An articulatory analysis of asymmetrically confusable consonants Free

Acoustic similarity and articulatory similarity do not always co-occur; a narrow range of acoustic outcomes can sometimes correspond to a comparatively wide range of articulatory parameters (Stevens, 1989). The articulatory parameters associated with a narrow window of acoustic events occasionally straddle a category boundary—some consonants (e.g., /t/ and /k/ in front vowel contexts) show acoustic (and perceptual) similarity with one another, despite differences in active articulator or place of articulation (Plauché, 2001). This study utilizes a corpus of vocal tract MRI (Sorensen et al., 2017) to relate the spatial dynamics of productions of /p/, /t/, /k/, /f/, and /θ/ to their acoustic properties. This analysis offers insight into why articulatorily dissimilar productions can show increased spectral similarity and how phonetic context conditions this similarity. For each segment production, five video frames were extracted—one frame before constriction release, and four after release. From each frame, a 30-point cross-sectional area function of the vocal tract airway was generated. Vocal tract parameters associated with the spectral properties of frication, aspiration, and stop bursts were analyzed. In phonetic contexts favoring misidentification, confusable segment pairs show greater similarity in vocal tract regions anterior to the primary constriction.