A sociophonetic investigation of Mexico City Spanish vowel weakening
Meghan Dabkowski, PhD Candidate
Recent developments in the phonological exploration of weakening processes have brought to the forefront the role of linguistic as well as social factors, and the importance of the interaction between the two. Additionally, vowel weakening tends to show more variability cross-linguistically than consonant lenition and this paper contributes to this line of research. Spanish vowels have been described in the literature as showing remarkable stability, and when compared to consonants, less subject to changes in their quality, duration, or voicing based on regional or social variation (Hualde 2005, Quilis 1981, Quilis and Esgueva 1983). However, there is a growing body of evidence suggesting that vowels in fact do show variation. The variety of Spanish spoken in Mexico City shows evidence of a vowel weakening process by which primarily unstressed vowels are variably shortened, devoiced, produced with weak voicing, or elided. A similar phenomenon has been documented in several varieties of Andean Spanish as well (Delforge 2008a, 2008b, 2009, 2012, Gordon 1980, Hundley 1983, Lipski 1990). Early research on vowel weakening in Mexican Spanish describes it as a highly variable process that primarily affects unstressed vowels adjacent to a voiceless consonant, especially /s/. The examples below in (1), taken from Canellada de Zamora and Zamora Vicente 1960 and Matluck 1952, show orthographic representations of devoicing and deletion, with superscript vowels representing devoiced variants, and the apostrophe representing apparent deletion of the target vowel.
(1) dientes ‘teeth’ dientes, dient’s
manos ‘hands’ manos
casa ‘house’ casa
amistad friendship amistad
Early research on this variety relied on impressionistic descriptions (Boyd-Bowman 1952, Lope Blanch 1963, Matluck 1952), and thus, the current study is concerned with conducting systematic instrumental measurements and analyses.
The data for the current study include recorded speech from 10 male and 10 female residents of Mexico City, and are part of a larger corpus of sociolinguistic interviews collected for a comprehensive sociophonetic study of the phenomenon. A token consists of any single vowel not adjacent to another vowel or semivowel. 150 tokens per speaker were analyzed, for a total of 3000. Tokens were analyzed acoustically in Praat (Boersma and Weenink 2016), and two measurements were taken: vowel duration, and voicing duration. Vowel duration was measured from the preceding consonant’s release to the onset of the following consonant or pause. Voicing duration was calculated by measuring the duration of the voicing bar in the spectrogram.
The acoustic analysis carried out for this study showed a gradient process: instead of simple presence or absence of voicing, many tokens show weak or partial voicing. In the majority of tokens analyzed, full voicing is found from the beginning of the vowel to somewhere between the midpoint and the end. Weak voicing is characterized by a lower intensity in the waveform, and a lighter voice bar. The presence of frication, or energy in the higher frequencies, distinguishes devoiced segments from weakly voiced segments when other aspects of the acoustic signal are not clear indicators of voicing. A token with a duration more than 50% shorter than the average length for that target vowel in the same stress condition, is classified as shortened.
A generalized linear mixed effects regression model was built that included speaker as a random variable and the fixed variables stress, target vowel, voicing of preceding segment, voicing of following segment, position relative to stress (pre- or post-tonic), and speaker gender. Results indicated 1) that /a/ was significantly less likely to weaken, when compared to mid and high vowels; 2) unstressed vowels were significantly more likely to weaken than stressed vowels; and 3) a target vowel followed by a pause was significantly more likely to weaken, while a target vowel followed by a voiced consonant was significantly less likely to weaken. Additionally, there was an interaction between voicing of following segment and speaker gender, such that when a target vowel is followed by a voiced segment, a male speaker is significantly more likely to produce a weakened segment.
The results of this investigation show the importance of considering both linguistic and social factors in order to better understand phonetic and phonological variation. Results of the acoustic analysis indicating a gradient process will inform future research that offers a continuous statistical analysis of the phenomenon in order to better understand how phonetic detail contributes to phonology.