TL;DR
This study reveals that while nasal coarticulation in English makes vowels acoustically closer, listeners largely compensate for this effect when vowels are in a word context, leading to similar identification times as oral vowels. However, this compensation isn't perfect, resulting in slightly lower accuracy for nasalized vowels, especially in isolation.
What Problem Does This Paper Address?
When a vowel is spoken next to a nasal consonant (like 'm' or 'n'), it can become "nasalized" – its sound changes. This paper investigates how this nasal coarticulation affects the acoustic properties of vowels and, crucially, how listeners perceive these changes. Specifically, it explores whether this acoustic similarity between nasalized vowels makes them harder to distinguish, and whether the context of a word helps listeners overcome this potential perceptual challenge. This is important because understanding how listeners process such subtle acoustic variations is key to understanding the robustness and efficiency of speech perception.
Methods
The researchers used a combination of acoustic analysis and perceptual experiments.
- Acoustic Analysis: They analyzed the speech of a single native English speaker, measuring the first two formants (F1 and F2) of specific vowels (/a/, /ε/, /æ/, /ʌ/) in both nasal and oral consonant contexts. These measurements were then converted to the "Bark" scale, which better reflects human auditory perception. They calculated acoustic distances between vowel pairs to quantify similarity.
- Perception Study:
- Stimuli Creation: They created speech stimuli from recorded words, including minimal pairs differing only in the nasality of the final consonant (e.g., "bud" vs. "bun"). These words were mixed with Gaussian noise to increase task difficulty.
- Experimental Design: Participants performed a forced-choice identification task. Stimuli were presented in two contexts:
- Lexical Context: The full word was presented.
- Isolation Context: The vowel was excised from the word.
- Measurements: They recorded both identification accuracy and reaction times (RTs) for correct responses.
- Statistical Analysis: RT data was analyzed using a linear mixed-effects model.
Datasets
The study used speech data from one native English-speaking male speaker. The dataset consisted of recordings of eight real words for each of the four target low vowels, forming oral-nasal minimal pairs. Each word was repeated five times, resulting in 160 tokens for acoustic analysis. For the perception study, 192 stimuli were presented, randomized within blocks.
Key Insights and Ideas
- Acoustic Contraction: Nasal coarticulation acoustically brings vowels closer together in the vowel space, meaning they become more similar in sound. This effect is more pronounced for certain vowel pairs.
- Perceptual Compensation: Listeners are generally good at compensating for nasal coarticulation. When a nasalized vowel appears in a word with a nasal consonant, listeners attribute the acoustic changes to the consonant, effectively perceiving the vowel as "oral."
- Context is King: The presence of a lexical context is crucial for this compensation. When vowels are presented in isolation, the acoustic effects of nasalization are not easily attributed to a source, leading to slower identification.
- Effortless Compensation (Mostly): The reaction time data suggests that this compensation happens almost instantaneously and without significant additional processing effort when the vowel is in a word.
- Compensation Isn't Perfect: Despite the speed of compensation, accuracy data reveals that it's not always complete. Some residual effects of nasalization can lead to misidentification, particularly when the vowel is heard in isolation.
Key Results
- Acoustic Distance: Oral vowels were found to be acoustically further apart than nasalized vowels. For example, the distance between /a/ and /ʌ/ was reduced by 0.74 Bark in nasal contexts compared to oral contexts.
- Accuracy: Listeners were more accurate at identifying oral vowels (95%) than nasalized vowels (91%). Accuracy was also higher in word contexts (96%) than in isolation (91%). The lowest accuracy was observed for nasalized vowels in isolation (89%).
- Reaction Times (RTs):
- Overall, responses to nasalized vowels were slower (637 ms) than to oral vowels (593 ms).
- Responses to vowels in words were slower (623 ms) than to isolated vowels (607 ms).
- Crucially: Oral vowels in isolation were faster to identify than nasalized vowels in isolation (p<.001). However, there was no significant difference in identification time between nasalized and oral vowels when they were presented in a word context (p=.4).
Limitations and Future Work
- Single Speaker: The acoustic analysis was based on data from only one native English speaker. This limits the generalizability of the acoustic findings.
- Limited Vowel Set: The study focused on four specific non-high vowels that are prone to nasalization. Future work could explore a wider range of vowels.
- Incomplete Compensation: While compensation is largely effective, the accuracy data suggests it's not perfect. The authors suggest that the residual acoustic effects of nasalization can still lead to misidentification, especially when context is absent. Future research could explore the specific acoustic cues that listeners rely on for compensation and the factors that contribute to incomplete compensation.
- Generalizability of Contextual Effects: The study demonstrated the importance of lexical context. Further research could investigate how different types of linguistic contexts (e.g., surrounding consonants, prosody) might influence the perception of nasalized vowels.
Summary updated