Abstract:

Traditional models of American English phonology include a rule that coronal stops appear as flaps before unstressed vowels. Recent studies suggest that this flapping is better explained as the result of blending the articulatory specifications for the stops with neighboring vowels. This blending produces compromise articulatory positions for the stops, which cross over the perceptual boundary between stops and flaps. This paper tests this explanation against an x-ray microbeam corpus of American English stops. The occurrence of flapping was determined via transcriptions. Regression analyses indicate that acoustic parameters such as VOT and voicing occurrence better predict transcriptions than articulatory parameters such as tongue position and motion during closure. In addition, articulatory analyses indicate a variation from [t] to flap which is generally consistent with a blending account. However, the contextual influence from the consonants on neighboring vowels and various aspects of articulatory postures in the consonants themselves is more consistent with a reduction of the stop, rather than a blending with neighboring vowels. These results suggest that flapping results from a linguistic convention which allows hypoarticulation of unstressed items and a secondary result of some of these items straying across an acoustic boundary. [Supported by the NSF and NIDCD.]