Acoustic correlates of breathy sonorants in Marathi

Abstract

Breathy voiced sonorant consonants are typologically rare, more so than other non-modal sonorants (e.g. voiceless sonorants, which are widely attested in language families like Tibeto-Burman and Otomanguean). Similarly, they are more vulnerable to diachronic loss than voiceless sonorants (e.g. in Tibeto-Burman). The acoustic correlates of breathiness in sonorants have not been thoroughly investigated, and a question arises as to whether there is a tie between their acoustics and their typology: does the acoustic encoding of breathiness in sonorants contribute to their typological scarcity? The current study probes this question via instrumental acoustic analysis of breathy and modal obstruents and sonorants in Marathi, an Indic language. Measures standardly used to assess voice quality (F0, H1*-H2*, H1*-A1*, H1*-A2*, H1*-A3*, and Cepstral Peak Prominence) are reported. As expected, breathy voiced obstruents are associated with increased values in the H1-based measures and decreased values in Cepstral Peak Prominence (CPP), a measure which reflects the presence of noise in the signal. Sonorants show the same general trend, with higher H1-based measures and lower CPP associated with breathy than with plain sonorants, but the differences between plain and breathy consonants is greater in obstruents than in sonorants. Specifically, the acoustic correlates of breathy voice are diminished in post-sonorant contexts. It is proposed that phonemic breathy voice is not acoustically encoded as robustly in sonorant consonants as in obstruents, and that this helps explain the typology.