Arróniz, S. Stronger After the Fall: Compensatory Fricativization of /b d g/ after Coda /s/ Weakening. Linguistic Symposium on Romance Languages. Pontificia Universidad Católica Madre y Maestra, Santo Domingo (DR). September 2025.

Abstract

In Western Andalusian Spanish (WAS), two hallmark processes—coda /s/-lenition and the spirantization of voiced stops—interact in ways that remain understudied. Coda /s/ weakening, widely documented across Spanish varieties, can result in aspiration, glottalization, or elision (Bybee 2000; Hualde 2005; Lipski 1994, 1999; Luna 2010). In Andalusian Spanish, this reduction has been shown to condition strengthening in following voiceless stops (Torreira 2007; Ruch & Peters 2016), suggesting compensatory adjustments across segmental boundaries. Meanwhile, the spirantization of /b d g/—once described as producing fricatives—has more recently been reanalyzed as yielding approximants [β̞ ð̞ ɣ̞] in intervocalic position (Martínez Celdrán 1991, 2008; Carrasco et al. 2012; Hualde et al. 2011). Yet the effect of coda /s/-lenition on following voiced stops has received little empirical attention.

This study addresses that gap by examining whether the weakening or elision of /s/ systematically triggers fricativization of /b d g/—productions we term Compensatory Voiced Fricatives (CVFs). Data come from a controlled production experiment involving 31 native WAS speakers who completed a sentence completion task designed to elicit /b d g/ in both intervocalic and post-/s/ contexts. A total of 4,822 tokens were analyzed, excluding overt /s/ realizations to isolate compensatory effects. Acoustic measurements included duration, intensity ratio, spectral centroid, spectral spread, zero-crossing rate (ZCR), and harmonic-to-noise ratio (HNR).

Results reveal systematic differences between contexts. Segments following lenited /s/ exhibit significantly longer durations (~75 ms vs. ~40 ms), lower intensity ratios, and increased spectral energy dispersion. Spectral centroid and spread values are consistently higher in these tokens, reflecting increased high-frequency energy and constriction. Though ZCR and HNR showed more overlap, they still contributed to the overall contrast. Statistical modeling reinforces these findings. A random forest classifier and a generalized linear mixed-effects model (GLMM) identified duration, intensity ratio, and spectral features as the most predictive cues of fricative realization. Principal component analysis revealed clustering patterns along temporal and spectral dimensions, supporting the interpretation of a structured, gradient shift.

These results support a model in which /s/-lenition prompts compensatory fricativization of /b d g/, enhancing the perceptual salience of the coda position. Rather than being optional or idiosyncratic, CVFs emerge as systematic phonetic responses to segmental loss. By identifying clear acoustic correlates of these variants, this study contributes to broader discussions on spirantization, phonetic compensation, and the role of duration in signaling contrast in WAS.

Slides here