TL;DR
This paper investigates how subtle, sub-phonemic variations in vowel duration (microdurational variations) influence vowel perception in Estonian, a language with phonemic quantity distinctions. Surprisingly, they found that these microdurational cues do play a role, particularly for vowels that are acoustically similar, suggesting a universal mechanism for enhancing perceptual contrast.
What Problem Does This Paper Address?
In languages like Estonian and Finnish, vowel duration is primarily used to distinguish between short and long vowels (phonemic quantity). This leads to the expectation that vowel quality (e.g., the difference between "ee" and "eh") should be perceived independently of duration. However, previous research on non-quantity languages has shown that intrinsic vowel duration (the natural tendency for certain vowels to be longer or shorter) can influence vowel quality perception, especially when spectral cues are ambiguous. This paper aims to test whether this influence extends to quantity languages, specifically exploring if microdurational variations can enhance perceptual contrast for acoustically similar vowels.
Methods
The study employed a two-stage experimental design:
- Experiment 1: Establishing Category Boundaries: Synthetic vowel continua were created by interpolating between prototype formant values for five Estonian vowel pairs along the close-open axis (/i/-/e/, /y/-/ø/, /u/-/o/, /e/-/æ/, /o/-/a/). These stimuli, all with a fixed duration of 160ms, were presented to 10 native Estonian speakers. The goal was to determine each listener's individual perceptual boundary for each vowel pair, as this is crucial for creating ambiguous stimuli for the next stage.
- Experiment 2: Testing Microdurational Effects: For each subject, three stimuli representing the most ambiguous formant structures at their individual category boundary were selected. These stimuli were then synthesized with varying durations (60ms to 140ms in 20ms steps). Listeners were presented with ABX or BAX discrimination tasks, where they had to decide if a target stimulus (X) was closer to one prototype (A) or another (B). This setup allowed researchers to measure how changes in duration affected the perception of vowel quality for these ambiguous sounds.
Datasets
The study utilized synthesized stimuli based on formant values derived from the Estonian BABEL database [2]. No new datasets were introduced.
Key Insights and Ideas
- Duration Matters, Even in Quantity Languages: Contrary to initial expectations, microdurational variations do influence vowel quality perception in Estonian, even though duration is a primary cue for phonemic quantity.
- Acoustic Similarity is Key: The effect of microduration was most pronounced for high-mid vowel pairs (/i/-/e/, /y/-/ø/, /u/-/o/), which are acoustically closer to each other. For mid-low vowel pairs (/e/-/æ/, /o/-/a/), which have larger spectral differences, duration had little to no perceptual effect.
- Enhancing Perceptual Contrast: The authors hypothesize that microdurational cues act as a "secondary feature" to boost perceptual contrast when spectral differences are small. This helps listeners reliably distinguish between acoustically similar vowels.
- Potential Universal Mechanism: The findings suggest that the use of intrinsic duration to enhance perceptual contrast might be a universal strategy in human speech perception, regardless of a language's phonemic quantity system.
Key Results
- For high-mid vowel pairs, there was a clear and significant correlation between stimulus duration and perception: longer durations led to a higher likelihood of identifying the vowel as the more open one (e.g., /e/ over /i/). Pearson's correlation coefficients were very high (-0.98 to -0.99, p < 0.01).
- For mid-low vowel pairs, stimulus duration did not significantly affect vowel category perception. Pearson's correlation coefficients were low (e.g., -0.46 for /e/-/æ/ and 0.07 for /o/-/a/), with high p-values indicating no significant effect.
- Perceptual distances (measured in Bark) between high-mid vowel pairs were smaller (1.4-2.1 Bark) compared to mid-low vowel pairs (2.9-4.3 Bark), supporting the hypothesis that duration is used to bridge smaller perceptual gaps.
Limitations and Future Work
- Subject Variation: The study noted considerable variation among subjects in their perception scores, particularly at mid-low category boundaries. The context of stimulus presentation (ABX vs. BAX) also had an impact. These variations warrant further investigation.
- Finnish Subjects: The authors plan to conduct similar experiments with Finnish speakers to gather more evidence on the role of microdurational variations in vowel perception in quantity languages.
- Other Factors: The paper acknowledges that other factors might influence vowel quality perception and suggests future work to explore these.
Summary updated