Predictions of Laryngeal Realism for breathy-voiced stops: Counter-evidence from Bangla

Abstract

The Laryngeal Realism (LR) framework (Honeybone 2005, Beckman et al. 2013) argues that the breathy-voiced stops in the four-way stop systems are specified for multiple features: [voice] and [spread glottis] ([sg]). As a diagnostic of the specified features, Beckman et al. (2013) proposed that the acoustic cues for a specified feature be longer in slower speech. While predictions of the LR framework have been tested on two- and three-way stop systems using this technique, there has not been much on the four-way stop systems. This study presents evidence from Bangla and argues that the predictions of the LR framework apply variably for the breathy-voiced stops; [sg] is found to be the stronger candidate as the specified feature than [voice]. In slower speech, the cue for the breathy voicing (the lag-time) lengthened in the breathy-voiced stops. Contrarily, the acoustic cue for the [voice] feature (the prevoicing) was not affected by speech rate. Mixed-effects models confirmed that the effect of speech rate ...