Consonant voicing in the Buckeye corpus

Area of Science:

• Phonetics
• Linguistics
• Speech Science

Background:

• Previous research often claimed English voiced stops are unvoiced in many positions.
• These claims were frequently based on limited speech data and acoustic measurements.

Purpose of the Study:

• To investigate the actual prevalence of phonetic voicing in American English voiced stops and affricates.
• To analyze voicing patterns across different speech positions (initial, intervocalic, final).

Main Methods:

• Analysis of 53,690 voiced stop and affricate tokens from the Buckeye corpus of American English.
• Measurement of vocal cord vibration percentage throughout consonant duration for 39 speakers.

Main Results:

• Contrary to common belief, voiced stops are mostly voiced in final and intervocalic positions.
• Voicing in initial positions is typically below 50%, but a notable subset shows nearly 100% voicing.
• High initial voicing was observed irrespective of interspeaker variation.

Conclusions:

• The study challenges the traditional view of English voiced stops being largely unvoiced in non-intervocalic positions.
• Phonetic voicing in English stops is more prevalent than previously documented, especially in final positions.
• Further research is needed to explain the high prevalence of voicing in some initial stop tokens.