The acoustic realization of intervocalic /v/ and /v:/ in Italian: evidence of articulatory variability

Abstract This study examines the variation in the acoustic properties of the allophones of intervocalic /v/ and /vː/ in Italian, a typologically unusual case where voiced labiodental fricatives contrast phonemically in length. Through an acoustic analysis of read speech from 18 speakers across three regional varieties (Veneto, Roman, Calabrian), we investigate whether realization patterns are shaped by prosodic (phase position and stress placement) and indexical factors (regional variety and speaker sex). Results indicate that /v/ frequently (47%) surfaces as an approximant [ʋ]. In contrast with the principle of geminate ‘inalterability’, /vː/ also exhibits great articulatory variability, with 36% of tokens showing increased constriction including 17% of all tokens realized as a previously undocumented plosive-like labiodental [b̪ː]. Bayesian modelling reveals that approximant variants of /v/ occur more frequently phrase-finally than phrase-medially and that male speakers are more likely to produce more constricted allophones of /vː/. No consistent effects of regional variety were observed, suggesting that the variation may be widespread across Italy. Furthermore, phonemic length alone significantly predicts duration. Therefore, notwithstanding the enhanced phonetic differentiation between /v/ and /vː/, it seems most likely speakers rely primarily on duration to distinguish the pair.