Introduction
While the production of speech sounds varies significantly across English dialects, it is less clear if these variations extend to perception. For instance, North American English speakers primarily distinguish the vowels in "bit" and "beat" using spectral changes (formants) rather than vowel length. This study investigates whether listeners from different global dialects utilize the same acoustic cues when processing a manipulated North American vowel contrast.
Methodology
The study employed a forced-choice identification task using audio tokens synthesized from a Canadian English speaker's voice. Two primary acoustic features were independently manipulated:
- Vowel Quality (Formants F1/F2): Four steps of 50 Mel.
- Vowel Duration: Six steps of 30ms (ranging from 60ms to 210ms).
This resulted in a 24-token grid. Participants included 23 native English speakers living in Tokyo (9 from Australia, 14 from other regions including the UK, Ireland, New Zealand, and Singapore). Results were compared against a baseline of 24 North American listeners from a previous study.
Key Findings: Dialectal Cue Reliance
The research reveals that listeners apply the perceptual rules of their own dialect even when listening to a foreign one. A significant "Australian exception" was observed. In Australian English, the duration difference between tense and lax vowels is much more pronounced than in North American English. Because the North American formants did not align with Australian vowel spaces, these listeners relied heavily on duration as a fallback cue.
Results
The data showed a stark contrast in categorical bias:
- Formant Bias: 79% of North Americans and 78% of "Other" region listeners prioritized formants.
- Duration Bias: 78% of Australian listeners relied primarily on duration.
Regression analysis confirmed these findings. For North Americans, formants were the dominant predictor ($\beta = 0.814$), whereas for Australians, duration was the primary cue ($\beta = -0.716$).
Furthermore, reaction time (RT) data indicated that Australian participants found the North American tokens fundamentally ambiguous. They took significantly longer to classify words compared to other groups, suggesting that their auditory processing for vowels is deeply entrenched and relatively inflexible.
Conclusion and Limitations
The study suggests that listeners do not easily adjust their categorical boundaries to match a speaker's dialect in the short term.
Limitations: The sample size for non-Australian regions was small, necessitating their grouping. Additionally, the expat status of participants in Japan might have subtly influenced phonetic perception.
Future Work: These findings have critical implications for second-language (L2) acquisition. Research must account for the specific dialectal exposure of learners, as it shapes their ability to distinguish novel vowel contrasts.
Summary updated