Identifying acoustic cues for English voiced stop consonants with reverse correlation: a pilot study

Auditory reverse correlation is an experimental paradigm that allows researchers to uncover the acoustic cues a listener relies on during an auditory categorization task. The method comprises introducing random fluctuations into a stimulus, relating these variations to participants’ behavioral responses, and thereby identifying which acoustic features are most influential for a given perceptual decision. In a previous study, we applied this approach to seven phoneme discrimination tasks to identify the acoustic cues underlying the perception of French plosive consonants (Carranante, Cany, Farri, Giavazzi, Varnet, 2024, doi:10.1038/s41598-024-77634-w). Six vowel-consonant-vowel stimuli (/aba/, /ada/, /aga/, /apa/, /ata/, /aka/) were presented in seven 1-interval 2-alternative tasks, allowing us to study the perception of two phonetic features: place of articulation and voicing. We present here a pilot study aimed at extending this work to the perception of English voiced stop consonants. Two main changes were introduced to the paradigm of Carranante et al. First, the stimuli were productions of /aba/, /ada/, and /aga/ by a native English speaker, and all participants were native speakers of English. Second, a 1-interval 3-alternative task was implemented and the analysis was adapted to derive kernels from multiple-alternative experimental designs. These preliminary results confirm the critical role of first and second formant transitions (both consonant-vowel and vowel-consonant transitions) for the perception of English place of articulation. In contrast, high-frequency release bursts appear to carry weaker perceptual weights on average.