the spectrograph pattern becomes difficult. Formant asymmetries together with the specific Sonagraph pr eemphasis might cause systematic e r ror s N w t ro l spoakers /I / CR F1 F2 F3 culslne D /a/ pour me coucho~ Table I-A-IA F1 F2 F3