Intervocalic fricative perception in European Portuguese: An articulatory synthesis study

Abstract

This study examined the conditions under which European Portuguese (EP) intervocalic fricatives are perceived as voiced/voiceless with respect to varying voiced-to-devoiced fricative portions, durations, pharyngeal widths and glottal heights. Articulatory speech synthesis incorporating a sophisticated noise source model was used to produce /aCa/ sequences (C=/f v s z ʃ ʒ/) for two perception experiments (forced choice voiced/voiceless decision). This approach has the main advantage that all relevant articulatory settings can be manipulated independently. The proportion of fricative voicing was varied between fully devoiced to fully voiced; phoneme durations were varied corresponding to natural EP speech. Glottal height and pharyngeal width each varied amongst three different states. The results showed strong interaction between the cues phoneme duration and voicing maintenance. A voiced-to-devoiced ratio of only 25% was enough to guarantee robust perception of fricative voicing. Phoneme duration and place of articulation had a significant effect on listener decisions, but only for voiced-to-devoiced ratios of 25–50%. Vowel duration, pharyngeal width and glottis height had no significant effect. The study provides new evidence for cue-trading in fricative perception. Furthermore, new insights into laryngeal vs. supra-laryngeal gestural coordination are gained which may facilitate the development of fricative models.